A multi-objective approach to the resource constrained project scheduling

Abstract

Many examples where the resource constrained project scheduling framework is applicable for modelling and solving the scheduling problem under consideration can be found both in the literature and in practice. Recently, more realistic and accurate resource models and objectives have been developed. These models provide better means for coping with practical issues faced in real world applications. In the resource constrained project scheduling, the heuristics that are existent in the literature aim at either minimizing the tardiness or maximizing the Net Present Value (NPV). In a project network, the constraints of the resource constrained project scheduling problem include precedence constraints representing the technological network and resource limitations. The resource constraints used in resource constrained project scheduling models have recently become more descriptive of real life situations. It is assumed that there is a project due date that is contractual and causes a penalty for each tardy period. NPV is the exponential or discrete discounting of the cash flows accrued at the event occurrence times to the present. Tardiness in a project stands for the total number of late periods over the project due date. In real life, it is essential to make money while keeping the image of the company at its highest level on customers' eyes. Therefore, in this study new heuristics are generated and their performance according to NPV and tardiness criteria is statistically compared with that of the heuristics fromthe literature. The trade-off between NPV and tardiness is easily illustrated. The new heuristics are named as hybrid heuristics, because their priorities are based on the weighted combination of the two objectives, namely minimizing tardiness and maximizing NPV. The scheduling algorithm used in this study is iterative, making consecutive forwardlbackward scheduling passes. The advantages of using the dynamic activity time windows in pushing and pulling event times are shown.