障害物回避を行う冗長マニピュレータの逆運動学マップ

Abstract

It is a important to solve the inverse kinematics on control of the redundant DOF (degree of freedom) manipulator. Because of the inverse kinematics, there are many set of solutions of the joint angles which realize an arbitrary end-effector position, and its needs long time for selecting a best joint angle set. To solve this selection problem, we propose diffusion-based learning theory with error correction on the inverse kinematics mapping. In the previous work we have considered no obstacle in the working space of the manipulator. In this work we set an obstacle in the working space and made an obstacle avoidance map based on the inverse kinematics mapping. Our goal of this work is an obstacle avoidance control of the redundant DOF manipulators. To realize this goal we used an inverse kinematics mapping method in order to change posture of the manipulator to a given end point position. To evaluate usefulness of our system, we made some experimental simulations that the manipulator avoids obstacle while its end point traced given trajectory. Direct inverse kinematics calculation method and our inverse kinematics mapping method are compared. A manipulatability on our method was higher than that of the direct method.