A recommendation-based routing approach for solving uncertain vehicle routing problems
Abstract
Delivering fuel to agricultural machinery in fields is important to reduce various costs for fuel replenishment, leading to the efficiency improvement of agricultural operations. However, our preliminary investigation in China has shown that due to inefficient offline promotion, there are not enough fuel delivery orders, causing a profitability problem. To address the problem, a customer recommendation strategy was proposed, which recommends some nearby candidate customers to a fuel delivery trip launched by existing customers. The existing customers and candidate customers are agricultural machines that have or have not placed orders, respectively. To solve the uncertain routing problem defined by the strategy, a recommendation-based approach was designed. Firstly, to distinguish candidate customers, a clustering-based customer recommendation algorithm was proposed, which leverages the real-time location information of agricultural machines to cluster, rank, and categorize candidate customers. Then, to obtain an optimal route, a two-stage route planning algorithm was proposed, which generates an initial route in the route initialization stage and optimizes it in the route optimization stage. In either stage, clusters of candidate customers are selected and utilized in different ways so as to generate an optimal route in a feasible computation time. Finally, a simulation dataset was constructed based on wheat harvesters operated on a harvesting peak day in 2023, and the experiments on the dataset illustrated the applicability of the proposed strategy as well as the effectiveness of the designed approach, e.g., a significant increase in delivery profit and a dramatic decrease in travel distance.