A Gait Generation for Biped Robot Based on Artificial Neural Network and Improved Self-Adaptive Differential Evolution Algorithm

Abstract

Gait generation is very important as it directly effects on the quality of locomotion of biped robots. In point of mathematical view, a gait generation problem is considered as an optimization problem with constraints, it is readily engaged itself to Evolutionary Computation methods and solutions. This paper proposes a novel approach for gait pattern generation of the biped robot. This is to aim to enable the robot to walk more naturally and more stably in locomotion on flat ground. In this study, the approximated optimization method based on Artificial Neural Networks (ANNs) and Improved Self-Adaptive Differential Evolution Algorithm (ISADE) for a gait generation problem is mentioned. To evaluate the achievement of the proposed method, the robot was simulated by multi-body dynamic simulation software, Adams (MSC software, USA). Besides, the walking behavior of the robot is also considered in comparison with that of the human. The result shows that the new approach is an effective method for a gait generation of the biped robot.