Decision Making in Project Management: Multi Objective Extended Resource Constrained Project Scheduling


Researcher: Dr Elena Rokou
Year: 2010-2013
Affiliation: National Technical University of Athens
Primary supervisor: Assistant Professor Konstantinos Kirytopoulos

Thesis language: English

A holistic approach is proposed for defining the resource constrained project scheduling problem (RCPSP). The Thesis aim is to give a formulation of the project scheduling problem where all deterministic aspects that have been previously explored in the relevant literature are covered.

The goal is to provide a way to model and solve project scheduling problems as they actually are, without compromises other than the assumption that the given inputs are realistic. An appropriate mathematical formulation along with a concise solution process, covering both the single and multi-objective case, are presented. Based on this model an adaptive evolutionary algorithm is implemented to solve the unified version of the problem along with and Add in for MS Project to provide an easy to use interface to the project managers. The efficiency of the proposed approach is compared to existing implementations through a number of experiments. The experiments are grouped in two classes: in the first one the best known results from each variation and extension of the single objective RCPSP are compared to the results given by the proposed algorithm and in the second one the multi-objective approach is compared to the single-objective results given in the same test cases appropriately adapted. Finally, the application of the proposed approach in real situations is illustrated through a case study on a medium sized project (200 activities) taken from the GIS domain.

The results show that the usage of the holistic model doesn't affect the quality of results or the needed CPU-time when compared to the existing RCPSP formulations, whereas it adds the ability to describe more realistically any complex project scheduling problem. We overcome the raise of complexity and the infeasibilities by using penalty functions when relaxation of the constraints is needed. In the multi-objective case the algorithm is capable of providing multiple solution scenarios that are generated either based on the simple Pareto front or on a weighted approximation of it.



LinkedIn logopyrforosgood