A configuration dependent muscle model for the myoelectric control of a transfemoral prosthesis.

This paper presents the development of a torque-based myoelectric impedance controller for an active-knee transfemoral prosthesis. An anthropomorphically inspired agonist-antagonist impedance controller studied in a myoelectric elbow prosthesis is adapted for the knee joint. To parameterize the controller, regression analysis was applied to a recently updated lower-extremity neuromuscular simulation model that provides estimates of knee torque as a function of knee angle and neural activation. Initial results using a constant moment arm suggest physically unreasonable parameters and poor model performance, but the inclusion of an angle-dependent moment arm in the reduced-order muscle model enables good correlation with the high-order neuromuscular model. The resulting limb controller is tested using a 1-DOF active knee prosthesis donned by a non-amputee subject with an able-bodied adapter. Initial treadmill walking tests demonstrate the potential of this controller to enable effective myoelectric control of the prosthetic limb.