Context Awareness in Ambient Systems by an Adaptive Multi-Agent Approach

摘要

In the field of ambient systems, the dynamic management of user context is needed to allow devices to be proactive in order to adapt to environmental changes and to assist the user in his activities. This proactive approach requires to take into account the dynamics and distribution of devices in the user's environment, and to have learning capabilities in order to adopt a satisfactory behaviour. This paper presents Amadeus, an Adaptive Multi-Agent System (AMAS), whose objective is to learn, for each device of the ambient system, the contexts for which it can anticipate the user's needs by performing an action on his behalf. This paper focuses on the Amadeus architecture and on its learning capabilities. It proposes some promising results obtained through various scenarios, including a comparison with the Multilayer Perceptron (MLP) algorithm.