Research

Google Open-Sources MilleMiglia Logistics Generator

Google Research has open-sourced MilleMiglia, a C++ instance generator that creates realistic, privacy-preserving benchmarks to help researchers optimize complex middle-mile logistics networks.

Google Research1 day agoResearch
Image: Google Research

Google Research, in collaboration with academic partners at UniBrescia and ENPC Paris, has introduced MilleMiglia, an open-source C++ instance generator designed to simulate middle-mile logistics networks. While first-mile and last-mile delivery logistics have long been studied using vehicle routing problem models, the middle-mile segment—which handles bulk transport between regional distribution centers—has lacked standardized, high-quality public data due to corporate privacy concerns. MilleMiglia addresses this gap by generating realistic, privacy-preserving benchmarks.

The tool is written in C++ and utilizes Protocol Buffers for compact data serialization, allowing diverse constraints to be stored in a single file. MilleMiglia uses spatial clustering and gravity models to position distribution centers realistically, while generating shipments with realistic volume and weight distributions. Unlike traditional vehicle routing problems solved by tools like Google's OR-Tools or the Google Maps Platform Route Optimization API, middle-mile logistics involve multi-commodity flow problems on space-time graphs. The generator incorporates real-world constraints such as fixed vehicle schedules, distribution center throughput limits, and complex synchronization requirements.

For practitioners and researchers, MilleMiglia provides a range of benchmark instances, from small toy problems for testing exact algorithms to massive, continent-wide industrial datasets. It also supports machine learning applications by generating massive datasets to train routing algorithms. Google and its academic partners, including researchers Aymane Lotfi, Matteo Petris, Thibaut Cuvelier, and Claudia Archetti, are also developing a specialized solver and API tailored to these middle-mile operational challenges. By establishing a standardized benchmarking suite similar to what the Capacitated Vehicle Routing Problem Library does for traditional routing, the creators hope to spark new academic and industrial breakthroughs in supply chain efficiency.

This is our own summary of reporting by Google Research

More in Research