多目标优化的交互式定界搜索法

An Interactive Bound Search Method for Multiobjective Optimization

在多目标优化问题的研究中,绝大多数求解方法都基于获取决策者偏好信息的不同而假设,这大致可分为三类: (1)从决策者那里能获取全部偏好信息; (2)从决策者那里不能获取任何偏好信息; (3)从决策者那里能获取部分偏好信息。第一类方法可用效用函数表示决策者的全部偏好信息,将多目标优化问题归结为一个单目标优化问题,由于构造出一个恰当的效用函数表达式常常是很困难的。

In this paper, an interactive bound search method for multiobjective optimization is presented. When the analyst gives an efficient solution, the decision maker makes update idea. Then the analyst searches an improvement solution by the bound method based on this idea. This interactive process is continued until the preference of the decision maker accords with reality and a preferred efficient solution of the decision maker is found. In this method the interaction is practical and clear. The method requires less calculation and is helpful to solve flexible decision problems. The theoretical analysis and numerical examples show that the method is practical and efficient.