基于末端视觉导航的无人机机群最优路径规划

Optimal Path Planning of UAV Fleet Based on Terminal Vision Navigation

主要应用于未来城市空中交通,研究城市无人机路径规划系统,解决城市无人机群体协同最优路径规划以及无人机在城市空域运行安全所面临的首要问题——避障。利用具有启发式优化的蚁群算法建立初步模型,使用样条插值对路径进行优化使其适用于真实的飞行路径,解决了空间交叉点上无人机与无人机碰撞的问题;借鉴地面无人驾驶汽车基于视觉的防撞系统,设计了基于机器视觉的载人无人机辅助防撞系统,使用SqueezeNet深度神经网络与迁移学习相结合的方法,以飞鸟、建筑物和马路作为研究对象,最终经过验证得到准确率达98%以上的防撞系统。

Mainly used in urban air traffic in the future,this paper mainly studies the urban UAV path planning system to solve the primary problem of urban drone group collaborative optimal path planning and the safety of drones in urban airspace-obstacle avoidance.The ant colony algorithm with heuristic optimization is used to establish a preliminary model,and the spline interpolation is used to optimize the path to make it suitable for the real flight path,which solves the problem of the collision between the drone and the drone at the spatial intersection;A vision-based collision avoidance system for human-driving cars,a manned drone-assisted collision avoidance system based on machine vision is designed,using a combination of SqueezeNet deep neural network and migration learning,and flying birds,buildings,and roads as the research objects.In the end,an anti-collision system with an accuracy rate of over 98%was obtained after verification.