Real-Time Near-Optimal Feedback Control of Aggressive Vehicle Maneuvers

Abstract

Optimal control theory can be used to generate aggressive maneuvers for vehicles under a variety of conditions using minimal assumptions. Although optimal control provides a very powerful framework for generating aggressive maneuvers utilizing fully nonlinear vehicle and tire models, its use in practice is hindered by the lack of guarantees of convergence, and by the typically long time to generate a solution, which makes this approach unsuitable for real-time implementation unless the problem obeys certain convexity and/or linearity properties. In this paper, we investigate the use of statistical interpolation (e.g., kriging) in order to synthesize on-the-fly near-optimal feedback control laws from pre-computed optimal solutions. We apply this methodology to the challenging scenario of generating a minimum time yaw rotation maneuver of a speeding vehicle in order to change its posture prior to a collision with another vehicle, in an effort to remedy the effects of a head on collision. It is shown that this approach offers a potentially appealing option for real-time, near-optimal, robust trajectory generation.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Sep 03, 2014
Accession Number
AD1146228

Entities

People

  • Panagiotis Tsiotras
  • Ricardo Sanz-diaz

Organizations

  • Georgia Tech Research Corporation

Tags

Communities of Interest

  • Space

DTIC Thesaurus Topics

  • Aircrafts
  • Algorithms
  • Automobiles
  • Collision Avoidance
  • Computations
  • Control Systems
  • Control Systems Engineering
  • Control Theory
  • Data Science
  • Dynamics
  • Engineering
  • Friction
  • Gaussian Processes
  • Information Science
  • Maximum Likelihood Estimation
  • Model Predictive Control
  • Network Science
  • Safety Equipment
  • Simulations
  • Standards

Fields of Study

  • Computer science

Readers

  • Computational Fluid Dynamics (CFD)
  • Control Systems Engineering.
  • Operations Research