A globally convergent SQP-type method with least constraint violation for nonlinear semidefinite programming

TL;DR

This paper introduces a globally convergent SQP-type algorithm for nonlinear semidefinite programming that minimizes constraint violation and converges to a Fritz-John point, demonstrated through numerical experiments.

Contribution

The paper proposes a novel two-phase SQP-type method that guarantees convergence to a Fritz-John point with minimal constraint violation in nonlinear semidefinite programming.

Findings

Algorithm converges globally to a Fritz-John point.

Effective in reducing constraint violation.

Numerical experiments confirm robustness and efficiency.

Abstract

We present a globally convergent SQP-type method with the least constraint violation for nonlinear semidefinite programming. The proposed algorithm employs a two-phase strategy coupled with a line search technique. In the first phase, a subproblem based on a local model of infeasibility is formulated to determine a corrective step. In the second phase, a search direction that moves toward optimality is computed by minimizing a local model of the objective function. Importantly, regardless of the feasibility of the original problem, the iterative sequence generated by our proposed method converges to a Fritz-John point of a transformed problem, wherein the constraint violation is minimized. Numerical experiments have been conducted on various complex scenarios to demonstrate the effectiveness of our approach.