skip to main content
survey
Open access

Enabling Technologies and Techniques for Floor Identification

Published: 07 October 2024 Publication History

Abstract

Location information has initiated a multitude of applications such as location-based services, health care, emergency response and rescue operations, and assets tracking. A plethora of techniques and technologies have been presented to ensure enhanced location accuracy, both horizontal and vertical. Despite many surveys covering horizontal localization technologies, the literature lacks a comprehensive survey incorporating up-to-date vertical localization approaches. This article provides a detailed survey of different vertical localization techniques such as path loss models, time of arrival, received signal strength, reference signal received power, fingerprinting utilized by WiFi, radio-frequency identification (RFID), global system for mobile communications (GSM), long-term evolution (LTE), barometer, inertial measurement unit (IMU) sensors, and geomagnetic field. The article primarily aims at human localization in indoor environments using smartphones in essence. Besides the localization accuracy, the presented approaches are evaluated in terms of cost, infrastructure dependence, deployment complexity, and sensitivity. We highlight the pros and cons of these approaches and outline future research directions to enhance the accuracy to meet the future needs of floor identification standards set by the Federal Communications Commission.

1 Introduction

During the last decade, positioning and localization, tracking, and navigation systems witnessed exponential growth, both in academia and industry. Such broad growth is attributed to the extensive penetration and proliferation of mobile devices, especially smartphones, which provide high-end computing and a wide variety of embedded sensors. Besides, the ubiquity of wireless communication networks has played a key role in the large use of such systems. The use of these systems extends to several domains, including location-based services and on-the-go services like mobile advertising and retail analysis and emergency response services such as E911 calls and other human positioning services during natural disasters.
Positioning is important in both contexts: outdoor and indoor. Outdoor positioning has been an important research area for approximately two decades. Global navigation satellite systems (GNSSs), like global positioning system (GPS), serve as the unparalleled solution for outdoor positioning in the clear sky. However, outdoor positioning is a big challenge for GNSS-deprived areas such as densely built tall buildings in large city centers, canyons, and similar areas where the satellite signals are not available, attenuated by multipath, or absorbed, or the number of visible satellite signals is not enough. Against the outdoor environment, people spend most of their time in the indoor environment, approximately 80% to 90% [72, 133]. Similarly, most cellular connections and data requests come from indoor environments like offices, university campuses, train stations, airports, shopping malls, and so forth. An estimated 70% of calls and 80% of data requests originate from indoor environments. Consequently, recent years have seen exponential demand for precise and accurate indoor positioning technologies.
Various indoor positioning approaches have been devised in response to the wide demand for indoor positioning systems, and several positioning approaches have been presented. For example, radio-frequency identification (RFID), which was earlier used for tracking inventory in warehouses, has been adopted for human positioning and tracking. Similarly, the introduction of IEEE 802.11 paved the way for the wide use of wireless networks for indoor positioning due to its ubiquity. Along with the same directions, other technologies have been developed to enhance positioning accuracy, such as infrared (IR) and ultra-wideband (UWB) positioning. Despite the requirement for additional sensors and hardware for positioning and high installation cost, UWB offers centimeter-level accuracy. As a result, cheap positioning approaches without the dependency on additional infrastructure have been investigated in recent years, such as magnetic field-based positioning systems (MFPS). MPFS utilizes the earth’s natural phenomenon of magnetism, which is ubiquitous and can be leveraged to perform indoor positioning.
Predominantly, indoor positioning technologies that track assets and individuals indoors focus on only the two-dimensional (2D) positioning as many domains such as asset tracking, smart homes, public safety, and so forth require 2D position alone. However, in several domains, emergency response in particular demands an accurate three-dimensional (3D) indoor position given the requirements set by the US Federal Communications Commission (FCC) for emergency response services. In 2015, the FCC announced tough requirements for network operators to provide an indoor position with a horizontal accuracy of 50 meters for 40% to 80% of the emergency calls, which are to be ensured within 2 to 6 years [32]. Also, a provision for the vertical (floor-level) position has been approved, which will be provided within 6 years. On March 15, 2019, the FCC proposed a vertical location accuracy metric of plus or minus 3 meters relative to the handset for 80% of indoor wireless 911 calls. “The Commission tentatively concluded that such a location accuracy metric within three meters above or below the phone would be sufficiently accurate to identify the caller’s floor level in most cases and would be technically feasible under the time frames established in the Commission’s Enhanced 911 rules” [41].
To meet the future requirements of 2D/3D positioning, position-enabling technologies need to be rigorously tested in the field, including the floor of the user’s current position. Floor determination (vertical positioning) has become an important part of indoor positioning due to the prevalence of modern multifloor residential and office buildings. Users’ reliable mobility information is the key to emergency response services, wherein the vertical position helps locate and track the users easily in indoor environments. So, vertical positioning or floor determination has become the key element of the user’s position for E911 calls, and stringent requirements have been set by the FCC for network operators to provide such information. While many of the existing indoor localization technologies can be extended easily, especially those that rely on received signal strength (RSS) observations, to provide vertical position/floor information, several other technologies, such as MFPS, can be adapted to provide floor information. In addition, the introduction of densification of microcells in 5G and results reported from upcoming 6G that uses terahertz frequency have opened new possibilities for vertical positioning.
Before starting the discussion of different vertical localization technologies and techniques, Table 1 (given in Appendix A) provides a list of the acronyms used in this article. In addition, the following concepts are provided to clarify the scope of this study.
3D localization or positioning is to determine the location of an object or pedestrian in terms of x, y, and z coordinates, where x and y indicate the position in 2D coordinates. At the same time, z may refer to the height of the object/pedestrian or the altitude of its location, as shown in Figure 1.
Fig. 1.
Fig. 1. An illustration of 2D, 3D, and vertical positioning. 2D positioning provides x and y, and 3D positioning provides x, y, and z (height of phone from ground) coordinates. The vertical positioning (floor identification) provides x, y coordinates and floor of a device or user.
Vertical localization is the process of estimating the floor number in a multistory building where the pedestrian is currently located. Floor identification, floor determination, floor detection, and vertical localization are often interchangeably used in the literature.

1.1 Overview of Existing Indoor Localization Survey Papers

Since a rapid and growing attraction has been witnessed during the past decade due to the wide proliferation of mobile devices and the inception of applications such as LBS, a large number of techniques have been presented. As a result, a plethora of survey papers can be found in the literature that describe the fundamental characteristics of various technologies and evaluate the pros and cons of techniques that leverage those technologies [47, 54, 55, 67, 73, 81, 125, 140, 141]. Similarly, a great deal of survey papers can be found that discuss such techniques and lay the directions for future research for indoor positioning and localization [5, 15, 20, 33, 35, 42, 74, 75, 87, 96, 136, 148].
Position tracking technologies for user indoor position in construction environments are characterized in [67]. The effectiveness of wireless local area network (WLAN), UWB, and indoor GPS in terms of dynamic indoor position tracking is evaluated and compared. A survey of wireless sensor networks (WSNs) is provided in [125] with a compact discussion of positioning and localization approaches of active and passive targets. Mainly approaches and systems exploiting RSS are investigated, and recent trends in WSN imaging are discussed. [54] discusses pedestrian dead reckoning (PDR) approaches by dividing them into inertial navigation systems (INSs) and step-and-heading systems (SHSs). Specifically focusing on the PDR, the working mechanisms of zero velocity update (ZUPT), threshold-based step detection, and inertial measurement unit (IMU) sensors such as the accelerometer, gyroscope, and magnetometer are elaborated.
Survey [140] discusses the positioning approaches with respect to mobility. Sensors and approaches for measuring mobility are presented, and their influence on enhancing localization accuracy is envisioned using IMU sensors and RSS-based approaches. A survey of calibration-free indoor positioning systems is initiated in [55], which emphasizes emerging fingerprinting solutions to reduce the cost and labor of wardriving fingerprinting solutions. Evaluation of positioning approaches is performed based on scalability, robustness, security, complexity, and latency. [5] provides a survey of UWB indoor positioning technologies and covers different approaches of UWB positioning such as angle of arrival (AoA), time of arrival (ToA), time difference of arrival (TDoA), and RSS for line-of-sight (LOS) and non-line-of-sight (NLOS) cases. [33] provides insight into indoor positioning and localization approaches targeting mass-market applications. Such techniques and technologies are discussed concerning infrastructural cost, scalability, and provided position estimation accuracy. Reliability of indoor positioning, availability of indoor maps, and privacy issues are studied in [20] with respect to indoor LBS. Positioning approaches are discussed in terms of the market segment of LBS that relies on them. Zafari et al. [148] present a comprehensive discussion of several indoor localization techniques such as AoA, time of flight (ToF), round time of flight (RToF), RSS, and channel state information (CSI) with a focus on the working principles of different technologies such as WiFi, RFID, UWB, and Bluetooth.
Ashraf et al. [15] provide a comprehensive survey of indoor positioning and localization approaches that take advantage of smartphone built-in microelectromechanical system (MEMS) sensors, including WiFi, Bluetooth low energy (BLE), accelerometer, gyroscope, and magnetometer. Survey paper [74] reviews the evolution of indoor positioning and localization approaches with a particular focus on vision-based techniques, including scene recognition methods to help enhance positioning accuracy. Obeidat et al. [96] discuss indoor localization technologies that leverage wireless technology. The survey covers GNSS, inertial, magnetic, sound, optical, and radio-frequency-based indoor localization approaches such as AoA, ToA, RSS fingerprinting, and scene recognition. [60] discusses indoor positioning approaches that do not require a fingerprinting phase to locate the user. Both 2D and 3D positioning approaches are elaborated with a focus on WiFi, BLE, vision, geomagnetic field, PDR, UWB, RFID, and visible light-based approaches within the context of triangulation, simultaneous localization and mapping (SLAM), and crowdsourcing techniques. Similarly, a survey of horizontal positioning approaches regarding the fusion-based approaches is presented in [49]. It covers 2D and 3D positioning approaches based on WLAN, ZigBee, global system for mobile communications (GSM), visible light, UWB, and WSN technologies and includes the discussions of RSS, ToA, TDoA, and AoA, yet floor identification approaches are not covered. Despite plenty of survey papers about indoor positioning and localization approaches, none of these papers discuss vertical positioning, as shown in Table 2 (given in Appendix A), except for [75], which has dedicated one section for vertical positioning. Therefore, a survey dedicated to vertical positioning techniques and technologies is a task of significant importance. For this purpose, the working principles of each technology and procedures adopted in each approach or group of approaches are introduced for novice readers. A comprehensive survey of existing technologies is provided concerning their use for vertical positioning in the indoor environment, and their pros and cons are highlighted.

1.2 Why Survey of Floor Identification Is Needed

Floor identification holds specific and critical importance for first responders and emergency response teams. In addition, the following aspects urge a separate survey paper on floor identification approaches:
The FCC put in place the requirements for commercial mobile radio service (CMRS) providers to give the floor level. “CMRS providers also must deliver z-axis information in Height Above Ellipsoid. Where available to the CMRS provider, floor-level information must be provided in addition to z-axis location information. (47 CFR Section 9.10(i)(2)(ii)” [41].
Despite the clear definition for z-axis and floor level from the FCC, existing survey papers use them interchangeably and no dedicated survey is available that covers floor identification approaches.
Floor-level information is different from the z-axis information, as shown in Figure 2. Predominantly, the survey papers cover the z-axis localization approaches and floor identification is ignored, except for [75], which includes one section for floor identification approaches.
Fig. 2.
Fig. 2. 3D positioning versus the floor identification. A 3D positioning approach provides users position in x, y, and z (user of device height from ground) coordinates, while floor identification provides x and y coordinates with the floor level of a user or device.
Since the floor identification approaches are different from 3D localization approaches and no separate survey is available for floor identification approaches, a dedicated survey is desirable.
Research works in the literature use 3D positioning and floor identification interchangeably, which creates a fuss; however, the positioning approaches performing 3D positioning are very much different from the floor identification approaches. Although some 3D approaches do cover floor identification, it is different from 3D localization. Floor identification refers to the user’s floor number where he/she is located now, while the 3D position provides the x, y, and z axes, where the z-axis refers to the user’s height or the height of the device used for localization. The height of the user/device is estimated with reference to the floor.

1.3 Key Contributions of Current Survey

(1)
This study presents a comprehensive survey of indoor positioning and localization techniques and technologies with respect to floor identification. For this purpose, the literature between 2000 and 2021 has been selected, emphasizing the more recent approaches.
(2)
Several recent innovative technologies are discussed for their suitability to perform floor identification to meet the FCC’s stringent vertical position requirements regarding E911 calls. Inherent limitations, as well as the opportunities of such technologies, are extensively studied.
(3)
Due to the wide proliferation of smartphones and the increased availability of embedded MEMS sensors in smartphones, approaches leveraging smartphone sensors are covered concerning their usage for floor identification in multifloor environments. Smartphone sensors’ working principles and pros and cons are intensively discussed to provide a complete understanding to the users.
(4)
The prevalence of machine learning and deep learning approaches leads to data-driven solutions to enhance indoor positioning and localization positioning accuracy. Current challenges to adopting deep learning approaches for floor identification positioning and the potential of such approaches are also covered.
(5)
Emerging solutions to resolve the problems related to existing floor identification solutions and enhance their efficacy are given special consideration. Crowdsourcing solutions, hybrid solutions, and sensor fusion approaches are discussed in detail.
(6)
With the developments in cellular technology over the past few years, a large number of cellular-based approaches have been developed involving GSM and long-term evolution (LTE). In addition to discussing these approaches, probable enhancements from the upcoming fifth generation (5G) and sixth generation (6G) are thoroughly covered.

1.4 Classification of Literature

Classification of the literature is very important as it groups the studies under different categories and makes it easy for researchers to analyze and grab recent trends in a particular technology or domain. We have divided the existing literature on floor identification into four categories:
(i)
Wireless signal-based studies: All studies that utilize wireless communication technology for floor identification are discussed under this category. It includes path loss models, WiFi-based fingerprinting solutions, and studies utilizing cellular signals for fingerprinting.
(ii)
Barometer-based hybrid solutions: It incorporates studies that utilize barometer-based solutions, either as the sole solution or combining other technologies with barometers such as inertial measurement unit, Bluetooth, accelerometer, WiFi, map information, and PDR.
(iii)
Machine learning-based solution: The studies that utilize a machine learning model to determine the current floor of the user are placed into this category.
(iv)
Delay-sensitive approaches: All solutions that consider latency for floor identification are discussed in this category. For the major part, this category contains solutions utilizing LTE, 5G, and signal flight time-based floor identification approaches.
While the possibility of including one study in multiple categories is possible, for simplicity, we did not consider domain overlap. The classification of the literature is based on the procedure adopted for floor identification and not the infrastructure. For example, although a wireless network is necessary for barometer-based solutions, such works are placed under the barometer category.

1.5 Organization of the Survey

Keeping in view the applications of floor identification and available solutions, the floor identification solutions are divided into five broad categories. For the division of existing literature into different categories, the emphasis is placed on the procedure adopted for floor identification and not the infrastructure involved in floor identification. For example, wireless communication-based solutions include those approaches that utilize WiFi RSS as fingerprints, path loss models for discriminating between different floors, and so forth. Similarly, barometer-based solutions involve approaches involving a barometer as the main sensor that determines the current floor of a user.
Section 2: Wide deployment of WiFi access points (APs) indoors makes it an attractive solution for horizontal positioning and floor identification due to its ubiquity and pervasiveness. Section 2 discusses approaches that utilize WiFi signals for floor identification, including WLAN, and approaches using path loss models for cross-floor attenuation estimation. In addition, fingerprinting has become the most commonly used approach for indoor positioning, both horizontal and floor identification. Due to its being simple and easy, it has been adopted with several technologies such as barometer, WiFi, cellular signals, and so forth. Section 2 elaborates on the techniques that take advantage of existing infrastructure to devise fingerprinting solutions for floor identification.
Section 3: Several scenarios require high accuracy for floor identification including robotics, smart factory environments, and safety and emergency situations. Hybrid approaches tend to show better performance for indoor positioning, whereby the weaknesses of one technology are compensated by the other. We discuss barometer-based hybrid solutions for floor identification in Section 3 such as barometer+accelerometer, barometer+WLAN+map information, and camera with barometer. The barometer is widely deployed to provide floor information in indoor environments. Moreover, several other methods are adopted for floor identification, such as Bayesian models, neural networks (NNs), and support vector machines (SVMs). Techniques that leverage barometer data alone for vertical positioning are also discussed here.
Section 4: The generation of a large amount of data over the past few years led to the inception of the data-intensive solution to provide resilient and high-accuracy solutions. Recently, deep learning approaches have been widely used for positioning and localization with high accuracy. Deep learning-based floor identification solutions are discussed in Section 4. Such solutions include using deep neural networks (DNNs), Bayesian models, deep autoencoders (DAEs), and convolutional neural network (CNNs).
Section 5: Many real-life scenarios are time-sensitive and positioning delays can cause large human and financial losses, e.g., safety and emergency response scenarios. Recent developments in fast-speed cellular technology have offered new dimensions to solve the indoor localization problem with low latency. In this regard, LTE and 5G are leveraged to provide high-accuracy positioning and localization, and such solutions are covered in Section 5.
Section 6: Limitations associated with current floor identification technologies and their challenges are discussed in Section 6. Such discussions consider the technology’s inherent limitations, sensing devices, and the approach adopted for positioning.
Section 7: Future directions and probable working trends are discussed here by providing insights on improving the positioning performance of floor identification technologies.
Section 8: It concludes the survey article.

2 Wireless Communication-based Solutions for Floor Identification

This section covers the floor identification approaches that utilize wireless signals. Such solutions utilize already deployed wireless networks for floor identification. Similarly, since the majority of fingerprinting approaches also utilize the signals from wireless networks, they are also discussed in this section.

2.1 Floor Identification Using Path Loss Model

The floor identification using the wireless signals can also be achieved using path loss models. Traditionally, the known positions of the APs are used to calculate the path loss models. For this purpose, signal attenuation is utilized. Attenuation refers to the phenomenon of weakening of signal strength during transmission. The attenuation depends on several factors including the transmission distance, frequency, and obstacles in the transmission environment. Using the difference between the attenuation caused by walls and the floor, a specific floor can be identified. Figure 3 shows the flow of a typical path loss-based floor identification scheme.
Fig. 3.
Fig. 3. Flow of typical path loss-based floor identification.
Radio frequency is widely used for horizontal positioning, although the same is not true for floor identification. Recently, the wide deployment of 802.11 wireless network APs made them the first choice for indoor positioning and localization for academia and industry. The wide proliferation and pervasive availability of 802.11 indoors and the common use of wireless network-enabled mobile devices laid the foundations of ubiquitous positioning. Besides horizontal positioning, WiFi-based approaches have been utilized for floor identification as well. Such approaches are leveraged to provide both coarse and fine position information. For example, Ahmad et al. [3] present an approach, “becognition,” to identify the current floor of a pedestrian using the WiFi signals. The concept of signal coverage area (SCA) of a wireless AP is utilized to find the coarse location of the pedestrian, and an election algorithm is used for vertical positioning. SCA represents the area covered by a particular WiFi AP. SCAs collected from wireless APs are used to train and test vertical positioning performance in a five-floor building whereby a correct floor detection accuracy of 89% is reported in a dense environment. However, sparse deployment of wireless APs is expected to reduce this accuracy.
Wireless signals are used for floor identification based on the phenomenon that attenuation of signals is significantly different across vertical and horizontal directions on account of the materials used for the floor and walls. Figure 4 shows three important concepts: LoS, intra-floor attenuation, and inter-floor attenuation. In a wireless environment, where there is no obstacle between the transmitter and receiver, it is an LoS scenario. Obstacles, both fixed and moving, can hinder communication leading to the weakening of signals, called attenuation. When such obstacles attenuate the signal on the same floor, it is intra-floor attenuation. For example, temporary partitions of boards and cupboards can lead to intra-floor attenuation. The roofs/floor, on the other hand, are often made of concrete and cause inter-floor attenuation. The inter-floor attenuation is substantially high compared to intra-floor attenuation with the exception when the floors are made of wood. On average, floor attenuation is significantly higher than that of walls. To determine the path loss, the multi-wall-floor (MWF) path loss model is used.
\begin{equation} L=L_0+10n log(d)+\sum _{l=1}^I \sum _{k=1}^{K_{wi}}L_{wik}+\sum {j=1}^J \sum _{k=1}^{K_{fj}}L_{fjk}, \end{equation}
(1)
where \(L_0\) is path loss at 1 m, n is the power decay index, d is transmitter-receiver distance, \(L_{wik}\) and \(L_{fjk}\) are attenuation to wall type i and the \(k^{th}\) wall and attenuation to floor type j and the \(k^{th}\) floor, and I and J indicate the number of wall types and floor types, respectively.
Fig. 4.
Fig. 4. Path loss phenomenon for wireless-based vertical indoor positioning. Based on the concrete and roof thickness, a wireless signal is attenuated across different floors. This attenuation is used along with the APs’ known location to identify the current floor of a user.
A path loss model-based floor identification (ID) approach is presented in [113] where the floor ID and RSS values determined through the path loss model are used to identify specific floors. Simulation results demonstrate approximately 100% accuracy provided that AP deployment is sufficiently dense. Path loss model-based floor identification approaches, especially the deconvolution-based ones, tend to show similar performance to that of the fingerprinting approach with the advantage of reduced database size. For example, Shrestha et al. [114] compare the performance of several deconvolution-based path loss models with fingerprinting techniques for floor identification in the multifloor and multibuilding setups including university buildings and small and large shopping malls. Experimental results indicate that a comparable floor identification performance is achievable for university buildings with the highest accuracy of 84.50%. However, the same is not true for large shopping malls, where path loss models show poor performance. Similarly, a path loss-based approach is discussed by [39] where wireless sensor nodes are deployed in a multifloor environment for positioning. In an area of 10×10 m, four beacon nodes are used for floor identification. Results demonstrate that a root mean squared error (RMSE) of 2.5 m for floor positioning is possible, given the a priori information of interior structure and wall distribution.
Gupta et al. [50] present a path loss-based method that utilizes a maximum likelihood estimation approach to find the current floor of a pedestrian indoors. A probabilistic approach is followed where the probability of a pedestrian at a particular floor is calculated given the information of APs and received RSS values. Additionally, the concept of invisible APs is exploited to improve the positioning performance. Invisible APs are those APs that physically exist in a given environment but are not detected in the current scan. Results show that the false detection rate of the approach is up to 1% only, which can further be reduced to 0.33% if pressure sensor data are used with the maximum likelihood algorithm. A modified COST231 path loss model has been utilized in [8] to generate a dynamic radio map for a multifloor environment. Several parameters of the path loss models are recalculated, such as floor attenuation factor (FAF), path loss at 1 m distance from the AP, and random shadowing factor. Different radio maps are generated, each with a different number of APs for the same floor to analyze the influence of the number of APs on the positioning accuracy. In a five-floor building, floor identification accuracy of 98% can be achieved with a radio map generated using 24 APs when 15 test points are used on each floor.
The concept of sum-RSS is introduced in [90], where the signals from reference nodes (RNs) are used to find the floor of a pedestrian. The sum-RSS approach is based on the summation of RSSs from working RNs. To handle signal fluctuation, a confidence interval comparison is carried out using the mean of RSS summation. RNs are installed in a multistory building with four RNs on each floor, and floor identification is performed using the real-time RSS values from these RNs without the need to gather the fingerprints beforehand. To improve the positioning accuracy, outliers are removed, and summation for each floor is calculated using the received RSS, IDs, and location of RNs from each floor. Maximum accuracy of 98.67% can be achieved in the cases where at least four RNs are available. However, reducing the RNs to two would yield the highest accuracy of 91.33%. Nevertheless, the model requires prior knowledge of indoor infrastructure, including walls and other separations, and the location of installed RNs. Yi et al. [146] propose a multistory differential (MSD) algorithm to achieve floor identification where free space path loss (FSPL) and multistory path loss (MSPL) models are used. For this purpose, pairs of anchors are installed on different floors, and information on floor height and floor material can help estimate the change in MSPL for each floor. Results indicate that by deploying multiple pairs of anchors, a higher floor identification can be achieved.
Table 3 (given in Appendix A) provides a comparative overview of floor identification approaches based on path loss. While existing works report floor identification accuracy, the reliability and importance of reported accuracy vary with respect to indoor space, the complexity of indoor infrastructure, the number of floors, and so forth. There is no standard rule for reporting floor identification accuracy, which makes it very difficult to compare the performance of approaches utilized in different indoor settings. To provide a more fair comparison and report the confidence of the reported accuracy, we have formulated the following equation:
\begin{equation} Confidence=1-\frac{{\it 100-Reported accuracy (\%)}}{{\it Reported accuracy (\%)} \times (n_f-1)}, \end{equation}
(2)
where \(n_f\) indicates the number of floors for which the accuracy is reported.
Equation (2) aims at providing a standard measure to report confidence for floor identification accuracy, and its value varies between 0 and 1. It infers that 90% accuracy for a three-floor test bed and a three-floor test bed are not the same. Using the sample accuracy of 90% for the three and five floors yields 0.9444 and 0.9722 confidence scores, respectively.

2.2 Fingerprinting-based Solutions for Floor Identification

Floor identification is very important, not only to guide people to navigate easily in multistory buildings and malls but also to save people in case of emergency situations like fire, earthquake, and so forth. However, adding a dedicate setup for floor identification is expensive. Instead, it is more attractive to utilize the existing communication infrastructure for positioning and localization, which decreases the deployment cost. Nonetheless, it increases the dependence on the infrastructure. Such systems follow a fingerprinting technique where real-time/current features including RSS, magnetic data, and channel information are mapped against the pre-collected features of an indoor area.
The idea of fingerprinting refers to associating locations called RPs in an environment with unique characteristics, also known as fingerprints. Typically signal features are used as fingerprints, whereas a mobile device can be used to gather single or multidimensional features at RPs. These features should exhibit uniqueness that can be used to discriminate different places [65]. Among several possible candidates for fingerprints, multipath structure and the RSS have been regarded as the most common and accurate fingerprints for horizontal indoor positioning [19, 98]. Depending upon the indoor structure of buildings, power-delay profile can be different at different places, provided the carrier frequency is larger than 500 MHz. Similarly, due to the large deployment of WiFi APs for indoor environments, the RSS can be easily gathered and used as a fingerprint.
Fingerprinting includes an offline and an online phase. The former, also called the training phase, involves the fingerprint collection on dedicated RPs, prior to positioning. RPs can be random or in a grid form. On the other hand, the online phase, also called the testing phase, involves positioning the pedestrian using the real-time data from his/her mobile device and comparing it with the fingerprint database through a fingerprint-matching algorithm. One potential challenge for fingerprinting-based solutions is fingerprint collection that is carried out using wardriving. In wardriving, one or multiple people use mobile devices to collect fingerprints at RPs. This process is repeated for all RPs on one or multiple floors where floor identification is needed. Depending upon the size of the floor and number of floors, this process may take days or even weeks. The process carried out in a fingerprinting scheme is illustrated in Figure 5. Owing to the popularity of fingerprinting-based positioning, a large body of positioning approaches can be found; however, such approaches are predominantly focused on horizontal positioning. Having said that, such solutions can be easily extended to perform floor identification.
Fig. 5.
Fig. 5. A schematic diagram for WiFi-based fingerprinting. Blue circles on each floor represent the reference points where the fingerprints are collected. The data, traditionally RSS from the user in real time, is matched with these reference points to find his/her current floor. The location of the user is associated with the nearest fingerprint.

2.2.1 Fingerprinting Using WiFi APs.

WiFi fingerprinting solutions can be grouped into five categories with respect to the approach used for positioning: traditional fingerprinting, clustering approaches, solutions aiming at dynamic environments, heuristic-based solutions, and fingerprinting with pedestrian activity sensing. Traditional fingerprinting solutions using the RSS values received from WiFi APs become laborious and computationally complex when the size and number of floors grow large, resulting in poor performance. The size of the fingerprinting database grows larger with the increase in the size of indoor space and RPs. Matching real-time data with large databases requires a longer time and negatively affects flood detection accuracy. Clustering approaches focus on dividing the RSS into various groups based on the coordinate or RSS similarity, thereby reducing the computing time and enhancing accuracy. Dynamic solutions focus on leveraging real-time information to update radio maps and add additional people’s movement features in the positioning area. Heuristics-based approaches make use of empirical statistical characteristics of RSS values to enrich the feature vector. Deploying additional information on pedestrian activity through smartphone sensors such as accelerometer and gyroscope helps provide an accurate short-term position that enhances the positioning accuracy.
Fingerprinting solutions tend to be laborious and time-consuming. Expert surveyors wardrive the area to collect the fingerprints during the offline phase, making it a task of significant human resources for sizeable multifloor buildings. Wardrive is the process of collecting fingerprints at ground-truth points for the whole area intended for positioning. Wang et al. [130] present an approach to reduce the labor involved in labeling the training data. Although the data are collected on all the building floors, labeling is carried out for only one floor. The authors use the information that despite exhibiting different APs and RSS vectors, the floor structure for university campuses is almost similar. AP visibility across several floors is used to find the correlation between the data from different floors. Results show that co-embedding the data from different floors can be used for automatic labeling. Two models named ”the nearest floor algorithm” and ”the group variance algorithm” have been proposed in [7] to solve the floor identification problem using the WiFi signals. The former is a modified form of the famous k nearest neighbor (k-NN); however, unlike traditional k-NN, which is used for horizontal positioning, the modified algorithm is used to find the neighbors for positions across different floors. For training and testing, WiFi fingerprints are collected containing media access control (MAC) and RSS information. Due to the fluctuations in RSS due to dynamic factors, additional parameters are added to the group variance algorithm, such as range, variance, and RSS availability. Experimental results in a five-floor building show a floor detection accuracy of 86% and 72% with the nearest floor and group variance algorithms, respectively. A linear discriminant analysis (LDA)-based floor identification approach is proposed in [86] that leverages the one-versus-one (OvO) rule combined with a majority voting to predict the final position. WiFi RSS values collected at RPs during the offline phase are used for training the classifier. Experiments are performed to analyze the influence of the number of deployed APs on the positioning accuracy using four and six APs for three-floor buildings. Using six evenly distributed APs, a 99% floor identification accuracy is obtained.
While the use of RSS values alone is the predominant approach, deploying additional features from the received signals tends to show better results. For example, confidence interval sum-received signal strength (CIS-RSS) is used in [89] for floor detection using the RSS from WiFi APs in a multifloor environment. The sum of RSS shows the different distributions for each floor and can be utilized for positioning. Online RSS values are used to estimate the CIS-RSS using the RSS values from installed APs on each floor, and a 100% accuracy can be achieved at a 95% interval. Similarly, Han et al. [52] use a feature vector containing both RSS and MAC addresses of APs on each floor. In addition, the repetition of MAC addresses at each RP is estimated for different floors and recorded in the database as an additional feature. Later, clustering is performed based on MAC addresses for each floor to reduce the computational time and improve accuracy. The feature vector is compared with the already built database during the online positioning to find the most probable floor. Results show that using a cluster size of three, floor identification accuracy of approximately 90% can be achieved. The use of MAC address joined with its RSS across different floors helps to reduce the search space. In larger buildings with a few floors, an AP may only be visible to a few floors, and recording the MAC address for AP along with RSS and floor information can provide better results, as reported in [52].
Fingerprinting is a time-consuming process and becomes tedious where large buildings with tens of floors are involved. One possible way to reduce the time and effort involved in fingerprinting is crowdsourcing. Crowdsourcing involves multiple users in the data collection process, which are later integrated into a single database. Since it uses non-expert surveyors, data collection is substantially reduced, compared to traditional fingerprinting. Khaoampai et al. [66] introduce a similar concept, called ”fingerprint self-learning,” where a new fingerprint (one that is not present in the database) is added to the database with the pedestrian’s current floor. The user’s current floor is determined using the last known floor and activity trace, which is achieved using the accelerometer and barometer data from the smartphone. Using the proposed approach FloorLoc-SL, an 87% accuracy can be obtained on the smartphone.
Clustering is another potential solution to improve positioning accuracy in floor identification schemes. Clustering involves dividing the data into different groups based on the similarity of intra-cluster samples and differences of inter-cluster samples. For large indoor buildings with tens of floors, the size of the fingerprint database can grow exponentially, increasing the processing time and causing latency for real-time positioning systems. Clustering can play an important role in clustering fingerprints and reducing the search space, thus improving the response time and enhancing the positioning performance. Clustering for fingerprinting solutions can be based on 3D coordinates or AP based [18]. A clustering approach for floor identification is proposed in [34], which utilizes the concept of penalized logarithmic Gaussian distance (PLGD). The concept of penalizing involves various penalties for clusters showing poor performance. Experiments are performed in two buildings to analyze 3D and AP-based clustering performance. Results demonstrate that both 3D and AP-based clustering reduce the response time. However, 3D coordinate clustering enhances the positioning performance. Using the proposed approach, an accuracy of 97% can be obtained for floor detection. Along the same directions, a clustering-based fingerprinting approach is presented in [107], where a modified K-means clustering approach is used along with the weighted centroid localization (WCL) algorithm. In the weighted approach, the position of a device is computed as the weighted average of visible AP positions. Clustering is carried out for RSS fingerprints collected for each floor to use only the cluster heads (CHs) for localization. An average floor identification accuracy of 88.00% can be achieved involving two university buildings, one shopping mall, and one office environment. Despite the lower accuracy, fingerprint database size and consequent computational time of position estimation are reduced substantially.
Tiwari and Jain [120] propose using Fuzzy means clustering for WiFi data and statistical features from a barometer to perform floor detection. Initially, different buildings are identified using the WiFi majority rule, followed by floor identification through a barometer. Several statistical features from barometer data are used, including root mean square (RMS), Kurtosis, Skewness, peak-to-peak, Crest factor, shape factor, margin factor, and impulse factor. Floor identification accuracy of 98.21% is achieved with a complexity of \(O(1)\). An unsupervised clustering (K-medians and Kohonen layer) is adopted by Campos et al. [28] to group the collected RSS by imposing architectural constraints. Initially, the online collected RSS values are processed using principal component analysis (PCA) to reduce data dimensionality. A backpropagation artificial neural network (ANN) is used to find the floor of the mobile device. Using a window length of 25 with a combination of unsupervised and supervised clustering and database correlation methods (DCMs), 97% vertical positioning accuracy can be achieved.
Besides clustering, employing probabilistic approaches helps to increase the robustness of fingerprinting solutions. Owing to the lower processing and storage available on mobile devices, robust and lower computational complexity solutions are desired. Razavi et al. [108] analyze the performance of four robust algorithms with gathered RSS, including weighted centroid localization, nonlinear joint parameter estimation and trilateration, linear joint parameter estimation and multilateration, and deconvolution-based path loss estimator. Experiments are carried out in two university buildings, a six-floor mall, and a shopping center. Results reveal that a probability of 85.80% is achievable with a weighted centralized localization algorithm.
The major drawback of the fingerprinting solutions is updating the fingerprint as the APs may be replaced, be added, or experience presence failure. AP presence failure during the online RSS collection has a significant influence on floor detection accuracy. Maneerat et al. [88] present a solution to AP presence failure using a robust mean of sum (RMoS), which is based on the mean of the summation of the strongest RSS and uses the confidence interval comparison [59]. Within a five-floor environment, 95% accuracy is achieved with 40% RN failure. Despite the low cost and high accuracy of fingerprinting solutions, fingerprint databases become outdated due to changes in the indoor infrastructure including the addition of furniture, vending machines, computer equipment, and so forth. Similarly, RSS varies with different types of mobile devices. In addition, the influence of human mobility also affects RSS values. To counter the effects of human mobility and device heterogeneity, an adaptive indoor positioning system, DIPS, is presented in [9]. It comprises a dynamic radio map generator, RSS certainty, and people location component. Considering these elements, it integrates the people’s presence into the RSS map. Results suggest that incorporating this information into the radio map helps to obtain a 99% floor identification accuracy.
A heuristic approach based on the collected RSS for WiFi signals is presented in [23] to estimate the floor of a pedestrian. Four characteristics from the WiFi signals for a multifloor environment are used to determine the floor information: floors with a maximum count of signals, floor with maximum signal strength, floor with maximum average signal strength, and floor with maximum signal strength variance. Such observations are attributed to the attenuation of wireless signals through ceilings and floors. A maximum of 99.97% accuracy can be achieved using the proposed Locus system. Table 4 (given in Appendix A) provides characteristic features of WiFi and cellular-based fingerprinting solutions.
Owing to the complexity of indoor infrastructure and dynamic environments, positioning accuracy suffers from fingerprinting solutions. One way to enhance the performance of fingerprint-based approaches is combining the activity recognition from inertial sensors of mobile devices [105]. Providing information on an indoor structure like the position of stairs and elevators, main entrance, and height of floors further elevates the floor identification performance. Sun et al. [118] present a multistage floor identification approach. It involves a discriminative floor model using Fisher’s linear discriminant (FLD), pedestrian state recognition, and WiFi fingerprinting matching. A floor identification accuracy of 94.3% can be achieved with the proposed model.

2.3 Cellular-based Fingerprinting Solutions

Besides the WiFi-based fingerprinting, the signals from the cellular network can also be utilized for floor identification. Using the principle of offline and online phases of WiFi fingerprinting, the signals from different network operators are collected on RPs to be later used for online positioning. The process of fingerprint collection for cellular-based fingerprinting is similar to WiFi fingerprinting; the only difference is that in cellular fingerprinting, cellular signals are collected. Varshavsky et al. [124] present a GSM-based fingerprinting approach, SkyLoc, for floor identification. A GSM cell contains a number of physical channels concerning traffic expectations. Broadcast control channels (BCCHs) are used to formulate the fingerprints offline. Experiments performed in 9, 12, and 16 buildings show that a floor identification accuracy of up to 73% is achievable using GSM fingerprints on smartphones with 97% accuracy within two floors. Additionally, it is found that the selection of a proper feature set is very important to enhance the performance of positioning approaches. Similarly, a GSM-based fingerprinting approach is used in [97], where the cell IDs and RSS of up to 35 channels are stored as a fingerprint. Experiments are performed using RSS from a single strongest cell, 6 strongest cells, and 35 GSM channels, and performance is compared with the WiFi RSS fingerprint. Results demonstrate that using the RSS from 35 GSM channels, the floor identification accuracy of 89.08%, 97.01%, and 93.69% is obtainable for university, research lab, and residential areas, respectively. WiFi performs better in university and research lab areas with reinforced concrete roofs. However, the performance of WiFi is substantially degraded in a residential area, which has wood roofs. Gorak et al. [45] propose a three-step approach using GSM signal readings from seven strongest cells involving point localization using one measurement, central tendency filters, and MLP. Results show that a higher number of visible BTSs is associated with higher floor detection accuracy. A 64% accuracy of floor identification is achieved using the RSS from the seven strongest cells with an MLP aggregation scheme.
Wang et al. [132] present a cellular network-based fingerprint scheme where the received signal code power (RSCP) is used and the fingerprint database is gathered using the ray-tracing technique at Huawei Technologies Co., Ltd. Each measurement contains RSCP from seven cells and user equipment (UE) locations. Cell matching degree is used to enhance the positioning performance. A 0 m error is reported for floor identification at the 67th percentile and 7 m at the 90th percentile. The code division multiple access (CDMA) networks can also be used for positioning. However, the change in the transmission power to handle network fluctuations affects the signal intensity and limits the functionality of RSS measurement. Instead, Rehman et al. [122] leverage the signal delay to make the fingerprint database-based positioning approach, called CDMA indoor localization system (CILoS). Compared to the GSM fingerprinting, the CDMA approach shows better results with an 87% floor identification accuracy.
Similar to GSM, frequency modulation (FM) signals can also be used for fingerprinting-based floor identification. FM solutions are robust and less susceptible to dynamic environments including human mobility, multipath, and fading. Chen et al. [31] present an FM fingerprinting approach that uses RSS of FM signals. FM signals are further augmented with additional information from the physical layer, including signal-to-noise ratio (SNR), multipath, and frequency offset. Study shows that the FM and WiFi errors are uncorrelated and these technologies can be combined to obtain higher positioning performance. Experiments in multifloor buildings using FM and WiFi signals show an accuracy of 83%.

3 Barometer-based Hybrid Approaches for High-accuracy Floor Identification

3.1 Barometer-based Floor Identification

Predominantly, existing indoor positioning and localization approaches determine the 2D position of objects and pedestrians in indoor environments and consider that floor determining has already been solved. However, in real scenarios where pedestrians move in complex multifloor indoor environments such as universities, shopping malls, airports, and so forth, floor information plays a pivotal role in locating pedestrians. Owing to the importance of floor information, the FCC added the floor position requirement to E911 calls. As per [41], the floor position accuracy of 3 meters above or below (plus or minus 3 meters) the handset for 80% of wireless E911 calls has to be provided by network providers. For providing the floor location of a pedestrian indoors, of the several possible solutions, the barometer is of significant importance due to its wide use. Pressure sensors, also called barometers, witnessed significant progress during the past few decades regarding size, cost, and accuracy. However, the breakthrough came in the form of MEMS sensors in centimeters. Additionally, the large and rapid proliferation of smartphones containing a wide variety of MEMS sensors like barometer, lux meter, accelerometer, and gyroscope made it possible to utilize mobile devices for positioning.
Today, every smartphone has an embedded barometer that can provide real-time atmospheric pressure. Consequently, the use of barometer sensors has been widely adopted for altitude estimation and floor identification. A pressure sensor can be used to determine the altitude of a pedestrian as the pressure varies with altitude over a given time. Barometric pressure is the exerted force on a surface per unit area by the weight of air above that surface due to the atmosphere of Earth [80, 142]. This pressure is decreased with respect to an increase in altitude. The relationship of altitude and pressure is given by
\begin{equation} h=\frac{((\frac{P_o}{P})^{\frac{1}{5.257}-1})\times (T+273.15)}{0.0065}, \end{equation}
(3)
where h is the altitude, T is the temperature, P is the current atmospheric pressure, and \(P_o\) indicates the sea-level pressure and is given by 1,013.25 hecto Pascal (hPa) [22, 27]. Pa is the international system of units (ISU) for measuring pressure and is equal to 1 N/m\(^2\). However, in meteorology, hPa is used, which corresponds to 100 Pa. Additionally, the relationship defined by the International Civil Aviation Organization (ICAO) can be used to infer the altitude, which says that an 8.7 m change in altitude is associated with a 1 mbar change in pressure, and air pressure is dropped by 11.2 mbar for every 100 m increase in the altitude.
Barometer-based floor identification is in common use due to its simplicity, easy adaptation, and low cost. The use of barometric altimetry for floor identification can be achieved through a single-device system or a dual-device system, where the former refers to successively measuring the barometric altitude of two different vertical positions. In comparison, the latter means obtaining the barometric altitude of two positions concurrently. The single-device-based systems are deployed more often, and Figure 6 shows one such system where a single barometer is placed on a floor, called the ”reference floor.” For floor identification, the barometric data from the smartphone can be compared for altitude difference with respect to the reference floor. A floor identification approach is presented in [56], where a reference pressure sensor is used. For estimating the current floor of the pedestrian, the pressure value from the installed barometer and its floor information are used to find the difference in barometric pressure, which is later used to determine the altitude from the reference floor. This process, however, requires the floor height information of the buildings where the positioning takes place. Along the same lines, a smartphone barometer-based approach is proposed in [71], where a reference barometer sensor is installed on the third and second floors of two buildings. Instead of using the barometric pressure to estimate the altitude, the height of each floor from the reference floor is used to infer the barometric pressure, which is then used to compare with the real-time value measured from the smartphone. Experiments indicate that an 85.71% accuracy can be achieved using the smartphone barometric data and height information of floors.
Fig. 6.
Fig. 6. A single barometer-based floor identification system. A reference barometer (RB) is used on a known floor for measuring the atmospheric pressure. The reading is updated periodically to represent the recent reading. The atmospheric pressure from the user phone is measured and compared with that of RB. The change in the atmospheric pressure is used to calculate user height which can be used to estimate the user’s current floor.
Ye et al. [144] present a floor identification system, B-Loc, that can be used to locate the user at specific floors in real time. Barometric readings are recorded for different floors when the pedestrian travels up and down inside a building. Activities of changing floors are identified using change in altitude, as well as using the accelerometer data. Different traces of barometer data are preprocessed, and noise is removed before moving the data to a central database. Afterward, clustering is applied for a specific time period \(t_0\) to segment the data for different floors using the clustering using representatives (CURE) algorithm [48]. For locating the user, the real-time barometer data is compared with the stored data. However, the data in the database may become insignificant as the pressure changes with changes in environmental conditions over time. The saved data is updated with reference barometer values from the reference floor periodically to solve this problem. In addition, the knowledge of the initial floor of the pedestrian is used to update the saved barometric data. Experimental results in a 10-floor building indicate that flood identification accuracy of 98% can be obtained using the proposed B-Loc approach.
Using the reference barometer installed on a specific floor indoors is simple and efficient yet requires a connection to a central server where the barometric pressure value is updated periodically. For floor identification, the positioning device needs to communicate with the server to get the reference data, which introduces latency. An alternative to the reference barometer is to provide the pedestrian’s current floor information, which can be tedious if the user has to do it manually every time he/she enters a building. A solution to avoid manual entry is presented in [109], where the entry point of the pedestrian is estimated using the RSS fingerprint data of WiFi APs. Prerecorded RSS values are used to determine possible entry points to determine the current floor where the pedestrian is located, and then barometer data are used to track his/her floor. Furthermore, going-upstairs and going-downstairs events are detected to mark the floor transition, increasing the positioning accuracy. Floor transition indicates the user moving from one floor to another floor. Accuracy scores of 91.6% and 91.1% are reported for going up and downstairs and entry-level detection, respectively. In contrast, 80.0%, 85.0%, and 91.7% floor detection accuracy are reported using Samsung Note 2, Xiaomi 5s, and Huawei Mate 8, respectively, with an average accuracy of 85.6% for the proposed approach.
Xia et al. [135] propose using the multireference barometer floor positioning (MFBP) approach for floor identification, where multiple reference barometers are installed to enhance positioning accuracy. Using multiple reference barometers helps to reduce the difference in the measured value from the smartphones and determine a threshold for each floor. For removing the inherent noise in the data, an outlier detection algorithm is proposed as well. In addition to that, the study uses a compensation coefficient for the offsets from different smartphones. It is observed that despite different measured barometric values, the difference in these values is approximately constant. Experiments performed in various buildings using Samsung S III and Xiaomi Mi3 indicate that using a longer time window (5 seconds) can yield 98.62% and 92.89% accuracy for Mi3 and SIII, respectively. Although floor identification accuracy of 95.48% is possible with a smartphone placed in the pocket, higher accuracy is possible if the smartphone is handheld by the pedestrian. Similarly, various indoor environments tend to affect floor identification accuracy. Different accuracy scores of 95.75%, 90.48%, and 97.19% are reported for the office building, airport terminal, and shopping center, respectively, which indicates that buildings with large indoor areas show relatively low positioning accuracy with barometers.
Although floor detection using a barometer yields accurate short-term results, floor transition methods enhance the accuracy. It reduces the latency as position estimation can be avoided. A floor transition approach is proposed in [123] that leverages the barometer patterns for pedestrian moving-up and -down events in a multifloor environment. Experiments are carried out following different paths and using both stairs and elevators to determine the performance of the proposed approach. Results suggest that a 0.82 accuracy score for floor transition can be achieved. Transition detection can significantly improve the performance of floor identification approaches and does not require indoor settings, floor information, and surveys. Similarly, Muralidharan et al. present a floor transition approach in [92] and investigate the influence of using heterogeneous devices on floor detection. The proposed approach uses several different devices and performs floor identification in seven buildings with 5 to 18 floors. Experiments reveal that single-floor transition can be detected with 99.54% accuracy. However, as the number of floor transitions increases, the detection accuracy is reduced. Similarly, the floor transition detection accuracy is affected by using different elevators in various buildings due to the varying speeds of elevators. A summary of barometer-based floor identification approaches is provided in Table 5 (given in Appendix A).
Barometer-based floor identification approaches become meaningless for indoor environments where floor heights are irregular as the floor change is traditionally determined with respect to average floor height. Better positioning performance has been revealed by studies that utilize more than one technology, often called ”hybrid approaches.” Such fusion is a matter of convenience, availability of positioning sensors, and the complementary nature of technologies used for position. A schematic diagram of a hybrid system for floor identification is shown in Figure 7. It can be seen that the data from a wide variety of indoor installed and smartphone-embedded sensors can be utilized to enhance positioning accuracy.
Fig. 7.
Fig. 7. An indoor scenario with multiple technologies for sensor fusion. The barometer provides altitude but may contain erroneous readings, which can be compensated by an accelerometer sensor that can determine the user’s climbing-up or -down activities. The gyroscope and magnetometer provide direction information that can be used to enhance floor identification accuracy.

3.2 Barometer and Inertial Measurement Unit

Xu et al. [139] present a hybrid approach where the data from a barometer are combined with IMU sensors. Accelerometer data are used for stair detection and finding the number of landings for a pedestrian on the stairs to approximate the number of steps. To find the direction and avoid false landing detection, gyroscope data are used. Using a Bayesian network-based algorithm fed with barometer, accelerometer, and gyroscope data, an average of 99.36% floor identification accuracy is achieved in a dormitory, office, and commercial building. Incorporating additional data from smartphone-embedded sensors helps to provide fine-grained location information. A hybrid system is put forward in [115], where the data from the accelerometer and gyroscope are used to detect users’ activities of the escalator, elevator, and stairs used for moving up and moving down. Based on ZUPT, these modules are used to calculate the vertical distance of the pedestrian. GPS and other beacons are used to detect pedestrians entering a building. In addition, elevator riding, walking, and standing are detected using vertical acceleration and geomagnetic data. Of the 80 predictions for floor identification, only 8 have an error of 0.5 m, thereby providing 90% correct floor identification. Even the error of 0.5 m is not considered a floor-level error as the average height of the floor is regarded as 1.5 m to 3 m.
Along the same lines, Yu et al. [147] devise a positioning system by integrating the activity detection modules for stairs-up and stairs-down cases using the data from smartphone sensors. The entry point, defined as the floor where a pedestrian enters a building, is detected using a fingerprinting approach using WiFi. For managing the smartphone heterogeneity, historical data from three smartphones are used to find an offset value for different smartphones, which helps mitigate the influence of pressure data differences across smartphones. Using three different smartphones, a floor identification accuracy of 85.60% is obtained. A floor identification approach is devised by Ye et al. [145], where the focus is to leverage only the smartphone-embedded sensors and propose a system called FTrack. Instead of depending upon GSM, WiFi, or GPS infrastructure support, the study relies on smartphone accelerometer data to capture user encounters and trails. For encounter detection, RSS of Bluetooth is used to find the presence of multiple pedestrians in close proximity on the same floor. The Naive Bayes (NB) model is used to identify pedestrians’ states of moving up and down elevators or stairs. Mapping data are used to track the floor information, which contains the historical data of pedestrians’ movement inside a building. An accuracy of 97% is reported with 3-hour historical data in a 10-floor building.

3.3 Barometer, Bluetooth, and Accelerometer

Jeon et al. [63] leverage the data from multiple sensors, including Bluetooth received signal strength indicator (RSSI), accelerometer, barometer, and magnetometer, using the smartphone. Step detection and step length estimation are performed with accelerometer and magnetometer data, wherein the accumulation error is minimized with Bluetooth beacons installed on ceilings. Change in the floor position is estimated with the barometer data. With known floor height information, the accuracy of 95.9% is realized with the proposed approach. A data fusion approach is utilized in [152] to perform floor identification using BLE and MEMS data. Using BLE data, a fingerprint database is built by performing outlier detection and filtering, thus reducing the noise in BLE data. The floor identification is estimated using the barometer and geographical position information. The speed and direction of pedestrian walking are determined by using accelerometer, gyroscope, and magnetometer data with extended Kalman Filter (EKF). Results prove that the positioning error is smaller when both BLE and MEMS data are used with an RMS of 0.35 m on average.

3.4 Barometer, WiFi, and Map Information

Attributed to the higher positioning accuracy of hybrid approaches, study [128] performs sensor fusion for floor identification. A barometer is installed on the first floor, which serves as the reference barometer for finding accurate floor information. The data from the barometer is fused with building information to determine the floor of a pedestrian. In addition, a pattern-matching algorithm is adopted for WiFi data collected during the offline phase. The fusion of WiFi, barometer, and building information helps alleviate the positioning error while pedestrians stay in the elevator. RSS-based floor identification shows an average of 8% false floor detection, while fusion proves to be 100% correct in a three-floor building. Similarly, Wang et al. [129] perform multi-information fusion for floor identification, including pressure data, topological building map, and RSS values from WiFi APs. A topological map used in the study provides information about the position of walls, doors, and floor height for refining the positioning accuracy. Using a Bayesian fusion model tends to show better results than NN. Additionally, incorporating bias estimation and correction proves to show 100% accuracy.

3.5 Barometer, WiFi, and Accelerometer

Ramana et al. [106] propose a hybrid approach using the WiFi RSS and smartphone accelerometer data. RSS data are gathered during the offline phase, where the data are preprocessed to remove outliers. Positioning is carried out separately from RSS and accelerometer data. Accelerometer data determine the number of steps and time required to traverse between two consecutive floors, and this information is used for refining the positioning accuracy. Using four RNs, an 88.7% accuracy is achieved, while the accuracy can go up to 99.1% with six RNs in a seven-floor environment. In a similar fashion, a two-module positioning approach is introduced in [4], which comprises floor identification and floor change detection phases. Three models are trained on the WiFi RSS data, including k-NN, BP network, and K-means. For floor change detection, WiFi and barometer are fused. Incorporating floor change detection for elevators and stairs enhances the positioning accuracy by up to 99%. A WiFi and accelerometer-based positioning approach is presented in [93], where the initial floor is determined using WiFi RSS data. Experiments are performed using different orientations and holding styles of smartphones to validate the efficacy of the proposed scheme. A 98% accuracy is achieved, while positioning errors are experienced near landmarks such as hollow areas and stairs.
On account of the time involved in the wardriving of the environment for WiFi fingerprints, such solutions have been regarded as laborious, error prone, and time-consuming. Alternative solutions are devised to reduce the labor and time of fingerprinting. A WiFi fingerprint and accelerometer traces-based floor identification approach, F-Loc, is implemented in [143], where the fingerprints are automatically collected via crowdsourcing while a pedestrian is moving indoors. Movement patterns are classified using the accelerometer data of the smartphone, such as moving upward or downward. Change of a single floor and the number of floors changed during a specified time can be found using the time and vertical acceleration in the elevator. Experiments conducted in a shopping mall, office building, and hotel show an accuracy of 98.8% for floor identification.

3.6 Barometer and WiFi

A crowdsourcing-based WiFi fingerprinting approach is adopted in [112], where a system called BarFi is presented that leverages the WiFi and barometer data. A two-phase clustering method is utilized, which performs barometer-based hierarchical clustering and WiFi-based K-means clustering to train the RSS fingerprint floor map. Collected fingerprints contain both WiFi and barometer data, and clustering is performed to reduce the complexity and enhance performance. BarFi achieves an accuracy of 96.3% using three different smartphones in nine-floor buildings.
ZeeFi is presented in [46], which does not require the process of fingerprinting or floor-level heights, often required by a majority of the vertical positioning systems. It comprises three components, ground floor detection, floor association, and floor identification. Ground floor detection is carried out from the number of visible GNSS satellites and light intensity. A user must pass the entrance/exit of the building where the data for GNSS, WiFi, barometer, and light sensor are recorded. Pressure data collected at the ground level serves as the reference point to detect different floors and associate collected RSS values to a specific floor. A stacked autoencoder is trained on the collected fingerprints to determine the floor of a pedestrian. Experiments performed in a multifloor building indicate that a 98% accuracy is achieved.
A Bayesian approach is adopted by Zhao et al. [150] for utilizing the WiFi APs to determine the floor levels with uneven structure. Initially determined floor-level information is utilized to calibrate the pressure data for specific floors. This initialization and calibration help to overcome the misclassification using the WiFi data in hollow areas. Using WiFi data alone, accuracy scores of 92.3% and 90.1% are achievable for single and two devices, respectively. By incorporating additional factors such as slow pressure drift calibration, floor update with quick pressure change, altitude compensation for floor transition, and so forth for the barometer data, the positioning performance of barometer-based positioning is enhanced. The proposed hybrid approach, HYFI, can achieve an accuracy of 99% when the data from WiFi and the barometer are combined with single device experiments.
Li et al. [79] introduce the concept of RSS profile-based floor identification, where the RSS fingerprinting approach is integrated with the barometer data. The geometry for wireless signals from APs is leveraged to weigh the probability for different floors. Sensor fusion is performed using EKF with RSS, magnetometer, accelerometer, and gyroscope data. Several constraints are used for enhancing positioning performance, including motion prediction, heading constraint, height constraint, velocity constraint, and location constraint. Compared to the RSS positioning alone with an accuracy of 83.8%, the proposed hybrid approach can obtain an accuracy of 97.00% in a five-floor environment. Similarly, a Monte Carlo Bayesian inference algorithm is proposed in [53] to fuse the data from RSS and the barometer. Allan Deviation is utilized to characterize and remove the non-stationary errors in the barometric data. Fusion is performed using log-normal and uniform models to analyze their performance. On average, accuracy scores of 97% and 96% can be achieved for log-normal and uniform models, respectively, for experiments involving three different buildings. However, the performance is degraded for areas with no AP availability, such as washrooms, stairways, and so forth, and the fusion model relies totally on barometric pressure.

3.7 Barometer, WiFi, and Bluetooth

Owing to the impact of environmental changes on the pressure data, Ichikari et al. [57] decompose the atmospheric pressure into three components of change in altitude, global variations, and the device used for data collection. Instead of using the absolute value of pressure data, the relative change in the data is leveraged for altitude estimation. Similarly, for alleviating the influence of changes in environmental conditions like temperature, humidity, and so forth, periodic updating of these conditions helps increase the positioning performance. Finally, an offset value is determined using the historical data from multiple smartphones to mitigate the impact of device heterogeneity. A reference barometer and BLE beacon data from the indoor environment are also used to elevate the positioning performance. An accuracy of 91.84% is achieved with the proposed approach.
Lohan et al. [84] analyze the characteristics of three wireless networks, including 2.4 gigahertz (GHz), 5 GHz, and BLE, to perform floor identification. For wireless signals, already available APs are used, while BLE beacons are installed on walls and ceilings for positioning. Instead of using a fingerprint matching approach, a probabilistic approach is adopted where the weighted centroid approach and path loss model-based approach are adopted. The path loss model with floor loss and path loss model without floor loss are used. Experimental results suggest that the performance of the 5 GHz sample is poorer than both 2.4 GHz and BLE data. The highest accuracy of 96.87% is achieved using the path loss model with floor loss by 2.4 GHz signals, while the accuracy scores of 5 GHz and BLE are 59.82% and 89.86%, respectively.

3.8 Barometer, WiFi, Map Information, and PDR

PDR is an inexpensive positioning technique that is specifically used for indoor positioning. PDR provides a short-term relative position and serves as a complementary approach to elevate the performance of the positioning approaches [16]. It involves the use of an accelerometer, magnetometer, and gyroscope to locate a user. The accelerometer is used to measure the relative distance from the initial (or previous) position, the magnetometer provides direction information, and the gyroscope can provide a change in orientation. For the PDR-based position, smartphone-embedded sensors can be utilized. The use of signals limits the scalability of the positioning approaches and relies on the indoor infrastructure. On the other hand, using floor-plan building features increases scalability and avoids additional installations [95, 134]. Jaworski et al. [62] integrate WiFi RSS, barometric data, inertial sensors, and building information to perform floor identification. Indoor map information improves the positioning accuracy, such as walls, closed rooms, the position of stairs, and floor height information. A 3D particle filter is contrived that incorporates the information from these sources and provides the floor information. Results demonstrate that the 3D PF approach produces better results. In a similar fashion, Lo et al. [83] fuse the wireless signals, inertial sensing data, and building information through an SPF. A partitioning approach is introduced to divide the indoor space into floors, and each floor is further split into different logical units connected by passages. A signal weighting mechanism is devised to control signal drifting by incorporating user walking activities on plane floors, stairs, and elevators. The barometer is used for horizontal and stair motion detection, while the elevator motion is captured using accelerometer data through a curve-fitting method. Compared to 83.00% and 89.00% accuracy scores of nearest neighbors in signal space (NNSS) and Monte Carlo PF, respectively, the proposed SPF can achieve a floor identification accuracy of 94.00%.

3.9 Smartphone Camera and Geomagnetic Field

With high-definition embedded cameras in modern smartphones, its usage for positioning has largely been increased. Despite the computational complexity of image processing approaches and the lack of smartphones to carry out such tasks, images can be sent to a web server for further processing. A similar mechanism is adopted by Ashraf et al. [13], where a smartphone camera is utilized for floor identification. Images are captured at specific RPs on each floor for training a custom-defined CNN during the offline phase. During the real-time positioning, smartphone camera images are sent to a web server where the trained CNN can relate the captured images to a specific floor. Experiments conducted in a three-floor environment indicate a floor identification accuracy of 91.04%.
Barometer and wireless signal-based floor identification approaches largely depend on the installed/available infrastructure, and the performance is affected if part of that infrastructure is not available. Recently, positioning approaches rely on pervasive phenomena such as the geomagnetic field, which is a natural phenomenon and does not require additional infrastructure. A fingerprinting approach for geomagnetic field data is adopted in [17], where the data from smartphone sensors are used, including the magnetometer, accelerometer, and gyroscope. Geomagnetic field data are leveraged to locate the pedestrian on a specified floor. At the same time, inertial sensors are used to track the pedestrian’s activity of moving up and down stairways to enhance the positioning performance. Smartphone orientations are also tracked to transform the geomagnetic field data using a Naive Bayes model. Experiments performed in three buildings of the university and shopping mall indicate that 89.24% accuracy can be achieved using the geomagnetic field data, which can further be improved to 91.02% by incorporating a floor change detection module. An overview of hybrid floor identification approaches is given in Table 6 (given in Appendix A).

4 Machine Learning-based Approaches for Floor Identification

Considering the recent developments in the field of machine learning approaches, their use for indoor positioning tasks is not surprising. Primarily targeted for image processing tasks, deep neural networks achieve better results than traditional state-of-the-art approaches. Deep learning approaches can produce superior results if provided with a large amount of data for training. Given the amount of data required for training, these models can obtain stable and rich characteristics to solve the problem of low accuracy [44]. Deep learning approaches can be adopted for floor identification and extended on WiFi RSS-based solutions. The only difference for deep learning approaches lies in utilizing deep learning models such as CNN, DNN, and autoencoder (AE) to be trained on RSS fingerprint instead of k-NN and hierarchical models, which are dominant in traditional fingerprinting schemes. A schematic diagram for the deep learning framework is shown in Figure 8, where a DNN is trained on the WiFi RSS data from a multifloor building. A DNN has input, hidden, and output layers. Input layers take the input data and pass it on to the hidden layer. Each neuron in a hidden layer may be connected to one or many of the neurons in the coming layer. The number of hidden layers and number of neurons in each layer vary with respect to model complexity. The neurons carry weights, which are refined with each iteration to match the input to the prediction. The output layer is responsible for making the final prediction and the number of neurons in the output layer corresponds to the number of classes that the model predicts.
Fig. 8.
Fig. 8. A deep learning framework for floor identification using WiFi RSS. Training data containing different parameters of wireless signals like RSSI and AP addresses are fed into a deep learning model for training. The trained model is later used to predict the user’s floor using the data collected in real time.
Given the accuracy and ease of adaption for deep learning models for WiFi RSS, several approaches can be found in the literature for floor identification. Nowicki and Wietrzykowski [94] utilize a DNN with AE for estimating the floor identification of a pedestrian in the indoor environment. Owing to the difficulty of reduced higher-level features for machine learning, a stacked autoencoder (SAE) is fed with raw RSS measurements that reduce the input data’s dimensionality during unsupervised training. Diversified AE models have experimented with different feature vectors of 520, 256, 128, and 64. The best performance for floor identification comes from SAE 256-128-64 with classifier 128-128, which achieves 92% accuracy. Despite achieving a comparable accuracy to state-of-the-art approaches, the study [94] does not consider the hierarchy of the building, and loss and accuracy are calculated over flattened building-floor labels. Kim et al. [70] resolve this issue by proposing a custom structure with SAE and a classifier without any hidden layer. Different weights are assigned for buildings and floors to train the DNN. Experiments are performed in three different buildings using WiFi RSS data with the proposed DNN model. Experimental results indicate that DNN yields a floor identification accuracy of 97.18% when validated on 253 fingerprints. Similarly, Alitaleshi et al. [6] make use of a hierarchical extreme learning machine (ELM) with the WiFi fingerprinting technique. The collected fingerprints comprise RSS from pre-installed APs in a five-floor building. With the WiFi collected data, a test accuracy of 98.13% is reported for vertical positioning. Study [69] leverages a DNN model based on WiFi fingerprint and the hierarchical nature of floors for scalable floor identification. For handling scalability, DNN architecture based on multilabel classification is proposed. Results indicate that a 91.18% floor identification accuracy can be achieved on average.
A floor identification system, TrueStory, is proposed in [37] that utilizes several characteristics of deployed APs indoors. It comprises a power equalizer module and filtering and normalization modules, where the former module is used to find RSS offset values for APs owing to their power difference. Using offset values instead of absolute values helps to increase the positioning performance. The latter module filters out the weakly heard APs or outliers based on their deployment position. A multi-layer perceptron (MLP) is used to train and test the filtered RSS data where multiple weak learners are combined to make a strong model. Floor identification accuracy is 91.08%, 86.00%, and 84.00% for shopping mall and university buildings one and two, respectively. Song et al. [116] take advantage of CNN with WiFi fingerprinting data to perform floor identification. For this purpose, an SAE is used to extract the most appropriate features from the WiFi RSS data, which are later used to train one-dimensional CNN. SAE helps to reduce dimensionality while preserving the necessary feature information. A dropout layer between SAE and CNN is introduced to avoid overfitting. The highest positioning accuracy of 96.03% is achieved using the UJIIndoorLoc dataset, while the accuracy scores for UTSIndoorLoc and Tampere are 94.57% and 94.22%, respectively.
Combining the output from multiple classifiers tends to show better performance than individual classifiers. Consequently, several ensemble models are proposed for indoor positioning [11]. Qi et al. leverage an ensemble of ELM with WiFi RSS data for floor identification in [103]. Considering the complexity of the feature vector due to dense APs in the indoor environment, PCA feature extraction is used for dimension reduction on the training data. During the online positioning phase, real-time RSS is processed using PCA before the prediction with the trained model. Experiments performed in a seven-floor environment show an accuracy of 98.00%. A summary of approaches using deep learning models for floor identification is given in Table 7 (given in Appendix A).

5 Delay-sensitive Applications for Floor Identification

Several applications of floor identification are delay sensitive, where delays can lead to huge human and financial loss. For example, the emergency and safety response actions require time-sensitive information about the user’s current floor. For example, localizing robots and other autonomous machinery in a smart factory system requires millisecond response time. In addition, such services require high accuracy. Envisioned 5G and beyond networks with dense infrastructure can be utilized to provide low latency and high accuracy.
Besides WiFi RSS data, cellular data can also be utilized to perform floor identification. Contradictory to WiFi data containing RSS values alone, cellular data contains several parameters such as reference signal received power (RSRP), reference signal receiving quality (RSRQ), and signal to inference plus noise ratio (SINR). Zhang et al. [149] utilize measurement report data containing RSRP, RSRQ, and SINR for LTE cellular data with a deep autoencoder-long short-term memory (DAE-LSTM) model. DAE is used for removing noise and feature extraction, while LSTM is utilized for floor identification. Data collected during the offline phase is fed into the proposed model for training. Testing is carried out in a five-floor building, and DAE-LSTM provides 93.21% accuracy compared to 81.12%, 90.83%, and 91.87% from k-NN, PCA-SVM, and DAE-SVM models, respectively.
The ToA concept refers to the estimation of the absolute time of a signal transmitting a transmitter and reaching a remote receiver. The distance between the transmitter and receiver is calculated using the measures’ times and signal speed, which is usually the speed of light. Each distance calculation yields a circle, and the intersection of circles determines the location of the device or the user. For a 2D location, ToAs from at least three RPs are needed, while a 3D position requires four ToA measurements. While ToA is a commonly used localization approach, it requires perfect synchronization of transmitters. On the other hand, TDoA does not require the time when the signal is transmitted from the transmitter; it only requires the time when the signal has arrived and the signal traveling speed [111]. TDoA calculates the difference in the ToA of two RPs. Using the known position of the beacons, the distance is converted to hyperbola, and the intersection of hyperbola shows the user’s location. The observed time difference of arrival (OTDoA) approach makes use of evolved node Bs (eNBs) to estimate TDoA from one serving and two neighboring cells in the case of 2D positioning. OTDoA is a downlink positioning introduced in LTE Rel-9 [43]. The ToA from multiple eNBs is calculated by the UE to calculate the OTDoA. Each OTDoA determines a hyperbola, leading to the intersection of hyperbolas from multiple eNBs. At least three OTDoAs are needed to determine the 2D position and four OTDoAs for the 3D position.
Owing to the deployment of LTE heterogeneous networks, floor identification techniques based on enhanced-cell identification (E-CID) and radio-frequency pattern matching (RFPM) have been presented. In addition, TDoA approaches are also studied for floor identification. Similarly, OTDoA is also used where positioning reference signal (PRS) from at least three eNBs are required so that the time difference \(\tau\) can be estimated. The working principle of both TDoA and OTDoA is the same except for the calculation of \(\tau\), where eNB is regarded as the serving node while the rest are called neighboring nodes. A typical environment for TDoA and OTDoA is provided in Figure 9.
Fig. 9.
Fig. 9. A schematic diagram for OTDoA-based positioning. The OTDoA approach requires signals from at least three eNBs to calculate the time of arrival. ToAs are subtracted from the ToA of a reference eNB. Each TDoA is then used to draw a hyperbola, and the intersection of the hyperbola shows the user’s location.
Peral-Rosado, et al. [36] deploy an experimental LTE femtocell network to obtain a vertical position in a two-floor indoor environment. Four LTE femtocell BSs are deployed on each floor with a floor height difference of 3.24 m. A system bandwidth of 10 Hz and carrier center frequency of 2,625 MHz are used. Omnidirectional antennas are placed close to windows at 1.2 m from the ground. To evaluate the performance of E-CID for floor identification, the femtocell BSs are synchronized. Experimental results indicate that a floor identification accuracy of 69.88% is achieved using the E-CID approach with LTE femtocells.
A TDoA-based floor identification approach is contrived by Lee et al. [77], where a novel derivation of the hyperbola equation is presented. Three TDoA-based models are investigated for floor identification, including the TS-based algorithm, Chan and Ho’s (CH) algorithm, and the two-step-based algorithm. However, for floor identification, time information from at least four eNBs is required or when they have the same height. The proposed method, on the other hand, utilizes three types of 2D TDoAs. Using the data from only three eNBs, with different heights, a positioning accuracy of 0.44 m is achieved at the 67th percentile.
With reference to the superior performance of OTDoA-based horizontal positioning approaches using LTE cellular signals, an OTDoA-based novel floor identification approach is presented in [30]. A compressive sampling-based channel estimation method is proposed that leverages the sparsity of the wireless channel. A transmit beamforming scheme is introduced to estimate the elevation AoD. In the end, ToA and AoD are combined to determine the vertical position of the pedestrian in a multifloor indoor environment. Simulation results suggest that using only three eNBs, an error of less than 5 m is obtained at 90%. Contrary to [77], where eNBs are used at different heights, this study uses eNBs at the same heights.

6 Challenges in Floor Identification Approaches

6.1 Challenges of Wireless Communication-based Approaches for Floor Identification

WiFi-enabled floor identification is widespread due to utilizing the already deployed WiFi network in the indoor environment. As a result, such approaches are easy to adapt, simple, and cost-effective as no new infrastructure is required. However, radio signal-based floor identification such as WiFi, Bluetooth, and so forth is highly susceptible to building materials, wall separations and partitions, and floor plans. These factors make such approaches vulnerable to random noise, path loss, multipath interference, shadowing, human body absorption, and so forth. [21]. Consequently, buildings with uneven and irregular floor structures experience high positioning errors [150].
Multifloor environments face the problem of resembling signal space where similar vertical areas on adjacent floors and cross-floors may come across this problem. In certain situations, signal attenuation for single-floor separation may be very similar or even lower than the same-floor signal attenuation, which makes it very difficult to discriminate the adjacent floors [127]. Similarly, the performance of wireless approaches for residential areas suffers severe degradation as the roofs are made of wood, and the attenuation of walls may be very similar to that of the roof, which makes it almost impossible to discriminate between a wall and floor [97]. Using wireless signals for floor identification experiences poor performance for uneven floors and hollow structures usually found beside stairs and elevators [150].
Several factors affect the positioning performance for path loss models, including the underlying path loss models used for radio map generation with several of the parameters used in the model. In addition, path loss models require complete information regarding the indoor infrastructure, IDs, position of the APs, nature of walls, thickness of intra-floor walls and separations, inter-floor thickness, nature of concrete used for floors, FAF, and wall attenuation factor (WAF) [8, 85], which may vary from one environment to another. Also, such information may not be available for every environment due to safety reasons, like airports and bus terminals.
Dynamic environments pose a real threat to the accuracy of floor identification systems, such as the placement of objects in the indoor setting, human mobility, and door movement. Even the opening and closing of doors are associated with a change in the positioning performance [14, 24]. Human mobility highly influences wireless signal propagation, and distance calculation for path loss models is severely affected [137]. Also, the temperature changes in the indoor environment can influence the quality of wireless signals [138]. The noise associated with the propagation channel model introduces measurement uncertainty [100]. One important point that needs to be considered regarding wireless communication-based floor identification is that such approaches utilize existing infrastructure, which is primarily deployed for communication. Consequently, using an existing wireless communication network does not provide the desired results for floor identification.
Despite the wide usage of fingerprinting approaches, especially leveraging the already available WiFi network, several factors degrade their performance. Although the widespread proliferation of smartphones with built-in WiFi sensors is seen as a potential tool to achieve a ubiquitous position, a large range of smartphone companies and models is indeed a great challenge to achieve such goals. The heterogeneity of embedded WiFi sensors in smartphones is prone to receiving various RSS values at a particular position. RSS magnitude for the same AP may change from one device to another [99]. Similarly, the number of visible APs may change abruptly between consecutive WiFi scans. The distribution of RSS values varies with indoor location due to indoor infrastructure such as corridors, doors, and furniture. However, RSS values may be very similar even for distinct locations due to indoor settings of objects, hallways, and furniture [82]. In addition, RSS values change over time for several reasons, such as multipath fading, shadowing, change in indoor infrastructure settings, amd moving objects and people [126]. Even the opening and closing of doors during RSS collection is reported to cause changes in RSS values [24, 151]. In addition, the WiFi APs are deployed to increase the capability of the communication network, and positioning is not the primary objective of such deployment. Fingerprints collected with and without human mobility are significantly different [121]. Similarly, fingerprint collection with the surveyor moving and stationary exhibits different levels of noise [102]. To overcome such issues, a new and dense deployment of APs can be made; however, it is not feasible or affordable in practical settings [29].
Traditional floor identification approaches, which are based on WiFi, barometer, and IMU sensors, are suitable for small to medium places. However, in large places where the indoor structure is complex, PDR methods introduce positioning errors that are accumulated over time. Similarly, the barometer data experience fluctuations even while the pedestrian is on the same floor. Integration of RSS from WiFi and GSM networks proves to overcome positioning issues in large places [78]. Contradictory to infrastructure-based approaches using WiFi, techniques that use RSS vectors and FM RSS present a better solution, especially when dynamic scenarios are considered. Operating at a low-frequency range, FM signals are less susceptible to human mobility and small obstacles. Covering an area of hundreds of kilometers, FM signals are sufficiently stronger with good indoor penetration to be used as fingerprints for indoor positioning [31]. However, due to FM towers placed hundreds of kilometers away and high transmission power, nearby locations do not exhibit unique FM RSS, and grain localization is not feasible using FM signals alone. Research indicates that an outdoor accuracy of tens of meters is achievable with FM signals [51, 91]. Even so, they can be augmented with WiFi and other technologies to provide better vertical positioning accuracy than WiFi alone.
Above all, the labor and time involved in the offline fingerprinting phase become exponential when large buildings with tens of floors are involved. The underlying assumption of infrastructure-based solutions that indoor conditions remain similar during the offline and online phases is hardly practical. Even smaller changes in the indoors, such as changing the position of furniture or that of WiFi APs, drastically change the radio map, which heavily impacts the positioning performance.

6.2 Challenges of Barometer-based and Hybrid Approaches

Several factors limit the use of barometer data. Although the decrease in the pressure is associated with an increase in altitude, this association is not linear, at least for the cases when a pedestrian is walking between two consecutive floors, as pointed out in [123]. The study shows that the change is linear for elevators when moving from one floor to another; however, walking on stairs does not show a linear trend in the pressure data. Using barometer data for floor identification with a smartphone often involves additional limitations as heterogeneous smartphone models of different, as well as the same, companies show different readings even for the same place at the same floor level. This offset in barometer readings of inter- and intra-phone models is large enough to produce a difference of one to three levels. Similarly, different devices have different levels of inherent noise, and comparing the absolute value of different devices directly is not appropriate to determine the floor level even when a reference barometer is used [92]. Different RMS errors are associated with different devices. For example, Google Galaxy Nexus uses a Bosch barometer sensor and has an RMS error of 0.5 m, corresponding to an RMS pressure error of 0.06 hPa. In comparison, the Samsung Galaxy series with STM Electronics has an RMS error of 0.65 m, corresponding to an RMS pressure error of 0.08 hPa.
Despite the difference in absolute pressure, changes in pressure values for different values tend to be similar. However, such a change in pressure is not always associated with floor transition, as the change in pressure may occur while walking on the same floor. The readings from a reference barometer (fixed at a reference floor) and smartphone barometer are often different even when the pedestrian is walking on the same floor where the reference barometer is installed [129]. Another problem with using the barometer for floor transition detection comes with the use of elevators; i.e., the speed of the elevator can increase the positioning uncertainty when the user moves across multiple floors. Single-floor transition detection has high accuracy, given that the floor height is at least 1.6 m, while multiple floor transitions using an elevator introduce a higher floor identification error. Environmental factors also affect the barometric pressure data, where electrical noise, internal air conditioning or heating changes, and sudden air flow introduce noise or errors [110]. The selection of a working mode for the installed reference barometer is also vital to optimize the power consumption and is often set to 1 minute, affecting the positioning accuracy. Using externally powered barometers and reducing the refresh rate of barometers elevate the floor identification performance.
In addition to the challenges associated with wireless signals, one widely used technology for hybrid vertical positioning, MEMS sensors, and their related approaches have their own limitations. Despite being practical, PDR approaches are often complicated, difficult, and prone to error accumulation. Additionally, users’ complex actions in 3D space make it very challenging to integrate the data from multiple sensors. The major limitation of PDR approaches is drift accumulation over time [12], which requires periodic correction using WiFi, BLE, and similar other sensors.
Unlike the controlled environments used in the proposed approaches found in the literature, real-life scenarios are complicated, and users’ actions are complex, which challenges the accuracy of such approaches. Often, sudden movements of the pedestrian in the dynamic environment lead to additional challenges. Despite using indoor maps to complement PDR approaches, indoor maps of all the buildings may not be available at all or not in a form suitable to be used with the mobile app. Against the WiFi systems, BLE-based positioning is low power consuming, light, and low cost. However, owing to the radio-frequency propagation, it suffers drawbacks similar to WiFi. BLE tags are to be placed in the intended area of positioning, which is expensive and requires maintenance and dense deployment for the desired level of accuracy. Hybrid approaches involve sensor fusion, where the data from heterogeneous sensors are combined to elevate the positioning performance. For multisensor fusion, filters such as particle filter, Kalman filter, or EKF are used, which are computationally complex. The positioning process often involves pedestrians in a static or semi-static manner with a controlled speed, limiting the wide application of such approaches in real-world scenarios.

6.3 Disadvantages Associated with Machine Learning-based Approaches for Floor Identification

An important challenge for deep learning approaches is the availability of large datasets for training and testing to analyze the performance. Predominantly, the publicly available datasets are collected with respect to their use for fingerprints. Although such datasets can be utilized with deep learning approaches, the desired positioning accuracy may not be possible. Often collected by novice users, multifarious devices, and poorly designed collection scenarios, the data are noisy, inconsistent, and improper, affecting the learning process of deep learning approaches. Furthermore, feature extraction is not very well studied with deep learning methods, and often raw data are used, which deteriorates the performance of such approaches.
The maintenance of a large-scale fingerprint database is a continuous task. The efficacy of deep learning approaches depends heavily on the amount and quality of the collected data. Data collection often involves multiple users, and changing indoor environments requires re-calibration of fingerprints, which is laborious and time-consuming. Gathering the location of fingerprints is a long and tedious task, especially when covering a vast area of a large building from surveyors or through crowdsourcing. On the other hand, crowdsourcing has its own limitations, such as users submitting wrong data, device heterogeneity, and sensor integration. A rather more noteworthy issue is the deployment of deep learning models on smartphones owing to the limited computational power of smartphones. The deep learning models are currently trained on computers that serve as servers and communicate to the positioning device through WiFi or the cellular network, which introduces latency.

6.4 LTE and 5G Solutions

Approaches that take advantage of LTE and 5G signals are based on TDoA, AoA, AoD, OTDoA, and so forth. Although these approaches have been adopted in WiFi, UWB, and RFID-based localization, they proved to be more accurate when utilizing LTE and 5G signals. Time-based positioning approaches for LTE and later rely on PRS, where the PRS symbols are identified on the UE side from eNB, and PRS time difference is used to find the position. First of all, four eNBs are needed to find the vertical position, which may not be available at all indoor locations due to the signal propagation phenomenon. An alternative is to have one of three eNBs at different heights. However, traditionally eNBs are installed at the same height. One major challenge with LTE and 5G approaches is the higher NLOS probability in dynamic indoor environments. Current solutions, both simulation and field experiments, assume LOS scenarios for TDoA, AoA, AoD, and oTDoA, representing a small portion of indoor scenarios. With NLOS environments, the problems become much more complicated due to biased ToA measurements. In addition, perfectly synchronized eNBs are assumed for simulations, which is impossible for real-world scenarios where a clock synchronization difference of 100 ns can lead to a 30 m horizontal positioning error. Positioning accuracy is also affected for long eNB distances due to quantization and multipath error.

6.5 Transition versus Non-transition Approaches for Floor Identification

Floor identification is not a one-time task (identifying the floor at which the user currently is); rather, it is continuous, where the user’s current floor needs to be determined at particular time intervals (latency). Latency is a matter of need where different applications require different latencies. The user may be moving on the same floor or may change the floor during the floor identification process. The former is called non-transition, where the user stays on the same floor. It is also possible that the floor identification determines the individual floor only and does not consider floor change. The latter is called transition, where the user is moving between the floors and the approach has to determine the change in floor as well.
Transition refers to moving from one floor to another floor, both up and down. Let’s say that two pedestrians are moving on stairs: pedestrian 1 is moving from floor 1 to floor 2, while pedestrian 2 is moving from floor 2 to floor 3. In this case, for both pedestrians, it is the transition of one floor. The user may transit for multiple floors and the transition can be using stairs or elevators. In both cases, the floor identification approach should be able to determine the up and down transitions.
Predominantly, existing floor identification approaches consider identifying the user’s current floor only, and only a few approaches focus on continuous tracking of the user on different floors. For example, Table 8 (given in Appendix A) shows that only a few approaches [66, 109, 123, 144, 147] consider floor transition while performing floor identification, while the majority of the approaches [4, 6, 9, 13, 23, 28, 34, 37, 46, 53, 57, 63, 69, 70, 71, 79, 83, 84, 88, 92, 93, 94, 103, 106, 107, 108, 112, 115, 118, 120, 124, 128, 129, 139, 143, 145, 150] consider only single-floor identification. In addition, only those approaches that incorporate a multi-sensor approach where an accelerometer or barometer sensor is used that can detect device behavior can detect floor transition. Incorporating floor transition increases complication, and the performance of the floor identification approach is reduced, as can be observed from Table 8 (given in Appendix A). Even in the approaches that considered floor transition [66, 109, 123, 144, 147], it is not stated how many floor transitions are considered while determining the average accuracy of the approach as the increase in floor transition is expected to reduce floor identification accuracy.

6.6 Discussion on Efficiency of Floor Identification Approaches

A wide range of floor identification approaches have been proposed in the existing literature, harnessing the power of various technologies. While different approaches report exceptional accuracy, even touching 100% for some approaches, accuracy alone is not a good metric. The efficiency of a floor identification approach is an important parameter in this regard as different approaches have been tested in different places with different indoor settings. Similarly, the number of floors where approaches are tested are not the same, raising questions about the efficiency of an approach tested in a two-floor building with 100% accuracy versus one tested in an eight-floor building with 95% accuracy. To judge the approaches on the same scale, a confidence score is utilized, which considers both the accuracy and the number of floors where a particular approach is tested. Even so, it is not possible to determine the efficiency of these approaches unless they are tested in the same indoor environment, which is impossible. It is so because the complexity of the floors is different, as is the area of the indoor environment.
From existing approaches, it is observed that increasing the number of floors alone, without even considering complexity in terms of pathways, furniture, equipment, and so forth, would lead to reduced floor identification accuracy and increased latency. Various indoor environments are attributed to affect floor identification accuracy. Research [135] indicates that buildings with large indoor areas show relatively low positioning accuracy with barometers. Even more, if floor transition is incorporated with an increased number of floors, it will further affect its accuracy. Indoor environment complexity has more effect on those approaches that utilize wireless propagation signals. Barometer sensors, Bluetooth, IMU sensors, and so forth have been utilized for floor identification approaches; however, each of these technologies can be favorable for a particular environment. The wireless communication-based solutions have the following favorable environments
Existing communication infrastructure can be used for floor identification
Pervasiveness and wide proliferation
Wide use of wireless network-enabled mobile devices
When the attenuation of signals is significantly different across vertical and horizontal directions
Favorable for LoS scenario
Dense deployment of WiFi APs
Small to medium-sized indoor spaces
Even floor spaces
Indoor environments without open roof spaces
Despite many advantages of wireless signal-based floor identification approaches, they face the following challenges:
Collection of fingerprints in large places is laborious
Changes in WiFi AP settings require re-calibration
Signal loss due to fading, shadowing, and other obstacles
Degradation of signal strength over time
Uneven floor surface increases floor identification error
Require several parameters for better accuracy like floor height, roof thickness, wall material types, and so forth
Wireless signal-based solutions are crippled in situations where the indoor environment is complex and the floor surface is not even. Barometer data coupled with wireless signals can compensate for such scenarios. It can be used to determine user-carrying devices, thereby determining user activity like going up or down across floors. Similarly, using regular updates for barometric pressure can be utilized for better accuracy. However, barometer-based approaches assume a fixed floor height, and in spaces where the floor height is irregular, such approaches become useless. Installing a dedicated barometer on each floor would be a better solution for improved accuracy. Using hybrid approaches is more fruitful for complex indoor areas where a single technology does not provide good results. Barometric data are used with other technologies like Bluetooth [145], IMU sensors [152], WiFi [143], GNSS, and light sensor and smartphone camera [13], as well for obtaining better floor identification.
Machine learning approaches work well provided they are trained on large-sized datasets. Often, such approaches can provide generalized results with slightly affected accuracy. However, machine learning models are not extensively investigated in the context of floor identification approaches and suffer from scalability and complexity problems.
Putting it in a nutshell, no single technology is good enough to provide a universal solution for floor identification as each has its pros and cons, which necessitates hybrid models that incorporate multiple sensors. Each technology can be utilized in different scenarios. For example, wireless technology can be used for reduced cost and coarse to fine accuracy for even floor surfaces. Barometer-based solutions are good for simple indoor environments with fixed floor heights. Machine learning solutions can be a good option where larger data are available from multiple sensors. PDR is only used for detecting floor transitions. Hybrid solutions incorporating machine learning technology can be better solutions where a more accurate floor identification is required.

6.7 Selecting Complementary Technologies for Floor Identification

Literature analysis reveals that the use of a single technology for floor identification carries several limitations. In the past, many hybrid systems have been designed that leverage more than one technology to alleviate these limitations. However, the choice of combining multiple technologies is very important and is carried out from the perspective of the final objective. For example, reduced cost, improved latency, computational complexity, and enhanced floor identification accuracy are a few of the objectives for which a particular approach is designed. The following potential combinations are suggested concerning these objectives:
WiFi and Bluetooth: WiFi and Bluetooth are based on wireless propagation and suffer from similar challenges of signal fading, signal absorption, shadowing, and so forth. However, due to small coverage, Bluetooth can be used as a complementary technology to WiFi. Bluetooth provides a fine location compared to the coarse location for which WiFi is used, particularly in scenarios where the WiFi APs’ deployment is sparse. In addition, using the PDR approach additionally helps determine the short-term position of the user with respect to its previous position, which can improve floor identification accuracy. PDR approaches are particularly useful for determining floor transitions. Due to the mass production of Bluetooth beacons, this solution is cost-effective and provides good accuracy.
WiFi and IoT Sensors: It is observed that WiFi signals experience RSS fading over time, which requires updating the fingerprint databases, leading to additional maintenance costs. If not updated, a compromise on the floor identification accuracy has to be made. The same is true if a new WiFi AP is installed or a WiFi AP is moved or removed. In such scenarios, using IoT sensors is a potential solution that can sense RSS from WiFi APs and periodically send it to the server, which can update the fingerprint database automatically. Adding IoT sensors incurs additional costs; however, it can provide better accuracy.
Cognitive Approaches and LTE, 5G and Beyond: The deployment of 5G and beyond 5G communication, with wide bandwidth, holds special importance for more accurate floor identification. Using machine learning-based cognitive solutions that can learn and adapt can elevate their performance for floor identification. Particularly, dense deployment of eNBs can provide fine location information. Despite being expensive, this solution is significantly important for situations requiring reduced latency.
Barometer, WiFi, PDR, and Indoor Maps: Using barometer-based floor identification with WiF and PDR is yet another cost-effective solution. Adding indoor map information is expected to greatly improve floor identification, particularly when PDR is used for floor transition detection. Adding barometers on each floor can greatly mitigate the floor identification errors caused by the WiFi for open areas around stairs. This hybrid solution is also very useful for buildings with open areas inside or where the floor height is irregular.

7 Open Issues and Future Directions

This section provides future directions acquired from the literature on indoor positioning and localization approaches, as well as the authors’ insight on probable future research trends and desired features indispensable for floor identification technologies.

7.1 Wireless Network-based Floor Identification

To mitigate the influence of dynamic factors for wireless-based vertical positioning approaches, IoT sensors are becoming largely available today. They can be utilized for both catering to dynamicity and real-time updating of the fingerprinting database. The wide use of Internet-enabled mobile devices presents the potential of crowdsourcing, where each device is both utilizing and making the fingerprint while navigating in an indoor environment. Smartphone heterogeneity, diversification of embedded sensors, and complex user orientations are big hurdles to achieving an efficient crowdsourcing approach.
The accuracy of barometer-based floor identification solutions depends on the accuracy of the built-in barometer, which has a relative accuracy of \(\pm\)10 Pa (Pascal). On the other hand, commercial barometers are high resolution with up to \(\pm\)1 Pa accuracy and offer more accurate pressure data. With more and more temperature and pressure-enabled IoT devices, the large proliferation of commercial barometer sensors is anticipated, which can lead to greatly flourished floor identification solutions in the near future. Currently, mobile devices’ embedded barometers offer a sampling frequency of 1 to 20 Hz; however, at a higher sampling rate, the noise is also high.
Constantly growing and widely penetrating smartphones are a big challenge for crowdsourcing-based approaches. For tackling device heterogeneity, novel, intuitive, and self-learning frameworks are required to integrate the data from heterogeneous sources. Complex environments such as terminals and airports with hundreds of pedestrians accessing the positioning service simultaneously introduce additional latency for real-time positioning approaches. With the fast speed of 5G and now 6G on its way, latency is envisioned to be appropriately handled. Currently, there is no standard for the installation of MEMS built-in sensors for a smartphone; a standardization effort is highly recommended in this direction to alleviate the influence of smartphone heterogeneity on floor identification approaches.
While the existing wireless network is designed with communication alone in mind, it can be rearranged or its indoor deployment can perhaps be redesigned to optimize it for communication as well as positioning. Another potential solution could be the installation of additional sensors in the form of IoT sensors, which are cheap and easy to deploy and can substantially increase the positioning performance. Such sensors can also be used to automatically collect the RSS from WiFi APs and update the dataset periodically, thereby alleviating the fingerprint obsolete problem as well.

7.2 Floor Identification Using Barometer-based Hybrid Approaches

Predominantly, hybrid approaches leverage the signals of opportunity, i.e., wireless and PDR approaches based on MEMS sensors embedded in mobile devices. With a large deployment of WiFi APs operating at 2.4 GHz and 5 GHz, dual-band can be leveraged to enhance positioning accuracy. Using barometer and WiFi, hybrid approaches have been regarded as inappropriate and unsuitable for floor identification, especially where the floor structure is irregular and uneven. However, with the proliferation of Internet of Things (IoT) devices, barometers are expected to become cheap, accurate, energy efficient, and densely deployed, which can help to improve positioning performance. With improvements in BLE, the accuracy of current positioning applications can further be improved. The new BLE v.5.0 offers higher data rates, which are expected to improve further in the near future [117].
The volume of provided data will grow further, and utilizing machine and deep learning approaches can help enhance the positioning performance. Possibly the inclusion of direction-finding mechanisms in BLE such as AoA and AoD can help to improve the performance. For the most part, floor identification involves a static user/device, or the movement is very slow compared to real-life scenarios. As a matter of fact, the focus has to be shifted toward challenging scenarios for trajectory estimation using the data from multiple sensors of smartphones such as inertial data, WiFi, BLE, and pressure, in addition to indoor topological maps like floor plan maps with information of stairs, elevators, and walls. Unlike widely available outdoor maps from Google globally and several other companies locally for different countries, map information for indoor environments is not available. Developing a central server where such maps can be uploaded without compromising security is a need of time. Such maps can be leveraged to provide extra aid for hybrid approaches, where map information can uplift the positioning accuracy significantly. Besides human surveyors, robots equipped with light detection and ranging (LiDAR) and radar can be used to build indoor maps automatically.
Barometers with higher sampling rates are foreseen for upcoming mobile devices, which can play an important role in elevating barometer-based positioning performance. Current barometers are low error and noise tolerant primarily due to their size and cost considerations. With further advancement in data measurement, cleaning, and device miniaturization, we expect high noise-tolerant and precise measurement-based barometers in the near future.

7.3 LTE, 5G, and Beyond

With accurate AoD and ToA measurements in 5G networks, both can be combined to calculate a more precise position. The use of directional antenna arrays and wide bandwidth is another important foreseen opportunity. The introduction of device-to-device (D2D) communication has opened new avenues of indoor positioning where direct communications between UEs are possible to share range and angular measurement. Being within close proximity to eNBs, the probability of being LoS and having high SNR is higher, offering higher positioning accuracy. Multipath can be resolved using ray tracing to improve positioning accuracy in NLoS environments. Utilizing the map information in conjunction with multipath, additional information about UE location can be obtained [64]. With the addition of further use cases of indoor and urban microenvironments for future 6G networks where base stations are located on the walls and under roof ceilings, the probability of LoS grows higher, which can further improve the positioning performance. Similarly, the backhaul scenarios where base stations are placed on street lamp posts provide opportunities for a fine-grained position [1].
With D2D communication, (\(_2^N\)) links are available for N available devices in a given indoor environment. Both range and angle measurements will be made on UEs in future D2D communication. Such measurements are to be sent to the BS for position estimation, where the positioning of all UEs is simultaneously calculated using non-linear least squares. For non-linear least squares, optimization techniques are required to speed up the positioning without compromising on accuracy. A growing concern for future networks is privacy as the users will be tracked in real time. Although policies are already taking place where the users’ consent is required from Android to allow position sharing, transparency is still required where the user location is tracked in the back end. Users’ location sharing and control on opting out of tracking should be completely placed under users’ command. Similarly, users’ tracking information stored in the device and over the network must be safe from hackers’ access.

7.4 IoT-based Localization Approaches

The IoT paradigm is progressing at a growing speed and has been adopted in a large number of fields like health services, manufacturing, smart cities, and ambient living. In the IoT paradigm, various objects, sensors, and devices are connected through the Internet and can communicate through different protocols. Such sensors and devices are now manufactured at a lower cost and are embedded in home appliances like coffee machines, air conditioners, televisions, and so forth. Due to their lower cost, it is convenient to place many sensors in an indoor environment for positioning. For IoT communication, several communication protocols have been designed such as LoRaWAN [38], Sigfox [76], long-term evolution for machines (LTE-M) [26], and narrowband IoT (NBIoT) [101]. In the existing literature, LoRaWAN [25, 68], Sigfox [61, 131], LTE-M [40], and NBIoT [10] have been used for 2D indoor and outdoor localization with promising results. Several studies [2, 58, 104, 119] indicate that in a multifloor environment, LoRaWAN experiences signal loss with respect to LoRaWAN gateway and location of the device, suggesting the possibility of floor identification using LoRaWAN. Utilizing these technologies for floor identification could be a promising avenue.

7.5 Additional Sensors for Future Smartphones

With rapid advancements in sensor technology and the growing competition among smartphone giants like Samsung, Apple, and Huawei, more and more sensors are expected to be included in future smartphones. Like the introduction of the UWB sensor in the iPhone 11 Pro, which can be used for precise indoor positioning, future smartphones may include lasers and precise distance measurement technologies. Using these sensors, on-the-fly map building would become possible, which can greatly enhance the positioning accuracy for complex indoor structures. Currently, WiFi-based indoor positioning approaches cannot be utilized on iOS phones due to the restriction from Apple for providing the WiFi services information, which limits the wide application of these approaches. In the future, a common standard is expected from major smartphone companies to provide access to such information. Similarly, the services to provide a seamless transition of phone apps from Android to iOS and vice versa are not very common. Such applications would ensure easy transitions from one smartphone to another and reduce interoperability problems.

7.6 Cognitive Approaches

With the advancement in the smart city concept, where a large number of devices and sensors are interconnected, continuous streams of a large amount of data will be collected in real time. Predominately, existing floor identification approaches, including fingerprinting, barometer-based, and hybrid approaches, rely on pre-localization data collection. These existing floor identification approaches are not suitable for future smart cities. Designing machine learning- and deep learning-based solutions that can utilize real-time data could be a promising approach. Such approaches should be cognitive to learn from the new data and improve floor identification accuracy. For future, cognitive cities, these approaches can serve better for ambient assisted living, smart industrial environments, and so forth.

7.7 Privacy and Data Confidentiality

Predominately, existing approaches focus only on the positioning accuracy perspective, and the aspect of data confidentiality and user privacy is totally ignored. As discussed, the large use of mobile devices has led to potential solutions involving using these devices for floor identification. Several approaches involve data sharing, location sharing, and sharing other sensitive information. This information may be the target of adversaries, thus compromising data confidentiality and user privacy.

8 Conclusion

This study provides a comprehensive analysis of indoor floor identification techniques involving wireless communication networks, barometer solutions, machine learning-based solutions, and recent 5G and beyond approaches. Barometer-based floor identification approaches provide high accuracy, yet the discrepancies in barometer reading due to device heterogeneity, indoor map information, and lack of a standardized relationship between pressure and floor height are the biggest hurdles. Fingerprinting approaches using signals of opportunity, BLE, and GSM are easy to adapt and can be used on widely penetrated smartphones. Yet, dynamic environments involving human mobility, signal absorption, and shadowing degrade their performance severely. With the wide availability of embedded MEMS sensors in smartphones, they can be leveraged to provide additional information about pedestrians’ short-term position; activities of moving on elevators, stairways, and escalators; and device orientations, which improves the positioning performance of the hybrid approach substantially. Low error tolerance, drift accumulation over time, providing relative position, and the complex attitude of pedestrians give rise to several problems. Moreover, sensor fusion in the hybrid system involves filtering approaches that are complex and require higher computational power. Deep learning approaches require a large amount of annotated data and computing power to obtain higher accuracy. The complexity of resolving multipaths in NLOS scenarios for LTE and 5G is not yet obtained for real-world scenarios.
This survey describes several technologies such as WiFi, BLE, and barometer and techniques such as fingerprinting, OTDoA, and CSI used for floor identification. These technologies are described, and their working mechanism and basic elements are elaborated comprehensively. Challenges associated with each technology are incorporated, and the foreseen technological advancements, probable trends, and future discussed are outlined.

A Tables of Manuscript

Table 1.
2DTwo-dimensional3DThree-dimensional
5GFifth Generation6GSixth Generation
AEAutoencodersANNArtificial Neural Network
AoAAngle of ArrivalAoDAngle of Departure
APAccess PointBCCHsBroadcast Control Channels
BLEBluetooth Low EnergyBPBack Propagation
BTSBase Transceiver StationCDMACode Division Multiple Access
CHChan and Ho’sCHsCluster Heads
CILoSCDMA Indoor Localization SystemCISConfidence Interval Sum
CMRSCommercial Mobile Radio ServiceCNNConvolutional Neural Network
CSIChannel State InformationCUREClustering Using Representatives
D2DDevice to DeviceDAEDeep Autoencoder
dBmdecibel-milliwattsDCMDatabase Correlation Methods
DIPSAdaptive Indoor Positioning SystemDNNDeep Neural Network
E911Enhanced 911E-CIDEnhanced-Cell Identification
EKFExtended Kalman FilterELMExtreme Learning Machine
eNBEvolved Node BFAFFloor Attenuation Factor
FCCFederal Communications CommissionFLDFisher’s Linear Discriminant
FLDFisher’s Linear DiscriminantFMFrequency Modulation
FSPLFree Space Path LossGHzGiga Hertz
GNSSGlobal Navigation Satellite SystemsGPSGlobal Positioning System
GSMGlobal System for Mobile communicationshPahecto Pascal
ICAOInternation Civil Aviation OrganizationIMUInertial Measurement Unit
INSInertial Navigation SystemsIoTInternet of Things
IRInfraredISUInternational System of Units
KFKalman Filterk-NNk Nearest Neighbor
LBSLocation-based ServicesLDALinear Discriminant Analysis
LEDsLight-emitting DiodesLOSLine of Sight
LTELong-term EvolutionLTE-MLong-term Evolution for Machines
MACMedia Access ControlMEMSMicro-electro-mechanical Systems
MFBPMultireference Barometer Floor PositioningMFPSMagnetic Field Positioning System
MLPMultilayer PerceptronMSDMulti-story Differential
MSPLMultistory Path LossMWFMulti-Wall-floor
NBNaive BayesNLOSNon-Line of Sight
NNNeural NetworkNNSSNearest Neighbors in Signal Space
OTDoAObserved Time Difference of ArrivalOvOOne-versus-One
PaPascalPCAPrincipal Component Analysis
PDRPedestrian Dead ReckoningPFParticle Filter
PLGDPenalized Logarithmic Gaussian DistanceRFRadio Frequency
RFIDRadio-frequency IdentificationRFPMRadio-frequency Pattern Matching
RMoSRobust Mean of SumRMSRoot Mean Squared
RMSERoot Mean Squared ErrorRNsReference Nodes
RSCPReceived Signal Code PowerRSRPReference Signal Received Power
RSRQReference Signal Receiving QualityRSSReceived Signal Strength
RToFReturn ToFSAEStacked Autoencoder
SCASignal Coverage AreaSHSStep-and-Heading Systems
SINRSignal to Inference plus Noise RatioSLAMSimultaneous Localization and Mapping
SNSensor NodeSNRSignal-to-Noise Ratio
SPFSensor-assisted Particle FilterSVMSupport Vector Machine
TDoATime Difference of ArrivalToATime of Arrival
ToFTime of FlightTSTaylor Series
UEUser EquipmentUSUnited States
UWBUltra-widebandVLCVisible Light Communication
WCLWeighted Centroid LocalizationWLANWireless Local Area Network
WSNsWireless Sensor NetworksZUPTZero Velocity Update
Table 1. List of Acronyms Used in This Article
Table 2.
Ref.-YearTechnologies coveredHorizontal positioningFloor identification
WLANUWBBLEIRRFIDInertialVisionGSMFMLTEMagneticBarometer
[81]-2007
[67]-2009
[47]-2009
[125]-2011
[54]-2013
[141]-2013
[140]-2015
[55]-2015
[5]-2016
[35]-2016
[87]-2016
[136]-2016
[33]-2017
[20]-2017
[42]-2017
[75]-2018
[60]-2018
[148]-2019
[15]-2020
[49]-2020
[74]-2020
[96]-2021
Current
Table 2. Brief Overview of Existing Survey Papers on Indoor Positioning and Localization
Table 3.
Ref.ModelApproachAccuracy (# of floors)ConfidenceAdvantagesDisadvantages
[113]MWFSimulation99% (9)0.9987Feedback model enhances performanceFloor heights, wall thickness, and attenuation coefficients are required
[114]Multi-scope Path LossField experiments84.50% (3)0.9319Reduced database size, APs’ positions are not neededHigh positioning error for buildings with large openings
[39]Floor and wall attenuationSimulationAvg. 2.5 m (3)-Use of a priori informationLow accuracy for irregular floor structure
[50]Multi-floor Radio PropagationField experiments99% (2)0.9810Invisible APs enhance accuracyPerformance degradation in poor AP visibility
[8]MWMField experiments98% (3)0.9897Fast fingerprint generation using APs’ positionPoor performance within elevators and stairs
[90]RSS from RNsField experiments98.67% (3)0.9930High accuracy with sparse deployment of RNIrregular floor structures affects the performance
[146]FSPL & MSCField experimentsN/AN/ASparse APs based positioningCustom coefficients for walls and floors with fixed APs’ position
Table 3. Review of Floor Identification Approaches Based on Wireless Signals
Table 4.
Ref.FingerprintAccuracy (# of floors)ConfidenceContribution
[130]WiFi RSSN/A-Automatic data labeling using co-embedding with inter-graph similarity
[7]WiFi RSS86.00% (5)0.9593K Nearest floor-based positioning on multiple devices
[86]WiFi RSS99.00% (3)0.9940OvO majority voting using LDA for high accuracy
[89]WiFi RSS100% at 95% interval (3)0.9736Confidence interval sum-RSS for obtaining high accuracy
[52]RSS and MAC90.00% (3)0.9444Clustering based on MAC visibility at particular floors
[66]Dynamic fingerprint87.00% (4)0.9501Self-learning fingerprinting to generate real-time WiFi fingerprint
[34]AP clustering97.00% (4)0.98963D coordinate clustering for high accuracy and reduced computational time
[107]RSS clustering88.00% (4)0.9545Reduced fingerprint complexity via K-Means
[120]Barometric data clustering and WiFi98.68% (5)0.9954RMS, Kurtosis, Skewness, peak-to-peak, etc., to increase accuracy
[28]WiFi RSS97.00% (13)0.9974Unsupervised clustering, PCA, and architectural constraints for high accuracy
[108]WiFi RSS85.80% (4)0.9448Weighted centroid localization using collected RSS
[88]WiFi RSS95.00% (3)0.9736Dynamic environments with AP failure
[9]RSS, people location99.00% (3)0.9940Human mobility and device heterogeneity
[23]RSS-based heuristics99.87% (4)0.9951Use of average signal strength and signal strength variance
[118]RSS, indoor structure94.30% (6)0.9879Indoor structure such as the position of stairs and walls, etc.
[124]GSM73% (9), 97% within 2 floor (2)0.9537, 0.9690Fingerprint stability over time
[122]CDMA87% (5)0.9624CDMA signal delay does not vary over time
[31]FM83% (3)0.8975Tolerance to human mobility, orientation, and small objects
[97]GSM93.26% (3)0.9638Better performance with wooden roof structures
[132]Macrocell0 m at 67% and 7 m at 90%-BS-based fingerprint, cell matching degree
[45]GSM65% (6)0.8923Strongest seven cells with MLP
Table 4. Characteristic Features of WiFi and Cellular-based Fingerprinting Solutions
Table 5.
Ref.BuildingsFloorsDevicesPositioning mode/approachAccuracyConfidence
[56]Single2SingleReference floorN/AN/A
[71]Single5MultipleMultidel reference barometers85.71%0.9583
[144]Single10MultipleReference barometer, floor change detection98%0.9977
[109]Multiple5–8MultipleStairways activity detection, barometer calibration, entry point detection85.6%0.9759
[135]Single4–9MultipleOutlier removing, multireference barometer, sensor offset value95.48%0.9940
[123]Multiple3MultipleFloor transition82%0.8902
[92]Multiple13MultipleFloor transition99.54% for single floorN/A
Table 5. A Summary of Barometer-based Floor Identification Approaches
Table 6.
      
RefSensorsAccuracy (# of floors)ConfidenceAdvantagesDisadvantages
[139]Baro., Acc., Gyro.99.36% (5)0.9983High accuracy, robust stair detectionFloor height information, waist-mounted device, inappropriate for large and complex buildings
[115]Acc., Gyro., Mag.99% (4)0.9966Elevator, escalator, and stairway detection modulesRequire entry beacons, GPS, false positives for elevator and stairway
[147]Baro., Acc., WiFi85.60% (6)0.9663Stairs-up and stairs-down modules, offset value for heterogeneous smartphones, floor change detectionFingerprint database, low accuracy, pressure reading errors for stair modules, building information
[145]Acc., Bluetooth97% (10)0.9965Infrastructure independent, stairs-up and -down modulesInitial floor information and Bluetooth data are needed
[63]Baro., Bluetooth, Acc., Mag.95.9% (2)0.9570Step detection and step length estimation modules, resolves accumulation errorsBluetooth beacons and floor height information are required
[152]BLE, Acc., Gyro., Mag.Avg. 0.35 m (2)N/AError mitigation using BLE and MEMS sensor dataBLE fingerprint data are needed
[128]Baro., WiFi, Indoor info.99.5% (3)0.9970High accuracy in elevator error reduction using WiFi RSSReference barometer and building information is a must
[129]Baro., WiFi, Topological map99.5% (3)0.9970Bias estimation and correction for pressure dataTopological map and reference barometer are needed
[143]WiFi, Acc.98.8% (10)0.9985Crowdsourcing-based WiFi fingerprintComplex integration of crowdsourced data, device heterogeneity
[106]WiFi, Acc.99.1% (7)0.9984High accuracyInitial floor information is required for accelerometer data, WiFi fingerprint data
[4]WiFi, Acc.99% (4)0.9966High accuracy, reference barometer is not requiredFloor height and WiFi fingerprints are needed
[93]WiFi, Acc.98% (3)0.9896Smartphone orientation independent positioningWiFi fingerprints, landmarks are required
[112]WiFi, Baro.96.3% (4)0.9871Crowdsourced fingerprints, floor height is not requiredComplexity and errors in crowdsourced data integration, initial floor information
[57]WiFi, BLE, PDR91.84% (3)0.9555Offset value for heterogeneous devicesReference barometer, BLE beacons are required
[46]WiFi, Baro. light sensor, GNSS98.00% (8)0.9970Site survey for fingerprint collection is not requiredPedestrian must enter/exit at the ground entrance, cannot handle elevators and escalator
[150]WiFi, Baro.99.00% (8)0.9985High accuracy, robustCannot deal with elevators, escalators, stairs
[79]WiFi, Baro., PDR97.00% (4)0.9896High accuracy, robustFloor information, indoor structure information are required
[53]WiFi, Baro.97.00% (3)0.9845High accuracy, low computational complexityAccurate floor information is needed; poor performance in WiFi-denied areas
[84]WiFi, BLE96.87% (4)0.9892Low complexity, no fingerprintingCalculation of path loss model parameters is required for each environment
[83]WiFi, Baro., PDR, Map information94.00% (3)0.9680Works well in different environmental conditionsFingerprint and indoor map are required
[13]Smartphone camera91.04% (3)0.9507Infrastructure independentComputational complexity, image fingerprints are needed
[17]Smartphone magnetometer91.02% (3)0.9506High scalability, infrastructure independentGeomagnetic fingerprints are needed
Table 6. Overview of Hybrid Solutions for Floor Identification
Table 7.
Ref.ModelAccuracy (# of floors)ConfidenceAdvantagesDisadvantages
[94]SAE, DNN92.00% (4)0.9710Dimensionality reduction, easy adaptationLarge dataset is required; building hierarchy is not considered
[70]SAE, DNN97.18% (4)0.9903High accuracy, dimensionality reductionValidation dataset is small
[103]Ensemble ELM, PCA98.00% (7)0.9965High accuracy, reduced complexitySmall experiment area
[6]Hierarchical ELM98.13% (4)0.9936Simple, fast, and efficientValidation is not performed
[69]DNN91.18% (4)0.9677Scalability, low complexityStatic positioning, fingerprint maintenance problems
[37]MLP91.08% (6)0.9804Improved performance for dynamic environmentsPoor performance in sparse APs and complex environments
[116]SAE, CNN95.00% (16)0.9964Sparse RSS-based positioningStatic environments, with even floor structures
Table 7. A Summary of Approaches Using Deep Learning Models for Floor Identification
All studies utilized the WiFi RSS for floor identification.
Table 8.
Ref.Accuracy# of floorsTransitingRef.Accuracy# of floorsTransiting
[3]89%5No[7]86.00%5No
[113]99%9No[86]99.00%3No
[114]84.50%3No[89]100%3No
[39]Avg. 2.5 m3No[52]90.00%3No
[50]99%2No[66]87.00%4Yes
[8]98%3No[34]97.00%4No
[90]98.67%3No[107]88.00%4No
[71]85.71%5No[120]98.68%5No
[144]98%10Yes[28]97.00%13No
[109]85.6%5-8Yes[108]85.80%4No
[135]95.48%4-9No[88]95.00%3No
[123]82%3Yes[9]99.00%3No
[92]99.54%1No[23]99.87%4No
[139]99.36%5No[118]94.30%6No
[115]90%4No[124]73%9No
[147]85.60%6Yes[122]87%5No
[145]97%10No[31]83%3No
[63]95.9%2No[97]93.26%3No
[128]99.5%3No[45]65%6No
[129]99.5%3No[17]91.02%3No
[143]98.8%10No[94]92.00%4No
[106]99.1%7No[70]97.18%4No
[4]99%4No[103]98.00%7No
[93]98%3No[6]98.13%4No
[112]96.3%4No[69]91.18%4No
[57]91.84%3No[37]91.08%6No
[46]98.00%8No[116]95.00%16No
[150]99.00%8No[79]97.00%4No
[53]97.00%3No[84]96.87%4No
[83]94.00%3No[13]91.04%3No
Table 8. Comparison of Floor Identification Approaches Concerning Floor Transition

References

[1]
3rd Generation Partnership Project. 2021. Technical Specification Group Services and System Aspects.
[2]
Michiel Aernouts, Rafael Berkvens, Koen Van Vlaenderen, and Maarten Weyn. 2018. Sigfox and LoRaWAN datasets for fingerprint localization in large urban and rural areas. Data 3, 2 (2018), 13.
[3]
Uzair Ahmad, Brian J. D’Auriol, Young-Koo Lee, and Sungyoung Lee. 2008. Multi-floor semantically meaningful localization using IEEE 802.11 network beacons. IEICE Transactions on Communications 91, 11 (2008), 3450–3460.
[4]
H. J. Ai, M. Y. Liu, Y. M. Shi, and J. Q. Zhao. 2016. Floor identification with commercial smartphones in WiFi-based indoor localization system. International Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences 41 (2016), 573.
[5]
Abdulrahman Alarifi, AbdulMalik Al-Salman, Mansour Alsaleh, Ahmad Alnafessah, Suheer Al-Hadhrami, Mai A. Al-Ammar, and Hend S. Al-Khalifa. 2016. Ultra wideband indoor positioning technologies: Analysis and recent advances. Sensors 16, 5 (2016), 707.
[6]
Atefe Alitaleshi, Hamid Jazayeriy, and Seyed Javad Kazemitabar. 2020. WiFi fingerprinting based floor detection with hierarchical extreme learning machine. In 2020 10th International Conference on Computer and Knowledge Engineering (ICCKE’20). IEEE, 113–117.
[7]
Firas Alsehly, Tughrul Arslan, and Zankar Sevak. 2011. Indoor positioning with floor determination in multi story buildings. In 2011 International Conference on Indoor Positioning and Indoor Navigation. IEEE, 1–7.
[8]
Iyad H. Alshami, Noor Azurati Ahmad, and Shamsul Sahibuddin. 2015. A light WLAN radio map for floor detection in multi-floor environment localization. In 2015 9th Malaysian Software Engineering Conference (MySEC’15). IEEE, 135–139.
[9]
Iyad Husni Alshami, Noor Azurati Ahmad, Shamsul Sahibuddin, and Firdaus Firdaus. 2017. Adaptive indoor positioning model based on WLAN-fingerprinting for dynamic and multi-floor environments. Sensors 17, 8 (2017), 1789.
[10]
Anjali Anjali. 2021. NB-IoT for localization and target detection. In Principles and Applications of Narrowband Internet of Things (NBIoT’21). IGI Global, 105–126.
[11]
Imran Ashraf, Soojung Hur, Yousaf Bin Zikria, and Yongwan Park. 2021. Ensembling neural networks for user’s indoor localization using magnetic field data from smartphones. CMC-Computers Materials & Continua 68, 2 (2021), 2597–2620.
[12]
Imran Ashraf, Soojung Hur, and Yongwan Park. 2018. mPILOT-magnetic field strength based pedestrian indoor localization. Sensors 18, 7 (2018), 2283.
[13]
Imran Ashraf, Soojung Hur, and Yongwan Park. 2019. Application of deep convolutional neural networks and smartphone sensors for indoor localization. Applied Sciences 9, 11 (2019), 2337.
[14]
Imran Ashraf, Soojung Hur, and Yongwan Park. 2019. Indoor positioning on disparate commercial smartphones using Wi-Fi access points coverage area. Sensors 19, 19 (2019), 4351.
[15]
Imran Ashraf, Soojung Hur, and Yongwan Park. 2020. Smartphone sensor based indoor positioning: Current status, opportunities, and future challenges. Electronics 9, 6 (2020), 891.
[16]
Imran Ashraf, Soojung Hur, Muhammad Shafiq, Saru Kumari, and Yongwan Park. 2019. GUIDE: Smartphone sensors-based pedestrian indoor localization with heterogeneous devices. International Journal of Communication Systems 32, 15 (2019), e4062.
[17]
Imran Ashraf, Soojung Hur, Muhammad Shafiq, and Yongwan Park. 2019. Floor identification using magnetic field data with smartphone sensors. Sensors 19, 11 (2019), 2538.
[18]
Abdelkarim Ben Ayed, Mohamed Ben Halima, and Adel M. Alimi. 2014. Survey on clustering methods: Towards fuzzy clustering for big data. In 2014 6th International Conference of Soft Computing and Pattern Recognition (SoCPaR’14). IEEE, 331–336.
[19]
Paramvir Bahl and Venkata N. Padmanabhan. 2000. RADAR: An in-building RF-based user location and tracking system. In Proceedings IEEE INFOCOM 2000. Conference on Computer Communications. Nineteenth Annual Joint Conference of the IEEE Computer and Communications Societies (Cat. No. 00CH37064), Vol. 2. IEEE, 775–784.
[20]
Anahid Basiri, Elena Simona Lohan, Terry Moore, Adam Winstanley, Pekka Peltola, Chris Hill, Pouria Amirian, and Pedro Figueiredo e Silva. 2017. Indoor location based services challenges, requirements and usability of current solutions. Computer Science Review 24 (2017), 1–12.
[21]
Alan Bensky. 2016. Wireless Positioning Technologies and Applications. Artech House.
[22]
Mário N. Berberan-Santos, Evgeny N. Bodunov, and Lionello Pogliani. 1997. On the barometric formula. American Journal of Physics 65, 5 (1997), 404–412.
[23]
Preeti Bhargava, Shivsubramani Krishnamoorthy, Aditya Karkada Nakshathri, Matthew Mah, and Ashok Agrawala. 2012. Locus: An indoor localization, tracking and navigation system for multi-story buildings using heuristics derived from Wi-Fi signal strength. In International Conference on Mobile and Ubiquitous Systems: Computing, Networking, and Services. Springer, 212–223.
[24]
Mekuanint Agegnehu Bitew, Rong-Shue Hsiao, Hsin-Piao Lin, and Ding-Bing Lin. 2015. Hybrid indoor human localization system for addressing the issue of RSS variation in fingerprinting. International Journal of Distributed Sensor Networks 11, 3 (2015), 831423.
[25]
Federico Bonafini, D. Fernandes Carvalho, Alessandro Depari, Paolo Ferrari, Alessandra Flammini, Marco Pasetti, Stefano Rinaldi, and Emiliano Sisinni. 2019. Evaluating indoor and outdoor localization services for LoRaWAN in Smart City applications. In 2019 II Workshop on Metrology for Industry 4.0 and IoT (MetroInd4.0 & IoT’19). IEEE, 300–305.
[26]
Suresh R. Borkar. 2020. Long-term evolution for machines (LTE-M). In LPWAN Technologies for IoT and M2M Applications. Elsevier, 145–166.
[27]
B. J. Burton. 1993. Atmospheric pressure, and temperature aloft. Weather 48, 5 (1993), 141–147.
[28]
Rafael Saraiva Campos, Lisandro Lovisolo, and Marcello Luiz R. de Campos. 2014. Wi-Fi multi-floor indoor positioning considering architectural aspects and controlled computational complexity. Expert Systems with Applications 41, 14 (2014), 6211–6223.
[29]
Gayathri Chandrasekaran, Mesut Ali Ergin, Jie Yang, Song Liu, Yingying Chen, Marco Gruteser, and Richard P. Martin. 2009. Empirical evaluation of the limits on localization using signal strength. In 2009 6th Annual IEEE Communications Society Conference on Sensor, Mesh and Ad Hoc Communications and Networks. IEEE, 1–9.
[30]
Chih-Yu Chen and Wen-Rong Wu. 2016. Three-dimensional positioning for LTE systems. IEEE Transactions on Vehicular Technology 66, 4 (2016), 3220–3234.
[31]
Yin Chen, Dimitrios Lymberopoulos, Jie Liu, and Bodhi Priyantha. 2012. FM-based indoor localization. In Proceedings of the 10th International Conference on Mobile Systems, Applications, and Services. 169–182.
[32]
Federal Communications Commission. 2015. Wireless E911 location accuracy requirements. Ps Docket 07-114 (2015).
[33]
Alejandro Correa, Marc Barcelo, Antoni Morell, and Jose Lopez Vicario. 2017. A review of pedestrian indoor positioning systems for mass market applications. Sensors 17, 8 (2017), 1927.
[34]
Andrei Cramariuc, Heikki Huttunen, and Elena Simona Lohan. 2016. Clustering benefits in mobile-centric WiFi positioning in multi-floor buildings. In 2016 International Conference on Localization and GNSS (ICL-GNSS’16). IEEE, 1–6.
[35]
Pavel Davidson and Robert Piché. 2016. A survey of selected indoor positioning methods for smartphones. IEEE Communications Surveys & Tutorials 19, 2 (2016), 1347–1370.
[36]
José A. del Peral-Rosado, Michele Bavaro, José A. López-Salcedo, Gonzalo Seco-Granados, Pravir Chawdhry, Joaquim Fortuny-Guasch, Paolo Crosta, Francesca Zanier, and Massimo Crisci. 2015. Floor detection with indoor vertical positioning in LTE femtocell networks. In 2015 IEEE Globecom Workshops (GC Wkshps’15). IEEE, 1–6.
[37]
Rizanne Elbakly, Heba Aly, and Moustafa Youssef. 2018. TrueStory: Accurate and robust RF-based floor estimation for challenging indoor environments. IEEE Sensors Journal 18, 24 (2018), 10115–10124.
[38]
Mehmet Ali Ertürk, Muhammed Ali Aydın, Muhammet Talha Büyükakkaşlar, and Hayrettin Evirgen. 2019. A survey on LoRaWAN architecture, protocol and technologies. Future Internet 11, 10 (2019), 216.
[39]
Xiao Fan and Yoan Shin. 2013. Indoor localization for multi-wall, multi-floor environments in wireless sensor networks. In Proceedings of the 9th Advanced International Conference on Telecommunication. Citeseer, 23–28.
[40]
Wei Fang and Bin Ran. 2020. An accuracy and real-time commercial localization system in LTE networks. IEEE Access 8 (2020), 120160–120172.
[41]
Federal Communications Commission. 2021. Indoor location accuracy timeline and live call data reporting template. https://www.fcc.gov/public-safety-and-homeland-security/policy-and-licensing-division/911-services/general/location-accuracy-indoor-benchmarks
[42]
André Filipe Gonçalves Gonçalves Ferreira, Duarte Manuel Azevedo Fernandes, Andre Paulo Catarino, and Joao L. Monteiro. 2017. Localization and positioning systems for emergency responders: A survey. IEEE Communications Surveys & Tutorials 19, 4 (2017), 2836–2870.
[43]
Sven Fischer. 2014. Observed time difference of arrival (OTDOA) positioning in 3GPP LTE. Qualcomm White Paper 1, 1 (2014), 1–62.
[44]
Xingli Gan, Baoguo Yu, Lu Huang, and Yaning Li. 2017. Deep learning for weights training and indoor positioning using multi-sensor fingerprint. In 2017 International Conference on Indoor Positioning and Indoor Navigation (IPIN’17). IEEE, 1–7.
[45]
Rafał Górak, Marcin Luckner, Michał Okulewicz, Joanna Porter-Sobieraj, and Piotr Wawrzyniak. 2016. Indoor localisation based on GSM signals: Multistorey building study. Mobile Information Systems 2016, 1 (2016), 2719576.
[46]
Fuqiang Gu, Jorg Blankenbach, Kourosh Khoshelham, Jan Grottke, and Shahrokh Valaee. 2019. ZeeFi: Zero-effort floor identification with deep learning for indoor localization. In 2019 IEEE Global Communications Conference (GLOBECOM’19). IEEE, 1–6.
[47]
Yanying Gu, Anthony Lo, and Ignas Niemegeers. 2009. A survey of indoor positioning systems for wireless personal networks. IEEE Communications Surveys & Tutorials 11, 1 (2009), 13–32.
[48]
Sudipto Guha, Rajeev Rastogi, and Kyuseok Shim. 2001. Cure: An efficient clustering algorithm for large databases. Information Systems 26, 1 (2001), 35–58.
[49]
Xiansheng Guo, Nirwan Ansari, Fangzi Hu, Yuan Shao, Nkrow Raphael Elikplim, and Lin Li. 2019. A survey on fusion-based indoor positioning. IEEE Communications Surveys & Tutorials 22, 1 (2019), 566–594.
[50]
Pankaj Gupta, Sachin Bharadwaj, Sthanunathan Ramakrishnan, and Jaiganesh Balakrishnan. 2014. Robust floor determination for indoor positioning. In 2014 20th National Conference on Communications (NCC’14). IEEE, 1–6.
[51]
Andreas Haeberlen, Eliot Flannery, Andrew M. Ladd, Algis Rudys, Dan S. Wallach, and Lydia E. Kavraki. 2004. Practical robust localization over large-scale 802.11 wireless networks. In Proceedings of the 10th Annual International Conference on Mobile Computing and Networking. 70–84.
[52]
Litao Han, Li Jiang, Qiaoli Kong, Ji Wang, Aiguo Zhang, and Shiming Song. 2019. Indoor localization within multi-story buildings using MAC and RSSI fingerprint vectors. Sensors 19, 11 (2019), 2433.
[53]
Fahimul Haque, Vahid Dehghanian, Abraham O. Fapojuwo, and John Nielsen. 2018. A sensor fusion-based framework for floor localization. IEEE Sensors Journal 19, 2 (2018), 623–631.
[54]
Robert Harle. 2013. A survey of indoor inertial positioning systems for pedestrians. IEEE Communications Surveys & Tutorials 15, 3 (2013), 1281–1293.
[55]
A. K. M. Mahtab Hossain and Wee-Seng Soh. 2015. A survey of calibration-free indoor positioning systems. Computer Communications 66 (2015), 1–13.
[56]
Chengkai Huang, Gong Zhang, Zhuqing Jiang, Chao Li, Yupeng Wang, and Xueyang Wang. 2014. Smartphone-based indoor position and orientation tracking fusing inertial and magnetic sensing. In 2014 International Symposium on Wireless Personal Multimedia Communications (WPMC’14). IEEE, 215–220.
[57]
Ryosuke Ichikari, Luis Carlos Manrique Ruiz, Masakatsu Kourogi, Takeshi Kurata, Tomoaki Kitagawa, and Sota Yoshii. 2015. Indoor floor-level detection by collectively decomposing factors of atmospheric pressure. In 2015 International Conference on Indoor Positioning and Indoor Navigation (IPIN’15). IEEE, 1–11.
[58]
Winfred Ingabire, Hadi Larijani, Ryan M. Gibson, and Ayyaz-UI-Haq Qureshi. 2022. LoRaWAN based indoor localization using random neural networks. Information 13, 6 (2022), 303.
[59]
Raj Jain. 1990. The Art of Computer Systems Performance Analysis: Techniques for Experimental Design, Measurement, Simulation, and Modeling. John Wiley & Sons.
[60]
Beakcheol Jang and Hyunjung Kim. 2018. Indoor positioning technologies without offline fingerprinting map: A survey. IEEE Communications Surveys & Tutorials 21, 1 (2018), 508–525.
[61]
Thomas Janssen, Maarten Weyn, and Rafael Berkvens. 2017. Localization in low power wide area networks using wi-fi fingerprints. Applied Sciences 7, 9 (2017), 936.
[62]
Wojciech Jaworski, Paweł Wilk, Paweł Zborowski, Witold Chmielowiec, Andrew YongGwon Lee, and Abhishek Kumar. 2017. Real-time 3d indoor localization. In 2017 International Conference on Indoor Positioning and Indoor Navigation (IPIN’17). IEEE, 1–8.
[63]
Je-sung Jeon, Youngsun Kong, Yunyoung Nam, and Kangbin Yim. 2015. An indoor positioning system using Bluetooth RSSI with an accelerometer and a barometer on a smartphone. In 2015 10th International Conference on Broadband and Wireless Computing, Communication and Applications (BWCCA’15). IEEE, 528–531.
[64]
Ojas Kanhere and Theodore S. Rappaport. 2021. Position location for futuristic cellular communications: 5G and beyond. IEEE Communications Magazine 59, 1 (2021), 70–75.
[65]
Hassan A. Karimi. 2013. Advanced Location-based Technologies and Services. Taylor & Francis.
[66]
Kornkanok Khaoampai, Kulit Na Nakorn, and Kultida Rojviboonchai. 2015. FloorLoc-SL: Floor localization system with fingerprint self-learning mechanism. International Journal of Distributed Sensor Networks 11, 11 (2015), 523403.
[67]
Hiam M. Khoury and Vineet R. Kamat. 2009. Evaluation of position tracking technologies for user localization in indoor construction environments. Automation in Construction 18, 4 (2009), 444–457.
[68]
Kyungki Kim, Sining Li, Milad Heydariaan, Nour Smaoui, Omprakash Gnawali, Wonho Suh, Min Jae Suh, and Jung In Kim. 2021. Feasibility of LoRa for smart home indoor localization. Applied Sciences 11, 1 (2021), 415.
[69]
Kyeong Soo Kim, Sanghyuk Lee, and Kaizhu Huang. 2018. A scalable deep neural network architecture for multi-building and multi-floor indoor localization based on Wi-Fi fingerprinting. Big Data Analytics 3, 1 (2018), 1–17.
[70]
Kyeong Soo Kim, Ruihao Wang, Zhenghang Zhong, Zikun Tan, Haowei Song, Jaehoon Cha, and Sanghyuk Lee. 2018. Large-scale location-aware services in access: Hierarchical building/floor classification and location estimation using Wi-Fi fingerprinting based on deep neural networks. Fiber and Integrated Optics 37, 5 (2018), 277–289.
[71]
Seong-Sik Kim, Jae-Won Kim, and Dong-Soo Han. 2017. Floor detection using a barometer sensor in a smartphone. In 2017 International Conference on Indoor Positioning and Indoor Navigation (IPIN’17).
[72]
Neil E. Klepeis, William C. Nelson, Wayne R. Ott, John P. Robinson, Andy M. Tsang, Paul Switzer, Joseph V. Behar, Stephen C. Hern, and William H. Engelmann. 2001. The National Human Activity Pattern Survey (NHAPS): A resource for assessing exposure to environmental pollutants. Journal of Exposure Science & Environmental Epidemiology 11, 3 (2001), 231–252.
[73]
Hakan Koyuncu and Shuang Hua Yang. 2010. A survey of indoor positioning and object locating systems. International Journal of Computer Science and Network Security 10, 5 (2010), 121–128.
[74]
Jayakanth Kunhoth, AbdelGhani Karkar, Somaya Al-Maadeed, and Abdulla Al-Ali. 2020. Indoor positioning and wayfinding systems: A survey. Human-centric Computing and Information Sciences 10, 1 (2020), 1–41.
[75]
Christos Laoudias, Adriano Moreira, Sunwoo Kim, Sangwoo Lee, Lauri Wirola, and Carlo Fischione. 2018. A survey of enabling technologies for network localization, tracking, and navigation. IEEE Communications Surveys & Tutorials 20, 4 (2018), 3607–3644.
[76]
Alexandru Lavric, Adrian I. Petrariu, and Valentin Popa. 2019. Long range Sigfox communication protocol scalability analysis under large-scale, high-density conditions. IEEE Access 7 (2019), 35816–35825.
[77]
Kyunghoon Lee, Wonjun Hwang, Hyunseok Ryu, and Hyung-Jin Choi. 2017. New TDOA-based three-dimensional positioning method for 3GPP LTE system. ETRI Journal 39, 2 (2017), 264–274.
[78]
Kejiong Li, John Bigham, Eliane L. Bodanese, and Laurissa Tokarchuk. 2013. Location estimation in large indoor multi-floor buildings using hybrid networks. In 2013 IEEE Wireless Communications and Networking Conference (WCNC’13). IEEE, 2137–2142.
[79]
You Li, Zhouzheng Gao, Zhe He, Peng Zhang, Ruizhi Chen, and Naser El-Sheimy. 2018. Multi-sensor multi-floor 3D localization with robust floor detection. IEEE Access 6 (2018), 76689–76699.
[80]
Guangwen Liu, Masayuki Iwai, Yoshito Tobe, Dunstan Matekenya, Khan Muhammad Asif Hossain, Masaki Ito, and Kaoru Sezaki. 2014. Beyond horizontal location context: Measuring elevation using smartphone’s barometer. In Proceedings of the 2014 ACM International Joint Conference on Pervasive and Ubiquitous Computing: Adjunct Publication. 459–468.
[81]
Hui Liu, Houshang Darabi, Pat Banerjee, and Jing Liu. 2007. Survey of wireless indoor positioning techniques and systems. IEEE Transactions on Systems, Man, and Cybernetics, Part C (Applications and Reviews) 37, 6 (2007), 1067–1080.
[82]
Hongbo Liu, Jie Yang, Simon Sidhom, Yan Wang, Yingying Chen, and Fan Ye. 2013. Accurate WiFi based localization for smartphones using peer assistance. IEEE Transactions on Mobile Computing 13, 10 (2013), 2199–2214.
[83]
Chi-Chung Lo, Ting-Hui Chiang, Tsu-Kuang Lee, Ling-Jyh Chen, and Yu-Chee Tseng. 2018. Wireless location tracking by a sensor-assisted particle filter and floor plans in a 2.5-D space. In 2018 IEEE Wireless Communications and Networking Conference (WCNC’18). IEEE, 1–6.
[84]
Elena Simona Lohan, Jukka Talvitie, Pedro Figueiredo e Silva, Henri Nurminen, Simo Ali-Löytty, and Robert Piché. 2015. Received signal strength models for WLAN and BLE-based indoor positioning in multi-floor buildings. In 2015 International Conference on Localization and GNSS (ICL-GNSS’15). IEEE, 1–6.
[85]
Matthias Lott and Ingo Forkel. 2001. A multi-wall-and-floor model for indoor radio propagation. In IEEE VTS 53rd Vehicular Technology Conference, Spring 2001. Proceedings (Cat. No. 01CH37202), Vol. 1. IEEE, 464–468.
[86]
Juan Luo, Zhenyan Zhang, Chun Wang, Chang Liu, and Degui Xiao. 2019. Indoor multifloor localization method based on WiFi fingerprints and LDA. IEEE Transactions on Industrial Informatics 15, 9 (2019), 5225–5234.
[87]
Halgurd S. Maghdid, Ihsan Alshahib Lami, Kayhan Zrar Ghafoor, and Jaime Lloret. 2016. Seamless outdoors-indoors localization solutions on smartphones: Implementation and challenges. ACM Computing Surveys (CSUR) 48, 4 (2016), 1–34.
[88]
Kriangkrai Maneerat, Kamol Kaemarungsi, and Chutima Prommak. 2016. Robust floor determination algorithm for indoor wireless localization systems under reference node failure. Mobile Information Systems 2016, 1 (2016), 4961565.
[89]
Kriangkrai Maneerat and Chutima Prommak. 2014. An enhanced floor estimation algorithm for indoor wireless localization systems using confidence interval approach. International Journal of Electronics and Communication Engineering 8, 7 (2014), 1182–1186.
[90]
Kriangkrai Maneerat, Chutima Prommak, and Kamol Kaemarungsi. 2014. Floor estimation algorithm for wireless indoor multi-story positioning systems. In 2014 11th International Conference on Electrical Engineering/Electronics, Computer, Telecommunications and Information Technology (ECTI-CON’14). IEEE, 1–5.
[91]
Vahideh Moghtadaiee, Andrew G. Dempster, and Samsung Lim. 2011. Indoor localization using FM radio signals: A fingerprinting approach. In 2011 International Conference on Indoor Positioning and Indoor Navigation. IEEE, 1–7.
[92]
Kartik Muralidharan, Azeem Javed Khan, Archan Misra, Rajesh Krishna Balan, and Sharad Agarwal. 2014. Barometric phone sensors: More hype than hope! In Proceedings of the 15th Workshop on Mobile Computing Systems and Applications. 1–6.
[93]
Khanh Nguyen-Huu and Seon-Woo Lee. 2021. A multi-floor indoor pedestrian localization method using landmarks detection for different holding styles. Mobile Information Systems 2021, 1 (2021), 6617417.
[94]
Michał Nowicki and Jan Wietrzykowski. 2017. Low-effort place recognition with WiFi fingerprints using deep learning. In International Conference Automation. Springer, 575–584.
[95]
Henri Nurminen. 2017. Estimation algorithms for non-gaussian state-space models with application to positioning. (2017).
[96]
Huthaifa Obeidat, Wafa Shuaieb, Omar Obeidat, and Raed Abd-Alhameed. 2021. A review of indoor localization techniques and wireless technologies. Wireless Personal Communications 119 (2021), 289–327.
[97]
Veljo Otsason, Alex Varshavsky, Anthony LaMarca, and Eyal De Lara. 2005. Accurate GSM indoor localization. In International Conference on Ubiquitous Computing. Springer, 141–158.
[98]
Kaveh Pahlavan and Prashant Krishnamurthy. 2001. Principles of Wireless Networks. Vol. 1. Prentice Hall PTR.
[99]
Jun-geun Park, Dorothy Curtis, Seth Teller, and Jonathan Ledlie. 2011. Implications of device diversity for organic localization. In 2011 Proceedings IEEE INFOCOM. IEEE, 3182–3190.
[100]
Paolo Pivato, Luigi Palopoli, and Dario Petri. 2011. Accuracy of RSS-based centroid localization algorithms in an indoor environment. IEEE Transactions on Instrumentation and Measurement 60, 10 (2011), 3451–3460.
[101]
Sakshi Popli, Rakesh Kumar Jha, and Sanjeev Jain. 2018. A survey on energy efficient narrowband Internet of Things (NBIoT): Architecture, application and challenges. IEEE Access 7 (2018), 16739–16776.
[102]
Francesco Potortì, Sangjoon Park, Antonino Crivello, Filippo Palumbo, Michele Girolami, Paolo Barsocchi, Soyeon Lee, Joaquín Torres-Sospedra, Antonio Ramón Jiménez Ruiz, Antoni Pérez-Navarro, Germán Martín Mendoza-Silva, Fernando Seco, Miguel Ortiz, Johan Perul, Valerie Renaudin, Hyunwoong Kang, Soyoung Park, Jae Hong Lee, Chan Gook Park, Jisu Ha, Jaeseung Han, Changjun Park, Keunhye Kim, Yonghyun Lee, Seunghun Gye, Keumryeol Lee, Eunjee Kim, Jeong-Sik Choi, Yang-Seok Choi, Shilpa Talwar, Seong Yun Cho, Boaz Ben-Moshe, Alex Scherbakov, Leonid Antsfeld, Emilio Sansano-Sansano, Boris Chidlovskii, Nikolai Kronenwett, Silvia Prophet, Yael Landay, Revital Marbel, Lingxiang Zheng, Ao Peng, Zhichao Lin, Bang Wu, Chengqi Ma, Stefan Poslad, David R. Selviah, Wei Wu, Zixiang Ma, Wenchao Zhang, Dongyan Wei, Hong Yuan, Jun-Bang Jiang, Shao-Yung Huang, Jing-Wen Liu, Kuan-Wu Su, Enq-Shiou Leu, Kazuki Nishiguchi, Walid Bousselham, Hideaki Uchiyama, Diego Thomas, Atsushi Shimada, Rin-Ichiro Taniguchi, Vicente Cortés Puschel, Tomás Lungenstrass Poulsen, Imran Ashraf, Chanseok Lee, Muhammad Usman Ali, Yeongjun Im, Gunzung Kim, Jeongsook Eom, Soojung Hur, Yongwan Park, Miroslav Opiela, Adriano Moreira, Maria João Nicolau, Cristiano Pendão, Ivo Silva, Filipe Meneses, AntÓnio Costa, Jens Trogh, David Plets, Ying-Ren Chien, Tzu-Yu Chang, Shih-Hau Fang, and Yu Tsao. 2020. The IPIN 2019 indoor localisation competition–Description and results. IEEE Access 8 (2020), 206674–206718.
[103]
Guowen Qi, Yi Jin, and Jun Yan. 2018. RSSI-based floor localization using principal component analysis and ensemble extreme learning machine technique. In 2018 IEEE 23rd International Conference on Digital Signal Processing (DSP’18). IEEE, 1–5.
[104]
Peter J. Radcliffe, Karina Gomez Chavez, Paul Beckett, Justin Spangaro, and Conrad Jakob. 2017. Usability of LoRaWAN technology in a central business district. In 2017 IEEE 85th Vehicular Technology Conference (VTC Spring’17). IEEE, 1–5.
[105]
Valentin Radu and Mahesh K. Marina. 2013. HiMLoc: Indoor smartphone localization via activity aware pedestrian dead reckoning with selective crowdsourced WiFi fingerprinting. In International Conference on Indoor Positioning and Indoor Navigation. IEEE, 1–10.
[106]
Kopparapu Venkata Ramana, Niu Jianwei, Muhammad Ali Abdul Aziz, and Mir Yasir Umair. 2016. A robust multi-cue blending-based approach for floor detection. In 2016 13th International Bhurban Conference on Applied Sciences and Technology (IBCAST’16). IEEE, 647–653.
[107]
Alireza Razavi, Mikko Valkama, and Elena-Simona Lohan. 2015. K-means fingerprint clustering for low-complexity floor estimation in indoor mobile localization. In 2015 IEEE Globecom Workshops (GC Wkshps’15). IEEE, 1–7.
[108]
Alireza Razavi, Mikko Valkama, and Elena Simona Lohan. 2016. Robust statistical approaches for RSS-based floor detection in indoor localization. Sensors 16, 6 (2016), 793.
[109]
Chao Ruan, Min Yu, Xiaona He, and Binbin Song. 2018. An indoor floor positioning method based on smartphone’s barometer. In 2018 Ubiquitous Positioning, Indoor Navigation and Location-Based Services (UPINLBS’18). IEEE, 1–9.
[110]
Angelo Maria Sabatini and Vincenzo Genovese. 2013. A stochastic approach to noise modeling for barometric altimeters. Sensors 13, 11 (2013), 15692–15707.
[111]
Nasir Saeed, Haewoon Nam, Tareq Y. Al-Naffouri, and Mohamed-Slim Alouini. 2019. A state-of-the-art survey on multidimensional scaling-based localization techniques. IEEE Communications Surveys & Tutorials 21, 4 (2019), 3565–3583.
[112]
Xingfa Shen, Yueshen Chen, Jianhui Zhang, Landi Wang, Guojun Dai, and Tian He. 2015. BarFi: Barometer-aided Wi-Fi floor localization using crowdsourcing. In 2015 IEEE 12th International Conference on Mobile Ad Hoc and Sensor Systems. IEEE, 416–424.
[113]
Jian Shi and Yoan Shin. 2013. A low-complexity floor determination method based on WiFi for multi-floor buildings. In Proceedings of the 9th Advanced International Conference on Telecommunications, Vol. 2328. Citeseer.
[114]
Shweta Shrestha, Jukka Talvitie, and Elena Simona Lohan. 2013. Deconvolution-based indoor localization with WLAN signals and unknown access point locations. In 2013 International Conference on Localization and GNSS (ICL-GNSS’13). IEEE, 1–6.
[115]
Wonsang Song, Jae Woo Lee, Byung Suk Lee, and Henning Schulzrinne. 2014. Finding 9-1-1 callers in tall buildings. In Proceeding of IEEE International Symposium on a World of Wireless, Mobile and Multimedia Networks 2014. IEEE, 1–9.
[116]
Xudong Song, Xiaochen Fan, Chaocan Xiang, Qianwen Ye, Leyu Liu, Zumin Wang, Xiangjian He, Ning Yang, and Gengfa Fang. 2019. A novel convolutional neural network based indoor localization framework with WiFi fingerprinting. IEEE Access 7 (2019), 110698–110709.
[117]
Petros Spachos and Konstantinos Plataniotis. 2020. BLE beacons in the smart city: Applications, challenges, and research opportunities. IEEE Internet of Things Magazine 3, 1 (2020), 14–18.
[118]
Lin Sun, ZengWei Zheng, Tao He, and Fei Li. 2015. Multifloor Wi-Fi localization system with floor identification. International Journal of Distributed Sensor Networks 11, 7 (2015), 131523.
[119]
Ekaterina Svertoka, Alexandru Rusu-Casandra, Radim Burget, Ion Marghescu, Jiri Hosek, and Aleksandr Ometov. 2022. LoRaWAN: Lost for localization? IEEE Sensors Journal 22, 23 (2022), 23307–23319.
[120]
Sushil Tiwari and Vinod Kumar Jain. 2018. Smartphone based improved floor determination technique for multi-floor buildings. In Proceedings of the World Congress on Engineering, Vol. 1.
[121]
Joaquín Torres-Sospedra, Antonio R. Jiménez, Adriano Moreira, Tomás Lungenstrass, Wei-Chung Lu, Stefan Knauth, Germán Martín Mendoza-Silva, Fernando Seco, Antoni Pérez-Navarro, Maria João Nicolau, et al. 2018. Off-line evaluation of mobile-centric indoor positioning systems: The experiences from the 2017 IPIN competition. Sensors 18, 2 (2018), 487.
[122]
Waqas ur Rehman, Eyal De Lara, and Stefan Saroiu. 2008. CILoS: A CDMA indoor localization system. In Proceedings of the 10th International Conference on Ubiquitous Computing. 104–113.
[123]
Salvatore Vanini and Silvia Giordano. 2013. Adaptive context-agnostic floor transition detection on smart mobile devices. In 2013 IEEE International Conference on Pervasive Computing and Communications Workshops (PERCOM Workshops’13). IEEE, 2–7.
[124]
Alex Varshavsky, Anthony LaMarca, Jeffrey Hightower, and Eyal De Lara. 2007. The Skyloc floor localization system. In 5th Annual IEEE International Conference on Pervasive Computing and Communications (PerCom’07). IEEE, 125–134.
[125]
Federico Viani, Paolo Rocca, Giacomo Oliveri, Daniele Trinchero, and Andrea Massa. 2011. Localization, tracking, and imaging of targets in wireless sensor networks: An invited review. Radio Science 46, 5 (2011), 1–12.
[126]
Brieuc Viel and Mikael Asplund. 2014. Why is fingerprint-based indoor localization still so hard? In 2014 IEEE International Conference on Pervasive Computing and Communication Workshops (PERCOM WORKSHOPS’14). IEEE, 443–448.
[127]
P. Vlach, B. Segal, J. LeBel, and T. Pavlasek. 1999. Cross-floor signal propagation inside a contemporary ferro-concrete building at 434, 862, and 1705 MHz. IEEE Transactions on Antennas and Propagation 47, 7 (1999), 1230–1232.
[128]
Hui Wang, Henning Lenz, Andrei Szabo, Uwe D. Hanebeck, and Joachim Bamberger. 2006. Fusion of barometric sensors, WLAN signals and building information for 3-D indoor/campus localization. In Proceedings of International Conference on Multisensor Fusion and Integration for Intelligent Systems (MFI’06). 426–432.
[129]
Hui Wang, Andrei Szabo, Joachim Bamberger, and Uwe D. Hanebeck. 2008. Simultaneous multi-information fusion and parameter estimation for robust 3-D indoor positioning systems. In 2008 IEEE International Conference on Multisensor Fusion and Integration for Intelligent Systems. IEEE, 445–450.
[130]
Hua-Yan Wang, Vincent W. Zheng, Junhui Zhao, and Qiang Yang. 2010. Indoor localization in multi-floor environments with reduced effort. In 2010 IEEE International Conference on Pervasive Computing and Communications (PerCom’10). IEEE, 244–252.
[131]
Shie-Yuan Wang, Yu-Hsiang Cheng, and Jenn-Hwan Tarng. 2019. Improving the localization accuracy for Sigfox low-power wide area networks. In 2019 IEEE Symposium on Computers and Communications (ISCC’19). IEEE, 1–6.
[132]
Yao-hui Wang, Hu Li, Xin-long Luo, Qi-ming Sun, and Jin-nan Liu. 2014. A 3D fingerprinting positioning method based on cellular networks. International Journal of Distributed Sensor Networks 10, 7 (2014), 248981.
[133]
Benedict W. Wheeler, Ashley R. Cooper, Angie S. Page, and Russell Jago. 2010. Greenspace and children’s physical activity: A GPS/GIS analysis of the PEACH project. Preventive Medicine 51, 2 (2010), 148–152.
[134]
Oliver Woodman and Robert Harle. 2008. Pedestrian localisation for indoor environments. In Proceedings of the 10th International Conference on Ubiquitous Computing. 114–123.
[135]
Hao Xia, Xiaogang Wang, Yanyou Qiao, Jun Jian, and Yuanfei Chang. 2015. Using multiple barometers to detect the floor location of smart phones with built-in barometric sensors for indoor positioning. Sensors 15, 4 (2015), 7857–7877.
[136]
Jiang Xiao, Zimu Zhou, Youwen Yi, and Lionel M. Ni. 2016. A survey on wireless indoor localization from the device perspective. ACM Computing Surveys (CSUR) 49, 2 (2016), 1–31.
[137]
Lisheng Xu, Cong Feng, Ying Wang, Yupeng Yao, and Max Q.-H. Meng. 2010. Variation of exterior telemetry links of capsule antenna ingested in human body. In 2010 8th World Congress on Intelligent Control and Automation. IEEE, 2269–2272.
[138]
Lisheng Xu, Feifei Yang, Yuqi Jiang, Lei Zhang, Cong Feng, and Nan Bao. 2011. Variation of received signal strength in wireless sensor network. In 2011 3rd International Conference on Advanced Computer Control. IEEE, 151–154.
[139]
Zhengyi Xu, Jianming Wei, Jinxin Zhu, and Weijun Yang. 2017. A robust floor localization method using inertial and barometer measurements. In 2017 International Conference on Indoor Positioning and Indoor Navigation (IPIN’17). IEEE, 1–8.
[140]
Zheng Yang, Chenshu Wu, Zimu Zhou, Xinglin Zhang, Xu Wang, and Yunhao Liu. 2015. Mobility increases localizability: A survey on wireless indoor localization using inertial sensors. ACM Computing Surveys (CSUR) 47, 3 (2015), 1–34.
[141]
Zheng Yang, Zimu Zhou, and Yunhao Liu. 2013. From RSSI to CSI: Indoor localization via channel response. ACM Computing Surveys (CSUR) 46, 2 (2013), 1–32.
[142]
Haibo Ye, Kai Dong, and Tao Gu. 2018. Himeter: Telling you the height rather than the altitude. Sensors 18, 6 (2018), 1712.
[143]
Haibo Ye, Tao Gu, Xianping Tao, and Jian Lu. 2014. F-Loc: Floor localization via crowdsourcing. In 2014 20th IEEE International Conference on Parallel and Distributed Systems (ICPADS’14). IEEE, 47–54.
[144]
Haibo Ye, Tao Gu, Xianping Tao, and Jian Lu. 2016. Scalable floor localization using barometer on smartphone. Wireless Communications and Mobile Computing 16, 16 (2016), 2557–2571.
[145]
Haibo Ye, Tao Gu, Xiaorui Zhu, Jinwei Xu, Xianping Tao, Jian Lu, and Ning Jin. 2012. FTrack: Infrastructure-free floor localization via mobile phone sensing. In 2012 IEEE International Conference on Pervasive Computing and Communications. IEEE, 2–10.
[146]
Ping Yi, Minjie Yu, Ziqiao Zhou, Wei Xu, Qingquan Zhang, and Ting Zhu. 2014. A three-dimensional wireless indoor localization system. Journal of Electrical and Computer Engineering 2014, 1 (2014), 149016.
[147]
Min Yu, Feng Xue, Chao Ruan, and Hang Guo. 2019. Floor positioning method indoors with smartphone’s barometer. Geo-spatial Information Science 22, 2 (2019), 138–148.
[148]
Faheem Zafari, Athanasios Gkelias, and Kin K. Leung. 2019. A survey of indoor localization systems and technologies. IEEE Communications Surveys & Tutorials 21, 3 (2019), 2568–2599.
[149]
Yongliang Zhang, Lin Ma, Bin Wang, and Danyang Qin. 2020. Building floor identification method based on DAE-LSTM in cellular network. In 2020 IEEE 91st Vehicular Technology Conference (VTC Spring’20). IEEE, 1–5.
[150]
Fang Zhao, Haiyong Luo, Xuqiang Zhao, Zhibo Pang, and Hyuncheol Park. 2015. HYFI: Hybrid floor identification based on wireless fingerprinting and barometric pressure. IEEE Transactions on Industrial Informatics 13, 1 (2015), 330–341.
[151]
Zengwei Zheng, Yuanyi Chen, Tao He, Fei Li, and Dan Chen. 2015. Weight-RSS: A calibration-free and robust method for WLAN-Based indoor positioning. International Journal of Distributed Sensor Networks 11, 4 (2015), 573582.
[152]
Mu Zhou, Bin Wang, Zengshan Tian, Jiacheng Wang, and Qiankun Zhang. 2017. A case study of cross-floor localization system using hybrid wireless sensing. In 2017 IEEE Global Communications Conference (GLOBECOM’17). IEEE, 1–6.

Index Terms

  1. Enabling Technologies and Techniques for Floor Identification

    Recommendations

    Comments

    Information & Contributors

    Information

    Published In

    cover image ACM Computing Surveys
    ACM Computing Surveys  Volume 57, Issue 1
    January 2025
    984 pages
    EISSN:1557-7341
    DOI:10.1145/3696794
    • Editors:
    • David Atienza,
    • Michela Milano
    Issue’s Table of Contents

    Publisher

    Association for Computing Machinery

    New York, NY, United States

    Publication History

    Published: 07 October 2024
    Online AM: 17 July 2024
    Accepted: 11 July 2024
    Revised: 29 June 2024
    Received: 12 June 2022
    Published in CSUR Volume 57, Issue 1

    Check for updates

    Author Tags

    1. Indoor localization
    2. location-based services
    3. floor identification
    4. 5G

    Qualifiers

    • Survey

    Funding Sources

    • Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education

    Contributors

    Other Metrics

    Bibliometrics & Citations

    Bibliometrics

    Article Metrics

    • 0
      Total Citations
    • 722
      Total Downloads
    • Downloads (Last 12 months)722
    • Downloads (Last 6 weeks)241
    Reflects downloads up to 18 Jan 2025

    Other Metrics

    Citations

    View Options

    View options

    PDF

    View or Download as a PDF file.

    PDF

    eReader

    View online with eReader.

    eReader

    Login options

    Full Access

    Media

    Figures

    Other

    Tables

    Share

    Share

    Share this Publication link

    Share on social media