In this chapter, we survey the various ways in which context-aware pervasive computing applications are likely to exploit and manage location information; we use this understanding to debate whether a universal location-management infrastructure should store location information in a topology-dependent (symbolic) or topology-independent (geometric) format. Our analysis of both location-aware and location-independent applications reveals three important points: (1) different systems and prototypes use a wide variety of location-resolution technologies, (2) a significant number of location-based applications are primarily interested in resolving the location of a mobile node only relative to the connectivity infrastructure, and (3) obtaining geographical location coordinates requires varying levels of hardware that are absent in many pervasive devices. We thus conclude that the universal location-management infrastructure should manipulate location information primarily in a structured, symbolic form. In cases where the geographical coordinates are needed, they may be obtained through the use of access-specific technologies or via appropriate mapping.
We then consider the objectives of pervasive computing and enumerate the desirable features of a universal location-management infrastructure. In particular, we believe that location prediction, location translation, signaling optimality, and location privacy are four "must-haves" in a practical pervasive infrastructure. While the problem of location privacy is beyond our current scope, we consider the problem of location prediction and signaling optimality in greater detail. We explain how the LeZi-update algorithm uses adaptive learning to optimize the signaling associated with location update and paging in a symbolic domain. By treating the movement of a mobile device as a sequence of strings generated according to a stationary distribution, the algorithm is able to efficiently store a mobile's entire movement history, and also predict future location with asymptotically optimal cost. We finally turn to the problem of location translation, and give an overview of our ongoing development of a hierarchical LeZi-update that permits efficient translation of location profiles between heterogeneous systems.
Our immediate plans for future work include the development and performance testing of the hierarchical LeZi-update algorithm. We are interested also in the problem of efficiently translating between symbolic and geometric coordinates in practical systems. A great deal of work also is needed to standardize protocols for location fusion and translation in real-life environments. We hope that the findings of this chapter serve as a useful starting point for the design and specification of formats for specifying user location, the architecture of location databases, and the development of intelligent location-reporting protocols.