具有混杂智能特性的多自主体网络动态演化与协调控制研究

The cooperative work mode of multi-agent groups containing one or more agents operated by human under network environment, plays an irreplaceable important role in economic development, national defense construction, scientific research and people's lives. This kind of multi-agent group is called human-robot hybrid multi-agent networks, which have some novel properties, such as heterogeneity, different construction, hybrid intelligent feature (the intelligence of different agents is divergent in thousands of ways), strong constraint (for example, the personnel security restrictions), etc. It is urgent to establish some appropriate theories and methods, which can achieve superiority complementation, coordination and optimization of the hybrid intelligence of different agents in the dynamic evolution of multi-agent networks, to provide guidance for the design, development and implementation of human-robot hybrid multi-agent networks. Based on several typical practical problems of the representative human-robot hybrid multi-agent networks, the research scope of this project will include the following contents: building the model of human-robot hybrid multi-agent networks by analyzing the hybrid intelligent feature and the interaction of the agents with or without human's participation; studying the dynamic evolution, coordination control and optimization scheduling of the human-robot hybrid multi-agent networks by considering the formation control, attitude synchronization, optimal coverage and optimal confrontation under the multi-agent network framework, establishing the theories and methods of coordinative evolution and optimization design of the human-robot hybrid multi-agent networks with heterogeneity, different construction, hybrid intelligent feature, strong constraint. All the results obtained in this project will be verified by numerical simulation, software coding and physical test.

网络环境下有人参与的多自主体群协同工作模式在经济发展、国防建设、科学探索和人民生活中发挥着不可替代的重大作用。这类多自主体群称为有人主体/无人主体群混合网络，该网络除有明显的异质异构特点外，还具有混杂智能（即个体智能千差万别）和强约束（如人员安全限制）等新特性。在多自主体网络动态演化中混杂智能如何优势互补协调优化，迫切需要建立相应的理论与方法，以指导实际混合群集系统的设计、开发与实现。项目拟从几类典型的有人主体/无人主体混合群集系统的实际问题出发，分析有人主体/无人主体的混杂智能特性和交互规律,建立有人主体/无人主体混合群模型；在多主体网络框架下结合队形控制、姿态同步和最优覆盖、最优对抗等典型问题，研究建立具有异质异构、混杂智能和强约束特性的混合多主体网络动态演化、协调控制与优化设计的理论方法，并对所得结果进行数值仿真、软件模拟和实物验证。

四年来，该项目组成员按计划开展了该项目的研究，完成了计划预期的研究任务，取得了预期的研究成果。该项目结合有人无人混合多主体网络，提出和建立了人机混合多主体群的典型模型，揭示了混杂智能网络的动态演化特性，建立了混杂智能群体的协同优化机制。具体包括结合典型人机混合多主体群的应用背景，根据人机混合多主体群异质、异构、混杂智能和强约束等特性，建立了能充分反映有人主体特点与地位的人机混合多主体群网络的模型。针对队形控制、姿态同步、人员安全保障等复杂协同任务，分析了人机混合多主体群的动态演化规律，设计了相应的协同控制算法。针对最优覆盖、路径规划等复杂的优化调度问题，设计了充分发挥有人主体和无人主体各自优势的分布式优化算法。项目的研究成果对于认识混杂智能网络的演化规律，控制设计复杂智能群体具有重要的理论和实际意义。在国内外刊物和会议上发表论文共39篇，其中在 Automatica、IEEE Transactions on Neural Networks and Learning Systems、IEEE Transactions on Systems, Man, and Cybernetics等SCI源刊上发表31篇。此外由Springer出版了著作“Introduction to Hybrid Intelligent Networks”。