无人机集群通信网络抗毁性拓扑模型和算法研究

In recent years, multi-UAV systems have attracted wide attention in military and civilian fields. The rapid mobility of UAVs and the complicated environment condition put forward high requirements for the survivability of multi-UAV communication networks. At present, the research on the multi- UAV network mainly focuses on physical layer protocol, routing protocol, resource optimization and so on. There is not much progress in the research of the survivable network design. This project studies the static preventive survivable design and the dynamic restorative survivable design of the multi-UAV network respectively based on the wireless physical interference model. Considering all related factors including the power of UAVs, the capacity of UAV links and wireless interference, this study optimizes the survivable design of the multi-UAV network as well as the use of network resources and the network performance. For the model with large-scale variables and constraints, and the dynamic distributed real-time decision-making problems, this project studies centralized algorithms based on the decomposition technology and distributed algorithms based on multi-agent reinforcement learning method, which provides theoretical and practical guidance for the application of the survivable design of multi-UAV networks.

近年来，无人机集群系统在军事和民用领域引起了广泛的关注。无人机的快速移动性以及执行任务环境的复杂性对无人机集群通信网络的抗毁性提出了较高的要求。目前关于无人机集群网络的研究主要集中在物理层协议、路由协议、资源优化等方面，在抗毁方面的研究还没有太多的进展。本项目在引入无线物理干扰模型的基础上，分别对无人机集群静态的预防性网络抗毁问题与动态的恢复性网络抗毁问题进行优化建模，充分考虑无人机节点功率、链路容量以及无线干扰等因素的影响，在优化抗毁设计的同时，尽可能节省网络资源，提高网络性能。对于实时性要求不高的静态抗毁设计模型，为应对大规模变量和约束的模型的求解问题，研究基于规划分解技术的集中式求解方法。对于实时要求较高的动态抗毁性拓扑维护模型，运用多智能体强化学习技术，实现无人机的自主分布式决策进而维护较高的网络抗毁性与网络性能，为无人机集群系统通信网络的抗毁性设计提供理论和实践指导。

随着智能技术的发展，无人机集群应用已经在多个领域展现出来。然而，实际环境的复杂性给无人机集群通信链路的稳定性带来挑战，进而影响集群整体运行能力。本项目对无人机集群通信网络抗毁性网络拓扑设计进行研究，首先对网络抗毁与网络优化相关概念和指标进行介绍，进而分别对静态网络抗毁性和动态网络抗毁性问题进行建模分析。针对静态网络抗毁性模型，设计了基于规划分解算法的模型求解方法，并进一步对耗时较高的子模型设计启发式求解方法，提高模型求解的速度。针对动态网络抗毁性问题，基于多智能体强化学习方法进行建模，并对完全信息和不完全信息条件下的网络抗毁性设计了多智能体强化学习训练方法。最后，在模拟环境中，对无人机集群通信网络进行模拟，并对所提出的模型和算法进行验证。实验结果表明，静态和动态抗毁性模型和算法能够有效根据网络运行状态计算出抗毁性的网络拓扑。该项目的研究对未来无人机集群通信网络抗毁性提供了理论和算法支撑，将能有有力促进其在各类复杂场景中的运用。