DEP-RL: Embodied Exploration for Reinforcement Learning in Overactuated and Musculoskeletal Systems

Abstract

Muscle-actuated organisms are capable of learning an unparalleled diversity of dexterous movements despite their vast amount of muscles. Reinforcement learning (RL) on large musculoskeletal models, however, has not been able to show similar performance. We conjecture that ineffective exploration in large overactuated action spaces is a key problem. This is supported by the finding that common exploration noise strategies are inadequate in synthetic examples of overactuated systems. We identify differential extrinsic plasticity (DEP), a method from the domain of self-organization, as being able to induce state-space covering exploration within seconds of interaction. By integrating DEP into RL, we achieve fast learning of reaching and locomotion in musculoskeletal systems, outperforming current approaches in all considered tasks in sample efficiency and robustness. * See https://sites.google.com/view/dep-rl for videos. Code will be published at a later date.Preprint. Under review.

SAR: Generalization of Physiological Dexterity via Synergistic Action Representation

“…Further, human physiology is multiarticular, meaning that there exist many-to-one and one-to-many relationships between muscles and joints (e.g., the flexor digitorum profundus muscle controls over 3 joints for each digit of the hand) [27]. Critically, the human hand is also overactuated, meaning that there are more muscle forces than degrees of freedom (i.e., approximately 39 muscles collectively control 23 joints in the human hand) [49].…”