Integration de la planification des operations et de la gestion du revenu pour le transport ferroviaire de marchandises

This thesis studies the integration of decision processes based on revenue management with operations planning policies of a rail freight carrier. Services companies in this field must face complex logistic operations such as: (1) block planning and car routing, (2) train routing, makeup and scheduling, (3) yard management, (4) locomotive assignment, (5) empty car repositioning, (6) crew scheduling. Even if freight rail transportation problems have been thoroughly analyzed in the past, they typically consider only a subset of the operations mentioned. The size and the complexity of combining the identified problems have prevented the development of integrated methodological approaches. Moreover, few researchers have tackled the challenge of proposing revenue management tools for freight rail applications despite their major impact in air transportation for instance.Once the research perspectives are identified, we present some formulations describing the proposed integration according to the chosen planning level, which could be either operational or tactical. These help to highlight the combined management decisions related to the development of the operating plan and those arising from tariff policies dictated by the company. Through bilevel programming we study user behavior in response to different pricing strategies while considering network capacity resulting from the operational policies of the carrier.We then suggest an analysis of the main characteristics and properties of the model combining pricing and capacity management at the operational level. To do this, two pricing policies are discussed. We also present valid inequalities that strengthen significantly the formulation. The identified properties and inequalities are subsequently exploited for solving the model with the Cplex optimization software. The validation of the developed methodology is carried on a class of instances generated in order to realistically portray the context studied.Finally, we discuss the development of heuristic approaches for the problem by proposing two families of algorithms. The first takes advantage of the relationship between the two pricing policies stated and is based on the relaxation of constraints binding rates between them. These are then reintroduced gradually. The procedure seeks, by this principle, the identification of the structure of an optimal solution to the original problem. Finally, we present a primal-dual heuristic which solves, in an alternating fashion, sub-problems which respectively consider the primal or dual constraints of the second level problem. The heuristics are compared among themselves and with the results obtained by the exact approach.First of all, we describe the current state of research for the two fields presented in order to clearly establish the problem that will be addressed. We therefore review the main scientific contributions and introduce the fundamental concepts that will enable us to position ourselves in relation to recent publications.