(648e) Assessing Project Schedule Resilience Against Activity Outcome Uncertainty

The resource constrained project scheduling problems (RCPSP) constitute a ubiquitous class of optimization problems for project scheduling with broad application in many industrial settings. Uncertainty affects the deterministic project schedules in a number of ways, with uncertainty in activity durations and resource availabilities being the most well-studied ones in the existing literature [1]. Motivated by the pharmaceutical R&D environment setting, we study the multi-mode RCPSP under a general form of uncertainty, namely activity outcome uncertainty. There exist discrete uncertain outcomes for specific research activities in the pharmaceutical drug development process, where recourse actions of extending or repeating a particular activity, or even a sequence of its preceding activities as well, might be required to ensure the quality of the project deliverables. Whereas the uncertainty in activity outcomes inherently has the most significant impact on the performance of a given schedule and could lead to major disruptions in planned performance, the majority of the studies do not explicitly address such a source of uncertainty [2].

We acknowledge that the multi-stage nature of the decision-making process for RCPSP with activity outcome uncertainty, where decisions at different time points in the horizon are made in-between realizations of various activity outcomes. However, as is often the case that typical scheduling horizons for RCPSP extend to the order of tens or hundreds of periods, and compounded by the fact that adapting a project schedule constitutes a highly combinatorial complex decision, we also recognize that using a formal stochastic programming or robust optimization methodology to achieve uncertainty-aware schedules with exact guarantees can be quite intractable in all but the smallest of instances. Primarily, there exist two heuristic methodologies dealing with RCPSP under uncertainty in the literature, namely “stochastic scheduling” and “proactive-reactive scheduling” [3]. The objective of the former is to find a policy that minimizes the expected project objective value. The solution approach could be viewed as a completely reactive process where the policy defines the recourse actions to be taken at certain decision moments, with different classes of policies having been studied in the literature [4-6]. On the other hand, the proactive-reactive scheduling approach generates a baseline schedule before applying reactive policies, which overcomes the serious disadvantage of the stochastic scheduling approach of not having one. The goal of proactive-reactive scheduling is to thus construct a solution that is as stable as possible against uncertainty, where both the solution robustness and quality robustness are considered [7].

We propose a novel tree-based search algorithm to simulate the reactive decision-making process in the context of a rolling horizon, which assesses the resilience of a project scheduling solution against various types of activity outcome uncertainties. We will demonstrate the effectiveness of the proposed algorithm by extending benchmark instances from the PSPLIB [8] to encapsulate various forms of uncertainty pertinent to the RCPSP context. In addition, we will apply our methodology on a larger dataset from a real-life industrial case study stemming from the pharmaceutical R&D setting [9, 10]. Finally, reactive procedures will be presented that can increase the quality robustness of the given baseline schedule facing activity outcome uncertainty.

References:

[1] Davari, Morteza. "Contributions to complex project and machine scheduling problems." (2017).

[2] Hazir, Oncu, and Gunduz Ulusoy. "A classification and review of approaches and methods for modeling uncertainty in projects." International Journal of Production Economics (2019): 107522.

[3] Davari, Morteza, and Erik Demeulemeester. "The proactive and reactive resource-constrained project scheduling problem." Journal of Scheduling 22.2 (2019): 211-237.

[4] Stork, Frederik. "Stochastic resource-constrained project scheduling." (2001).

[5] Ashtiani, Behzad, Roel Leus, and Mir-Bahador Aryanezhad. "New competitive results for the stochastic resource-constrained project scheduling problem: exploring the benefits of pre-processing." Journal of Scheduling 14.2 (2011): 157-171.

[6] Li, Haitao, and Norman K. Womer. "Solving stochastic resource-constrained project scheduling problems by closed-loop approximate dynamic programming." European Journal of Operational Research 246.1 (2015): 20-33.

[7] Herroelen, Willy. "Project scheduling—Theory and practice." Production and operations management 14.4 (2005): 413-432.

[8] Kolisch, Rainer, and Arno Sprecher. "PSPLIB-a project scheduling problem library: OR software-ORSEP operations research software exchange program." European journal of operational research 96.1 (1997): 205-216.

[9] Viswanath, Shekhar, Steve Guntz, Jon Dieringer, Shankarraman Vaidyaraman, Hua Wang, and Chrysanthos E. Gounaris. "An ontology to describe small molecule pharmaceutical product development and methodology for optimal activity scheduling. " Under Review (2020).

[10] Wang, Hua, Jon Dieringer, Steve Guntz, Shankarraman Vaidyaraman, Shekhar Viswanath, Nikolaos H. Lappas, Sal Garcia-Munoz, and Chrysanthos E. Gounaris. "Portfolio-wide Optimization of Pharmaceutical R&D Activities Using Mathematical Programming. " Under Review (2020).