It is well-known that experience can lead to increased efficiency, yet this is largely unaccounted for in project scheduling. We consider project scheduling problems where the duration of activities can be reduced when scheduled after certain other activities that allow for learning relevant skills. Since per-period availabilities of renewable resources are limited and precedence requirements have to be respected, the resulting optimization problems generalize the resource-constrained project scheduling problem. We introduce four constraint programming formulations that incorporate the alternative learning-based job durations via logical constraints, dynamic interval lengths, multiple job modes, and a bi-objective reformulation, respectively. To provide tight optimality gaps for larger problem instances, we further develop five lower bounding techniques based on model relaxations. We also devise a destructive lower bounding method. We perform an extensive computational study across thousands of instances based on the PSPlib to quantify the impact of project size, potential learning occurrences, and learning effects on the optimal project duration. In addition, we also compare formulation strength and quality of the obtained lower bounds using a state-of-the-art constraint programming solver.
View A Computational Study of Constraint Programming Approaches for Resource-Constrained Project Scheduling with Autonomous Learning Effects