1: \begin{abstract}

2: This paper considers the regularization continuation method and the

3: trust-region updating strategy for the nonlinearly equality-constrained optimization

4: problem. Namely, it uses the inverse of the regularization quasi-Newton matrix

5: as the pre-conditioner to improve its computational efficiency in the well-posed

6: phase, and it adopts the inverse of the regularization two-sided projection

7: of the Hessian as the pre-conditioner to improve its robustness in the

8: ill-conditioned phase. Since it only solves a linear system of equations at

9: every iteration and the sequential quadratic programming (SQP) needs to solve a

10: quadratic programming subproblem at every iteration, it is faster than SQP.

11: Numerical results also show that it is more robust and faster than

12: SQP (the built-in subroutine fmincon.m of the MATLAB2020a environment

13: and the subroutine SNOPT executed in GAMS v28.2 (2019) environment).

14: The computational time of the new method is about one third of that of fmincon.m

15: for the large-scale problem. Finally, the global convergence analysis of the new

16: method is also given.

17: \end{abstract}

18: