1 Introduction
Supplying the 38 factories of Renault with raw materials represents a major economic challenge. The contractualized transportation plan with third-party logistics providers must bedefined by respecting operational or network constraints. The first step in defining the inboundtransportation plan is to decide for each flow whether it will be direct or indirect, knowing itsforecast demands. In the literature, this decision problem is called service network design [1]or network planning in the literature of cross-docking [2]. The following decisions consist infinding the routes when the determined volumes do not complete a full vehicle. These routesare composed of multiple loading points (suppliers) and one or more unloading points (factorydocks). This problem is called the vehicle routing problem (VRP), extensively studied in theliterature [3].
2 Problem description
The Renault inbound supply chain consists of supplying its factories with raw materials byseveral hundred of geographically dispersed suppliers. The high number of suppliers increasesthe complexity and cost of operating the inbound supply chain. Thus, a transportation planmust be tailored to the inbound logistic where supplies meet demands at the factories, whileseeking to minimize the total logistic costs. The transportation plan is defined based on theforecast demand of factories, the geographic location of suppliers and factories and Renault'soperational constraints. It includes tactical and operational decisions. The tactical decisionsdefine the transportation network design for several upcoming weeks, i.e. the flows of supplies.There are two types of flows, (i) Direct and (ii) Indirect. The direct flow (Milk-run and mono-suppliers) is a route with one or more loading point to one or more unloading points. Theindirect flow is a pick-up from several loading points, to a cross-dock, then from the cross-dock to one or more unloading points. Since the cross-dock is responsible in designing theroutes for the indirect flows, we are only interested in direct flows. The operational decisionsdetermine the routes for every supplier in direct flows, and their respective frequency, loadingand unloading points, and arrival day at the factory. We call it Fiches Caractéristique Circuit(FCC). The process of defining the transportation plan ends with the contractualization ofthe FCC with transporters. This work aims to develop an optimization tool to decrease thetransportation costs and to computerize the construction of the inbound transportation plan.
†Institute of Engineering Univ. Grenoble Alpes
3 Solution approach
The current process of defining Renault's upstream transportation plan is not optimal. Itis done manually through the expertise of Renault Transport Planners (RTP). Hence, we propose a three-phase algorithm, called Clustering-First-Routing-Second-Scheduling-Third. In theclustering phase, we propose three graph-based algorithms, the minimum clique cover problem(MCCP), the minimum dominating set problem (MDSP), and the minimum weighted spanning tree (MWST) to partition suppliers into clusters. Based on these algorithms, differentheuristics were developed to construct the clusters respecting two main constraints: the maximum distance inter-suppliers, and the maximum number of suppliers per cluster. In the routingphase, a mixed integer linear programming model (MIP) was developed to identify the type offlows. The objective function minimizes the total transportation cost that includes the variabletransportation cost, the fixed cost per vehicle and stop, the cross-dock cost, and the directcost for mono-suppliers, while respecting Renault's operational constraints. In the schedulingphase, a MIP was developed to assign routes to days of the week to minimize the maximumnumber of daily routes, while satisfying the constraint of smoothing routes over the week.
4 Computational experiments
We tested the efficiency of the proposed algorithm on several real-world use cases definedby Renault key users. The number of suppliers per use case varies between tens to hundredsof suppliers. At first, the computational experiments consisted in finding the best clusteringalgorithm and, then, in comparing the existing approach with our solution algorithm. Forthis, a process was defined to verify the cost of the tactical planning through the operationalplanning. The tactical planning defines the inbound transportation plan for a defined planninghorizon. The operational planning algorithms maximize the loading of the defined vehicles inthe tactical planning by grouping together to be consumed demands and bringing them forwardto upfront vehicles, by respecting the allowable days that a demand can be delivered before itsconsumption day at the factory. The two solutions are compared through various KPIs.
5 Conclusion & Future work
A three-phase sequential algorithm was developed for the construction of Renault inboundtransportation plan. Preliminary results show an improvement in the total transportation costs.For future work, an integrated approach can be developed and compared with the results ofthree-phase algorithm. Environmental aspects and inventory optimization can be considereddirectly while defining the transportation plan. To make the operational planning algorithmswork better, the tactical planning should anticipate those algorithms, and especially to integratethe concept of DLO (Optimized Delivery Request) at the factory that allows a demand to bedelivered before its consumption date. Finally, the computational experiments can be extendedto other Renault factories to ensure the robustness of the proposed approaches.
Références
[1] G. Guastaroba, M. G. Speranza, and D. Vigo. Intermediate Facilities in Freight Transportation Planning : A Survey. Transportation Science 2016 50 :3, 763-789
[2] Buijs, Paul and Vis, Iris and Carlo, Héctor. Synchronization in Cross-Docking Networks : A Research Classification and Framework. European Journal of Operational Research 239(3) :593–608
[3] Thibaut Vidal and Gilbert Laporte and Piotr Matl. A concise guide to existing and emerging vehicle routing problem variants. European Journal of Operational Research, Volume 286, Issue 2, 2020,