A decomposition approach to solve the individual railway crew Re-planning problem

Abstract

Crew re-planning is an important and difficult task in railway crew management. In this paper, we establish a path-based model solving the Individual Crew Re-planning Problem (ICRP). The individual indicates that we focus the problem on specific (non-anonymous) crew members, considering their roles (leader and cabin crew) and qualifications. This problem is inspired by the crew planning problem faced in Chinese high-speed railway operations. To generate feasible paths, we construct a multi-layer time-space connection network and develop a heuristic algorithm. To decrease the complexity and scale of the model, we decompose the ICRP into two sub-problems (for leaders and for cabin crew members respectively) which can be solved in sequence. In addition, we develop a Lagrangian relaxation (LR) algorithm to get valid paths quickly for both sub-problems. We combine the LR algorithm with solving the restricted decomposed models to get a good quality solution for the studied ICRP problem. We test our methods on several real-world instances from Chinese high-speed railways. The computational experiments show that our LR algorithm with a decomposition strategy can solve the decomposed models in a relatively short computation time compared to solving the original model directly, while obtaining (near-)optimal solutions for all instances.