Agent based middleware infrastructure for autonomous context-aware ubiquitous computing services

Middleware for ubiquitous and context-aware computing entails several challenges, including the need to balance between transparency and context-awareness and the requirement for a certain degree of autonomy. In this paper we outline most of these challenges, and highlight techniques for successfully confronting them. Accordingly, we present the design and implementation of a middleware infrastructure for ubiquitous computing services, which facilitates development of ubiquitous services, allowing the service developer to focus on the service logic rather than the middleware implementation. In particular, this infrastructure provides mechanisms for controlling sensors and actuators, dynamically registering and invoking resources and infrastructure elements, as well as modeling of composite contextual information. A core characteristic of this infrastructure is that it can exploit numerous perceptual components for context acquisition. The introduced middleware architecture has been implemented as a distributed multi-agent system. The various agents have been augmented with fault tolerance capabilities. This middleware infrastructure has been exploited in implementing a non-obtrusive ubiquitous computing service. The latter service resembles an intelligent non-intrusive human assistant for conferences, meetings and presentations and is illustrated as a manifestation of the benefits of the introduced infrastructure.