A benchmark suite for liner shipping network design problems and a parser for the instances in c++
========
This is the benchmark suite for liner shipping network design described in
"Brouer, Berit D.; Alvarez, J. Fernando; Plum, Christian Edinger Munk; Pisinger, David; Sigurd, Mikkel M. A base integer programming model and benchmark suite for liner shipping network design. forthcoming in Transportation Science 2013"
and the technical report:
"Løfstedt, Berit; Alvarez, Jose Fernando; Plum, Christian Edinger Munk; Pisinger, David; Sigurd, Mikkel M. An integer programming model and benchmark suite for liner shipping network design. - Kgs. Lyngby : DTU Management, 2010 (p. 42) In: DTU Manangement Engineering report ; 19.2010"
The benchmark suite originates from publicly available data and past data for Maersk Line. External data sources are: “Distances between ports - Pub. 151”, National Imagery and Mapping Agency (www.marcon.com/library/resources/DBP/DBP.html). The Hamburg Index, Vereinigung Hamburger Schiffsmakler und Schiffsagenten, (www.vhss.de). “Charter Rates 2000-2010”, Alphaliner charter rates 2000-2010, (www.alphaliner.com). “Container Freight Rate Insight”, Drewry Shipping Consultants, (www.drewry.co.uk). Hapag-Lloyd Freight rate quotes (www.hapaglloyd.com/en/schedules/interactive.html). The data was last viewed in January 2011.
There are 7 instances for which each Demand and Fleet file are specific. The first line in all files are the headings.
Ports and distances are for all instances and contains all ports although they are not all part of every instance.
Ports are identified by their UNLOCODE where applicable.
Two distance files exist: dist_dense.csv which is an all to all distance file with drafts and canal traversal indication where applicable.
dist_sparse.txt which constitute a port with its nearest ports and way points. A distance from any two ports is done through routing of the corresponding way points.
Due to the generation of the distances via NIMA there is a 5% error margin on the distances.
Fleet information is gathered in the file fleet_data.csv and contains all data on the 6 vessel types, that are used throughout the 7 instances.
For every instance a fleet file provides a type and a quantity for the vessel types available in the instances.
It is possible to generate 3 instances (high, medium and low) related to different capacity/demand relation from every of the 7 instances.
To generate the high and low instances TC rates and quantity of vessels are adjusted according to the fleet_data.csv and fleet_$INSTANCE$.csv according to the following rules:
To generate a HIGH capacity case you need to: Multiply TC Rates by 0.8 and round to nearest thousand. Multiply Quantity by 1.2 and round to nearest integer.
To generate a LOW capacity case you need to: Multiply TC Rates by 1.4 and round to nearest thousand. Multiply Quantity by 0.8 and round to nearest integer.
for more information on each cost parameter and data creation and source please consult the technical report.
We strongly urge you to stay informed about the LINERLIB project by following us on GitHub.
Acknowledgements: We would like to thank Maersk Line, Network and Product for valuable insights and access to real life data for the creation of the benchmark suite. In particular we would like to thank Niels Madsen for valuable comments and suggestions for improvement. We also thank Jørgen Harling and Niels Rasmussen, Network and Product, Maersk Line. This project would not have been possible without their support.
We are grateful to Vereinigung Hamburger Schiffsmakler und Schiffsagenten, Drewry Shipping Consultants, Alphaliner.com and National Imagery and Mapping Agency for providing data to construct the benchmark suite data.
This project was supported in part by The Danish Strategical Research Council under the ENERPLAN project.
======== HISTORY: Version 1.0 released August 16. 2013
Version 1.1 released April 2014 Duplicate entries of RULED removed lifting and port call cost in ports.csv rounded of to nearest integer readme files clarifies rounding of TC rates to nearest thousand as explained in the original article