Self-Adaptive Model Generation for Ambient Systems

Ambient systems are composed of many interacting entities, and their behaviour is constantly changing. Under these conditions, static models are insufficient to understand and control such systems. In this paper, we investigate the possibility to generate real-time dynamic model of an ambient system. For this, we present AMOEBA, a multi-agent system designed to address this problem. It is based on a set of cooperative mechanisms from the Adaptive Multi-Agent System theory. Experiments on simulated physical systems highlight the interesting properties of AMOEBA.