Abstract
The optimal closed-loop quantized control is derived for the linear-quadratic-Gaussian formulation and shown to be separable in estimation, control, and quantization. The optimal quantizer is time-varying and minimizes a quadratic distortion measure with weighting matrix dependent upon the solution to the matrix Riccati equation. The optimal cost-togo is shown to be the sum of the cost-to-go for the optimal continuous-valued control solution and a term reflecting the quantizer distortion.

This publication has 7 references indexed in Scilit: