一种求解非线性约束优化问题的无罚函数无滤子的方法

A Method without a Penalty Function or a Filter for Nonlinear Constrained Optimization

借助于强次可行方向法的思想和滤子法的思想,给出了一种求解非线性约束优化问题的无罚函数无滤子的方法.方法借助于广义投影技术产生搜索方向,直接通过原目标函数和约束违反度函数作为搜索函数来产生步长,有效地避免了消耗计算成本的恢复阶段.最后在适当的假设条件下,给出了算法的全局收敛性和有效性.

In this paper,we present an algorithm for nonlinear constrained optimization based on the strongly sub-feasible direction method and the filter method.The proposed approach can avoid using neither a penalty,nor a filter.The search direction is yield via the generalized projection technique,and the step-size is produced through the original objective function and constraint violation function as search function.Therefore,this method is effective to avoid the restoration phase which is necessary for the general filter methods.To a certain extent,the computation cost is simplified.Under some reasonable assumptions,the method can be proved to be globally convergent.