grubhub / mdrplib

Meal Delivery Routing Problem Test Instance Library

Meal Delivery Routing Problem Test Instances

This repository contains problem files used to test models for solving the Meal Delivery Routing Problem and a solution evaluator. The sample instances have been provided by Grubhub and Georgia Tech's H. Milton Stewart School of Industrial & Systems Engineering to advance research on the subject.

The public_instances folder contains 240 instances using anonymized data from meal delivery operations at Grubhub. These were derived from a set of 10 seed instances through the following procedure.

• Reducing their size. This defines three subsets of instances:
  • o100: no size reduction
  • o50: 50% reduction in number of orders and courier hours, sampling the order set directly.
  • r50: 50% reduction in number of orders and courier hours, sampling the restaurant set and keeping all orders from selected restaurants.
• Changing the courier schedule. This defines two subsets:
  • s1: schedule based on historical courier shifts.
  • s2: schedule with "optimized shifts" (subject to having about the same, and never more, courier hours than the historical counterpart)
• Increasing the travel speed. This defines two subsets:
  • t100: original travel times
  • t75: travel times are multiplied by 0.75
• Increasing the preparation times of all orders. This defines two subsets:
  • p100: original preparation times
  • p125: preparation times are multiplied by 1.25

Each instance is then labeled by concatenating the labels of the sets to which it belongs. For example: 1o50s2t100p125 is the instance derived from seed 1, by reducing its size by 50% on the order set, using an optimized courier schedule, original travel speed, and longer preparation times; and seed 1 is represented as 1o100s1t100p100.

The folder MDRP_code contains the solution evaluator script. Meal Delivery Routing: The Grubhub Instances provides a complete description of the Meal Delivery Routing Problem and the test instance set.

mdrplib's Issues

Missing data

Hello,

I am willing to reproduce your results for academic purposes, but I have some issue doing so. To the best of my knowledge, the data files for listings 5-7 (2.1 Solution evaluation) are missing as well as means to produce them.

Could you please tell me if I am missing something from the picture?

Best regards,

Natalia

Please add the ability to bypass Google Maps Validation

Look, I don't know any other way to reach an actual developer at grubhub. Your support staff doesn't care at all...

I live in an area where Google Maps isn't fully up to date. It's a new community, new roads, new houses, etc.

I've been here 6 months, but unfortunately Google Maps is not omnipotent and I can't force Google Maps to suddenly get 100% accurate...I understand that you use Google Maps for address checking...

Yet somehow GrubHub is literally the only service that I cannot get at my house because of your address verification system and the fact that it has no way to manually validate, or use GPS.

My road exists on Google Maps, if you put in my address a pin drops 5ft from my house... But because Google Maps doesn't yet have street numbers for my road, your system refuses to validate the address and I can't place any orders...

I can get UPS, FedEx, USPS, Uber, Lyft, etc. Why can't I get GrubHub?

I can't possibly be the only person this affects, think of how many thousands of new communities are built in the US every year?