Using and Determining Location in a Context-Sensitive Tour Guide

W e began our work on the Lancaster Guide project in 1997 with the aim of exploring the issues arising from developing and deploying context-sensitive applications. Specifically, we aimed to develop and deploy a context-sensitive tour guide for the city of Lancaster, UK. Motivated by the work on CyberGuide at Georgia Tech, we envisioned a system that could provide tailor-made tours for visitors by adapting its behavior to changes in a user’s location. For example, the system would generate a different tour depending on a visitor’s interests, location, available time, financial limitations, mobility constraints, and local weather conditions. We also wanted the system to adapt tours as these conditions changed. In contrast to the CyberGuide project, we adopted a network-centric approach. In Guide, the systems obtain information through a high-speed wireless network deployed throughout the target city. Our work explored the role of the network in supporting such location-based applications and provided data for developing future networks with similar characteristics. In addition, we decided to contrast our work with existing projects by developing a system that members of the general public could use. In this way, we hoped to gain feedback from real end users that would contribute to future applications and system designs. After we implemented the Guide system in 1999, we ran a series of field trials involving members of the general public. When we give talks describing our work, researchers and engineers who are developing similar systems often ask us questions that focus on two topics, which together form the basis for the subject matter of this article. The first topic relates to our choice of positioning technology—beacons that broadcast using an IEEE 802.11 wireless network combined with user input. The most obvious choice for obtaining position information is GPS, but GPS is not necessarily the correct choice for location-based applications. The second topic concerns techniques for generating custom tours for electronic city-guide systems. Guide generates these custom tours by taking into account multiple contextual triggers and user preferences. In practice, producing good tours and, indeed, assessing the quality of a tour are difficult tasks. While our analysis of techniques for producing custom tours is somewhat specific to the city-guide domain—which is itself an area of intense activity—we believe that the majority of our work is relevant to location-based systems in general.