Utility Based Optimal Task Scheduling Problem in a Multi-agent System

A multi-agent system (MAS) has been built based on characteristics of emergency system, in which task schedulable and maximum task scheduling problem for the MAS has been discussed in our previous work. However, the mentioned task scheduling method both supposed that all tasks were equally important and agents have the same capability in playing a certain role, and evaluated the quality of task scheduling only by the number of accomplished tasks, which ignored the differences among tasks or Agents that conflicts with our real life. In order to make the task scheduling be more close to actual process of emergencies, the utility based optimal task scheduling problem (UOTSP) is proposed in this paper. The method assigns each task with a utility to represent the earnings from accomplishing it and adds preference on each Agent to distinguish the Agent’s preferences and roleplay abilities. The UOTSP problem is proved to be NP complete by reducing MAX-S to it in polynomial time, and two approximate algorithms (GTS and RTOM) are proposed to solve the UOTSP problem approximately. The experiment results show that both GTS and RTOM can get excellent approximate solutions, and the time performance of GTS is wonderful which makes it suitable for the emergency applications of MAS (emergency system).