Reflective Nested Simulations Supporting Optimizations within Sequential Railway Traffic Simulators

Abstract

This article describes and discusses railway-traffic simulators that use reflective nested simulations. Such simulations support optimizations (decision-making) with a focus on the selection of the most suitable solution where selected types of traffic problems are present. This approach allows suspension of the ongoing main simulation at a given moment and, by using supportive nested simulations (working with an appropriate lookahead), assessment of the different acceptable solution variants for the problem encountered—that is, a what-if analysis is carried out. The variant that provides the best predicted operational results (based on a specific criterion) is then selected for continuing the suspended main simulation. The proposed procedures are associated, in particular, with the use of sequential simulators specifically developed for railway traffic simulations. Special attention is paid to parallel computations of replications both of the main simulation and of supportive nested simulations. The concept proposed, applicable to railway traffic modelling, has the following advantages. First, the solution variants for the existing traffic situation are analyzed with respect to the feasibility of direct monitoring and evaluation of the natural traffic indicators or the appropriate (multi-criterial) function. The indicator values compare the results obtained from the variants being tested. Second, the supporting nested simulations, which potentially use additional hierarchic nesting, can also include future occurrences of random effects (such as train delay), thereby enabling us to realistically assess future traffic in stochastic conditions. The guidelines presented (for exploiting nested simulations within application projects with time constraints) are illustrated on a simulation case study focusing on traffic assessment related to the track infrastructure of a passenger railway station. Nested simulations support decisions linked with dynamic assignments of platform tracks to delayed trains. The use of reflective nested simulations is appropriate particularly in situations in which a reasonable number of admissible variants are to be analyzed within decision-making problem solution. This method is applicable especially to the support of medium-term (tactical) and long-term (strategic) planning. Because of rather high computational and time demands, nested simulations are not recommended for solving short-term (operative) planning/control problems.