In order to solve the problem of free gait planning of hexapod robot in complex and rugged terrain, the discrete gait model of hexapod robot is constructed based on the discretization of single foot step distance of hexapod robot. The stability of the robot is analyzed by using the longitudinal stability margin, and the corresponding gait planning strategy is designed. The complex gait planning problem is transformed into the scheduling optimization problem among the position states in each oscillation period. Based on the discrete gait model designed in this paper, a free gait planning method based on DQN (deep Q-learning) algorithm is designed with the average stability margin of the robot as the optimization objective, and its experience replay mechanism is improved. This algorithm optimizes the logic of the experience replay part, reduces the low value iterations and greatly improves the convergence speed. The experimental results show that compared with traditional method, the improved DQN algorithm can plan a free gait with fast convergence, high stability margin and strong adaptability.
Free Gait Planning of Hexapod Robot Based on Improved DQN Algorithm
2020-10-14
215365 byte
Conference paper
Electronic Resource
English
Double-gait hexapod robot and double-gait hexapod robot crawling method
European Patent Office | 2023
|Hexapod robot fault-tolerant gait planning method based on reinforcement learning
European Patent Office | 2023
|Free fault-tolerant gait planning method and device for hexapod robot and storage medium
European Patent Office | 2020
|European Patent Office | 2023
|