Mobile Robots Path Planning Based on Dynamic Movement Primitives Library

This paper proposes a general framework to establish the dynamic movement primitives library (DMPL) for a mobile robot path planning in unknown environment. Based on DMPL, the mobile robot can autonomously compute a smooth path from the initial position to the target points while avoiding any existed obstacles. Path planning based on DMPL consists of two phases, the library building and its application. Before the library is building, the workspace of the mobile robot is divided into multiple sectors through a unique sampling technique. Then, using a joystick, a user operates the mobile robot moving from start to any sample point, simultaneously recording the states such as position, velocity and acceleration. The primitives will be extracted from the recorded state sequence and the learned weights will be stored in the DMPL. In the second phase, the DMPL is used online to supply the path planning decision. Depending on the searching criteria, the minimum distance between the robot and the destination, the optimal primitive will be chosen carefully from the library established before. Several simulations with MATLAB show that the method based on the DMPL is effective for mobile robot path planning, and the generalization to new targets can be achieved well by using the DMPL.