Performance evaluation of object localization based on active radio frequency identification technology
Introduction
As a result of the mandates from government agencies and large organizations, such as the U.S. Department of Defense and Wal-Mart, requiring shipment of goods identifiable by radio frequency identification (RFID) tags, and the benefits this technology can bring, RFID technology is being widely adopted for tracking and tracing purpose in an unprecedented pace. A major technical advantage for RFID is that the communication between the reader and tag does not rely on line-of-sight. This makes it a viable solution of higher efficiency to replace barcodes in identifying and tracking goods, devices, animals, and even human beings. Applications of RFID technology have been reported in various industries, including manufacturing industry [1], [2], food industry [3], transportation and logistics [4], [5], agricultural industry [6], and pharmaceutical stores and healthcare facilities [7].
In the traditional applications of RFID technology, information collection and processing is limited to the general existence of objects in the reading range of an RFID reader. However, in addition to the “Yes or No” existence information, the accurate location of an object is required for many situations, such as finding missing items in a warehouse [8], locating equipments in construction sites [9], avoiding collision between vehicles [10], and rescuing victims in underground mines [11]. In fact, a variety of techniques have been developed for localization purpose, which are based on radio frequency (RF), ultrasound, or infrared technology [12]. Adding the localization devices to an existing RFID application could be challenging due to the geometrical constraints—the device may not fit into the object if it is not compact enough. Also, these techniques may not be applicable for some applications in nature. For example, a global position system (GPS) is hardly effective in most indoor environments, and a video localization system cannot be used in most warehouses due to the frequent blockage of light by the structures. Even without these technical obstacles, introducing a localization technique to an RFID application will still incur additional cost. Therefore, the most cost effective way is to exploit RFID technology itself for localization. Indeed, RFID technology is technically feasible to position objects. This is because radio signal strength indicator (RSSI) data received by the RFID reader reflects the distance information between the tag and the reader. A stronger signal usually means a shorter distance. For RFID systems lack of the capability of RSSI data collection, the readrate information more or less represents distance information [13]. By combining RSSI or readrate data with the localization methods developed in literature, the location of tagged objects can be estimated. It should be noted that active RFID is usually preferred over passive RFID for localization purpose because the former has a longer communication range to cover application area and is less sensitive to environment noises and disruptions.
RFID localization is essentially an instance of the general RF-based localization. As such, it can use the methods developed for other RF-based localization techniques, such as wireless local area network (WLAN) localization. Nevertheless, how to choose a localization method for a particular RF signal source and/or an application remains a challenge. It is unclear that, for active RFID technology, how the method selection, and also how the parameter tuning in each method will affect the results. In this study, we evaluate the performances of three major localization methods based on the experimental RFID data collected in a realistic environment. Not only the feasibility of each method is investigated, but also the results of the methods are compared in terms of localization accuracy and precision. The rest of the paper is organized as follows. A brief background introduction on the three methods is provided in Section 2. The experimental set-up and environment, and data collection schemes are described in Section 3. Then the results and discussion are presented in Section 4. Finally, conclusions are drawn in Section 5.
Section snippets
Localization algorithms
The localization methods reported in literature, such as multi-lateration, nearest-neighbor, Bayesian inference, proximity, and support vector machines, can be classified into two groups. The first group of methods estimates the position of an object in two steps. The first step is to estimate the corresponding distance for an RF signal reading based on an established relation between distance and RF signal reading, such as time-of-arrival (TOA) or RSSI data. The second step is to compute the
Experiment
We conducted the data collection experiment in the lobby area of the Memorial Union at North Dakota State University. The rectangular test area has the dimension of 27.43 m × 12.19 m. The major obstacles in the area include six pillars with a cross-section area of 0.37 m2 and others, such as floor lamps and furniture. The set-up includes an RX201 RFID reader and ten TG800 asset RFID tags made by Wavetrend Technologies. The tags were fixed on tables at the same height, and the reader was connected to
Result and analysis
As mentioned above, data collection at reference nodes is for model training, which is to provide inputs for localization computation on target nodes. To investigate whether it is necessary to re-collect data at reference nodes and re-train the model every time RSSI readings are collected at target nodes, we studied the evolution of RSSI signal of the RFID tags in the test environment. Fig. 2 shows the RSSI readings of three tags for a period of 50 h with the presence of little traffic. It can
Conclusion
Using RFID devices to localize objects is innovative, and it can become an attractive add-on feature for numerous applications in which RFID tagging is already mandated. In this paper, we have presented the first-time comparison study on the performances of active RFID localization using three localization methods. Unlike the common simulation approach reported in literature, we collected the experimental data in a realistic environment to ensure the validity of the study. In the experiment,
Junyi Zhou received the BS degree in mechanical engineering and MS degree in management science and engineering from Southeast University, China, in 1998 and 2003 respectively. From July 1998 to August 2000, he worked as a process engineer in China. From April 2003 to July 2006, he was a lecturer at Anhui University of Technology, China. Currently he is a doctoral candidate in the Department of Industrial and Manufacturing Engineering at North Dakota State University. His research interests
References (23)
- et al.
A proximity-based method for locating RFID tagged objects
Advanced Engineering Informatics
(2007) - et al.
Statistical theory for location fingerprinting in wireless LANs
Computer Networks
(2005) - et al.
Distributed localization in wireless sensor networks: a quantitative comparison
Computer Networks
(2003) - M. Baudin, A. Rao, RFID applications in manufacturing. Available online at: http://www.mmt-inst.com (accessed...
- et al.
Mobile RFID—a Case from Volvo on innovation in SCM
RFID technologies: supply-chain applications and implementation issues
Information System Management
(2005)What does RFID do for the consumers?
Communication for the ACM
(2005)- et al.
RFID enabled manufacturing: fundamentals, methodology and applications
International Journal of Agile Systems and Management
(2006) - et al.
Some methods for privacy in RFID communication
Security in Ad-hoc and Sensor Networks
(2005) IBSS launches healthcare tracking
RFID Journal
(2005)
Cited by (17)
A strain-sensing based scheme for indoor localization: Analysis, algorithm, and demonstration
2014, Measurement: Journal of the International Measurement ConfederationCitation Excerpt :As a result, the importance of selecting an appropriate position sensor has increased. Various solutions currently exist for indoor positioning, navigation, and communications, such as GPS localization [1], inertial navigation system (INS) [2], ultrasonic localization [3], IR [4], Bluetooth [5], radio frequency identification (RFID) [6], and Wi-Fi signals [7,8]. However, due to restrictions of position timing, position accuracy, and complex indoor environments, a flawless positioning technique has not yet been achieved.
A hybrid forecasting approach applied to wind speed time series
2013, Renewable EnergyCitation Excerpt :The PACF plays a very important role in recognizing the non-stationarity of wind speed series and determining the lag orders. The SVM is a useful methodology and has been used extensively for classification, regression and pattern recognition and successfully applied to time-series prediction with satisfactory prediction results in various fields [28–33]. Single SVM shows many unique advantages in addressing small samples and nonlinear or high dimension pattern recognition problems, but it still has difficulties in extracting favorable forecasting results for wind speed series.
Fine tuning support vector machines for short-term wind speed forecasting
2011, Energy Conversion and ManagementCitation Excerpt :Besides, some new methods based on artificial intelligence techniques have also been developed and investigated such as artificial neural networks [10–13] and fuzzy logic [14,15]. In recent years, support vector machines (SVM), a novel and powerful machine learning tool, has been successfully used for time-series prediction with satisfactory prediction results in various fields [16–21]. SVM is an elegant and highly principled learning method for a feed forward network design with a single hidden layer of units that are usually nonlinear [22].
Error analysis of non-collaborative wireless localization in circular-shaped regions
2010, Computer NetworksRFID platform for construction materials management
2023, International Journal of Construction ManagementRFID Based Motion Direction Estimation in Gate Systems
2018, Proceedings of the 2018 IEEE 22nd International Conference on Computer Supported Cooperative Work in Design, CSCWD 2018
Junyi Zhou received the BS degree in mechanical engineering and MS degree in management science and engineering from Southeast University, China, in 1998 and 2003 respectively. From July 1998 to August 2000, he worked as a process engineer in China. From April 2003 to July 2006, he was a lecturer at Anhui University of Technology, China. Currently he is a doctoral candidate in the Department of Industrial and Manufacturing Engineering at North Dakota State University. His research interests include optimization in wireless sensor networks, modeling in wind energy, and RFID applications. He is a member of IIE, IEEE, INFORMS, and SMTA.
Jing Shi received his Ph.D. in industrial engineering from Purdue University, West Lafayette, IN, USA in 2004. He is currently an Assistant Professor in the Department of Industrial Engineering at North Dakota State University, Fargo, ND, USA. His research interests include wireless sensor network/RFID, computer vision, structure health monitoring, and mathematical modeling for renewable energy and health care systems. He has authored more than 30 papers on these topics, which have been published in referred journals and conference proceedings. Dr. Shi is also a member of IEEE, IIE, and ASME