Using VRPSolver to efficiently solve the Differential Harvest Problem

Abstract

Precision farming in viticulture raises challenging combinatorial issues such as the Differential Harvest Problem. This problem might appear at first as similar to a simple Capacitated Vehicle Routing Problem but it exhibits problem-specific constraints that make this problem much harder to solve and which are discussed in this article. Our objective was to develop efficient exact methods using column generation and VRPSolver™ based models. In order to investigate possibilities of hybridization, two other models were designed: a constraint programming model and a local search model using LocalSolver™ . In order to reach good scalability, new valid inequalities and a hybrid solving scheme are proposed. Extensive experiments were performed both on simulated and real data sets. The results are discussed from operational cost point of view and also performance.