Localization of tagged inhabitants in smart environments M. Javad Akhlaghinia, Student Member, IEEE, Ahmad Lotfi, Senior Member, IEEE, and Caroline Langensiepen School of Science and Technology Nottingham Trent University Nottingham, United Kingdom Mohammadjavad.akhlaghinia@ntu.ac.uk Abstract In this paper, tagging techniques are employed to identify monitored inhabitants in smart environments. The aim of the work is to recognize the presence of the tagged inhabitants in different areas of the environment. RSSI based localizing agents are used in separate areas to measure the distance of the tagged persons from the reader agents. Then a clustering method is applied to the readings from the reader agents to find the data cluster of each area representing the occupied area by the tagged inhabitants. In this paper two different technologies including active RFID and ZigBee wireless technology are compared in distance based and area based experiments. It is shown that the accuracy of occupied area detection is depended on RSSI accuracy factors such as receiver sensitivity, data rate, layout of the environments, and obstacles in the smart environment Keywords- Localizing, occupancy, tagging, active RFID, WSN, ZigBee, radio signal strength, intelligent environment. I. INTRODUCTION mart environments are designed to provide the S inhabitants with facilities such as energy saving, safety, security and convenience. The prediction of inhabitants behaviors plays an important role in creating a smart environment. Therefore, for better prediction the behavior of the inhabitants should be monitored for long durations. Occupancy monitoring of different areas is one of the behaviors which can be used as an important measure [1]. Occupancy detection in an absolute single-occupant environment is feasible by implementing a Wireless Sensor Network (WSN) of motion detection and door contact sensors [2, 3]. But, due to the sensitivity of motion detectors to every single movement of living objects, they cannot identify the inhabitant presence in the monitored areas in a multiple-occupant environment or a single-occupant environment in the presence of visitors. Absolute single-occupancy environments are rather unlikely in the real world. For instance, the situation in which a single person lives in an apartment without having any visitors is not very likely to happen in reality. However, finding a solution for a single-occupant environment is a relevant approach towards finding a solution for an environment with a more complicated situation. As a resolution to multiple occupancy scenarios, a multiple occupancy problem is simplified into a number of single occupancy problems [13] by applying the single occupancy scenario to each inhabitant and distinguishing them by tagging as depicted in Figure 1. Multiple-Occupant Scenario Single-Occupant Single-Occupant Scenario Scenario Single-Occupant Single-Occupant Scenario Scenario Tagging Figure 1 A Multiple Occupancy Scenario derived from Single Occupancy Scenarios. In this paper, instead of finding the exact location of inhabitants, localization addresses the occupancy of different areas (e.g. inhabitant in bedroom, kitchen, lounge, or bathroom) in the monitored environments. This paper is an attempt to investigate tagging technologies and examine how they can help to reduce the problem of localization in multiple occupancy scenarios in practice. Hence, two active Radio Frequency Identification (RFID) products and a ZigBee WSN product are chosen and compared in a series of distance based and area based experiments. In addition, a clustering method is applied to the data collected in area based experiments to improve the accuracy of occupancy detection in different areas. In this paper, section II is a review of current tagging and localization technologies. Section III explains the Regional Clustering Scheme. Experimental results are shown in section IV, and concluding remarks are drawn in the final section. II. TAGGING AND LOCALIZATION TECHNIQUES The tagging of mobile nodes in order to identify and track them has been one of the considerations in creating smart environments [4, 5]. The tagged node can be a person, asset, or other objects [6]. Wireless technology has spread due to the ease of installation, reduction in resources, and provision of mobile communication. This has attracted research towards WSN in smart environments. Currently, most of the
tagging and localizing technologies are based on Received Signal Strength Indicator (RSSI) as explained below: Received Signal Strength Indicator: In wireless technology, RSSI is defined as the strength or the quality of radio signal detected at the receiver side. By measuring this quality factor, the distance between transmitter and receiver can be measured approximately. The distance between transmitter and receiver is in inverse relationship with the transmitted signal s RSSI as given by expression 1. An example of this is depicted in Figure 2. RSSI 1 2 (1) d Therefore, this relation has been used in many localization techniques. Basically, RSSI-based localization has the advantage of using the data communication infrastructure for the localization of mobile nodes. Despite its advantages, RSSI has its limitations due to the physical characteristics of radio signal propagation including multipath propagation and signal fading [5]. Some localization techniques employ hybrid technologies such as ultrasound to reduce these limitations and improve the effectiveness of the RSSI-based localization [4]. Figure 2 - Distance-based Received Signal Strength Indicator. Radio Signal Strength Identification is dependent on following factors: 1. Transmission Power: Transmission power means how far a radio signal can travel. The RSSI-distance measuring resolution decreases for higher powered transmitted signals. 2. Hindrance: The signal strength received by the receiver is affected by environmental s. Metal objects or thick walls can decrease the signal quality remarkably. 3. Receiver Sensitivity: Receiver sensitivity means how faint a Radio Frequency (RF) signal can be detected by the receiver. The lower the power level that the receiver can successfully process, the better the receive sensitivity. Higher sensitivity receivers can provide higher resolution of RSSI distance measurements. 4. Data rate: The accuracy of the RSSI distance measurement is dependent on the baud rate of the wireless communication. Lower baud rates can result in higher resolution RSSI-distance measurements. 5. Interference: Interference of the tag radio signal with other radio signals can decrease the quality of RSSI-distance measurements. There are two major technologies for tagging a person inside an environment; namely, Radio Frequency Identification (RFID) [7] and Localizing WSN [9, 10]. These two technologies which are based on the RSSI are as follows: RFID tagging: An RFID tag consists of a small microchip and an antenna. In passive RFID technology, tags have no source of energy (e.g. batteries) and the range of detection is not more than 20-30 centimeters. Due to the range required, passive RFID is not a suitable technology for the application of occupancy detection in an ambient intelligent environment. In contrast, in active RFID technology, each tag has a source of energy and the range of detection is up to several meters [8]. Any active RFID tag broadcasts a unique beacon periodically. This beacon can be received by active RFID readers around. By using the RSSI technique with active RFID technology, the distance of the RFID tag from readers can be measured approximately. Tagging in WSN: For localizing a mobile node in wireless sensor networks, two different approaches are available including range-free localization [11] and rangebased localization [12]. In the range-free localization which is used in large scale WSNs (normally in environments including indoor and outdoor spaces) the location of the mobile node is determined by tracing the nodes hopped by the data packet or beacon transmitted from a mobile node. Therefore, the localization resolution is low for range-free localization in WSNs. Range-based localization works using the RSSI technique to find the distance of a mobile node from a fixed node. This approach is more applicable for small WSNs with a small number of nodes. As we are interested in the occupancy of monitored areas, it is not required to deal with the exact location of the mobile node (i.e. node s coordinates). So, the limitations in RSSI technique should not affect our occupancy detection mechanism significantly. In the next section, it will be explained how a regional clustering method is used for occupancy detection of inhabitants.
III. REGIONAL CLUSTERING For occupancy detection of the tagged inhabitants in a living/working environment, an RSSI enabled tag broadcasts a beacon every few seconds. This beacon is received by the readers installed in different areas. Therefore, the signal strength of the beacon received by a reader should represent the distance between tag and reader. A prototype environment is shown in Figure 3. Figure 3 - Readers and a mobile tag in a prototype environment. Due to the possible interference in the monitored environment, in order to reduce their error effects, a regional clustering scheme is proposed as follows: 1- Installation of each reader in the centre of correspondent area for a balanced coverage of the whole area, 2- Installation of readers based on the regions instead of areas in the case of imbalanced coverage, 3- Clustering monitored areas to reduce the occupancy detection error. Clustering: Using Fuzzy C-Means clustering [14], a number of clusters equivalent to the number of monitored areas will be found. This approach is a regional approach as explained below. Suppose that the tagged inhabitant is present or moving in Area 1 in Figure 3. A statistically reasonable number of RSSI readings will be taken when the occupant is in Area 1. These readings are taken by the readers installed in all areas respectively the reader in Area1 ( r 11 ), Area2 ( r 12 ), Area3 ( r 13 ), and Area4 ( r 14 ). [ r11 r12 r13 r 14 ] (2) In expression 2, the first index show the area occupied by the tagged inhabitant and the second index is the reader index. So, for a four area layout (Figure 3), an RSSI readings matrix can be created (expression 3). r11 r12 r13 r14 r21 r22 r23 r24 Rij = r31 r32 r33 r34 (3) r41 r42 r43 r44 Each element in the reading matrix can contain a number of readings but in different locations of the same area. In regional clustering, a Fuzzy C-Means clustering is applied to the RSSI readings matrix resulting in a number of clusters equivalent to the number of monitored areas. In the case of our prototype environment, four clusters will be derived with the centers in expression 4. c11 c12 c13 c14 c21 c22 c23 c24 Cij = c31 c32 c33 c34 (4) c41 c42 c43 c44 In Fuzzy C-Means clustering, after finding cluster centers, the distance of a new reading which is the degree of belonging to clusters ( u k ) will be found. The sum of the elements in degree of belonging matrix should be 1 as shown in expression 5. i uk = 1 0 uk 1 (5) k = 1 The cluster centers in expression 4 can be used to identify the presence of the tagged inhabitant in one of the areas. This happens by comparing the new readings with cluster centers and finding the maximum similarity between them i.e. the maximum element in belonging matrix Max[ u k ]. IV. EXPERIMENTAL RESULTS In order to compare RSSI tagging technologies for the purpose of occupancy detection, two active RFID products (WaveTrend and Syris) and a ZigBee product (XBee) were chosen based on their specification details and compared in a series of comprehensive experiments. The experiments were conducted based on three sets of conditions: i. Distance based with no obstacles (Line of sight between tag and reader), ii. Distance based with an obstacle, iii. Area based in a multiple area home environment. A. Distance Based Experiments without obstacle This experiment was conducted on a basketball pitch without any obstacles between the reader and the tag. The experiment reveals how the RSSI in the experimented products is affected by the tag s distance from the reader. For the accuracy of data, fifty readings were recorded at every meter distance between the tag and the reader.
In the following figures (Figure 4-9), the vertical axis shows the radio signal strength (0-255) detected by the reader. The data points in the graph are the average of the readings, the bars are the standard deviation of the readings and the curve is the trend of the signal drawn for a two degree moving average estimator. Depicted in Figure 6, the XBee product follows the RSSIdistance equation as far as six meters. After six meters the products shows a little rise in signal strength and cannot be used for measuring further distances. B. Distance Based Experiments with obstacle This experiment was performed in a basketball pitch with an obstacle between a mobile tag and the reader. The obstacle contained metal, plastic, and wood materials with a thickness of forty centimeters. For accuracy, fifty readings were recorded at each reading point. Figure 4 - WaveTrend RSSI-Distance graph with no obstacles between reader Depicted in Figure 4, the WaveTrend product follows the RSSI-distance equation for the first five meters where the signal strength decreases as the distance between tags and RFID reader increases. After five meters the product shows unreliable RSSI behavior and cannot be used for measuring distance of more than five meters. Figure 7 WaveTrend RSSI-Distance graph with obstacle between reader Figure 7 shows the WaveTrend product follows the RSSIdistance equation for the first six meters. Figure 5 - Syris RSSI-Distance graph with no obstacles between reader In Figure 5, signal strength of the Syris product follows the RSSI-distance equation for the first three meters but further distances cannot be measured by this product. Figure 8 - Syris RSSI-Distance graph with obstacle between reader Figure 8 shows the signal strength of the Syris product decreases according to the RSSI-distance equation for the first five meters. Figure 6 - XBee RSSI-Distance graph with no obstacles between reader Figure 9 - XBee RSSI-Distance graph with obstacle between reader
As shown in Figure 9, the signal strength of the XBee product decreases according to the RSSI-distance equation for the first five meters. A summary of the results for distance based experiments is shown in Table 1. Product Table 1 - Summary of distance based experiments. Range without (m) Error without (unit) Range with (m) Error with (unit) WaveTrend 5 4.24 6 1.79 Syris 3 6.68 3 3.97 XBee 6 2.18 5 3.42 C. Area Based Experiment This experiment was performed to determine how well the RSSI tagging recognizes the occupancy of different areas in an indoor environment such as a residential apartment. The experiment was conducted in a four area apartment with the layout shown in Figure 10. Clusters 1- Tag in Bedroom1 2- Tag in Bathroom 3- Tag in Kitchen 4- Tag in Bedroom2 Table 2 - Active RFID (WaveTrend) cluster centres. Reader1 (Bedroom1) Reader2 (Bedroom2) Reader3 (Kitchen) Reader4 (Bathroom) 224.30 242.73 217.81 224.14 234.81 232.55 192.27 231.20 200.41 228.66 253.38 190.50 213.04 243.35 200.00 194.63 The centers of clusters for the XBee tag s RSSI readings are shown in table 3. In Figure 12, 4-dimensional data from four readers is represented in 2-dimension plots. Data points on the plot area show the readings and the numbers on the plot are cluster centers. Clusters Table 3 - XBee cluster centres. Reader1 (Bedroom1) Reader2 (Bedroom2) Reader3 (Kitchen) Reader4 (Bathroom) 1- Tag in Bedroom1 2- Tag in Bedroom2 3- Tag in Bathroom 4- Tag in Kitchen 189.70 185.11 184.42 184.34 180.94 186.06 175.04 183.46 183.17 184.33 180.02 201.10 169.04 174.12 185.85 178.99 Figure 10 - Layout of the apartment in the third experiment. All four areas in the experimented apartment were equipped with a reader in the center of their ceilings. In this experiment, RSSI tag readings were collected from all four readers. By moving the mobile tags in twenty different locations in each area, the readings are collected for eighty different points. After data collection, a fuzzy c-means clustering method is applied to the collected data to group them in four clusters. These clusters should represent the presence of the tag (i.e. tagged inhabitant) in each area (Bedroom 1, Bedroom2, Bathroom, and Kitchen). The centers of clusters for the active RFID tag s RSSI readings are shown in Table 2. In Figure 11, 4-dimensional data from four readers is represented in 2-dimensional plots. Data points on the plot area show the readings and the numbers on the plot are cluster centers. It can be summarized from the first two distance based experiments that the accuracy and reliability of the RSSIdistance measuring varies over the distance between tag and reader and is affected by many interfering factors. Despite the uncertainty in the RSSI-distance measuring, experiment 3 shows that by using multiple readers and applying clustering techniques the RSSI capability of both active RFID and ZigBee WSN can be used to identify the occupied areas by the tagged inhabitants.
Figure 11 - WaveTrend readings and cluster centre (x1: bedroom1 reader, x2: bedroom2 reader, x3: kitchen reader, and x4: bathroom reader). Figure 12 ZigBee readings and cluster centre (x1: bedroom1 reader, x2: bedroom2 reader, x3: kitchen reader, and x4: bathroom reader).
V. CONCLUSIONS The possibility of tagging inhabitants was investigated in this paper. Using radio signal strength of active RFID and ZigBee WSN technologies, a series of comprehensive comparative experiments were conducted to examine the quality and accuracy of RSSI in measuring the distance. To reduce the uncertainty of the distance measurement; firstly, occupancy detection of separate areas were suggested and secondly, a regional clustering scheme was proposed. By applying the regional clustering to the RSSI data in the area based experiment, it was shown that despite the uncertainty in the distance measuring capability of RSSI enabled wireless devices, the area occupancy detection can be performed satisfactory. The facility of the occupancy detection can provide smart environments with the means of monitoring occupancy pattern in an inhabited environment. Therefore, the pattern of occupancy for each inhabitant can be extracted for use in automation and other predictive purposes such as security, safety, convenience and energy saving. REFERENCES [1] S. Mitra and T. Acharya, Data mining multimedia, soft computing, and bioinformatics, Wiley, 2003, [2] M. Javad Akhlaghinia, Ahmad Lotfi, Caroline Langensiepen, and Nasser Sherkat, Occupant behaviour prediction in ambient intelligence computing environment, International Journal of Uncertain Systems, Vol.2, No.2, 2008, pp. 85-100, [3] M. Javad Akhlaghinia, Ahmad Lotfi, Caroline Langensiepen, and Nasser Sherkat, Fuzzy Predictor model in an Inhabited Intelligent Environment, in proceeding of IEEE World Congress on Computational Intelligence (WCCI), Hong Kong, 2008, pp. 939-946, [4] Won-Suk Jang, Miroslaw J. Skibniewski, A Wireless Network System for Automated Tracking of Construction Materials on Project Sites, Journal of Civil Engineering and Management, Vol.14, No.1, 2008, pp. 11-19, [5] D. Lymberopolos, Q. Lindsey, and A. Savvides, An Empirical Analysis of Radio Signal Strength Variability in IEEE 802.15.4 Network using Monopole Antenna, Technical Report 050501, Embedded Networks and Applications Lab (ENALAB), Yale University, USA, 2006, [6] Ron Weinstein, RFID: A Technical Overview and Its Application to the Enterprise, IT Professional, vol. 7, no. 3, pp. 27-33, May/Jun, 2005, [7] Klaus Finkenzeller, RFID Handbook: Fundamentals and Applications in Contactless Smart Cards and Identification, Wiley, 2003, [8] L. M.NI, Y. Liu, Y. Cho Lau, and A. P.Patil, LANDMARC: Indoor Location Sensing Using Active RFID, Journal of Wireless Networks, Springer Netherlands, Vol.10, No.6, November 2004, [9] Neal Patwari, et al, Locating the Nodes Cooperative localization in wireless sensor networks, IEEE Signal Processing Magazine, July 2005, pp. 54-69, [10] M. Rudafshani and S. Datta, Localization in Wireless Sensor Networks, in proceeding of International Conference on Information Processing in Sensor Networks (IPSN), USA, 2007, pp. 51-60, [11] Tian He, et al., Range-free localization schemes for large scale sensor networks, in proceeding of the 9 th international conference on mobile computing and networking, USA, 2003, pp. 81-95, [12] Bram Dil, Stefan Dulman, and Paul Havinga, Range-based localization in mobile sensor networks, Wireless Sensor Networks, Springer Berlin/Heidelberg, 2006, pp. 164-179. [13] M. Javad Akhlaghinia, Ahmad Lotfi, Caroline Langensiepen, and Nasser Sherkat, Occupancy Monitoring in Intelligent Environment through Integrated Wireless Localizing Agents, in Proc. of the 2009 IEEE Symposium on Intelligent Agents, 30 March-2 April 2009, pp. 7. [14] Sadaaki Miyamoto, Hidetomo Ichihashi, Algorithms for Fuzzy Clustering: Methods in C-Means Clustering with Applications, Springer,2008.