Assessment of Inventory and Transportation Collaboration in a Logistics Marketplace

Abstract

The globalization of businesses and the consequent competition introduce significant pressures on logistics systems. One such key demand on the role of logistics within organizations is the increased emphasis on time-based competition, i.e. the speedy manufacture, delivery to markets and the servicing required, as a result of increasing dependence on the ability of organizations rapidly and efficiently to deliver customer adapted products worldwide. The growing need for transparent, flexible, and easily adjustable logistics services has fostered the creation of digital brokerage platforms that match a variety of logistics demands with supply, widely known as an Electronic Logistics Marketplace (ELM). In this thesis, a comparative study was conducted for three distribution strategies with different levels of inventory collaboration. The Multi-Depot Vehicle Routing Problem (MDVRP) with supply and demand constraints were formulated for each strategy, and the mathematical models thus created were tested using random datasets generated on varied levels of customer dispersion. It was determined that the model representing the strategy of full inventory and distribution collaboration resulted in the least cost in all cases with an average savings of 76.70% as compared to the model with no collaboration. Since General Algebraic Modeling System (GAMS) was limited to small-sized problems for the models, an adaptation of the Variable Neighborhood Search (VNS) metaheuristic was developed and coded using C++ programming language. The heuristic was tested on several datasets, and had a deviation of 2% - 8% from the optimal solutions. The algorithm was further analyzed by testing larger sets, and it returned solutions for 90 nodes within 3200 seconds. Furthermore, a sensitivity analysis was conducted to assess the effect of changing some key input parameters on the total logistics costs. The input parameters were changed one at a time in a one-way sensitivity analysis study, and the effect of parameters’ variation on the total logistics cost was observed. Overall, the most influential inputs are the number of customer nodes, error in forecasted demand, last-mile cost per km, and vehicle capacity.