Comparative analysis of global optimization-based controller tuning methods for an exoskeleton performing push recovery

Abstract

In this study an exoskeleton is considered. A simplified mathematical model for the exoskeleton is given, and the conditions for it to remain vertically balanced are discussed. The paper focuses on push recovery control. One of the known control strategies for this motion is discussed, and a feedback controller is presented. The controller has 10 tunable coefficient, and the aim of the paper was to make a comparative analysis of two well known global optimization techniques, particle swarm optimization (PSO) and genetic algorithm as applied to the task of tuning this controller. The obtained results show that PSO has higher quality performance, while requiring few function evaluations. This agrees with the results of a number of previous comparative studies of these two methods. Also a tuning method based on Sobol sequences is studied. It showed better performance than both mentioned methods, which suggests that it might be useful for controller tuning applications.