Using priorities to simplify behavior coordination

Abstract

Real-world behavior-based robot control problems require the coordination of a large number of competing behaviors. However, coordination becomes increasingly difficult as the number of competing behaviors increases. Behavior hierarchies address this difficulty, but do not solve it since high-level behav-iors still require state information of lower-level behaviors. Abstracting the state of lower-level behaviors through priorities re-moves this restriction, which should allow the behavior hierarchy to better scale. However, whether or not this abstraction negatively impacts the effectiveness of the hierarchy is unknown. In this work, both the quality of priority-based behavior hierarchies and their ease of development are evaluated. This is done by using grammatical evolution to learn how to coordinate low-level behaviors to ac-complish a task. I show that not only do priority-based behavior hierarchies perform just as well as standard hierarchies, but that they promote faster learning of solutions that are better suited as components in larger hi-erarchies. 1.