System identification and control optimization of an active prosthetic knee in swing phase

A DC motor is mounted to a Mauch SNS prosthetic knee to obtain an active prosthetic knee. Evolutionary optimization and derivative-based optimization are used to identify system parameters, and to tune a proportional-integral-derivative (PID) controller for knee ankle tracking during swing phase. A Kalman filter is used to estimate knee angle velocity on the basis of the measured knee angle for feedback to the controller. The performance of the optimization algorithms are evaluated based on integral square error (ISE) between experiment and simulation for the system identification problem, and tracking ISE for the control problem. Results show that for system identification, particle swarm optimization (PSO) gives better results than sequential quadratic programming (SQP) and biogeography-based optimization (BBO). Then PID controller optimization is performed while considering nine different shank lengths. BBO achieves the best average overall ISE, and PSO shows the fastest convergence. Finally, we see that increasing shank length results in an increase in the optimal proportional gain of the controller and a decrease in the optimal derivative and integral gains.