This paper proposes a new continuous-time linear mixed model for rough cut capacity planning (RCCP) adapted to tackle various tactical project planning scenarios. RCCP models are typically designed for the early phases of projects to decide the work package's execution intensities for each project planning period. The goal is to solve large-scale, complex instances while ensuring both optimality and efficient resolution times. We propose modifications to the constraints of one of the best models in this field to improve its resolution. Recognizing that overlap between two consecutive periods adversely affects the performance of the MIP solver, we introduce minimal, fictitious boundaries between these periods. Additional strategies, derived from analyzing initial constraints and solver behaviour, ensure solution optimality. The objectives considered are: minimising the costs of using external resources by setting the project's end date (Time-driven) and minimising the makespan (Resource-driven). The bi-objective case is also addressed. The improved model reduces the average number of dual simplex iterations of CPLEX by 72%. Moreover, only 0.7% of large instances remained unresolved with the new model compared to nearly 30% of instances with the previous best model. The fast resolution of a single-objective problem opens up opportunities for multi-criteria approaches, making the model adaptable to more complex needs in practice.
Read full abstract