Published In Plays as Effective Multiagent Plans Enabling Opponent-Adaptive Play Selection

Coordinated action for a team of robots is a challenging problem, especially in dynamic, unpredictable environ-ments. Robot soccer is an instance of a domain where well defined goals need to be achieved by multiple ex-ecutors in an adversarial setting. Such domains offer challenging multiagent planning problems that need to coordinate multiagent execution in response to other agents that are not part of our team plans. In this work, we introduce the concept of a play as a multiagent plan that combines both reactive principles, which are the fo-cus of traditional approaches for coordinating robot ac-tions, and deliberative principles. We further introduce the concept of a playbook as a method for seamlessly combining multiple team plans. The playbook provides a set of alternative team behaviors which form the basis for our third contribution of play adaptation. We de-scribe how these concepts were concretely implemented in the CMDragons robot soccer team. We also show empirical results indicating the importance of adapta-tion in adversarial or other unpredictable environments