Application of Optimization Methods to Predict Performance of a Vibrotactile Balance Prosthesis

System identification and modeling methods were employed to investigate how subjects use orientation information provided by a vibrotactile balance prosthesis. Previous results showed systematic, frequency-dependent changes in the dynamic responses to postural perturbations due to surface tilts as a function of prosthesis feedback parameters. These results could be modeled by a relatively simple feedback control model with the contribution from the prosthesis feedback being dependent on the relative proportion of sway position and velocity information encoded by the prosthesis. Results presented here identify candidate "cost functions" that predict this dependency on prosthesis feedback information. The most accurate prediction was obtained from a cost function that included a weighted combination of the root-mean-square values of body sway jerk (3rd derivative of angular body sway) and angular body sway