Key Features

A portable implementation of the Reduced Feasible Sequential Quadratic Programming (RFSQP) algorithm, a superlinearly convergent algorithm for directly tackling optimization problems with: Multiple competing linear/nonlinear objective functions (minimax). linear/nonlinear inequality constraints, linear/nonlinear equality constraints.

The algorithm contains special provisions for

	Maintaining ``semi-feasibility'' of each iterate.
	Efficiently handling problems with many ``sequentially related'' objectives and/or constraints.

RFSQP features

	Availability of RFSQP source code, thus both can be easily integrated into user's own program.
	Ability to search for an initial feasible point satisfying all linear constraints and nonlinear inequality constraints.
	Ability to generate iterates satisfying all linear constraints and nonlinear inequality constraints (mandatory for many applications), starting from a feasible point.
	Ability to improve objective function after each iteration if there are no nonlinear equality constraints.
	Ability to handle multiple competing objective functions (minimax problems).
	Finite difference approximation to gradients if exact gradients are not provided by the user.
	Global convergence and fast (superlinear) local convergence.
	Special scheme to efficiently handle problems with many more objectives or/and constraints than variables (finely discretized semi-infinite problems).
	Ability to use "warm starts" to speed up computation of search directions, when appropriate quadratic programming software is available; see "What is RFSQP".
	Availability of original authors to discuss any theoretical and practical issues related to RFSQP.

Back to the top

Send mail to webmaster@aemdesign.com with questions or comments about this web site.
Last modified: February 21, 2005