Parallel Hybrid 2-Opt Flower Pollination Algorithm for Real-Time UAV Trajectory Planning on GPU

Abstract. The development of autonomous Unmanned Aerial Vehicles (UAVs) is a priority to many civilian and military organizations. An essential aspect UAV autonomy the ability for automatic trajectory planning. In this paper, we use parallel Flower Pollination Algorithm (FPA) deal with problem's complexity compute feasible quasi-optimal trajectories fixed-wing UAVs in complex 3D environments, taking into account vehicle's flight properties. global optimization algorithm improved addition 2-opt local search providing significant improvement. proposed planner implemented parallelized on multicore processor (CPU) using OpenMP Graphics Processing Unit (GPU) CUDA resulting 9.6x 68.5x speedup respectively compared sequential implementation CPU. Index Terms—Flower Algorithm, Unit, Parallel Programming, Trajectory Planning, Vehicle.