Learning Robot Manipulation Tasks With Task-Parameterized Semitied Hidden Semi-Markov Model

摘要: In this letter, we investigate the semitied Gaussian mixture models for robust learning and adaptation of robot manipulation tasks. We make use spatial temporal correlation in data by tying covariance matrices model with common synergistic directions/basis vectors, instead estimating full each cluster mixture. This allows reuse discovered synergies different parts task having similar coordination patterns. extend approach to task-parameterized hidden semi-Markov autonomous changing environmental situations. The planned movement sequence from is smoothly followed a finite horizon linear quadratic tracking controller. Experiments encode whole body motion simulation, valve opening pick-and-place via obstacle avoidance tasks Baxter robot, show improvement over standard much less parameters better generalization ability.