Year: 2024
Type: article
Abstract: When human acquire physical skills (e.g., tool use) from experts, we tend to first learn from merely observing the expert. But this is often insufficient. We then engage in practice, where we try to emulate the expert and ensure that our actions produce similar effects on our environment. Inspired by this observation, we introduceCombiningIMitation andEmulation for MotionRefinement(CIMER) – a two-stage framework to learn dexterous prehensile manipulation skills fromstate-onlyobservations. CIMER's first stage involvesimitation:simultaneously encode the complex interdependent motions of the robot hand and the object in a structured dynamical system. This results in a reactivemotion generationpolicy that provides a reasonablemotion prior, but lacks the ability to reason about contact effects due to the lack of action labels. The second stage involvesemulation:learn amotion refinementpolicy via reinforcement that adjusts the robot hand's motion prior such that the learnedobjectmotion is reenacted. CIMER is bothtask-agnostic(no task-specific reward design or shaping) andintervention-free(no additional teleoperated or labeled demonstrations). Detailed experiments with prehensile dexterity reveal that i) imitation alone is insufficient, but adding emulation drastically improves performance, ii) CIMER outperforms existing methods in terms of sample efficiency and the ability to generate realistic and stable motions, iii) CIMER can either zero-shot generalize or learn to adapt to novel objects from the YCB dataset, even outperforming expert policies trained with action labels in most cases.
Institution: Georgia Institute of Technology
Language: English
Subfield: Control and Systems Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W4401634211