Behavior-based cognitive control for mobile robots

In this paper, behavior-based cognitive control for special shape mobile robots is studied thoroughly by using reinforcement learning idea. The study case considered is that a cross mobile robot with the shape of a Greek cross "+", which is equipped with a perception system for sensing environment, passes through a crack whose width is less than the width and length of robot body. This means that the robot can never pass through the narrow crack unless it knows how to take advantage of its self special shape intelligently. The objective of our work is to investigate how to automatically generate a trajectory for the mobile robot based on its own behavior. In the paper we first briefly introduce a behavior-based cognitive control system developed in Windows platforms, by which we can design and study various behavior-based cognitive control schemes for different shape mobile robots. Secondly, we discuss the definitions of state and action space for the behavior-based cognitive control, and then investigate a learning algorithm using reinforcement learning idea. Thirdly, we propose a behavior-based cognitive control architecture. Finally, in order to prove the validity and efficiency of the developed behavior-based control strategy, we present some simulation results for the cross mobile robot.