Learning Task-Parameterized Skills From Few Demonstrations

Moving away from repetitive tasks, robots nowadays demand versatile skills that adapt to different situations. Task-parameterized learning improves the generalization of motion policies by encoding relevant contextual information in task parameters, hence enabling flexible executions. However, training such a policy often requires collecting multiple demonstrations To comprehensively create situations is non-trivial thus renders method less applicable real-world problems. Therefore, with fewer demonstrations/situations desirable. This paper presents novel concept augment original dataset synthetic data for improvements, allows task-parameterized few demonstrations.