Robot Obstacle Avoidance Controller Based on Deep Reinforcement Learning

As the core technology in field of mobile robots, development robot obstacle avoidance substantially enhances running stability robots. Built on path planning or guidance, most existing methods underperform with low efficiency complicated and unpredictable environments. In this paper, we propose an method a hierarchical controller based deep reinforcement learning, which can realize more efficient adaptive without planning. The controller, multiple neural networks, contains action selector runner consisting two network strategies single actions. Action selectors each strategy are separately trained simulation environment before being deployed robot. We validated wheeled More than 200 tests yield success rate up to 90%.