Memory-based control for recognition of motion environment and planning of effective locomotion

Abstract

Points out the importance of memory-based schemes for motion control of robots. In memory-based control schemes, all values concerning robot motion are stored in memory of digital computers instead of modeling the robot motion and the environment around the robot. The authors propose a memory-based control scheme which is suitable for locomotion of robots in unknown environment. At the first step of the proposed control scheme, a robot recognizes the unknown environment through test motions which mean actual movement of the robot. As a result, the robot has knowledge of the motion environment as a kind of map in the memory of digital computers. At the second step, by using path planning techniques in the environment maps obtained by the test motions, the robot forms a motion pattern which can realize effective locomotion. The effectiveness of the proposed memory-based control scheme is demonstrated through two experimental results in which robots move in the water and on the ground.<>