Abstract:

In practice， multiple projects may be located in different locations， therefore， the global resource transfer times cannot be neglected when making multi-project scheduling plans. Besides， project activities have different execution modes， activity durations can be shortened by increasing the budget. Both global resource transfer times and budget constraints are considered and the time/cost trade-off problem in decentralized multi-project scheduling is proposed.Based on a multi-agent system， a two-stage model including local scheduling model and global coordination model is constructed， each project agent （PA） independently schedules the corresponding single project under the allocated budget constraints to minimize the project makespan， and the coordination agent （CA） uses a coordination mechanism to eliminate budget competition conflicts to minimize average project delay （APD）. The local scheduling is the discrete time/cost trade-off problem under budget constraints （DTCTP-B）， a new genetic algorithm is designed to solve it. The sequential game negotiation mechanism is employed to coordinate the allocation of the total budget among multiple projects and make the global resource transfer plans.To evaluate the performance of the proposed algorithm， 45 instances are selected from MPSPLIB and converted into the problems studied in this paper. Experimental results show that compared with the heuristic method， the method proposed in this paper is more effective and can further reduce APD. The results of parameter analysis show that both problem size and global resource conflict degree have influences on the APD， and the influence of the global resource conflict degree is greater. Taking instance 2 of problem subset MP30_5 as an example， the influence of total budget input on APD is analyzed and discussed.

Key words: decentralized multi-project scheduling, resource transfers, budget allocation, time/cost trade-off, sequential game-based negotiation mechanism