GCTD3: Modeling of Bipedal Locomotion by Combination of TD3 Algorithms and Graph Convolutional Network

In recent years, there has been a lot of research using reinforcement learning algorithms to train 2-legged robots move, but are still many challenges. The authors propose the GCTD3 method, which takes idea Graph Convolutional Networks represent kinematic link features robot, and combines this with Twin-Delayed Deep Deterministic Policy Gradient algorithm robot move. very effective in graph-structured problems such as connection joints human-like robots. method shows better results on motion trajectories bipedal compared other Gradient, Soft Actor Critic. This is implemented model six independent joint coordinates through Robot Operating System Gazebo simulator.