Rational coordination in multi-agent environments

We adopt the decision-theoretic principle of expected utility maximization as a paradigm for designing autonomous rational agents, and present a frame work that uses this paradigm to determine the choice of coordinated action. We endow an agent with a specialized representation that captures the agen t's knowledge about the environment and about the other agents, including i ts knowledge about their states of knowledge, which can include what they k now about the other agents, and so on. This reciprocity leads to a recursiv e nesting of models. Our framework puts forth a representation for the recu rsive models and, under the assumption that the nesting of models is finite , uses dynamic programming to solve this representation for the agent's rat ional choice of action. Using a decision-theoretic approach, our work addre sses concerns of agent decision-making about coordinated action in unpredic table situations, without imposing upon agents pre-designed prescriptions, or protocols, about standard rules of interaction. We implemented our metho d in a number of domains and we show results of coordination among our auto mated agents, among human-controlled agents, and among our agents coordinat ing with human-controlled agents.