Reorganization of Logistics Network

Abstract

In Chap. 8, the stochastic programming model for the logistics network design and the efficient solution method are shown. The traditional expected cost model and the Conditional Value at Risk (CVaR) model are compared in numerical experiments. Selection of facilities to be used in logistics network design is a very important decision item with regard to the supply chain. Facilities maintenance, sale, and relocation must be taken into consideration in a logistics network reorganization problem concerning the use of existing facilities and new facilities. Reorganization of the network requires considerable capital investment. It is necessary to make the optimal decision from a long-term view. However, as customer demand is not predictable, it is difficult to optimize the network. Decision-making on relocation of a facility takes place at a strategic level. This decision drives other operational decisions according to the standards, and thus, careful discussions are required. According to Ballou (Inf Syst Front 3:417–426, 2001), restructuring of a logistics network has the effect of reducing logistics costs by 5–15%. In this chapter, we consider a logistics network model considering demand uncertainty, and we propose a mathematical planning model using the stochastic programming method. Variables related to warehouse maintenance, integration, opening are called first stage variables, and variables related to transport determined under demand scenario are referred to as second stage variables. We propose a Conditional Value-at-Risk (CVaR) minimization model developed by Rockafellar and Uryasev (J Risk 2:21–41, 2000) employing the CVaR risk measure, and we apply the solution using the L-shaped method. First, divide the original problem into a first stage problem and a second stage problem. We solve the second stage problem using the solution obtained by solving the first stage problem, and approximates the objective function when it is feasible. Then, we compare the results with cases using the conventional expected cost minimization model. In numerical experiments, it is expected that the algorithm of this research is more effective than the conventional branch and bound method when the number of scenarios increases. The CVaR minimization model shows that the expected cost rises slightly more than the expected cost minimized model, while the worst case scenario can reduce the cost.