Learning point-to-point movements on an elastic limb using dynamic movement primitives

Compliance is an important feature for robots in human–machine interaction to increase the safety of humans during such interactions. While active compliance is online adaptable and easy to control its dynamic response is determined by the sampling rate and the response rate of the controller. In contrast, passive compliance is inherent to the system. It responds naturally fast to any perturbation exerted on the robot independently of the controller frequency. This paper introduces an extension to the DMP framework which facilitates the generation of trajectories for passive compliant joint drives using reinforcement learning. The compliance of the limb is preserved entirely during motion as well as at the goal position. The proposed approach is evaluated with a simulation of a 2 DOF robot limb with passive compliant joint drives. Experiments are presented for point-to-point movements. The results demonstrate that the proposed approach is capable of generating trajectories for point-to-point movements for the gear-side of a compliant limb such that the drive-side follows a desired trajectory.