Strategic freight transportation network models often lack some kinds of refinements. In the case of railway transport flows, models typically onlycompute one or more shortest paths, which may not correspond to the actual lines. The main objective of this study is to propose a methodology thatmakes it possible to take railway lines into account and, by extension, railway services (frequencies) in a multi-modal freight transportation network model. The presentation of transportation assignments is obviously notthe main topic of this study. Assignment methods are looking for a way tomodel the distribution of traffic over a network according to a set of constraints, notably related to transport capacity, time and cost. Traditional assignment models are based on cheapest path algorithms. They use networks in which every road (or railway, waterway, ...) can be represented by a weighted edge. The total cost of a trip is simply the sum of the costs associated to each link along the path. However, trains have to follow a predefined planned route, which can differ from the shortest or cheapest one, possibly not identified by shortest path algorithms. As a consequence, the cost of the computed path can be lower than the cost on the real route.Beside the fact that the flows are not correctly rendered on the network,the total costs for railway transport are often underestimated and, in multi-modal models, the market share for railway transport can be overestimated. The line concept can be defined as an ordered sequence of links and nodes along a path. In this definition, the origin node of each link must coincide with the destination node of the preceding link. Generally, the route followed by a train coincides partially or totally with a set of lines. Note that the line concept can obviously be used in passengers transportmodels, in order to model bus lines for instance. Railway lines are seldom taken into account in large strategic multi-modal network models, because it is often claimed that their implementation only has a limited impact on the results. The real reason is probably that not many, if any, multi-modal models are able to conveniently mix free flows (trucks for instance) and line flows. Lines are however modelled in most tactical or operationalrailway models. The proposed methodology is based on virtual networks, inwhich a virtual link is created with a specific cost for each particular transport operation performed along the transportation chain. This paper presents an improved algorithm that is able to generate virtual networks that explicitly take into account the defined lines. To experiment the method on a real network, a credible reference scenario is proposed, and the impact of the implementation of the new methodology is estimated and commented. For the covering abstract see ITRD E145999
Abstract