CAREER: Strategic decision-making for communication and control in decentralized systems

该项目旨在为动态分散系统中的通信和控制开发一个战略决策框架。集中式动态决策问题在工程、经济和数学文献中有着悠久的历史。然而，许多现代系统涉及在不确定和动态环境中做出决策的代理网络。这种分散式系统出现在不同的应用领域：基础设施系统，如电力，交通和通信网络，传感和监控系统，如自动驾驶汽车或机器人团队，以及互联设备网络，如物联网。这种系统的特点是存在多个代理/决策者，他们的环境信息不完整，相互通信的能力有限，并且仍然必须做出可能影响整个系统的决策。这些系统必须处理权力下放的两个关键方面：㈠信息的权力下放，㈡决策的权力下放。信息和决策的分散化产生了通信的必要性。这种沟通可能是有限的、不可靠的和不完美的。决策者需要适应他们的沟通环境。另一方面，沟通本身可能需要决策，例如决定何时沟通、与谁沟通以及沟通什么信息。因此，在分散系统中，沟通和决策是密切相关的。拟议的研究的总体目标是调查这种相互作用的通信和控制/决策中出现的分散系统在动态和不确定的环境中运行。拟议研究的技术和社会影响将通过在分散系统中开发新的通信和控制理论工具来实现，这些工具将影响许多应用和行业，包括：传感和监控系统，机器人技术，智能电网，通信网络，城市基础设施维护和网络物理安全。该项目的教育影响将通过培训研究生和通过学校访问和/或校园活动与当地高中学生接触来实现。拟议研究的重点将是战略决策，这需要确定代理人的决策策略，使他们能够对真实的-时间信息展开事件在他们的环境中，以优化分散的动态系统的性能。拟议的工作将解决在不确定和动态环境中运行的分散系统的控制和决策的基本问题。特别是，它将探讨以下领域的决策策略的表征和计算：分散随机控制与控制器之间的不可靠和不完美的通信，极大极小决策和分散系统中的控制问题，实时编码的联合优化，通信调度和分散控制的控制策略。这项研究的成果将有可能成为进一步了解动态分散系统中信息、通信和决策相互作用的重要一步。

This project aims to develop a strategic decision-making framework for communication and control in dynamic decentralized systems. Centralized dynamic decision-making problems have a long history in engineering, economics and mathematics literature. Many modern systems, however, involve networks of agents making decisions in an uncertain and dynamic environment. Such decentralized systems arise in diverse application domains: infrastructure systems like power, transportation and communication networks, sensing and surveillance systems like teams of autonomous vehicles or robots, as well as networks of inter-connected devices such as the internet of things. Such systems are characterized by presence of multiple agents/decision-makers that have incomplete information about their environment, limited abilities to communicate with each other and that still have to make decisions that may affect the overall system. These systems have to contend with two key aspects of decentralization: (i) Decentralization of information, and (ii) Decentralization of decision-making. The decentralization of information and decision-making creates the need for communication. This communication may be limited, unreliable and imperfect. Decision-makers need to adapt to their communication environment. On the other hand, communication itself may require decision-making such as deciding when and with whom to communicate and what information to communicate. Thus, communication and decision-making are intimately coupled in decentralized systems. The overarching goal of the proposed research is to investigate this interplay of communication and control/decision-making that arises in decentralized systems operating in dynamic and uncertain environments. The technological and societal impacts of the proposed research will come through the development of new theoretical tools for communication and control in decentralized systems that will impact many applications and industries including: sensing and surveillance systems, robotics, smart grid, communication networks, urban infrastructure maintenance, and cyber-physical security. The educational impact of the project will come through the training of graduate students and engagement with local high-school students through school visits and/or on-campus events.The focus of the proposed research will be on strategic decision-making which requires identification of decision strategies for agents that allow them to respond to real-time information about unfolding events in their environment in order to optimize the performance of a decentralized dynamic system. The proposed work will address fundamental questions regarding control and decision-making in decentralized systems operating in uncertain and dynamic environments. In particular, it will explore characterization and computation of decision strategies in the following domains: decentralized stochastic control with unreliable and imperfect communication among controllers, minimax decision and control problems in decentralized systems, joint optimization of real-time coding, communication scheduling and control strategies for decentralized control. The fruits of this research will potentially constitute an important step in the further understanding of the interplay of information, communication and decision-making in dynamic decentralized systems.