Mapping the State-of-the-Art in Vehicle Routing Instance Characterization

The effectiveness of optimization algorithms for the Vehicle Routing Problem (VRP) is intrinsically linked to the underlying structure of problem instances. Characterizing these structures through feature engineering is essential for tasks such as algorithm selection, performance prediction, and instance generation. This paper presents a systematic mapping of the VRP instance characterization literature. Following a protocol based on PRISMA guidelines, we analyzed 50 primary studies published over the last 30 years. Our results reveal a multifaceted research ecosystem, yet one hampered by a significant replicability gap, with only 51\% of studies providing sufficient artifacts for validation. We contribute a unified taxonomy of feature extraction methods and a bibliometric overview that highlights current trends and critical methodological gaps in the field.