instances/README.md

TABLE OF CONTENTS

  1. Benchmark instances (how they look like, details on the format, ...)
  2. Other files

BENCHMARK INSTANCES

GENERATION

All the instances have been generated with the netgen R package (see https://github.com/jakobbossek/netgen).

FILE NAMING

The file naming underlies the following pattern: vrp_{morphed}_n{#customers}[_cl{#clusters}][_mo{morphing.coefficient}][_dyn{percent.of.dynamic.customers}][_r{repetition}].csv

All the stuff wrapped with [ and ] is optional. Examples for filenames are definitely helpful.

Examples

Note: Moreover we append another index right before the file extension to differentiate between different Poisson-process realizations, e.g., vrp_n50_dyn10_r6.csv

CSV-FORMAT

We differentiate between the 'minimal' and the 'full' format. The latter is only interesting for instance reconstruction in the netgen R package.

Minimal format

The instances are stored as comma-separated files. Each row in the file contains the customer ID, the x- and y-coordinate in euclidean space and the request time of the customer in exactly this order. Keep in mind, that an arrival time of 0.00 is assigned to 'static customers' as well as 'depots'. By convention the first line is stated as the start depot in the context of the VRP and the second line is the final depot. The file contains no header information.

Full format

The full is similar to the minimal format. The main differences are: the full format contains header information and the full format contains additional columns like (cluster-) membership, types, lower and upper bound of the bounding box.

Publication format

Instances are stored in TSPlib file format.



jakobbossek/dynvrp documentation built on Jan. 19, 2020, 9:53 p.m.