A three-stage decomposition algorithm for decentralized multi-project scheduling under uncertainty. (October 2021)

• Record Type: Journal Article
• Authors: Liu, Dongning; Xu, Zhe; Li, Feifei
• Is Part Of: Computers & industrial engineering. Volume 160(2021)
• Publication Date: 2021-10
• Subjects: Multi-project scheduling -- Decentralized scheduling -- Resource disruption -- Priority rule; Engineering -- Data processing -- Periodicals; Industrial engineering -- Periodicals; 620.00285
• DOI: 10.1016/j.cie.2021.107553
• Languages: English
• ISSNs: 0360-8352

Abstract

Highlights: A three-stage decomposition algorithm with task scoring mechanism is proposed. Our algorithm is generic for generating and repairing baseline schedule. Our algorithm outperforms several existing algorithms on baseline schedules. Our algorithm repairs disrupted plans efficiently even for large instances. Abstract: Decentralized management of multiple projects is a common practice in areas such as maintenance service, supply chain management, and globalized manufacturing. Nevertheless, the resolution of resource conflicts among multiple projects caused by management independence is a challenging task in generating a multi-project baseline schedule. Various unexpected events may also affect the project execution phase, leading to resource disruptions that can cause the baseline schedule to become infeasible and necessitate its repair. Here, we study the decentralized multi-project scheduling problem subject to global resource disruption, which is NP-hard. To obtain high-quality baseline schedules and effectively repair disrupted schedules, we formulated a three-stage decomposition model. Then, we developed a three-stage decomposition algorithm with a task-scoring mechanism (TDA-TS), which is generic for the processes of schedule generation and repair. Proposed new priority rules were developed that used information about the local schedules and were organized as TDA-TS to further improve its performance. A comprehensive experiment was conducted on the Multi-Project Scheduling Problem LIBrary dataset. The experimental results indicated that the proposed algorithm outperformed several existing centralized and distributed algorithms in the generation of baseline schedules and that it could effectively repair disrupted schedules even for large instances. Furthermore, we saw that the newly proposed priority rule embedded in TDA-TS significantly outperformed other priority rules in repairing the disrupted schedules. The results of this research are valuable for managers who need to effectively repair disrupted schedules in a decentralized multi-project environment facing various disruptions.