Multi agent systems for flexible and robust Bayesian information fusion

Abstract

This panel position paper discusses advantages and challenges of multi agent fusion systems (MAS) with respect to the modeling flexibility and fusion reliability. We argue that the MAS paradigm in combination with rigorous modeling and inference methods can facilitate design of theoretically and technically sound fusion systems. This is illustrated with the help of a MAS approach to Bayesian fusion which supports robust and efficient situation assessment in crisis management settings. However, the MAS paradigm might be ill-suited for certain domains and information fusion theories. In addition, implementation of multi agent fusion systems for real world applications can be theoretically as well as technically very challenging. We identify several research topics which address these challenges.