Location based mobile computing - a tuplespace perspective

Abstract

Location based or "context aware" computing is becoming increasingly recognized as a vital part of a mobile computing environment. As a consequence, the need for location-management middleware is widely recognized and actively researched. Location-management is frequently offered to the application through a "location API" (e.g. JSR 179) where the mobile unit can find out its own location as coordinates or as "building, floor, room" values. It is then up to the application to map the coordinates into a set of localized variables, e.g. direction to the nearest bookshop or the local timezone. It is the opinion of the authors that a localization API should be more transparent and more integrated: The localized values should be handed to the application directly, and the API for doing so should be the same as the general storage mechanisms. Our proposed middleware for location and context management is built on top of Mobispace. Mobispace is a distributed tuplespace made for mobile units (J2me) where replication between local replicas takes place with a central server (over GPRS) or with other mobile units (using Bluetooth). Since a Bluetooth connection indicates physical proximity to another node, a set of stationary nodes may distribute locality information over Bluetooth connections, and this information may be retrieved through the ordinary tuplespace API. Besides the integration with the general framework for communication and coordination the middleware offers straightforward answers to questions like: Where is node X located? Which nodes are near me? What is the trace of node Y?