Analysis and Control of Stochastic Systems Using Semidefinite Programming over Moments

Andrew Lamperski, Khem Raj Ghusinga, Abhyudai Singh

Research output: Contribution to journalArticle

Abstract

This technical note develops a unified methodology for probabilistic analysis and optimal control design for jump diffusion processes defined by polynomials. The statistical moments of these systems can be described by a system of linear ordinary differential equations. Typically, however, the low-order moments depend on higher order moments, thus requiring an infinite system of equations to compute any moment exactly. Here, we develop a methodology for bounding statistical moments by using the higher order moments as inputs to an auxiliary convex optimal control problem with semidefinite constraints. For steady-state problems, the auxiliary optimal control problem reduces to a static semidefinite program. The method applies to both controlled and uncontrolled stochastic processes. For stochastic optimal control problems, the method gives bounds on achievable performance and can be used to compute approximately optimal solutions. For uncontrolled problems, both upper and lower bounds on desired moments can be computed. While the accuracy of most moment approximations cannot be quantitatively characterized, our method guarantees that the moment of interest is between the computed bounds.

Original languageEnglish (US)
Article number8472813
Pages (from-to)1726-1731
Number of pages6
JournalIEEE Transactions on Automatic Control
Volume64
Issue number4
DOIs
StatePublished - Apr 1 2019

Fingerprint

Stochastic systems
Random processes
Ordinary differential equations
Polynomials

Keywords

  • Optimal control
  • nonlinear systems
  • stochastic processes

Cite this

Analysis and Control of Stochastic Systems Using Semidefinite Programming over Moments. / Lamperski, Andrew; Ghusinga, Khem Raj; Singh, Abhyudai.

In: IEEE Transactions on Automatic Control, Vol. 64, No. 4, 8472813, 01.04.2019, p. 1726-1731.

Research output: Contribution to journalArticle

@article{f338aecf685f4eb29ca31f9fb2ecf90b,
title = "Analysis and Control of Stochastic Systems Using Semidefinite Programming over Moments",
abstract = "This technical note develops a unified methodology for probabilistic analysis and optimal control design for jump diffusion processes defined by polynomials. The statistical moments of these systems can be described by a system of linear ordinary differential equations. Typically, however, the low-order moments depend on higher order moments, thus requiring an infinite system of equations to compute any moment exactly. Here, we develop a methodology for bounding statistical moments by using the higher order moments as inputs to an auxiliary convex optimal control problem with semidefinite constraints. For steady-state problems, the auxiliary optimal control problem reduces to a static semidefinite program. The method applies to both controlled and uncontrolled stochastic processes. For stochastic optimal control problems, the method gives bounds on achievable performance and can be used to compute approximately optimal solutions. For uncontrolled problems, both upper and lower bounds on desired moments can be computed. While the accuracy of most moment approximations cannot be quantitatively characterized, our method guarantees that the moment of interest is between the computed bounds.",
keywords = "Optimal control, nonlinear systems, stochastic processes",
author = "Andrew Lamperski and Ghusinga, {Khem Raj} and Abhyudai Singh",
year = "2019",
month = "4",
day = "1",
doi = "10.1109/TAC.2018.2872274",
language = "English (US)",
volume = "64",
pages = "1726--1731",
journal = "IEEE Transactions on Automatic Control",
issn = "0018-9286",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "4",

}

TY - JOUR

T1 - Analysis and Control of Stochastic Systems Using Semidefinite Programming over Moments

AU - Lamperski, Andrew

AU - Ghusinga, Khem Raj

AU - Singh, Abhyudai

PY - 2019/4/1

Y1 - 2019/4/1

N2 - This technical note develops a unified methodology for probabilistic analysis and optimal control design for jump diffusion processes defined by polynomials. The statistical moments of these systems can be described by a system of linear ordinary differential equations. Typically, however, the low-order moments depend on higher order moments, thus requiring an infinite system of equations to compute any moment exactly. Here, we develop a methodology for bounding statistical moments by using the higher order moments as inputs to an auxiliary convex optimal control problem with semidefinite constraints. For steady-state problems, the auxiliary optimal control problem reduces to a static semidefinite program. The method applies to both controlled and uncontrolled stochastic processes. For stochastic optimal control problems, the method gives bounds on achievable performance and can be used to compute approximately optimal solutions. For uncontrolled problems, both upper and lower bounds on desired moments can be computed. While the accuracy of most moment approximations cannot be quantitatively characterized, our method guarantees that the moment of interest is between the computed bounds.

AB - This technical note develops a unified methodology for probabilistic analysis and optimal control design for jump diffusion processes defined by polynomials. The statistical moments of these systems can be described by a system of linear ordinary differential equations. Typically, however, the low-order moments depend on higher order moments, thus requiring an infinite system of equations to compute any moment exactly. Here, we develop a methodology for bounding statistical moments by using the higher order moments as inputs to an auxiliary convex optimal control problem with semidefinite constraints. For steady-state problems, the auxiliary optimal control problem reduces to a static semidefinite program. The method applies to both controlled and uncontrolled stochastic processes. For stochastic optimal control problems, the method gives bounds on achievable performance and can be used to compute approximately optimal solutions. For uncontrolled problems, both upper and lower bounds on desired moments can be computed. While the accuracy of most moment approximations cannot be quantitatively characterized, our method guarantees that the moment of interest is between the computed bounds.

KW - Optimal control

KW - nonlinear systems

KW - stochastic processes

UR - http://www.scopus.com/inward/record.url?scp=85054239504&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85054239504&partnerID=8YFLogxK

U2 - 10.1109/TAC.2018.2872274

DO - 10.1109/TAC.2018.2872274

M3 - Article

VL - 64

SP - 1726

EP - 1731

JO - IEEE Transactions on Automatic Control

JF - IEEE Transactions on Automatic Control

SN - 0018-9286

IS - 4

M1 - 8472813

ER -