Controller design by dynamic output sensing for linear partly-quantized feedback control systems

Abstract

In this paper the problem of stabilizing a linear timeinvariant system with quantized feedback is addressed. We focus on systems with rectilinear quantization where only some of the outputs are quantized either by limited sensor precision or data rate constraints and the rest of the output is sent to the controller without quantization. We call these systems partly-quantized feedback systems. Aside from the rectilinear boxes that we know output resides in, we determine moving hyperplanes in output space such that exact value of output has to lie on these planes too. Using the intersection of rectilinear boxes and hyperplanes, we find the best estimate of output to be used in linear feedback controllers and hence we reduce the maximum quantization error and find a smaller invariant region around origin. A simulation of the proposed controller is also presented.