Sequential Decomposition of Sequential Dynamic Teams:Applications to Real-Time Communication and Networked Control Systems.

[摘要] Optimal design of multi-agent sequential teams is investigated in this thesis. Asystematic methodology is presented to convert the search for an optimal multistagedesign into a sequence of nested optimization problems, where at each stepthe best decision rule of a agent at a given time is search. This conversion is calledsequential decomposition and it drastically simplifies the search of optimal solutionfor both finite and infinite horizon problems. The main idea is as follows. A statesufficient for input-output mapping of the system is identified. A joint probabilitymeasure on this state is an information state sufficient for performance evaluation.This information state evolves in time in a deterministic manner depending on thechoice of decision rules of the agents. Thus, these information states are a controlledMarkov process where the control actions are the decision rules of the agents. Theoptimal control of the time-evolution of these information states results in a sequentialdecomposition of the problem. Applications of this methodology to real-timecommunication and optimal feedback control over noisy communication channelsis also investigated.

[发布机构] University of Michigan

[效力级别] Real-time Communication

[关键词] Sequential Teams;Real-time Communication;Networked Control Systems;Zero-delay Communication;Decentralized Control;Non-classical Information Structures;Electrical Engineering;Engineering;Electrical Engineering: Systems