Autonomous localized path planning algorithm for UAVs based on TD3 strategy

Aim: ing at the current local path planning methods for UAVs, the problem of weak portability due to the large influence of different controllers on its planning effect and the weak autonomous decision-making ability of UAV path planning, an autonomous local path planning algorithm for UAVs based on the TD3 policy update is proposed. On the basis of not changing the UAV's native controller, the decision-making model is designed to help the UAV realize local path planning work in unfamiliar environments based on environmental information. The model filters and extracts the UAV sensor, camera image information, and lidar information, inputs them into the reinforcement learning agent, outputs high-level command actions to the robot cooperative operating system, and transmits the movement commands from the operating system to the UAV, realizes the local path planning task of the UAV, and carries out the test experiments based on the training results. The test results show that the autonomous local path planning algorithm for UAVs based on deep reinforcement learning can effectively realize the autonomous local path planning task of UAVs, and the success rate reaches 93% under the interference of no obstacles, and 92% in the environment with obstacles. The success rate and portability of UAV implementation of path planning are improved.