Path Planning for Fixed-Wing Unmanned Aerial Vehicles in Complex Terrains

This paper proposed a global path planning algorithm integrating intelligent sampling and dynamic constraints for multiple fixed-wing unmanned aerial vehicles (multi-FUAVs) operating in complex environments. Classical sampling-based algorithms suffer from low efficiency and struggle to satisfy dynamic constraints. To overcome these limitations, an enhanced rapidly-exploring random tree (RRT*) algorithm using a point cloud neural network-optimized sampling strategy combined with cosine annealing and artificial potential field (APF) refinement for efficient and safe scattered path searching. Subsequently, the kinematic constraints are converted into mathematical optimization problems with minimum control effort (MINCO) trajectory representation for generating smooth and flyable trajectories. Extensive experimental evaluations across 500 complex environment test cases demonstrate that the proposed APF-NIRRT* algorithm achieves a 57% reduction in average iteration count and 30% shorter path lengths, while strictly complying with dynamic and obstacle-avoidance constraints, thereby validating the effectiveness and practical applicability for real-world FUAV deployment in complex terrains.