A Reinforcement Learning Framework for Maximizing the Net Present Value of Stochastic Multi-work Packages Project Scheduling Problem

Yaning Zhang, Xiao Li (), Yue Teng, Qiping Shen and Sijun Bai
Additional contact information
Yaning Zhang: The Hong Kong Polytechnic University
Xiao Li: The University of Hong Kong
Yue Teng: The Hong Kong Polytechnic University
Qiping Shen: The Hong Kong Polytechnic University
Sijun Bai: Northwestern Polytechnical University

Chapter Chapter 51 in Proceedings of the 28th International Symposium on Advancement of Construction Management and Real Estate, 2024, pp 733-756 from Springer

Abstract: Abstract Project scheduling to maximize net present value (NPV) poses a significant challenge due to the inherent complexities associated with large-scale projects comprising multiple work packages and uncertain task durations. Existing scheduling methods encounter difficulties in effectively maximizing NPV when confronted with multi-work package projects characterized by stochastic task duration distributions. In light of this problem, this paper proposes a three-level reinforcement learning (TRL) framework aimed at addressing these challenges. To determine resource allocation for each work package within the project, the TRL framework leverages human empirical decision-making at the resource assignment level. At the work package level, a Priority Experience Replay Dueling Double Deep Q-Network (PER-DDDQN) is trained. This PER-DDDQN incorporates a graph embedding method, enabling it to maximize the expected NPV for each work package. The graph embedding method facilitates the determination of the work package's scheduling state, while the PER-DDDQN governs the scheduling of task start times within the work package. Furthermore, at the project level, work packages are scheduled using the same principles employed at the work package level to maximize the expected NPV of the entire project. Numerical experiments conducted on adapted case projects provide evidence that the TRL framework surpasses existing heuristics in achieving higher NPV for most work packages. Moreover, the TRL framework yields a minimum improvement of 26.68% in the maximum expected NPV of the entire project compared to the heuristic method employed in this study. This research contributes significantly to the enhancement of cash flow management in large-scale projects characterized by multiple work packages. Additionally, it opens up possibilities for the integration of reinforcement learning technology within the field of construction project management.

Keywords: Reinforcement learning; Net present value; Work package; Project scheduling (search for similar items in EconPapers)
Date: 2024
References: Add references at CitEc
Citations:

There are no downloads for this item, see the EconPapers FAQ for hints about obtaining it.

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:spr:lnopch:978-981-97-1949-5_51

Ordering information: This item can be ordered from
http://www.springer.com/9789819719495

DOI: 10.1007/978-981-97-1949-5_51

Access Statistics for this chapter

More chapters in Lecture Notes in Operations Research from Springer
Bibliographic data for series maintained by Sonal Shukla () and Springer Nature Abstracting and Indexing ().