Effects of Muscle Properties and Motor-Unit Differentiation on Virtual Reflex for Biomimetic Prosthetic Hand.

Jiayue Liu,Reid Bolding,Qi Luo,Tong Wang,Jie Zhang,Chuanxin Minos Niu,Ning Lan

DOI: https://doi.org/10.1109/RCAR47638.2019.9044121

2019-01-01

Abstract:Prosthetic hands are often rejected in part due to lack of compliant manipulation in grasp control. Such capability of compliant control is achieved in humans through the neuro-musculo-skeletal properties. Previous work developed a neuromorphic model that enabled human-like reflexes, which provided a testbed for a new strategy for prosthetic control. The previous model could maintain the joint angle after a torque perturbation, but it lacked differentiation in sizes of muscle fiber, thus the development of force was not smooth under Hennemanu0027s size-principle. Here we matched the differentiation of muscle fibers to the size of motoneurons, which allowed us to examine how reflex performance may be affected by varying motor-unit distribution. In a virtual prosthetic hand controlled by the virtual reflex, simulations showed that an enlarged population of small motor units tended to overdamp the hand, and delay the recovery of hand opening. We also tested how post-perturbation trajectories were affected by refining the force-length property of the muscle model. Results showed that the refined model with smooth force-length curve reduced overshoot and recovery time. This study is the first step towards constructing a complete “virtual reflex” to be integrated in a biomimetic controller for prosthetic hands.