휴머노이드 로봇을 위한 양팔 경로 생성에 대한 연구

Abstract

The domestic study for humanoid robot is mainly focused on walking and running etc. And the study for Robot Arms is emphasizing in motion that hold goods, but study that robot arms avoids obstacle autonomously is unprepared. And most study for robot arms is focused on the motion of one arm and is not considered about the motion of two arms at the same time. The ultimate target in this paper is to suggest efficient path planning algorithm about work that is given about 13 DOF upper body with two arms of humanoid robot Maru-1 and to demonstrate it through 3D simulation that the real robot is modeled equally. For this target, in this paper, kinematics of the robot upper body is analyzed to apply 13 degree of freedoms two arms of robot and body to path generation algorithm and inverse kinematics is analyzed to get each joint angle about the inputted target object of robot two arms. And merits and demerits of each algorithms is reviewed through investigation and study about classical path planning algorithm and the theory of sampling based algorithm to be possible to apply to multi-DOF system is analyzed. There are RRT (Rapidly-exploring Random Tree) and PRM (Probabilistic Raodmap) in sampling based algorithm. RRT based algorithm is analyzed in priority in this paper. RRT define free configuration space and choose random possible configuration repeatedly and connect these configuration from initial configuration to goal configuration. This method can reduce unnecessary calculation that is used to construct road map in first step unlike PRM and is effective to find the path in configuration space of nonholonomic system and kinodynamic system. In this paper, the acceptance of these algorithm is demonstrated through the result of 3D simulation.