A Study on Applying Interactive Multi-objective Optimization to Multiagent Systems
A Study on Applying Interactive Multi-objective Optimization to Multiagent Systems
复制标题
DOI：
--
复制
发表时间：
2017
影响因子：
2.8
通讯作者：
T. Matsui
中科院分区：
文献类型：
--
作者：
T. Matsui
多智能体系统中的约束优化问题已作为决策和资源分配的基本问题进行了研究。当每个代理对某个问题有自己的特定兴趣时，该任务被定义为具有不对称函数的多目标优化问题。由于多目标问题在一般情况下具有多个帕累托最优解，因此使用适当的标准来选择优选的解决方案。交互式方法是多目标优化问题的求解方法，其中基于用户指定的参数反复修改解。虽然交互式方法是为单个用户开发的，但通过用代理组替换单个用户，这样的框架可以被视为多代理系统上的分析或游戏的基础。为了研究这种方法，本文研究了一个类似于交互式愿望水平方法的框架，该框架采用标准化标准来解决代理之间的不公平问题，并评估交互的示例案例。
Constraint optimization problems in multiagent systems have been studied as fundamental problems of decision making and resource allocation. When each agent has its own specific interest in a problem, that task is defined as a multiobjective optimization problem with asymmetric functions. Since multi-objective problems have a number of Pareto optimal solutions in general cases, a preferred solution is selected using an appropriate criterion. The interactive methods are approaches of solution methods for multi-objective optimization problems, where a solution is repeatedly modified based on userspecified parameters. While interactive methods are developed for single users, such a framework can be considered as the base of analysis or games on multiagent systems by replacing single users with sets of agents. To study such an approach, this paper investigates a framework that resembles the interactive aspiration level methods employing a scalarization criterion that addresses unfairness among agents, and evaluates an example case of interaction.