A Preliminary Test of Modified Applying Methods of Feedback Error Learning for FES Control

The Feedback Error Learning (FEL) was found to be applicable to joint angle control by FES in our previous study. However, the inverse dynamics model (IDM) in the FEL-FES controller was not learned appropriately in some cases. In this paper, 4 methods of applying the FEL to FES control were compared in controlling 1-DOF movement of the wrist joint stimulating 2 muscles through computer simulation. It was considered that the learning by the FEL was not achieved appropriately because positive stimulation intensity between the minimum and the maximum intensities was only used for the IDM and very small output values of the IDM produced very small change of artificial neural network (ANN) connection weights. Computer simulation results suggested that the learning of the IDM would be improved by considering the IDM output range with setting the minimum ANN output value in calculating ANN connection weight change.