Exploiting invariant structure for controlling multiple muscles in anthropomorphic legs: An inspiration from electromyography analysis of human pedaling

Motor redundancy is a fundamental problem inherent to the inverse problem of motor planning in a muscle-activated endoskeleton such as that of a human and an anthropomorphic musculoskeletal robot. This paper proposes a biologically inspired control framework for multiple redundant muscles, based on the invariant structure in the group of agonist-antagonist coactivation. The theoretical framework explains that the invariant structure represents the balance of the muscle mechanical impedances and reveals that the invariant structure is identical to the reference frame in the muscle space for motor representation at the endpoint. We tested our approach using a typical locomotion task - pedaling - and validated our approach by analyzing electromyography signals during human movement; we applied the extracted invariant structure to anthropomorphic musculoskeletal legs. Our study provides a novel insight into the interpretation of muscle synergies.