Hybrid Control of Autonomous Systems with Mixed-Integer Quadratic Optimization

Abstract

This research proposal offers new solution methods for hybrid control of autonomous systems with mixed-integer quadratic optimization. In particular, it describes new approaches for deriving strong relaxations and scalable algorithms for their convex quadratic optimization formulations with indicator variables. The main idea is to decompose the convex quadratic objective function in these formulations into a sum of convex quadratics defined over low-rank matrices and to derive strong convex relaxations for them. Separation and duality approaches will be investigated to find the best decompositions into low-rank blocks. Low-order relaxations and special algorithms will be developed for scalable implementations. The resulting models and algorithms will be tested on hybrid control systems through extensive computational experiments.

Document Details

Document Type
DoD Grant Award
Publication Date
Mar 11, 2020
Source ID
N000142012076

Entities

People

  • Alper Atamtürk

Organizations

  • Office of Naval Research
  • United States Navy
  • University of California Regents

Tags

Readers

  • Computational Fluid Dynamics (CFD)
  • Linear Algebra
  • Robotics and Automation.

Technology Areas

  • Autonomy
  • Autonomy - Autonomous System Control