A reconfiguration algorithm for the maintenance of collective commitments

Abstract

The objective of this research is to define a rigorous methodology to be of help in cooperative problem solving (CPS) based on the BDI paradigm. We construct an approach sensitive to the variety of features that multiagent systems have to meet in a dynamic environment. We base our analysis on the four-stage model of (Wooldridge and Jennings, 1996), containing the consecutive stages of potential recognition, team formation, plan formation and team action. These stages are viewed as levels of an abstract architecture. This architecture is profitable in clarifying the nature of dependencies between agents involved: some of them do domain problem solving, others are responsible for the proper organization of CPS. Because of the unpredictable environment, all stages have a dynamic character and require application of methods reflecting this. When defining the levels we abstract from particular methods and algorithms that realize level-associated goals, but instead formulate their final results and associate them with the appropriate individual, social, and collective motivational attitudes.