Functional neuromuscular stimulation for joint control using nonlinear model predictive control and extended kalman filtering

In this work we present a control algorithm for restoring movements in paralyzed limbs through functional neuromuscular stimulation (FNS). We employed a nonlinear model predictive controller (NMPC) in conjunction with an Extended Kalman Filter (EKF). Both the NMPC and EKF were equipped with a neuromuscular feline limb model. The internal muscle states were estimated from noisy kinematic data by the EKF and inputted into the NMPC control policy. We evaluated the FNS-based control algorithm via numerical simulation by making a feline model of paralysis follow different commanded trajectories. Besides achieving low tracking errors, the algorithm execution time met the real-time constraint due to the use of analytical Jacobians.