Investigation of hybrid optimization methods to evolve effective gaits of a hexapedal robot

摘要

With the understanding that an efficient optimization method is crucial to evolve effective gaits of a walking robot, this work investigates several integrations of well known optimization techniques, including Taguchi method, particle swarm optimization algorithm, and Nelder-Mead simplex method. Four benchmark nonlinear optimization problems are chosen for performance comparison. Numerical results demonstrate the superiority of the Taguchi method that requires only limited number of trials to achieve minimization goals. The method is then implemented experimentally in the search of effective phase difference and cycle time of a six-legged walking robot.