A fantastic outcome of the Amazon Last Mile Routing Challenge is the public release of a whopping 9,184 practical instances of the asymmetric traveling salesman problem (ATSP). This includes 6,112 training instances, together with 3,072 instances that were used to evaluate codes in the competition. For each instance, the data include point-to-point travel times, latitude and longitude of each stop, stop service times and delivery time windows, size and of number of packages to be delivered at each stop, sort zone IDs, and the actual sequence of stops that was followed by the van driver.
An overview of the data set can be found on the 2021 Amazon Last Mile Routing Research Challenge Dataset page of the Registry of Open Data on AWS. A detailed description of the data items and format is given on the Data Structures page of the competition GitHub site.
To download the data you will need to install the (freely available) AWS Command Line Interface. Daniel Merchán provided the following example, showing the command to download the data in a Mac terminal after the aws interface is installed.
aws s3 sync --no-sign-request s3://amazon-last-mile-challenges/almrrc2021/ ./almrrc2021/
The data are available under the Creative Commons Attribution-NonCommercial License.
We thank Daniel Merchán (Senior Research Scientist at Amazon) and Julian Pachon (Director WW Last Mile Science at Amazon) for their work in releasing the Amazon data. This is a major contribution to the research community.