Eyal Shlomo Shimony

A common solution to multi-agent decision problems is to commit a team of collaborating agents to a joint plan. Once committed, any deviation from the plan by an agent, is hazardous. Hence, in many such systems, agents ignore potential beneficial actions not in the plan, (even if such " opportunistic" actions increase the expected utility of the team). We model the (stochastic) tradeoff of such opportunistic actions vs. continued commitment to the joint plan, and address this issue as a formal decision problem under uncertainty. We use a modified version of the AWOL model, presented in an earlier paper, showing how it works in the context of a simplified version of the Unreal Tournament (TM) computer game. Empirical results suggest that it is faster to compute the solution to the AWOL abstraction than to solve the problem in the original domain.