Multi UAVs Path Planning Based on Adversarial Training Under Wind Disturbance

Aiming at the problem that traditional algorithms ignore the existence of disturbing factors in the actual flight process of unmanned Aerial Vehicle (UAV), which leads to unsatisfactory path planning, a genetic algorithm based on adversarial training is proposed for solving the path planning problem of UAV under wind disturbance. Firstly, the UAV paths under wind disturbance are modelled, and the fitness function of the flights of multi UAVs is defined under the constraints. Secondly, the wind disturbance and the adversarial population are introduced to co-evolve with the candidate population for adversarial training. Finally, simulation experiments are carried out, and the experimental results show that the improved algorithm improves 45.96%, 43.32%, 45.90% and 25.14% in the four indexes of optimal value, worst value, average value and variance respectively compared with the original algorithm, and the planned paths are more stable and consume less energy, which verifies the effectiveness of the improved algorithm.