UAV real-time path planning in 3D multidynamic obstacle environments

This paper proposes an Improved Interfered Fluid Dynamical System (IIFDS) method for Unmanned Aerial Vehicle (UAV) path planning that integrates Deep Reinforcement Learning (DRL) with the Interfered Fluid Dynamical System (IFDS). The proposed IIFDS method adaptively adjusts parameters to accommodate complex Three-Dimensional (3D) environments. Simulation results demonstrate that the IIFDS with Proximal Policy Optimization (IIFDS-PPO) outperforms the IIFDS with Deep Deterministic Policy Gradient (IIFDS-DDPG) and Soft Actor-Critic (IIFDS-SAC) across comprehensive performance metrics. Compared to traditional IFDS method, IIFDS-PPO achieves over one-third performance improvement while generating significantly smoother paths. Remarkably, even in intricate 3D multi-dynamic obstacle environments, the proposed IIFDS-PPO method enables instantaneous path planning through its pre-trained policy.