报告题目：Optimal decentralized control and optimization based on optimal control

报告人：张焕水教授（山东科技大学）

报告时间：6月18日  16：00-17:00

报告地点：数学楼301室

摘要：The decentralized LQ control has a long history of study, and majority of the works is contingent on the scenario of history-control sharing. In the case of no history-control sharing, the problem poses a challenge, and the solution in general cannot be obtained explicitly because it is generally nonlinear. For these reasons, research effort has been by and large focused on special systems such as those with a nested structure, or sub-optimal solutions. By proposing a method based on linear optimal control, we derive an analytical solution of decentralized controllers with feedback of a linear optimal estimator for cases including information inclusion and delayed information sharing. The decentralized LQ control for deterministic systems with general information pattern is also investigated. We show that the controller approaches to the optimal centralized controller if the observer is stable. The approach is further shown to have the advantage to be readily extendable to study consensus control problems, leading to optimal consensus controllers with fast converge speed. In particular, when applied to solve nonlinear optimization problems, the approach circumvents effectively the weaknesses of gradient descent (slow convergence) and Newton iteration (instability and divergence).