Managing Uncertainties and Disruptions in Resource Constrained Project Scheduling

Download files
Access & Terms of Use
open access
Copyright: Chakrabortty, Ripon
Altmetric
Abstract
The resource constrained project scheduling problem (RCPSP) has been an important research topic over the past few decades, mainly because of its applicability to many real-world situations. Over the last few decades, research on resource constrained project scheduling has focused on the development of mathematical programming based approaches for the generation of a nominal schedule under a deterministic environment. However, in practical situations, the parameters of projects are vulnerable to uncertainty, change or disruption, which necessitates that the initial nominal schedule must be revised. Hence, it can be said that in spite of substantial research on traditional RCPSPs, the improvement of existing approaches and the development of new techniques are still needed to cope with realistic situations, such as multiple execution modes, stochastic activity durations and unreliable resource breakdowns. This thesis focuses on the study of disruption and uncertainty management in RCPSPs. Four different types of disturbances or uncertainties are explored for single mode RCPSPs, namely (1) unreliable resource availabilities or breakdowns, (2) imprecise activity duration uncertainties, (3) stochastic activity durations, and (4) time-varying resource requirements and availabilities with stochastic durations. Except time-varying requirements and imprecise duration uncertainties, both unreliable resource and stochastic activity durations are also considered for scheduling multi-mode RCPSPs. While considering resource breakdowns, this thesis focuses on generating and analysing different reactive decisions after the occurrence of a disruption on a real-time basis. On the contrary, for activity duration uncertainties, this thesis attempts to develop different proactive or uncertainty tolerant schedules by executing different newly proposed heuristics. For solving uncertainty or disruption based RCPSPs, this thesis proposed several heuristic and traditional optimization algorithms. The heuristics proposed in this thesis are: (1) variable neighbourhood search based heuristic for generating robust schedules due to unreliable resources, (2) six different robust optimization based heuristics for handling imprecise uncertainties in activity durations, (3) greedy randomized local search heuristic for generating uncertainty tolerant schedules from stochastic activity durations, (4) variable neighbourhood search based local search heuristic for generating uncertainty tolerant schedules from time-varying resource parameters and stochastic activity durations, and (5) evolutionary local search based heuristic for generating uncertainty tolerant schedules from stochastic activity durations for multi-mode RCPSPs . In real-life situations, multiple disturbances, as a series, can happen at any time and at any stage of the system. While proposing reactive schedules, this thesis considers multiple disturbances, one after another in a series, that may or may not affect the plans revised after previous disturbances. If a new disturbance occurs during the recovery time window of another, a new revised plan which considers the effects of both disturbances must be derived. Accordingly, as this is a continuous process, some extended mathematical models and solution approaches are developed for both single and multi-mode RCPSPs, to deal with a series of disturbances on a real-time basis. To judge the performance of all proposed heuristics and exact algorithms, extensive simulation-based analyses have been conducted for different scenario-based schedules. For most cases, for validating the proposed approaches, benchmark instances from the Project Scheduling Library (PSPLIB) are considered. The superiority of proposed heuristic algorithms over others has also been demonstrated though an exhaustive comparison with many state-of-the-art heuristic and meta-heuristic approaches. Sensitivity analysis is carried out for different model parameters and disruption scenarios. For justifying the heuristic for time-varying resource requirements and stochastic activity durations, this thesis employed a standard set of 5760 and 7200 self-generated test instances with 30 and 120 activities respectively. From those exhaustive experimental studies, it has been shown that the developed approaches are statistically more beneficial in comparison to other existing approaches.
Persistent link to this record
Link to Publisher Version
Link to Open Access Version
Additional Link
Author(s)
Chakrabortty, Ripon
Supervisor(s)
Sarker, Ruhul
Essam, Daryl
Creator(s)
Editor(s)
Translator(s)
Curator(s)
Designer(s)
Arranger(s)
Composer(s)
Recordist(s)
Conference Proceedings Editor(s)
Other Contributor(s)
Corporate/Industry Contributor(s)
Publication Year
2017
Resource Type
Thesis
Degree Type
PhD Doctorate
UNSW Faculty
Files
download public version.pdf 3.14 MB Adobe Portable Document Format
Related dataset(s)