An Order-Based Approach to Mission-Oriented Autonomous Robot Control: Managing Complexity, Merging Multiple Plans, and Performance Analysis Given Partial Probabilistic Information

摘要

This paper formulates a framework for mission-oriented robot control. The benefit of this approach is multifacet. First, it suggests a promising direction for managing the complexity of encoding control policies. Second, it establishes a natural connection between feedback control and planning. Third, it allows the analysis of the time-to-mission-accomplishment distribution with almost minimum probabilistic information. This framework has been applied to the domain of robot soccer as a test of effectiveness. The resulting system competed at par in a simplified game setting against the strategy of a former championship team in the International RoboCup Competition.