基于改进蒙特卡洛树搜索的无人机目标分配与突防决策方法

Multi-UVA target allocation and penetration decision based on an improved Monte Carlo tree search method

针对多无人机任务规划问题,在多种约束与机动动作下,进行目标分配和突防决策统一建模与优化求解方法研究。首先,基于无人机自身优势、目标威胁以及突防概率分别建立目标分配优化函数和突防决策优化函数;然后,利用线性加权法将两者融合,形成多无人机协同任务规划统一目标函数;其次,在强化学习框架下,分阶段构建协同任务规划的状态空间和动作空间,并根据统一目标函数设计奖励函数;提出一种改进的蒙特卡洛树搜索强化学习算法,在统一目标函数最大收益下实现对无人机目标分配和突防决策问题的求解;最后,通过对比仿真实验验证所提出的方法的时效性和最优性。研究结果表明:相较于传统方法,所提出的方法在提升收敛程度的同时,将训练时间减少了15%。

In order to solve the problem of multi-UAV(unmanned aerial vehicle)mission planning,the unified modeling and optimization method of target assignment and penetration decision was studied with consideration of a variety of constraints and maneuvers.Firstly,based on each UAV's own advantages,target threat and penetration probability,the target allocation optimization function and penetration decision optimization function were established,respectively.Then,the linear weighting method was used to combine the above two functions to form a unified objective function for multi-UAV cooperative mission planning.Secondly,by applying the reinforcement learning technique,the state space and action space of collaborative task planning was constructed by stages,and the reward function was designed according to the unified objective function.In the meanwhile,a brand-new improved Monte Carlo tree search algorithm was proposed to solve the UAV target assignment and penetration decision problems simultaneously with the maximum benefit of the unified objective function.Finally,the effectiveness of the proposed method was verified by comparative simulation experiments.The results show that compared to traditional methods,the proposed method reduces training time by 15%and improves convergence.