Collision Avoidance for a Car-like Mobile Robots using Deep Reinforcement Learning

摘要

—The applications of mobile robots are more and more diverse and extensive. The motion planning of the mobile robots should be considered in aspect of effectiveness of the navigation, and collision-free motion is essential for mobile robots. In addition, dynamic path planning of unknown environment has always been a challenge for mobile robots. Aiming at navigation problems, this paper proposes a Deep Reinforcement Learning (DRL) based path planning algorithm which can navigate nonholonomic car-like mobile robots in an unknown dynamic environment. The output of the learned network are the robot’s translational and angular velocities for the next time step. The method combines path planning on a 2D grid with reinforcement learning and does not need any supervision. The experiments illustrate that our trained policy can be applied to solve complex navigation tasks. Furthermore, we compare the performance of our learned controller to the popular approaches. Keywords— Deep reinforcement learning, path planning, , artificial neural network, mobile robot, autonomous vehicle