Robot path planning based on deep reinforcement learning

Abstract

Q-learning algorithm based on Markov decision process as a reinforcement learning algorithm can achieve better path planning effect for mobile robot in continuous trial and error. However, Q-learning needs a huge Q-value table, which is easy to cause dimension disaster in decision-making, and it is difficult to get a good path in complex situations. By combining deep learning with reinforcement learning and using the perceptual advantages of deep learning to solve the decision-making problem of reinforcement learning, the deficiency of Q-learning algorithm can be improved. At the same time, the path planning of deep reinforcement learning is simulated by MATLAB, the simulation results show that the deep reinforcement learning can effectively realize the obstacle avoidance of the robot and plan a collision free optimal path for the robot from the starting point to the end point.