Self-Improving Behavior Arbitration

During the last few years, and in an attempt to provide an efficient alternative to classical methods to designing robot control structures, the behavior-based approach has emerged. Its success has largely been a result of the bottom-up development of a number of fast, tightly coupled control processes. These are specifically designed for a particular agent-environment situation. This new approach, however, has some important limitations because of its lack of goal-directedness and flexibility. In earlier work we presented a model for an architecture that would deal with some of these problems. The architecture bases on two levels of arbitration, a local level which enables the robot to survive in a particular real world situation, and a global level which ensures that the robot's reactions be consistent with the required goal. In this paper the emphasis is put on the local arbitration. We show how the local priorities can be computed and learnt and present simulation results.