Design of Anti-Interference Path Planning for Cellular-Connected UAVs Based on Improved DDPG

The flight and communication security of the Cellular-Connected Unmanned Aerial Vehicles (UAVs) is an important and popular research direction. Due to the complexity of the environmental space, UAVs face a complex and ever-changing task space. In recent years, reinforcement learning has rapidly advanced and widely applied in complex scenarios path planning problems. However, due to the discrete action space, their accuracy is limited. To address aforementioned problems, a new method for UAV path planning based on Deep Reinforcement Learning has been proposed in this paper. Specifically, this paper adopts an improved DDPG method with Actor-Critic framework, which can improve the accuracy. To further enhance the algorithm's precision and training speed, this paper introduces Post-Decision State method, which leverages experience for prediction to optimize the training results and enable UAVs to adapt to the ever-changing environment. Simulation experiments have proved that the improved method can increase training speed and make significant improvements in path performance.