An Improved Trajectory Learning Algorithm Based on DMPs

Trajectory learning has always been a research hotspot in the field of intelligent robotics. This paper focuses on the analysis and optimization of the trajectory learning algorithm based on dynamic movement primitives. Firstly, the shortcomings of the internal core gaussian basis function of DMPs algorithm are studied emphatically. To solve the problem that the learning trajectory is not accurate enough, this paper proposes a uniform selection method for the center of gaussian basis function inside DMPs. Secondly, the learning trajectory modified by the optimization method is compared with that of the original method by simulation. Finally, the accuracy of trajectory learning is quantified by a method for calculation the trajectory shape error. It is proved that the more the number of gaussian basis function, the higher the accuracy is when the basis function is uniformly activated. The comparison of trajectory shape errors and simulation results both verified the correctness of trajectory learning optimization method. The proposed method is helpful to improve the robot’s man-machine cooperation ability and the accuracy of trajectory learning.