Autonomous navigation method of robot in dynamic environment based on proximal policy optimization

In response to the problem that traditional navigation algorithms have poor adaptability when dealing with complex working environments of robots, this paper studies the robot autonomous navigation method based on deep reinforcement learning (DRL). Compared with traditional navigation methods, the proposed method is more robust and more adaptable to dynamic environments. Firstly, the speed and position of the possible obstacle behavior patterns in the robot working scene are analyzed, and the obstacle model is constructed by demarcating the dangerous area. Secondly, the proximal policy optimization (PPO) is used to design a reward function based on the obstacle dangerous area to define the robot's goal, and the robot is trained to continuously interact and learn with the complex dynamic simulation environment to obtain the maximum reward. Finally, the optimal strategy for path planning in complex environments is found to achieve autonomous navigation of the robot in complex environments. Experimental tests show that the success rate of fixed target navigation reached 100%, and the success rate of random target navigation reached 70.6%.