Resource Management for Boundedly Optimal Agent Societies

In this paper, we propose a generalization of the bounded rationality paradigm to multi-agent systems (MAS) in which self- interested agents work on a common goal. A hierarchical resource adaption scheme is developed that links the micro-level of a hy- brid and layered agent architecture to the macro-level of the corre- sponding society. The introduction of societal structures provides the key for breaking down the highly complex search for optimal solu- tions into a multi-staged process. Each of the social stages integrates the complementary benefits of simple and complex decision making based on the device of abstract resources. These are unified repre- sentations of quantitative environmental and architectural constraints or interdependencies between subordinate problem-solvers. The pre- sented framework is integrated into an existing multi-agent system and thus applied to a series of applications. To clarify our approach, we refer to a transportation telematics domains.