Development of a New Robust Stable Walking Algorithm for a Humanoid Robot Using Deep Reinforcement Learning with Multi-Sensor Data Fusion

The difficult task of creating reliable mobility for humanoid robots has been studied decades. Even though several different walking strategies have put forth and performance substantially increased, stability still needs to catch up expectations. Applications Reinforcement Learning (RL) techniques are constrained by low convergence ineffective training. This paper develops a new robust efficient framework based on the Robotis-OP2 robot combined with typical trajectory-generating controller Deep (DRL) overcome these limitations. consists optimizing trajectory parameters posture balancing system. Multi-sensors used parameter optimization. Walking optimized using Dueling Double Q Network (D3QN), one DRL algorithms, in Webots simulator. hip strategy is adopted Experimental studies carried out both simulation real environments proposed Robotis-OP2’s algorithm. results show that performs more stable than It thought will be beneficial researchers studying field locomotion.