Concurrent Probabilistic Motion Primitives for Obstacle Avoidance and Human-Robot Collaboration

The paper proposed a new method to endow a robot with the ability of human-robot collaboration and online obstacle avoidance simultaneously. In other words, we construct a probabilistic model for human-robot collaboration primitives to learn the nonlinear correlation between human and robot joint space and Cartesian space both based on interaction trajectories from the demonstration. This multidimensional probabilistic model not only helps to infer robot collaboration motion depending on the human action by the correlation between human and robot in joint space but also convenient to conduct robot obstacle avoidance reverse kinetics from cartesian space via the correlation between them. Specifically, as for the latter, a modulation matrix is established from the obstacle form to automatically generate robot obstacle avoidance trajectory in Cartesian space. Obstacle avoidance in the human-robot collaboration experimental is investigated, and its simulation results verify the feasibility and efficiency of the algorithm.