Plan Recognition for Real-World Autonomous Robots: Work in Progress

An agent operating in the real world must be able to coordinate its activities with those of other agents. Traditionally, work it/multiagent coordination has assumed that agents can communicate about their intentions or that they are coordinated through the efforts of a third party. In many environments, however, reliance on communication or on a coordinating agent is infeasible due to the unpredictable nature of the environment or to negative side-effects of communication. We have developed a multiple-resolution hierarchical scheme by which an agent can use observations to infer the high-level goals of other agents. The research domain to which we have applied our scheme is coordinated motion and navigation among multiple robots, both in simulation and in the real-world. Our hierarchical scheme makes probabilistic plan recognition possible in this domain, and it helps to further identify and begin solving crucial issues in plan recognition in physical domains.