Fuzzy Real-Time Multi-objective Optimization of a Prosthesis Test Robot Control System

This paper investigates the fuzzy real-time multi-objective optimization of a combined test robot/transfemoral prosthesis system with three degrees of freedom. Impedance control parameters are optimized with respect to the two objectives of ground force and vertical hip position tracking. Control parameters are first optimized off-line with an evolutionary algorithm at various values of walking speed, surface friction, and surface stiffness. These control parameters comprise a gait library of Pareto-optimal solutions for various walking scenarios. The user-preferred Pareto point for each walking scenario can be selected either by expert decision makers or by using an automated selection mechanism, such as the point that is the minimum distance to the ideal point. Then, given a walking scenario that has not yet been optimized, a fuzzy logic system is used to interpolate in real time among control parameters. This approach enables automated real-time multi-objective optimization. Simulation results confirm the effectiveness of the proposed approach.