Minimum output variance control for FSN models: Continuous-time case (Q5950073)

In control theory, the most widely used model for noise signals is additive noise taken independent of the signal intensity, which is unrealistic in practice. The authors propose finite signal-to-noise ratio models for linear stochastic systems in which the noise intensity is related to the signal covariance in proportion to the noise-to-signal ratio. Computation errors are treated similarly. The objective is to design a state feedback control law such that the closed-loop system is mean square asymptotically stable and the output variance is minimized. The minimum output variance control problem is formulated and solved for finite signal-to-noise models. Necessary and sufficient conditions for optimality are proposed. An optimal control law is proposed that involves a positive definite solution to a Riccati type equation. A numerical algorithm for the solution to this Riccati type equation is given. Convergence of the algorithm is investigated. The proposed finite signal-to-noise optimal controller achieves the maximum accuracy in finite control steps. A relation to information quality, the maximum accuracy, and the minimum control effort are obtained.