Abstract:The focus of this paper is on linear system identification in the setting where it is known that the underlying partially-observed linear dynamical system lies within a finite collection of known candidate models. We first consider the problem of identification from a given trajectory, which in this setting reduces to identifying the index of the true model with high probability. We characterize the finite-time sample complexity of this problem by leveraging recent advances in the non-asymptotic analysis of linear least-square methods in the literature. In comparison to the earlier results that assume no prior knowledge of the system, our approach takes advantage of the smaller hypothesis class and leads to the design of a learner with a dimension-free sample complexity bound. Next, we consider the switching control of linear systems, where there is a candidate controller for each of the candidate models and data is collected through interaction of the system with a collection of potentially destabilizing controllers. We develop a dimension-dependent criterion that can detect those destabilizing controllers in finite time. By leveraging these results, we propose a data-driven switching strategy that identifies the unknown parameters of the underlying system. We then provide a non-asymptotic analysis of its performance and discuss its implications on the classical method of estimator-based supervisory control.

On Necessary and Sufficient Conditions for Identifiability and Identification of Switching Dynamical Networks

Identifiability and Identification of Switching Dynamical Networks: A Data-Based Approach

On-line identification of power system dynamic equivalent: Part One: Models and Identifiability

On the Identifiability of Switching Dynamical Systems

Reliability of a Class of Nonlinear Systems under Switching Random Excitations

A least-square method for non-asymptotic identification in linear switching control

State identification for a class of uncertain switched systems by differential neural networks

On Identification Of Parameterized Switched Linear Systems

Identifiability in Dynamic Acyclic Networks with Partial Excitations and Measurements

A Novel Recursive Approach for Online Identification of Continuous‐time Switched Nonlinear Systems

A Two-stage Identification Method for Switched Linear Systems

Nonlinear Network Identifiability with Full Excitations

Identification of a Class of Singular Boolean Control Networks

Unidentifiability of System Dynamics: Conditions and Controller Design

Network Identifiability from Intrinsic Noise.

Simultaneous Recursive Identification of Parameters and Switching Manifolds Identification of Discrete-Time Switched Linear Systems

Local Identifiability of Networks with Nonlinear Node Dynamics

Sparse Bayesian Learning for Switching Network Identification

Necessary and Sufficient Conditions for Controllability and Observability of Switched Impulsive Control Systems

Identification of non-causal systems with arbitrary switching modes

Identification of Switched Linear System Based on Dynamic Programming