FES Adaptive Feedback Control using Neural networks with PID

FES control is difficult for musculoskeletal system’s time-variability and dynamic nonlinearity. It often needs good performance control models and complex control methods to get better tracking precision, also the high performance control model is hard to build. Neural network (NN) can learn complex nonlinear mappings so it can be used to identify control model on-line. At the mean while, PID control is a widely used feedback control method for it simplicity. However, its parameters need to be confirmed before it using in a control system. By using the NN identifying results, the self-adaptive PID feedback control strategy can be modified in real-time. This will improve the wrist tracking precision. Some testing experiments done on the able-bodied subject showed the control method was robust and stability.