Autonomous navigation for indoor mobile robots based on reinforcement learning

The algorithm of building a map based on SLAM (simultaneous localization and mapping) in indoor environment and using the map for autonomous localization and navigation is mature and has more successful cases applied in practical scenarios, but the cost of building a map is still very expensive, we try to use mapless method to achieve autonomous navigation of robots, according to the human in unknown environment can clearly reach the target location because human in We try to analogize human decision making thinking from the direction of intelligent decision making to guide mobile robots to achieve autonomous navigation, and reinforcement learning is an important algorithm to achieve intelligent decision making. In this paper, We use reinforcement learning methods from the original images acquired by the vision sensors for the robot to learn the optimal decision navigation method from the initial position to the target position. The experiment showed good results