Robot Obstacle Avoidance Controller Based on Deep Reinforcement Learning

Figure 11

Large angular deviation for obstacle avoidance rather than crossing gaps. Agents run similar actions in the simulation environment.