Optimization of Walking Control of a Biped Robot using Differential Evolution

This paper presents a solution to the problem of optimizing parameters (gains) of a walking control scheme of a biped robot with feet. The problem is addressed using Differential Evolution, a generic optimization method not dependent on mathematical properties of the fitness function. For the optimization of control parameters, we propose and compare three different mono-objective fitness functions and two different encodings of optimization variables that were evaluated in simulation. We have introduced an explicit management of important constraints like joint limits, center of pressure and energy consumption, in a mono-objective formulation. The explicit management of these constraints has allowed to obtain improved results than those obtained by an expert.