Controlling unknown linear dynamics with almost optimal regret (Q7006399)

This paper studies in great detail control strategies for a time-invariant, linear system with the noise denoted by \(d\,W(t) \) and a single unknown parameter denoted by \(a\). The noise is defined by a stochastic process with a family of random variables. Several variants of the optimal control problem are presented. In Variant I as ``Classical control'', \(a\) is known; the optimal strategy is defined and a limit is imposed on the control. The authors have used this assumption to advance their goal in the article; in the modern control, we do not need such assumptions for a time-invariant linear system with the known \(a\). In the second type, parameter \(a\) is unknown and the optimal strategy is proposed. Same goes for other variants. In the remainder of the article, which is very long, different strategies are introduced and studied. In short, it is advisable to read and investigate this long article for those who research in adaptive control theory, optimal control system with unknown dynamics, stochastic optimization, and optimization methods by using random variables.