Design of Closed-Chain Multi-Legged Robot with Master-Slave Legged Mechanism Based on Deep Reinforcement Learning

Closed-loop linkage mechanisms offer advantages such as low power requirements and high rigidity,which are widely applied in the legged mechanism design of legged robots.However,obtaining the optimal link parameters that achieve the desired foot trajectory is the key issue in the dimensional synthesis of closed-chain legged mechanisms.Inspired by human gait characteristics,the study proposes a master-slave legged mechanism for closed-chain multi-legged robots to improve its walking stability.A dimension synthesis method for obtaining optimal linkage parameters is proposed combined with deep reinforcement learning to enhance the efficiency of mechanism design.Simulation results show that when the crank speed of the legged mechanism is 80rpm,the proposed closed-chain multi-legged robot with master-slave legged mechanism can reduce the body fluctuation range by 34.3% and decrease the average motor torque by up to 45.46%.The multi-legged robot prototype can successfully complete obstacle-surmounting of 48mm,slope climbing of 30°,and In-situ steering tasks,further validating the effectiveness and applicability of the method proposed in this paper.