Top repository - CVRPLIB

The best repository is CVRPLIB. This repository contains Capacitated Vehicle Routing Problem instances in an easy accessible way.

Taxonomy of test-instances is explained in Eduardo Uchoa et al. (2017). Below is a summary of the notation used for the instances.

Number of customers \(n\)
Number of trucks \(k\)

Depot positioning

Central (C) depot in the center of the grid, point (50 0,50 0).
Eccentric (E) –depot in the corner of the grid, point (0,0).
Random (R) –depot in a random point of the grid.

Customer positioning

Random (R) –all customers are positioned in random points of the grid.
Clustered (C) – a number S of customers that will act as cluster seeds is picked from an uniform discrete distribution UD[3,8]. Next, the S seeds are randomly positioned in the grid. The seeds will then attract, with an exponential decay
Random-clustered (RC) - half of the customers are clustered by the above described scheme, the remaining customers are randomly positioned.

Demand distributions

Unitary (U) –all demands have value 1.
Small values, large CV (1–10) – demands from UD[1,10].
Small values, small CV (5–10) – demands from UD[5,10].
Large values, large CV (1–100) –demands from UD[1,100].
Large values, small CV (50–100) –demands from UD[50,100].
Depending on quadrant (Q)
Many small values, few large values (SL)

Other repositories

Smartana

This contains several new instances and mirrors test-instances from other repositories. Go to download folder

A collection of test-instances prepared for the multiple Travelling Salesmen Problem (mTSP). These are available as Matlab files. They were used to gain insights about the Balanced-Dynamic (BD) mTSP. Go to download folder

Some of the above mentioned test-instances were used for:

Garn, W., 2020. Closed form distance formula for the balanced multiple travelling salesmen. arXiv preprint arXiv:2001.07749.
Garn, W., 2020. Balanced dynamic multiple travelling salesmen: algorithms and continuous approximations. arXiv preprint arXiv:2008.12063.