Proximal Limited-Memory Quasi-Newton Methods for Scenario-based Stochastic Optimal Control

Ajay Kumar Sampathirao, Pantelis Sopasakis, Alberto Bemporad, Panagiotis Patrinos

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

3 Citations (Scopus)

Abstract

Stochastic optimal control problems are typically of rather large scale involving millions of decision variables, but possess a certain structure which can be exploited by first-order methods such as forward-backward splitting and the alternating direction method of multipliers (ADMM). In this paper, we use the forward-backward envelope, a real-valued continuously differentiable penalty function, to recast the dual of the original nonsmooth problem as an unconstrained problem which we solve via the limited-memory BFGS algorithm. We show that the proposed method leads to a significant improvement of the convergence rate without increasing much the computational cost per iteration.
Original languageEnglish
Title of host publicationIFAC World Congress
Pages11865-11870
Number of pages6
ISBN (Electronic)2405-8963
DOIs
Publication statusPublished - Jul 2017
Event20th World Congress of the International Federation of Automatic Control 2017 - Toulouse, France
Duration: 10 Jul 201714 Jul 2017

Conference

Conference20th World Congress of the International Federation of Automatic Control 2017
Abbreviated titleIFAC 2017
Country/TerritoryFrance
CityToulouse
Period10/07/201714/07/2017

Keywords

  • Quasi-Newtonian methods
  • Convex optimization
  • Large-scale optimization
  • Parallelization
  • Stochastic optimal control
  • Stochastic Model Predictive Control
  • GPGPU

Fingerprint

Dive into the research topics of 'Proximal Limited-Memory Quasi-Newton Methods for Scenario-based Stochastic Optimal Control'. Together they form a unique fingerprint.

Cite this