Finite-time Sample Complexity Analysis of Least Square Identifying Stochastic Switched Linear System

Negin Musavi, Geir E. Dullerud

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

In this paper, we examine the high-probability finite-time theoretical guarantees of the least squares method for system identification of switched linear systems with process noise and without control input. We consider two scenarios: one in which the switching is i.i.d., and the other in which the switching is according to a Markov process. We provide concentration inequalities using a martingale-type argument to bound the identification error at each mode, and we use concentration lemmas for the switching signal. Our bound is in terms of state dimension, trajectory length, finite-time gramian, and properties of the switching signal distribution. We then provide simulations to demonstrate the accuracy of the identification. Additionally, we show that the empirical convergence rate is consistent with our theoretical bound.

Original languageEnglish (US)
Title of host publication2023 62nd Annual Conference of the Society of Instrument and Control Engineers, SICE 2023
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages488-493
Number of pages6
ISBN (Electronic)9784907764807
DOIs
StatePublished - 2023
Externally publishedYes
Event62nd Annual Conference of the Society of Instrument and Control Engineers, SICE 2023 - Tsu, Japan
Duration: Sep 6 2023Sep 9 2023

Publication series

Name2023 62nd Annual Conference of the Society of Instrument and Control Engineers, SICE 2023

Conference

Conference62nd Annual Conference of the Society of Instrument and Control Engineers, SICE 2023
Country/TerritoryJapan
CityTsu
Period9/6/239/9/23

Keywords

  • Identification
  • Machine learning
  • Sample complexity
  • Switched linear systems

ASJC Scopus subject areas

  • Artificial Intelligence
  • Automotive Engineering
  • Control and Systems Engineering
  • Control and Optimization
  • Instrumentation

Fingerprint

Dive into the research topics of 'Finite-time Sample Complexity Analysis of Least Square Identifying Stochastic Switched Linear System'. Together they form a unique fingerprint.

Cite this