A Theory of Satisficing Control. Part II: Multiple-Agent Control

Multiple agent control problems have proven difficult to solve using classical control methods partly because first, in non-cooperative problems different agents may disagree about what constitutes ``optimality'' and second, even when agents cooperate, an optimal solution may be computationally unsolvable. Satisficing solutions present an alternative approach based on a procedurally rational decision methodology. In problems which are inherently ``rationality limited'', a perspective frequently encountered in multiple agent problems, procedurally rational decisions are required. We present a design mechanism for generating multi-agent satisficing controllers that not only performs well in cooperative problems, but also allows agents to make decisions with partial information in competitive environments. We demonstrate the power of the satisficing solution methodology for cooperative and non-cooperative control problems.