Data-Driven Modeling for Dual Retrospectve Cost Adaptive Control

Abstract

We propose to apply principles and techniques of DDDAS to adaptive control of complex systems.In particular, the focus is on dual retrospective cost adaptive control (DRCAC), whichis applicable to command following and disturbance rejection for discrete-time systems. DRCACrequires minimal modeling information, and the objective of this proposal is to developand demonstrate ecient and eective DDDAS-inspired identication methods that can be implementedonline to obtain this modeling information. One of the key challenges is to perform thedata-driven modeling with sucient speed and accuracy that the adaptive control algorithm canobtain the required modeling information. We will develop and demonstrate techniques for thisproblem based on quasi-closed-loop identification methods, where the goal is to determine probingsignals that are minimally disruptive but sufficiently persistent to obtain the required modelinginformation. The concurrent implementation of data-driven modeling and adaptive control is achallenging but important problem for control technology with diverse applications to vehicles andprocesses.

Document Details

Document Type
DoD Grant Award
Publication Date
Apr 09, 2018
Source ID
FA95501810171

Entities

People

  • Dennis S. Bernstein

Organizations

  • Air Force Office of Scientific Research
  • United States Air Force
  • University of Michigan

Tags

Fields of Study

  • Computer science

Readers

  • Adaptive Control and Estimation with Uncertainty in Dynamic Systems.
  • Data Mining and Knowledge Discovery.
  • Distributed Systems and Data Platform Development