Revista de Matemática Aplicada e Computacional

An Interior-Point Method for Nonlinear Constrained Optimization Problem with Trust-Region Mechanism

Abstract

Bothina El-Sobky, Gehan Ashry and Yousria Abo-Elnaga

We introduced an algorithm to solve a Non Linear Constrained Optimization (NLCO) problem in this paper. This algorithm follows Das’s idea of Newton’s interiorpoint method that uses a diagonal matrix of Coleman and Li for NLCO problems. A Trust-Region (T-R) mechanism is used to globalize the algorithm. This algorithm follows Byrd and Omojokun’s idea of step decomposition. It is a successful idea to overcome the difficulty of having an infeasible quadratic T-R sub problem and converts the quadratic T-R sub problem into two unconstrained T-R sub problems.

A global convergence theory of the algorithm is studied under five standard assumptions. This algorithm is different and maybe simpler than similar ideas such that the global convergence theory is not depending on the linear independence assumption on the gradients of the constraints.

Some numerical tests are stated to indicate that the algorithm performs effectively and efficiently in pursuance.