How many people attended President Obama's inauguration and heard his speech? All sorts of conflicting statistics were beamed around the world by different reporters. Some said a million people, others that 1.6 million attended. This is a difference of 600,000, which is more than the entire population of Antwerp, the second largest city in Belgium. It must be admitted that it is hard to count people in open areas. Bluetooth tracking, a neat technique for following and analysing moving objects both indoors and out, might provide a solution.

Bluetooth is a radio communication technology developed in the middle ‘90s as a substitute for data cables and infrared (IR) communication. Nowadays it is integrated into the majority of new mobile phones, PDAs (Personal Digital Assistants), notebooks, cars, satellite navigation systems, and personal audio. The technology allows exchange of information with other Bluetooth-enabled devices. You might already be using this technology to call hands-free in your car or synchronise your Smartphone with your computer. As Bluetooth is based on a radio (broadcast) communication system, devices do not have to be in line of sight of each other. Another characteristic is limited range; depending on power class, a Bluetooth transceiver has a range of approximately 1 to 100 metres.

As a team of researchers in the Department of Geography, Ghent University, we are well experienced in modelling, analysing and visualising data about moving objects. From our point of view, Bluetooth tracking is just another tool in the toolbox for gathering positions of moving objects. As Bluetooth becomes more common in mobile devices, so the ratio of Bluetooth users increases and gives rise to more reliable tracking results and wider applicability. The technique is particularly useful when positions are required only at points of interest where positional accuracy is of marginal importance. Since every Bluetooth device can be uniquely identified by its MAC address, Bluetooth tracking has a competitive advantage over other detection techniques such as surveillance cameras and IR detectors. On the other hand, not everyone has a Bluetooth-enabled device, so only a sample is taken of a total population. Moreover, there is no absolute link between an individual and a Bluetooth device; people may pass devices between them. However, such instances may be considered exceptional. To make an estimation of the complete population the number of Bluetooth devices must be multiplied by the ratio of persons per discoverable Bluetooth device. This ratio can be estimated by making an additional independent count of everyone at a Bluetooth tracking point.

In essence, a simple Bluetooth scanner consists of two elements: Bluetooth sensor and computer unit. The sensor enables radio communication and the computer instructs the radio and processes the data. At the CartoGIS Cluster in the Department of Geography at Ghent we have built our own Bluetooth scanners. To this end, an embedded board was used as CPU and a Bluetooth USB dongle as sensor. The dongle is a class-two device with a range of approximately 10m. The embedded boards were equipped with Voyage Linux (an ad hoc-designed Linux distribution) and our own software for scanning, processing and storing data. We co-operated on a tracking project at the Rock Werchter festival near Brussels with two companies from the Netherlands: Flucon specialises in customer counting and developed additional Bluetooth scanners for this project, and Realworld Systems is a GIS consultant agency from which we received financial support.

A Bluetooth scanner continuously searches for discoverable devices. This means that the Bluetooth radio of the device must be allowed to broadcast signals. Many devices have the ability to switch this capability on and off. Whenever a Bluetooth-enabled device is within the detection range of a scanner, its MAC address is stored in a log file, together with a time-stamp. The data thus obtained tells us where and when a mobile device has been at a particular place. Obviously, to gather meaningful tracking data several scanners are required at different locations. Log files can be integrated in combination with knowledge about scanner location in order to generate a spatiotemporal database. The path of a moving device can now be reconstructed and visualised. Until today this has been a post-processing step, since data is stored locally on our Bluetooth scanners. However, once scanners are connected by a network to send their data directly to a central database server, real-time tracking becomes feasible. As this considerably increases the applicability of Bluetooth tracking, we aim to set up such systems in future projects.

Bluetooth tracking on the basis of MAC addresses does not violate privacy law. In fact, it simply makes use of a general Bluetooth function: scanning for nearby devices. Everyone is free to use this function, for instance when turning on a mobile phone in a public place. An MAC address is a unique identification linked to a Bluetooth chip, not to a person, let alone their phone number or home address. As one can always turn off one's Bluetooth radio, every individual freely chooses whether or not they wish to be tracked. Our approach captures only the MAC address. Hence, although it is possible to do so, we do not register the UFN (User Friendly Name). We are not even setting up a connection with the discovered device, as location-based services would require. Earlier testing has shown that people often change their UFN into their true name, a web page URL, a phone number, an e-mail address, or whatever they would like to share with the rest of the world.

The CartoGIS Cluster at Ghent is currently valorising its theoretical research concerning moving objects by elaborating concrete case-studies. Preparatory tests have been conducted on the university campus and in nearby streets, as well as in Ghent Sint-Pieters, the main railway station. The first real test case took place in July, at the Rock Werchter festival. Rock Werchter is one of Europe's biggest music festivals, and has several times received the Arthur Award for Best Festival. It is of particular interest in this context due to the extensive festival area and large number of visitors, about 80,000 per day. In this project, scanners were set up at 36 locations in the festival area, including entrances, cash points, food and beverage stands, speaker locations, and first-aid posts. Over the whole time of the festival, about 20,000 unique MAC addresses were logged, resulting in an extensive spatiotemporal dataset. Obviously, the reconstructed space-time paths of these 20,000 visitors will not be extremely detailed; however, we do strongly believe that the current methodology will give new insights into how, when and where people move around at festivals. After cleaning the dataset, the data will be visualised and analysed and finally interpreted in order to find answers. The first results are expected at the end of 2009, beginning of 2010. One thing is certain: the application possibilities of this technique are vast. Only consider safety issues at mass events, logistical management, tourist applications, transport issues, and location-based games.

The authors would like to thank Live Nation, organisers of Rock Werchter; the Department of Information Technology (UGent-IBBT); Flucon; Realworld Systems; and Roel Huybrechts, masters degree student.