The Genetic Algorithm For Solving Multi-objective Resource Constrained Multi-project Scheduling Problem

Resources constrained project scheduling problem(RCPSP)plays a key role in business management.The multi-objective resource constrained multi-project scheduling problem(MPCMPSP)is still a challenge to the company in operation and management under the complicated circumstances.There are resource sharing and competition considerations in the MRCMPSP.The key point to solve the MRCMPSP is to set each activity's start time,and meanwhile to meet their resource requirements,furthermore to optimize the objectives.In order to solve the problem,a mathematic model is proposed for the MRCMPSP with the activities' relationship and resource constraint taken into consideration.Based on the genetic algorithm,which is used for solving continuous multi-objective problem,a discrete reference-point based non-dominated sorting genetic algorithm(DNSGA-III)is proposed.The resource constrained project scheduling problem should be solved with three objectives,i.e.,project time,cost and resource moment deviation,are taken into account.The comparison between reference-point based non-dominated sorting genetic algorithm and crowding-distance based non-dominated sorting genetic algorithm are done both on a simple benchmark and a complex real construction project.And the results show that the mathematic model is correct and the discrete reference-point based non-dominated sorting genetic algorithm is valid and outperforms.A multi-project scheduling problem based on three single projects is constructed and several reasonable schedules are obtained by using of DNSGA-III.