Unidentifiability of System Dynamics: Conditions and Controller Design

TL;DR

This paper explores conditions under which system dynamics are unidentifiable and proposes a controller design method to ensure unidentifiability in linear systems, using Fisher Information Matrix analysis and low-rank controllers.

Contribution

It introduces a necessary and sufficient condition for dynamic unidentifiability and a controller design scheme for linear systems to achieve unidentifiability.

Findings

Derived the unidentifiability condition using Fisher Information Matrix.

Proposed a low-rank controller design for unidentifiability in linear systems.

Validated the approach through simulation results.

Abstract

How to make a dynamic system unidentifiable is an important but still open issue. It not only requires that the parameters of the systems but also the equivalent systems cannot be identified by any identification approaches. Thus, it is a much more challenging problem than the existing analysis of parameter identifiability. In this paper, we investigate the problem of dynamic unidentifiability and design the controller to make the system dynamics unidentifiable. Specifically, we first define dynamic unidentifiability by taking both system parameters and equivalent systems into consideration. Then, we obtain the necessary and sufficient condition for unidentifiability based on the Fisher Information Matrix. This condition is derived by analysis of the relationship between the unidentifiable parameters and the Hessian matrix of the system function. Next, we propose a controller design scheme for ensuring dynamic unidentifiability under linear system models. We prove that for controllable and observable linear time-invariant systems, the requirement of unidentifiability is equivalent to the requirement of a low-rank controller. Then, the low-rank controller design problem is solved by transforming it into a model order reduction problem. We demonstrate the effectiveness of our method by simulation.