Networked multi-agent cooperative control: theory and applications

网络化的多智能体系统是由多个通过通信网络相互连接的动态智能体组成的，它可以用来解决单个智能体难以解决或不可能解决的问题。应用实例包括：一组用于地雷搜索的移动的机器人;用于监视的无人驾驶飞行器;以及用于海底勘探的水下航行器。这些都是对保持加拿大在全球知识经济中的突出地位至关重要的关键领域。其主要目标是通过设计一种网络化的协同控制策略，使智能体的行为保持一致，即所谓的一致性问题。然而，通信网络和复杂的代理动态提出了巨大的挑战，控制工程师和设计师：（1）代理之间的多速率信息传输，（2）网络引起的随机延迟和数据包丢失，（3）代理的异构动态。现有的控制设计方法已经跟不上技术的步伐-大多数协作控制理论和实践忽略了上述实际的通信约束，这可能会显着降低性能。 该研究计划将有助于对这种在复杂的非结构化环境中和不可靠的通信网络上运行的网络化多智能体系统的基本理解和统一设计框架。本研究将整合通信和合作控制的方式，既将增加合作控制的性能和能力，参与代理在复杂的任务一起工作。该研究方案将填补理论与实践之间的差距，为控制工程师提供网络化多智能体系统协同控制分析与综合的新工具，建立一种新的解决实际约束的统一范式。此外，该计划将通过技术和行业相关的研究培训使研究生受益匪浅。

A networked multi-agent system, being composed of multiple interacting dynamic agents connected via communication networks, can be used to solve problems which are difficult or impossible for an individual agent. Application examples include: a team of mobile robots for land mine search; unmanned aerial vehicles for surveillance; and, underwater vehicles for undersea exploration. These are all critical areas of importance to maintaining Canada's prominence in the global knowledge-based economy. The primary objective is to make the agents behave consistently, the so called consensus problem, by designing a networked cooperative control strategy. However, communication networks and complex agent dynamics present tremendous challenges to control engineers and designers: (1) the multirate information transmission among agents, (2) network-induced random delays and packet losses, and (3) the heterogeneous dynamics of agents. Existing control design methodology has not kept pace with the technology - most cooperative control theory and practice ignore the aforementioned practical communication constraints, which may significantly degrade performance. The research program will contribute to the fundamental understanding and unified design framework of such networked multi-agent systems that operate in complex, unstructured environments and over unreliable communication networks. This research will integrate communication and cooperative control in ways that will both increase the cooperative control performance and the ability of involved agents to work together in complex tasks. The proposed research program will fill the gap between theory and practice; it will provide control engineers with new tools for analysis and synthesis of the cooperative control for networked multi-agent systems; it will establish a novel unified paradigm addressing practical constraints. Furthermore, the program will greatly benefit graduate students through technology- and industry-relevant research training.