中国科学院自动化研究所机构知识库

Human can accomplish manipulation with high precision, flexibility and robustness. As a musculoskeletal system naturally possesses superiorities of flexibility and robustness, realizing human-like manipulation with the musculoskeletal system is investigated in this paper. However, the redundancy, coupling and strong nonlinearity of musculoskeletal system will bring many challenges for control. Therefore, based on strategies of human manipulation and muscle coordination, a biologically inspired control scheme is proposed. First, we apply a strategy based on attractive region in environment to plan the manipulation under environmental constraints in the task space. Then, we extract muscle synergies as movement primitives and design an iterative learning controller to coordinate these synergies to accomplish complex manipulation. The muscle synergy based iterative learning controller can overcome the difficulty of constructing an explicit model of the sophisticated musculoskeletal system and improve motion learning by transforming the high dimensional muscle excitation into a lower dimensional space. With the combination of strategy based on attractive region in environment and muscle synergy based controller, we preliminarily realize human-like manipulation with high precision, flexibility and robustness. The effectiveness of our control scheme is verified in a peg-in-hole assembly task. (C) 2019 Elsevier B.V. All rights reserved.