A multi-agent system for DRCMPSP based on auction and critical chain

Abstract

This paper develops a multi-agent system based on auction and critical chain (MAS/AC) for decentralised resource constrained multiple project scheduling problem (DRCMPSP). In this system, parallel projects are to be scheduled simultaneously under the consideration of multi-objectives, where the allocation of global resources is independent from the allocation of activities and local resources through a multi-agent system. The critical chain method (CCM) is applied to adjust the schedule and local resource constraints of each single project. Subsequently, multiple objectives of different projects are considered in the process of calculating bid price. Moreover, the distribution of global resources is conducted by AA employed combinatorial auction. Through the interactive communication between AA and PAs, parallel projects could be scheduled simultaneously with respect to their different characteristics and interests, and the benefits of the whole system could be ensured meanwhile. An experimental example together is applied to demonstrate the efficiency and superiority of MAS/AC.