Behavioral Programming with Hierarchy and Parallelism in the DARPA Urban Challenge and RoboCup

Abstract

Research in mobile robotics, unmanned systems, and autonomous man-portable vehicles has grown rapidly over the last decade. This push has taken the problems of robot cognition and behavioral control out of the lab and into the field. In such situations, completing complex, sophisticated tasks in a dynamic, partially observable and unpredictable environment is necessary. The use of a Hierarchical State Machine (HSM) for the construction, organization, and selection of behaviors can give a robot the ability to exhibit contextual intelligence. Such ability is important for maintaining situational awareness while pursuing important goals, sub-goals, and sub-sub goals. Using the approach presented in this paper, an assemblage of behaviors is activated with the possibility of competing behaviors being selected. Competing behaviors are then combined using known mechanisms to produce the appropriate emergent behavior. By combining hierarchy with parallelism we present an approach to behavior design that balances complexity and scalability with the practical demands of developing behavioral systems for use in the real-world. The effectiveness of merging our hierarchical arbitration scheme with parallel fusion mechanisms has been verified in two very important landmark challenges, the DARPA Urban Challenge autonomous vehicle race and the International RoboCup robot soccer competition.