A study of representations for resource constrained project scheduling problems using fitness distance correlation

Abstract

Four representations for resource constrained project scheduling problems (RCPSPs) are studied by making use of the fitness landscape analysis technique. The fitness distance correlation (FDC) measure is used to analyze the landscapes. In the experiments, the study on the benchmark problems J30 is first presented to investigate which distance metric is more suitable for calculating FDC for RCPSPs. Then, the benchmark problems Patterson, J30, and J60 are used to evaluate the effect of the four encodings on the performance of evolutionary algorithms. Finally, a standard genetic algorithm is applied to verify the predictions made by the FDC. To the best of our knowledge, this is the first work on using FDC to study different encodings for RCPSPs. © 2013 Springer-Verlag.