多重威胁下的无人机自主避障航迹规划

无人机作为一种新兴的无人作战力量和不可或缺的民用设备,现已渐渐融入到国家安全和社会发展中的各个方面,航迹规划是保障无人机顺利完成既定任务的核心环节.为解决规划空间存在诸多静态和动态威胁的实时航迹规划问题,提出了一种基于滚动时域的无人机自主避障航迹规划方法.首先将航迹规划模型构建为单目标函数优化问题,根据无人机简化运动学模型和约束条件,采用滚动优化策略生成最优航迹序列;然后对最优航迹序列之间的航迹再一次采用滚动优化策略产生子序列,综合考虑威胁和飞行约束,利用负梯度下降法搜索航路点,采用遗传算法对子序列进行规划;最后经反复滚动迭代优化可得近似全局最优航迹,同时利用贝塞尔曲线对航迹进行处理,使其表征实际的飞行航迹.实验仿真结果表明:验证了模型的合理性和方法的有效性;具有良好的威胁规避能力并能规划出一条光滑航迹;与全局规划方法相比,该方法减少了收敛时间,实时性更强,能够快速、鲁棒地收敛到近似全局最优解.

UAV autonomous obstacle avoidance path planning under multiple threats

As an emerging unmanned combat force and indispensable civilian equipment, unmanned aerial vehicles (UAVs) have gradually been integrated into all aspects of national security and social development. Path planning is the core link to ensure that UAVs successfully complete the established task. In order to solve the problem of real-time path planning with many static and dynamic threats in the planning space, a method of autonomous obstacle avoidance path planning with receding horizon is proposed. Firstly, the path planning model was constructed as a single objective function optimization problem. According to the simplified kinematic model and constraints of the UAV, the receding horizon optimization strategy was used to generate the optimal path sequence. Then, the receding horizon optimization strategy was also used to generate sub-sequences for the trajectories between the optimal path sequences. Considering the threat and flight constraints, the negative gradient descent method was used to search the waypoint, and the genetic algorithm was used to plan the sub-sequences. Finally, the approximate global optimal path was obtained by repeated receding iterative optimization, and the trajectory was processed by bezier curve to represent the actual flight path. The experimental simulation results show that the model is reasonable and the method is effective. Meanwhile, it also has good threat avoidance ability and can plan a smooth path. Compared with the global planning method, the proposed method reduces the convergence time, has stronger real-time performance, and can converge to the approximate global optimal solution quickly and robustly.