Sequential sensor fusion-based W-DDPG gait controller of bipedal robots for adaptive slope walking
Ping-Huan Kuo,Jun Hu,Kuan-Lin Chen,Wei-Hsin Chang,Xin-Yu Chen,Chiou-Jye Huang
DOI: https://doi.org/10.1016/j.aei.2023.102067
IF: 8.8
2023-06-21
Advanced Engineering Informatics
Abstract:Walking is a critical topic to the research and development of humanoid robots . However, humanoid robots in action often encounter diverse terrain, which impedes their ability to walk. A slope is a major challenge for humanoid robots. To address this challenge, a gait control algorithm based on wavelet transform and Deep Deterministic Policy Gradient was proposed in this study to enable robots to correct gait parameters in real time through continuous self-learning. In the algorithm, the signals of the three-axis accelerometer and three-axis gyroscope measured by an inertial measurement unit are decomposed and analyzed through wavelet transform to improve the learning outcomes . Robots were trained to walk on slopes with various angles to verify the proposed algorithm, which was also used to optimize the gait parameters of the central pattern generator. The experimental results revealed that the proposed algorithm is superior and more stable than other advanced algorithms. The feasibility and practicality of this algorithm were also verified in the experiments.
engineering, multidisciplinary,computer science, artificial intelligence