The difference and unity of irregular LQ control and standard LQ control and its solution

Irregular linear quadratic control (LQ, was called Singular LQ) has been a long-standing problem since 1970s. This paper will show that an irregular LQ control (deterministic) is solvable (for arbitrary initial value) if and only if the quadratic cost functional can be rewritten as a regular one by changing the terminal cost $x'(T)Hx(T)$ to $x'(T)[H+P_1(T)]x(T)$, while the optimal controller can achieve $P_1(T)x(T)=0$ at the same time. In other words, the irregular controller (if exists) needs to do two things at the same time, one thing is to minimize the cost and the other is to achieve the terminal constraint $P_1(T)x(T)=0$, which clarifies the essential difference of irregular LQ from the standard LQ control where the controller is to minimize the cost only. With this breakthrough, we further study the irregular LQ control for stochastic systems with multiplicative noise. A sufficient solving condition and the optimal controller is presented based on Riccati equations.

Optimization of Constrained Stochastic Linear-Quadratic Control on an Infinite Horizon: A Direct-Comparison Based Approach

In this paper we study the optimization of the discrete-time stochastic linear-quadratic (LQ) control problem with conic control constraints on an infinite horizon, considering multiplicative noises. Stochastic control systems can be formulated as Markov Decision Problems (MDPs) with continuous state spaces and therefore we can apply the direct-comparison based optimization approach to solve the problem. We first derive the performance difference formula for the LQ problem by utilizing the state separation property of the system structure. Based on this, we successfully derive the optimality conditions and the stationary optimal feedback control. By introducing the optimization, we establish a general framework for infinite horizon stochastic control problems. The direct-comparison based approach is applicable to both linear and nonlinear systems. Our work provides a new perspective in LQ control problems; based on this approach, learning based algorithms can be developed without identifying all of the system parameters.