Models for operational optimisation in a horizontal logistic cooperation. Gain sharing, incentives and multi-level objectives
16 May 2017
University of Antwerp - Stadscampus - Hof van Liere - W. Elsschotzaal - Prinsstraat 13 - 2000 Antwerp (route: UAntwerpen, Stadscampus
Prof. dr. Kenneth Sörensen
PhD defence Christof Defryn - Faculty of Applied Economics
Horizontal cooperation in logistics has received increased attention from the research community, especially over the last ten years. The existing contributions, however, are still scarce and scattered, and no solid frameworks or approaches for analysing and solving logistics optimisation problems in the context of horizontal cooperation exist. The research is mainly focused on proving the potential of horizontal logistics cooperation by means of simulation experiments or the reporting on actual case studies, and the definition and selection of an appropriate allocation rules for dividing the total coalition cost (or profit) among the collaborating partners. This dissertation is the first attempt towards the definition of a general framework for tackling logistics planning problems that involve multiple partners. Whereas existing contributions generally consider traditional (non-collaborative) logistics optimisation techniques to tackle collaborative problems, the first multi-partner optimisation techniques are developed in this thesis. More specifically, we focused on two main aspects: (i) the inclusion of allocation rules and (ii) a multi-objective approach with multiple levels of decision making.
The search for an appropriate allocation method is currently often solely based on fairness criteria. We showed, however, that these allocation rules are able to provide incentives to the individual partners. As these incentives might differ significantly for different methods, we argue that they should also be considered when deciding on an allocation rule. In other words, a good method should ensure that if a partner is willing to do something that is considered beneficial by the coalition, this action is rewarded by the allocation rule. The interaction between cost allocation, partner behaviour and the operational logistics solution is studied empirically, by reporting on an actual case study, and theoretically for the selective vehicle routing problem.
With respect to the second research question, we developed three solution models for tackling collaborative logistics problems with multiple levels of objectives: the coalition efficiency model, the partner efficiency model and an integrated model. These models differ in the level at which the optimisation procedure takes place. The coalition efficiency model was defined as a four-step procedure in which the logistics problem is first solved by taking only the coalition objective(s) into account. From the moment that a solution set is obtained for the coalition, the individual partner objectives are used solely to differentiate among these solutions and select only the subset of solutions that are not dominated according to the individual partner objectives (if such solutions exist). This is in contrast to the partner efficiency model, which gives full priority to the individual partner objectives. The resulting solution set returned by the partner efficiency model therefore guarantees Pareto efficiency with respect to all partner objectives without assuring that the coalition as a whole is performing efficiently. The integrated model combines the coalition efficiency model and the partner efficiency mode and is defined in a general way so it can be used for solving any collaborative vehicle routing problem. It consists of two subproblems, the Coalition Level Optimisation Problem (CLOP) and the Partner level Optimisation Problem (PLOP), which are solved sequentially. Both subproblems are linked through the definition of the acceptable region R(x), in which x represents the optimal solution of the CLOP.
Finally, this thesis also contributes to the research on the clustered vehicle routing problem, a recent vehicle routing variant in which customers are grouped into predefined clusters. The clustered vehicle routing problem can, e.g., be used to model parcel delivery in e-commerce applications, where the distribution area tends to be divided in so-called 'zones'. We propose a two-level, state-of-the-art algorithm that can compete with the existing approaches.