Fast and robust biped walking involving arm swing and control of inertia based on Neural network with harmony search optimizer

Abstract

In this paper a new model free approach is being presented to make walking humanoid robot faster and more stable. This approach presents a controller that is composed of nonlinear and linear neurons. We have used Matsuoka Neural oscillator, for nonlinear neurons, to generate rhythmic signal to control walking of robot. This controller gets feedback from output signal that helps to correct and generate better signals to improve stability of walking. In this approach the role of hands is considered to make walking of robot smooth and robustness. Also, this method controls the role of inertia to improve the speed and robustness of bipedal walking by refining signals and reducing the role of inertia. In this regard, a new population-based search algorithm that called harmony search is used to optimize the signals that are produced by the controller which control joints angel and other parameters. This method is implemented in simulated NAO robot and simulation result shows getting feedback from controller and considering the role of hands and inertia make walking more stable and faster.