Artificial rabbits optimization–based motion balance system for the impact recovery of a bipedal robot
Ping-Huan Kuo,Wei-Cyuan Yang,Yu-Sian Lin,Chao-Chung Peng
DOI: https://doi.org/10.1016/j.aei.2024.102965
IF: 8.8
2024-11-30
Advanced Engineering Informatics
Abstract:Research on the control of bipedal robots has predominantly focused on ensuring stability and balance during locomotion, often neglecting the robot's ability to respond to unexpected external disturbances. In the present study, an algorithm is proposed to enable humanoid robots to maintain balance when they experience external impacts. In evaluation experiments, a robot was placed on flat surfaces and sloped terrain, where it experienced impacts from five angles. To evaluate the robot's stability, data were collected before, during, and after each impact. The study utilized the artificial rabbits optimization (ARO) algorithm to optimize parameters and trained the robot's control model by using a five-layer multilayer perceptron (MLP) neural network. Notably, the joint use of ARO and MLP yielded computational savings relative to conventional reinforcement learning methods. The proposed hybrid approach allowed the robot to adapt quickly to external forces and maintain balance effectively. The findings of this research hold considerable promise for enhancing the practical applications of bipedal robots in real-world scenarios, where unpredictable forces or impacts are common. By improving a robot's ability to react dynamically and maintain balance, the proposed method enables humanoid robots to operate in highly challenging and dynamic environments, such as those associated with disaster response, industrial tasks, or everyday human interaction, without falling because of unexpected disturbances. Thus, the present study contributes to the field of humanoid robotics by addressing real-world challenges and providing a robust solution for impact resistance.
engineering, multidisciplinary,computer science, artificial intelligence