Trajectory Planning of UAV Based on Continuous Decision Reinforcement Learning

To address the issues of insufficient exploration efficiency and value estimation bias in reinforcement learning methods for Unmanned Aerial Vehicle (UAV) path planning, an improved Twin Delayed Deep Deterministic Policy Gradient (TD3) path planning algorithm incorporating continuous decision-making is proposed. A Reinforcement Learning (RL) training scheme based on continuous decision-making is designed, optimizing the RL architecture, exploration strategy, and reward function to enhance the efficiency and stability of drone navigation in high-dimensional complex environments. During training, an inertia-based exploration mechanism and a multi-objective reward function are introduced to accelerate the convergence of the navigation strategy and improve the average reward obtained. Finally, experiments conducted in a simulated environment and comparison between existing methods demonstrate the effectiveness of the proposed method in terms of path results, time consumption, and navigation stability.