Gradient-based trajectory planner of real-time obstacle avoidance for fixed-wing unmanned aerial vehicles in cluttered environments

Autonomous flight of fixed-wing Unmanned Aerial Vehicles (UAVs) in cluttered environments demands a trajectory planner that offers low computational overhead while ensuring high robustness. However, trajectory planning problems subject to nonlinear dynamics and obstacle avoidance constraints, especially with irregularly shaped obstacles, often impose prohibitive computational costs on existing algorithms. To achieve efficient and robust trajectory generation for fixed-wing UAVs, this paper proposes a multi-stage trajectory planning method leveraging B-spline curve representation and gradient-based optimization. Initially, the A* algorithm is modified to search for collision-free control points on an occupancy grid map, taking into account various irregular obstacles. Subsequently, the B-spline-based trajectory optimization problem is formulated, incorporating obstacle avoidance constraints via the Euclidean signed distance field. A minimum-curvature-limited trajectory refinement stage is designed with a curvature adjustment mechanism to ensure dynamic feasibility. Simulation results demonstrate that this planner achieves over an order of magnitude speedup compared to state-of-the-art optimization-based approaches in global planning scenarios. Furthermore, the method is implemented into a local replanning framework to validate its real-time obstacle avoidance ability, where re-planning consumes only tens of milliseconds even in cluttered environments. Trajectory tracking experiments in simulation and outdoor flight experiments with a physical fixed-wing UAV (Sky Surfer X8P) both verify the executability of the generated trajectories by the proposed planner.