Resource-constrained project scheduling with the objective of minimizing project duration (RCPSP) is one of the most studied scheduling problems. In this paper we consider the RCPSP with general temporal constraints and calendar constraints. Calendar constraints make some resources unavailable on certain days in the scheduling period and force activity execution to be delayed while resources are unavailable. They arise in practice from, e.g., unavailabilities of staff during public holidays and weekends. The resulting problems are challenging optimization problems. We develop not only six different constraint programming (CP) models to tackle the problem, but also a specialized propagator for the cumulative resource constraints taking the calendar constraints into account. This propagator includes the ability to explain its inferences so it can be used in a lazy clause generation solver. We compare these models, and different search strategies on a challenging set of benchmarks using the lazy clause generation solver chuffed and IBM CPLEX CP Optimizer, respectively. We close all but 8 of the open problems of the benchmark set, extend the benchmark set by instances with up to 500 activities, and show that CP solutions are highly competitive with existing Mip models of the problem.

中文翻译：

---

使用约束编程求解RCPSP / max-cal

以最小化项目工期（RCPSP）为目标的资源受限的项目调度是研究最多的调度问题之一。在本文中，我们考虑具有一般时间约束和日历约束的RCPSP。日历约束使某些资源在计划期间的某些天不可用，并在资源不可用时强制延迟活动执行。实际上，它们是由于，例如，公众假期和周末工作人员的空缺。所产生的问题是对优化问题的挑战。我们不仅开发了六个不同的约束编程（CP）模型来解决该问题，同时还要考虑到日历约束，专门针对累积资源约束进行传播。该传播器具有解释其推论的能力，因此可以在惰性子句生成求解器中使用。我们分别使用惰性条款生成求解器chuffed和IBM CPLEX CP Optimizer在具有挑战性的一组基准上比较了这些模型和不同的搜索策略。我们关闭了除基准测试集中所有未解决的问题之外的所有8个问题，将基准测试集扩展到最多包含500个活动的实例，并显示了CP解决方案与该问题的现有Mip模型具有高度竞争力。