Potential and Implications of Bluetooth Proximity- Based Tracking in Moving Object Research

-

Another key advantage of Bluetooth tracking is its applicability in indoor as well as outdoor environments, whereas conventional GPS positioning is impossible in indoor environments (Zeimpekis et al. 2003) .

3. How does Bluetooth tracking work?

Bluetooth was originally developed as a short-range communication technology between mobile devices, and is currently the de facto technology for easily sharing information (contacts, images, video) between devices in each other’s vicinity. In order to communicate between devices, both devices need to know which other devices are within their communication range. This is done by broadcasting a device inquiry. Such a scan typically lasts 10.24 seconds (Peterson et al. 2006) , and returns a list of devices identified by their MAC-address (i.e. a unique identifier linked to each Bluetooth device).

Generally speaking, there are two potential methods of using this technology as a tracking technology: proximity-based and multilateration of the signal strength. Since Bluetooth signals only propagate over a limited distance, the detection of a mobile device at a sensor with a known location implies that the device was within the communication range of the sensor. This method of extracting rough location information is generally known as proximity-based positioning. The theoretical communication range of the sensors used in our experiments (class 2) is around 10m. The actual range, however, which depends on various factors such as reflections and hardware quality, is usually somewhat higher (~30m). During the device inquiry, the received signal strength intensity (RSSI) at which a mobile device is detected can also be recorded. Because this intensity usually decreases with increasing distance, it is theoretically possible to get an estimate of the distance from the sensor by correlating the RSSI values with distance in prior experiments (Hossain and Soh 2007) . By estimating this distance from multiple sensors at different locations, one could theoretically multilaterate in order to get an estimation of the actual location of the mobile device (Awad et al. 2007) . Multilateration potentially offers more detailed motion information than proximity measurements, but remains problematic to date because of the imperfect correlation between RSSI and distance (Hossain and Soh 2007) . Therefore, we will focus on proximity-based tracking in the remainder of this paper.

4. Bluetooth proximity-based tracking data

Bluetooth scanners installed across the study area continuously scan for other devices. When a mobile device is detected, an ‘in’ registration with its MAC-address and a timestamp gets registered. Then, when this device is not detected anymore for at least 10.24 seconds (the duration of a scan cycle), an ‘out’ registration with another timestamp gets registered. In this way, it is possible to trace individual mobile devices because it is known where (location of the scanner) and when (‘in’ until ‘out’) a certain mobile device (MAC-address) was. The following is an extract of a log file of a Bluetooth scanner showing three mobile devices that were detected: 1246525539,0021080577xx,5898756,in 1246525544,0021080577xx,5898756,out 1246525429,0019B74FABxx,5243404,in 1246525575,0019B74FABxx,5243404,out 1246525558,001E3A5C31xx,5898756,in 1246525590,001E3A5C31xx,5898756,out The general format of a log line is: timestamp (unix format), MAC-address, device class code, in/out. The device class code contains information about what kind of mobile device (cell phone, smart phone, handsfree kit, etc.) is associated with the MAC-address.

5. Characteristics of Bluetooth proximity data

The special nature of proximity tracking data can be illustrated by Figure 1:

Where GPS tracking data (a) are practically continuous in both space and time, Bluetooth proximity data (b) are inherently discrete. Locations that are not covered by a scanner do not provide any movement information. Additionally, the location calculations in the GPS tracks are linked to a single point in time, whereas the proximity-based data consist out of time intervals during which the mobile device was detected somewhere.

Both types of tracking data are inherently characterized by a limited accuracy which diminishes the reliability of the measurements. Whereas the only uncertainty in GPS tracking data lies in the position estimation at each timestamp (depicted by the variable ellipses in Figure 1a), the uncertainty in Bluetooth proximity tracking data is more complex and embedded in both space and time. Due to the potential delay between a device entering the communication range and its detection, the ‘in’ and ‘out’ time registrations are never exact. The spatial uncertainty is actually twofold. First, in unconstrained space, a mobile device can only be assumed to be within a circular region around the sensor. Second, the actual communication range is not crisp but fuzzy where the likelihood of getting detected decreases as one moves away from the Bluetooth sensor (depicted by the double boundary of the cylinders in Figure 1b). Because the propagation of Bluetooth is highly susceptible to influencing factors such as reflections by obstacles there is also a chance that a device that is within range does not get detected.

6. Analysis potential

Hence, Bluetooth proximity tracking data are much coarser than GPS tracking data and cannot be converted into detailed geospatial lifelines (Hornsby and Egenhofer 2002) . As a consequence, some finer-grained motion attributes such as instantaneous speed, acceleration or motion azimuth cannot be extracted. This limits the application potential of existing analysis methods using such motion attributes to extract higherlevel patterns and knowledge (Laube et al. 2005) . Future queries such as in Future Temporal Logic (Wolfson et al. 1998) also become challenging. A more promising method to extract valuable information is sequence analysis (Shoval and Isaacson 2007) . Additionally, more general yet insightful indicators providing static information from one sensor can be readily extracted from the tracking data. An example of this is shown in Figure 2 which depicts the varying crowdedness around a Bluetooth sensor placed at the entrance of a rock festival attracting around 100.000 visitors per day. A typical pattern is visible where there is a gradual influx of visitors during the afternoon, followed by a much sharper efflux at night. Over the course of four days, a total number of around 23.000 mobile devices were detected by 36 sensors across the festival area (Van Londersele et al. 2009) .

5000  

A GIS for Moving Objects (GISMO) is currently developed in java for analyzing the data. A screenshot is shown in Figure 3.

7. Conclusion and open research questions

Bluetooth proximity-based tracking offers an alternative way of generating massive amounts of tracking data, but its true analysis potential remains hard to predict. Do completely new methods need to be developed, or can some of the current methods be adapted to support these data? Ultimately, how can we extract interesting patterns from these trajectories?

Acknowledgments

The research work of Mathias Versichele is funded by the Agency for Innovation by Science and Technology in Flanders (IWT).

References

Awad A , Frunzke T and Dressler F , 2007 , Adaptive Distance Estimation and Localization in WSN using RSSI Measures . In: Proceedings of the 10th Euromicro Conference on Digital System Design Architectures, Methods and Tools , 471 - 478 .

Hay S and Harle R , 2009 , Bluetooth Tracking without Discoverability . In: Location and Context Awareness , 120 - 137 .

Hornsby K and Egenhofer MJ , 2002 , Modeling moving objects over multiple granularities . Annals of Mathematics and Artificial Intelligence , 36 ( 1-2 ): 177 - 194 .

Hossain AK and Soh W , 2007 , A comprehensive study of Bluetooth signal parameters for localization . In: IEEE 18th International Symposium on Personal, Indoor and Mobile Radio Communications , Athens, Greece, 1185 - 1189 .

Laube P , Imfeld S and Weibel R , 2005 , Discovering relative motion patterns in groups of moving point objects . International Journal of Geographical Information Science , 19 ( 6 ): 639 - 668 .

Miller HJ and Han J , 2001 , Geographic data mining and knowledge discovery . Taylor & Francis , London, UK.

Peterson BS , Baldwin RO and Kharoufeh JP , 2006 , Bluetooth Inquiry Time Characterization and Selection . IEEE Transactions on Mobile Computing , 5 ( 9 ): 1173 - 1187 .

Ratti C et al., 2006 , Mobile Landscapes: using location data from cell phones for urban analysis . Environment and Planning B: Planning and Design , 33 ( 5 ): 727 - 748 .

Shoval N and Isaacson M , 2007 , Sequence alignment as a method for human activity analysis in space and time . Annals of the Association of American Geographers , 97 ( 2 ): 282 - 297 .

Van Londersele B , Delafontaine M and Van de Weghe N , 2009 , Bluetooth Tracking - a spy in your pocket . GIM International: 23 - 25 .

Wolfson O et al., 1998 , Moving Objects Databases: Issues and Solutions . In: Scientific and Statistical Database Management, International Conference on, Los Alamitos, CA, USA, 111 .

Zeimpekis V , Giaglis GM and Lekakos G , 2003 , A taxonomy of indoor and outdoor positioning techniques for mobile location services . ACM SIGecom Exchanges , 3 ( 4 ): 19 - 27 .