Collaborating agent communities for information fusion and decision making

Increasing number of smart sensors are being embedded in the environment to constantly sense and react to events. The applicability of sensors can be enhanced significantly by deploying collaborating software agents that meet the needs of dynamic applications. In this paper, we investigate the coupling of sensors and associated agents for real time information fusion and decision making in distributed and dynamic applications. Two major challenges for proactive and real time collaboration among agents are (I) heterogeneity of sensors, information representation and granularity and (2) fusion of uncertain, redundant, complementary and time sensitive information from various sensors. We propose a framework that efficiently solves the above two challenges through the use of communities of agents that cooperate in real-time to make intelligent and informed decisions using Bayesian network reasoning. We describe in detail the organization of the agent community and the process of community formation. We propose and demonstrate a learning based approach called LEVeL to effectively measure the confidence in cooperating agent observations.