Skip to main navigation Skip to search Skip to main content

A robust homotopy method for equality constrained nonlinear optimization

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

This paper presents a robust algorithm for solving nonlinear equality constrained optimization problems by introducing a globally convergent homotopy method. The first order necessary condition for the equality constrained optimization is expressed as an orthogonal condition between the performance gradient vector and the null space vectors of the constraint Jacobian matrix. By constructing a nonlinear system of equations consisting of the first order condition and the constraints, we utilize a probability-one homotopy method to search along the zero curve of the homotopy and solve for the optimal solution. Because the integrated homotopy solution never leaves the constrained surface, the proposed algorithm needs neither a penalty function nor Lagrangian multipliers for finding optimal solutions. The root solution histories are smooth and no artificial behaviors are introduced to force the satisfaction of the constraints. No Taylor expansions are introduced for either the performance index or the constraints so no problems arise because of approximated model assumptions. Several examples which could not be solved using 'fmincon', which is a nonlinear optimization function in MATLAB7.0.1, are presented to demonstrate very encouraging results from the proposed algorithm.

Original languageEnglish (US)
Title of host publication12th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference, MAO
PublisherAmerican Institute of Aeronautics and Astronautics Inc.
ISBN (Print)9781563479472
DOIs
StatePublished - 2008

Publication series

Name12th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference, MAO

ASJC Scopus subject areas

  • Aerospace Engineering
  • Mechanical Engineering

Fingerprint

Dive into the research topics of 'A robust homotopy method for equality constrained nonlinear optimization'. Together they form a unique fingerprint.

Cite this