Tsumura, Koji (The Univ. of Tokyo)

Approximation of Discrete Time Linear Systems via Bit Length of Memory

Scheduled for presentation during the Mini-Symposium "Networked control: Rate constraints and quantization effects" (TuP08), Tuesday, July 25, 2006, 17:00−17:25, Room I

17th International Symposium on Mathematical Theory of Networks and Systems, July 24-28, 2006, Kyoto, Japan

This information is tentative and subject to change. Compiled on May 27, 2023

Abstract

In this paper, we deal with approximation problem of discrete time linear systems by bit-memory systems. We consider bit-length for quantized state vector as the complexity of systems. Then, we propose an approximation method by the systems of short bit-length. On the other hand, the complexity of systems is the degree of state variables in ordinary system approximation. The advantages of this method are: 1) it clarifies the relationship between system dynamics and information, 2) it enables finer approximation compared with the ordinary one, 3) it enables on-line tuning of the system complexity according to its necessity. In this paper, we analyse the relationship between the bit length and approximation error, and the minimum bit length or its upper/lower bound for the given approximation specifications is induced.