Slack Time Allocation in Robust Double-Track Train Timetabling Applications

In response to various stochastic disturbances unfolding in the real-time decision environment, train planners need to construct robust train timetables to improve railroad service reliability. A stochastic optimization formulation is proposed to incorporate real-time uncertainty and dispatching policies into medium-term train timetabling decisions that aim to (1) minimize the total trip times in published timetables and (2) reduce the expected schedule delay under random departure times and segment running times. With a sequential solution procedure, this study focuses on solving a series of subproblems for individual trains, while the subproblem of optimizing the slack time allocation on different segments is reformulated as a stochastic time-dependent shortest path problem in a space-time expanded network. A branch-and-bound search algorithm is further developed to systematically evaluate partial solutions and eliminate dominated branches, with an embedded numerical approximation scheme that aims to accurately evaluate the expected direct and knock-on delay of individual trains. Several illustrative examples are used to demonstrate the proposed model and solution algorithms.