Deep-reinforcement-learning-based gait pattern controller on an uneven terrain for humanoid robots

Although conventional gait pattern control in humanoid robots is typically performed on flat terrains, the roads that people walk every day have bumps and potholes. Therefore, to make more similar humans, movement parameters of these should be modified allow them adapt uneven terrains. In this study, solve problem, reinforcement learning (RL) was used engage self-training automatically adjust their for ultimate control. However, RL has multiple types, each type its own benefits shortcomings. a series experiments were performed, results indicated proximal policy optimization (PPO), combining advantage actor-critic trust region optimization, most suitable method. Hence, an improved version PPO, called PPO2, used, experimental combination deep with data preprocessing methods, such as wavelet transform fuzzification, facilitated balance robots.