Accession Number:

AD1093314

Title:

Control And Learning Of Uncertain Dynamical Systems: Optimization, Sampling, And Regret

Descriptive Note:

Technical Report,13 Apr 2018,13 Oct 2019

Corporate Author:

University of Washington Seattle United States

Report Date:

2019-11-01

Pagination or Media Count:

25.0

Abstract:

This report shows that first order methods can be used to provide an effective bridge between optimal control theory and sample-based reinforcement learning. The work focuses on the linear quadratic regulator problem and Markov decision processes. Some of the results include a proof that gradient descent starting from a stabilizing policy converges to the globally optimal policy and an algorithm that provides nearly tight regret bounds for the control of a linear dynamical system with adversarial disturbances.

Subject Categories:

  • Cybernetics
  • Electrical and Electronic Equipment

Distribution Statement:

APPROVED FOR PUBLIC RELEASE