NLPSPR

Nonlinear programming problems with a large number of variables and constraints where the Jacobian and Hessian matrices are sparse.

The algorithm can be used to locate a feasible point for a system of nonlinear equality and/or inequality constraints. The algorithm is well suited for applications derived from discretized differential equations and boundary value problems.

NLPSPR is a sequential quadratic programming algorithm that uses an augmented Lagrangian merit function and a sparse quadratic programming algorithm based on the Schur complement approach of Gill, Murray, Saunders, and Wright. Sparse linear systems are solved efficiently using a multifrontal algorithm that implements a modified Cholesky decomposition for symmetric indefinite systems. The user must supply the sparse Jacobian and Hessian matrices, although this information can be computed efficiently using sparse finite differences which are implemented in a utility package that is also available. The software incorporates a reverse communication structure and is especially well suited for applications derived from discretized optimal control problems.

The software is written in Fortran 77, and uses portions of the BCSLIB mathematical subroutine library, including the BLAS and LINPACK.

Need more info?

Visit the official SOCS web page or contact:

Dr. John T. Betts or
Dr. Paul D. Frank
The Boeing Company
PO Box 3707, MS 7L-21
Seattle, WA 98124

Reference:

Betts, J.T. and Frank, P.D., ``A Sparse Nonlinear Optimization Algorithm,'' Journal of Optimization Theory and Applications, Vol. 82, No. 3, Sept. 1994.

[ Optimization Software Guide | OTC Home Page | NEOS Server | NEOS Guide ]