Path planning for fixed-wing UAVs based on expert knowledge and improved VFH in cluttered environments

Obstacle avoidance is a vital problem for flight safety. This paper proposes a fixed wing unmanned aerial vehicle (UAV) path planning algorithm for obstacle avoidance in cluttered environments. Two critical problems have been solved. One is the path planning, i.e., to detour the obstacles at a higher energy cost or to fly through the gaps between obstacles at a lower energy cost by evaluating the collision risk. The other is the optimization of smooth path and pose which is indispensable when the gaps are small. First, the UAV makes the decision of detouring or flying through gaps based on the safety threshold set by the algorithm and the density of the obstacles. Second, if the UAV decides to detour the obstacles, a path planning scheme is developed based on a Dubins-based improved vector field histogram (VFH) for collision avoidance satisfying the kinematic constraints of the fixed-wing UAV. If the UAV decides to fly through the gaps, a pose planner based on machine learning technique is developed, which integrates the human flight experience obtained from trials. The novelty of the proposed strategy is to make the human behavior understandable to machines and let the UAV mimic human behavior in different environments with good adaptability so that the complex nonlinear control and path planning problem can be solved efficiently. Finally, a human-machine interactive simulator is built up to verify the effectiveness of the proposed strategy, and the results show the stable performance in simulations.