Affective Action Selection and Behavior Arbitration for Autonomous Robots

摘要

In this paper we suggest an action selection and behavior arbitration scheme for autonomous robots, called AASBA (for "affective action selection and behavior arbitration"), which uses affective states to select the robot's behavior at any given time and pass control from the currently active behavior to the newly selected one. AASBA views action selection and behavior arbitration as an integral part of the agent architecture. The major strength of the proposed scheme is that it can be employed in a variety of agent architectures and that it allows for modifications and extensions of the agent architecture without the need of restructuring the overall control system. Preliminary evaluations of the AASBA scheme are performed with a sample implementation of a two-layered architecture (based on the AASBA scheme) on an autonomous robot.