基于自适应信息共享模型的无人机集群智能协同方法研究

The intelligent unmanned system represented by Unmanned Aerial Vehicles (UAV) has gradually become the mainstream and strategic emerging industries of science and technology. It has broad application prospects in military and civilian fields. With the cluster and large-scale development of UAV, under the influence of the task complexity and the environment uncertainty, intelligent coordination of UAV swarm is confronted with a larger theoretical challenge. It is mainly reflected in the adaptability of information sharing model of UAV swarm under uncertain environment, the growth of behavior control mechanism under complex dynamic environment and the efficiency of distributed coordination processing of the large scale UAV swarm. This project takes the UAV swarm as the carrier, and focuses on (1) adaptive information sharing model of UAV swarm based on the hyper-heuristic search framework, enhancing the adaptability of the model in the non-deterministic scenario; (2) the cooperative behavior control mechanism of UAV swarm based on group reinforcement learning algorithm, increasing the generalization and learning ability for coordination of UAV swarm; (3) parallel algorithm design of intelligent coordination for UAV swarm based on specialized intelligent hardware, improving the efficiency of the swarm coordination. The outcomes of this project are expected to be widely applied in the field of intelligent unmanned systems in the future, which not only has important scientific significance, but also has important engineering application value.

以无人机为代表的智能无人系统逐渐成为科技发展主流方向和战略新兴产业，在军事和民生领域都有着广泛的应用前景。随着未来无人机集群化、规模化发展，在任务复杂度以及环境非确定性等因素的影响下，无人机集群智能协同面临更大的理论方法挑战。主要体现在非确定环境下无人机集群信息共享模型的适应性、复杂动态环境中集群行为管控的可成长性以及大规模集群分布式协同过程高效性等方面。本项目以无人机集群为载体，重点研究（1）基于超启发式搜索框架的无人机集群自适应信息共享模型，增强模型的非确定场景适应性；（2）基于群体强化学习的协同行为管控机制，提升集群行为管控的泛化能力和学习推理能力；（3）基于专用智能硬件的无人机集群智能协同并行算法优化，提高集群协同的效率。本项目的研究成果有望在智能无人系统领域得到应用，不仅具有重要的科学意义，也具有重要工程应用价值。

信息共享是无人机集群协同行为管控的基础，如何通过制定自适应信息共享策略以提升集群协同效率，是该问题的关键。目前，无人机集群的信息共享通常以全连接通信或预设通信机制，易造成信息冗余和协同关系固化等缺陷。本项目结合无人机集群协同特点，建立了自适应信息共享模型，并在此基础上，设计实现了无人机集群协同行为管控机制，最后，面向目标搜索和跟踪典型场景，构建规模化无人机集群虚实结合仿真实验平台，验证所提出的方法。具体包括如下3个方面：（1）建立了包括无人集群环境的部分可观测的马尔科夫模型，自适应控制通信模块和自适应构建通信拓扑模块在内的无人机集群自适应信息共享的模型。并设计了自适应控制通信的门控机制以及通过预通信的方式自适应的建立个体之间的通信拓扑；（2）设计实现了基于历史经验的智能体协同行为策略管控，基于历史经验和信息对相关性的协同策略优化和基于面向无人机集群特性的自适应路由转发机制的协同行为管控方法；（3）基于Gazebo仿真平台，结合DJI M100无人机，NexFi MF5800自组网模块，构建了虚实结合仿真实验平台，面向无人机集群目标搜索和跟踪典型场景，验证所提出模型与方法。实验结果表明无人机集群通信冗余降低30%-60%，协同效率提升56%-70%，协同成果率提升17%-31%。.综上所述，本项目对无人机集群智能协同方法中涉及的科学问题进行了深入研究，有助于降低集群通信冗余、提升协同效率。