Over the last decades, the transportation sector has put an enormous effort into improving the efficiency of its operations. Algorithms developed in the Operations Research community for operational transportation planning problems, most notably vehicle routing problems, have contributed considerably in reducing the number of kilometers driven unnecessarily. Driven by the more recent trend towards sustainable (often referred to as “green”) supply chain management, the need for more efficient vehicle routing has only intensified.
When different transportation companies join a so-called horizontal logistic co-operation and agree to execute each other’s transportation orders when this benefits the total efficiency of the coalition, additional opportunities for optimization appear. The main motivations for companies to engage in a horizontal co-operation are a lower total logistic cost, improved resource and capacity utilisation, higher degree of sustainability (e.g., lower exhaust of greenhouse gases and other undesirable substances), as well as an increased service level (e.g., more frequent deliveries). By sharing the cost and benefits of the collaboration among all companies involved, a win–win situation is created.
An overlooked aspect of horizontal logistic collaboration is the importance of operational planning, most importantly vehicle routing. Clearly, a coalition has more opportunities for optimisation than an individual partner, but those opportunities need to be seized, which requires advanced planning algorithms. Moreover, operational planning in a horizontal logistics coalition is considerably more complex than stand-alone operational planning. Partly, this is due to the size of the optimisation problem (which is obviously much larger in a horizontal coalition). Partly this is due to the multi-partner character of the problem. After all, an individual partner is only willing to accept a solution if this leads to personal gains. Furthermore, it might be that all partners have different individual objectives they seek to optimise. Finding a good (preferably optimal) solution for the group of collaborating partners is therefore not straightforward.
The problem of construction a solution set that reflects a good compromise between the efficiency of the coalition as a whole and the individual objective(s) of each partner is the subject of the research performed by Christof Defryn and Kenneth Sörensen. As a result, the authors define two models that can be used to optimise the logistic operations within the context of horizontal co-operation.
In the coalition efficiency model, the coalition first defines a set of global coalition objectives, encompassing all objectives of all partners, then finds a solution or a set of non-dominated solutions for these global objectives, and finally divides the objectives (costs) back to the individual partners. The second option is to consider all individual partner objectives and find a set of non-dominated solutions for the (larger) set of partner objectives, without first aggregating them into coalition objectives. We call this approach the partner efficiency model.
Simulation experiments are performed on a Traveling Salesman Problem with soft time windows. The authors found that the coalition efficiency model is able to generate good quality solutions in relatively short calculation times. The method, however, can only be applied if the coalition is able to define a global set of coalition objectives upfront. Furthermore, only a very limited number of solutions is obtained by the coalition efficiency method and these solution tend to be not robust as they differ significantly when the algorithm is executed multiple times. The partner efficiency model, on the other hand, is able to provide the decision maker with a more complete Pareto front approximation, allowing a better understanding of the underlying trade-offs between the different objectives of the individual partners. Because of the increased complexity, due to the large number of objectives in the model, the better Pareto-front approximations come at the expense of very high calculation times.