Towards autonomous updating of world models in location-aware spaces

Abstract

Location-aware systems are systems where location information is used as a rich source of context to enable computers to interact and react to their environment. To date, research involving such systems in the indoor environment has concentrated on the development of cheap, realisable, accurate location systems rather than the continued maintenance of the associated world model needed to derive context. This paper characterises the problems relating to world models that have been observed in a real deployment of a location-aware system. It develops a framework that separates the task of synchronising the real and virtual worlds into two—first monitoring to identify coarse regions of inconsistency, and the second accurate updating of these regions. The paper details methods by which the monitoring can be achieved without the cost of deploying specialised sensors, but rather through collating and analysing human movement patterns and raw location data. It presents an overview of the framework for monitoring, an in-depth case study with experimental results from a deployed system, and a discussion of how accurate updates could be achieved once inconsistent regions are identified.