A Hidden Markov Model based smartphone heterogeneity resilient portable indoor localization framework

Abstract

Indoor localization is an emerging application domain that promises to enhance the way we navigate in various indoor environments, as well as track equipment and people. Wireless signal-based fingerprinting is one of the leading approaches for indoor localization. Using ubiquitous Wi-Fi access points and Wi-Fi transceivers in smartphones has enabled the possibility of fingerprinting-based localization techniques that are scalable and low-cost. But the variety of Wi-Fi hardware modules and software stacks used in today's smartphones introduce errors when using Wi-Fi based fingerprinting approaches across devices, which reduces localization accuracy. We propose a framework called SHERPA-HMM that enables efficient porting of indoor localization techniques across mobile devices, to maximize accuracy. An in-depth analysis of our framework shows that it can deliver up to 8× more accurate results as compared to state-of-the-art localization techniques for a variety of environments.

Introduction

The arrival of Global Positioning System (GPS) technology within smartphones has revolutionized the way we navigate in the outdoor world. Today, indoor localization technology holds a similar potential to disrupt the way we navigate within indoor spaces that are unreachable by GPS. An example scenario is localizing patients, staff, and equipment in large hospitals and assisted living facilities. Precise location information can allow first responders closest to a patient to be notified in emergencies. Some startups (e.g., Shopkick, Zebra) are also beginning to provide indoor localization services that can help customers locate products inside a store [1].

Unlike GPS for outdoor localization, no standardized solution exists for indoor localization. Therefore, a myriad of techniques have been developed that use various sensors and radio frequencies. Some commonly utilized radio signals are Bluetooth, ZigBee, and Wi-Fi [2]. Among these, Wi-Fi based indoor localization has been the most widely researched, due to its low setup cost and easy availability. Today, Wi-Fi access points are deployed in most indoor locales around the world and all smartphones support Wi-Fi connectivity.

Despite the advantages of Wi-Fi based indoor localization, there are also some drawbacks. Many prior solutions perform indoor localization by measuring Wi-Fi Received Signal Strength Indicator (RSSI) values and calculating distance from Wi-Fi Access Points (WAPs). These works assume that wireless signal strength reduces in a deterministic manner as a function of distance from a signal source (i.e., WAP). But Wi-Fi signals suffer from weak wall penetration, multipath fading, and shadowing effects in real-world environments, making it difficult to establish a direct mathematical relationship between RSSI and distance from WAPs. These issues have served as a motivation for using fingerprinting-based techniques. Fingerprinting is based on the idea that each indoor location exhibits a unique signature of WAP RSSI values. Due to its independence from the RSSI-distance relationship, fingerprinting can overcome some of the aforementioned drawbacks with Wi-Fi based indoor localization.

Fingerprinting is usually carried out in two phases. In the first phase (called offline or training phase), the RSSI values for visible WAPs are collected along indoor paths of interest. The resulting database of values may further be used to train models (e.g., machine learning-based) for location estimation. In the second phase (online or testing phase), the models are deployed on smartphones and used to predict the location of the user carrying the smartphone, based on real-time readings of WAP RSSI values on the smartphone.

A majority of the literature that utilizes fingerprinting employs the same smartphone for (offline) data collection and (online) location prediction [[3], [4], [5], [6], [7]]. This assumes that in a real-world setting, users would have access to the same smartphone as the one used in the offline phase. But today's diverse smartphone market, with various brands and models, largely invalidates such an assumption. In reality, the smartphone user base is a distribution of heterogeneous devices that vary in antenna gain, Wi-Fi chipset, OS version, etc. [8,[25], [26], [27], [28], [29], [30]].

Recent work has shown that the perceived Wi-Fi RSSI values for a given location captured by different smartphones can vary significantly [9]. This variation degrades the localization accuracy of conventional fingerprinting. Therefore, there is a need for portable and device heterogeneity-aware fingerprinting techniques. In this paper, we present a lightweight Wi-Fi RSSI fingerprinting framework for Smartphone Heterogeneity Resilient Portable localization with Hidden Markov Models (SHERPA-HMM) that is portable across smartphones with minimal accuracy loss. The novel contributions of our work are:

• We conduct an in-depth analysis of Wi-Fi fingerprinting across smartphones to emphasize the importance of device heterogeneity-resilient indoor localization;

• We formulate the indoor localization problem as a Hidden Markov Model (HMM) that utilizes heterogeneity resilient metrics for user path prediction;

• We design the SHERPA-HMM framework for portable Wi-Fi fingerprinting-based indoor localization; SHERPA-HMM employs a lightweight software-based approach to combine noisy fingerprints over distinct smartphones and pattern matching/filtering to improve location accuracy;

• We evaluate SHERPA-HMM against state-of-the-art localization techniques, across a variety of Android-based smartphones that are used for indoor localization along paths in real buildings.