Arm Trajectory Generation Based on RRT* for Humanoid Robot

In this paper, an arm trajectory generation method based on the Rapidly-Exploring Random Tree Star (RRT*) is proposed for humanoid robot. The RRT* is one of anytime motion planning algorithms. The RRT* adopts the three fundamental components from the RRT, the preceding version of RRT*: the state variables, local planner, and cost function. The end effector of humanoid robot is positioned on the desired point by manipulating the joint angles of the arm, which are the state variables. The Minimum-jerk method is applied as a local planner for more realistic trajectory and the local planner fulfills collision detection test. The cost taken to transit between two points is defined as the sum of angle differences of motor corresponding to the two points. Also, there has been the need for real time control and it is taken care of by introducing a multi-thread system. The arm under control initiates motioning, once the first trajectory that meets the target zone is constructed. While the arm is on the move, the RRT* continuously updates the trajectory. The effectiveness of the proposed method is demonstrated by simulating the 7 DOF robot arm, which has been performed under two environments: the obstacle-free and obstacle-constrained cases. Simulation is carried out for the humanoid robot, Mybot-KSR, developed in the RIT Lab., KAIST.