Motion Generation Based on Contact State Estimation Using Two-stage Clustering

Many contact-rich tasks in factories are still executed manually because it is often difficult for robots to adapt variable environmental conditions. To incorporate the adaptability of humans environment into robots, a machine learning technique called from demonstration (LfD) has been well studied. A popular approach automating complex involves selecting appropriate action several segmented motions, movement primitives. In this study, we propose method autonomously recovery and correcting trajectory when task determined have failed based on force. Although conventional technologies can divide motions time series, they unable recognize presence failures detail response slight variations. Therefore, two-stage clustering method, which consists segmentation trajectories labeling generate actions failures. The proposed was able perform even cases 20 mm position error by accurately trajectory.