一种序列线性方程组滤子算法的全局收敛性

Global Convergence of Sequential Systems of Linear Equations(SSLE) Filter Algorithm

提出了一种不可行序列线性规划滤子方法,只需求解2个具有相同系数矩阵的线性方程组以得到搜索方向,在一定程度上克服了序列二次规划方法的缺点并提高了计算效率.算法中使用了χ-有效集.给出了该算法的全局收敛性证明,并给出了数值结果说明该算法的有效性.

The paper presents a new filter algorithm based on the sequential systems of linear equations, and infeasible QP-free method. The new algorithm just needs to solve two systems of linear equations with the same nosingular coefficient matrix and avoids large computational amount and infeasiblility of quadratic subproblems, which are the defects of the existing sequential quadratic programming type algorithms. The X-active set procedure is used to improve the computational effect, and the global convergence of the proposed algorithm is given. Finally, the paper presents the numerical results to validate the efficiency of this new algorithm.