Scribd
Upload
140 views

TSPLIB in XML Format

The document describes a new XML data format for TSPLIB instances of travelling salesman problems. It provides the data from TSPLIB instances in a uniform XML structure, making it easier to use the data and generate new instances. An example XML file is shown representing an undirected graph instance, with tags for the problem type, graph vertices and edges, and edge costs. The format can be extended to other related problems and non-complete or multigraphs.

Uploaded by

daskir
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
140 views

TSPLIB in XML Format

The document describes a new XML data format for TSPLIB instances of travelling salesman problems. It provides the data from TSPLIB instances in a uniform XML structure, making it easier to use the data and generate new instances. An example XML file is shown representing an undirected graph instance, with tags for the problem type, graph vertices and edges, and edge costs. The format can be extended to other related problems and non-complete or multigraphs.

Uploaded by

daskir
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

TSPLIB in XML format

To facilitate the usage of the classical TSPLIB by Gerhard Reinelt, we


now offer the data of the well-known test instances also in an XML data for-
mat, both for the symmetric and the asymmetric travelling salesman prob-
lem. Thus, the different data representations of the current instances are
transformed into a uniform style. This should be helpful also for using the
data as a basis for generating instances with additional constraints or other
properties. All files have the suffix .xml
The data format should be easy to use together with the XML parsers pro-
vided for the most common programming languages. An example for using
a parser is given in the included program contained in the file ”Sample.zip”.
A validation file “TSPConfiguration.xsd” is also provided.

Data format For illustration consider the undirected graph in Figure 1.

0
2 1
2
1 4
1 2
2 2
2 1

2 3
2
costs

Figure 1: An instance of the Symmetric travelling salesman problem.

Listing 1 shows the corresponding XML-file. Lines No. 1 and 2 define


the parameters of the XML file – e.g. the version and the encoding. The
following tag (line No. 3) specifies the problem type and has to be “travel-
lingSalesmanProblemInstance”.
The next entries define the name, source and the description of the TSP
instance. The tag doublePrecision states the number of decimal digits after
the comma of the edge costs which are read from the file. The tag ignoredDig-
its represents the number of remaining decimal digits which have to be ignored
when comparing two costs, in the sense that 10−(doubleP recision−ignoredDigits)

1
yields 0 in the computation. Note that all these tags have to be defined and
that their relative order is not allowed to vary.
Finally, the weighted graph is defined by the tag graph (line No. 9). Each
vertex (tags vertex) contains the list of all emanating edges. These edges
are defined by the tags edge (see e.g. the line No. 11). Each edge has an
attribute cost (always written as a real number) and contains the identifier of
the other vertex incident to that edge. The vertex identifiers are defined by
their relative order (starting from 0). Note that the cardinality of the vertex
set is defined implicitly by the number of vertex tags contained in the file. If
the graph is undirected, all edges are defined twice (from i to j and from j
to i) having the same costs.
1 <?xml version="1.0" encoding="UTF-8"
2 standalone="ignoredDigitsno" ?>
3 <travellingSalesmanProblemInstance>
4 <name>Small 1</name>
5 <source>Rostislav Stanek</source>
6 <description>A small instance.</description>
7 <doublePrecision>15</doublePrecision>
8 <ignoredDigits>5</ignoredDigits>
9 <graph>
10 <vertex>
11 <edge cost="2.000000000000000e+00">1</edge>
12 <edge cost="1.000000000000000e+00">2</edge>
13 <edge cost="2.000000000000000e+00">3</edge>
14 <edge cost="1.000000000000000e+00">4</edge>
15 </vertex>
16 <vertex>
17 <edge cost="2.000000000000000e+00">0</edge>
18 <edge cost="2.000000000000000e+00">2</edge>
19 <edge cost="2.000000000000000e+00">3</edge>
20 <edge cost="2.000000000000000e+00">4</edge>
21 </vertex>
22 <vertex>
23 <edge cost="1.000000000000000e+00">0</edge>
24 <edge cost="2.000000000000000e+00">1</edge>
25 <edge cost="2.000000000000000e+00">3</edge>
26 <edge cost="1.000000000000000e+00">4</edge>
27 </vertex>
28 <vertex>
29 <edge cost="2.000000000000000e+00">0</edge>

2
30 <edge cost="2.000000000000000e+00">1</edge>
31 <edge cost="2.000000000000000e+00">2</edge>
32 <edge cost="2.000000000000000e+00">4</edge>
33 </vertex>
34 <vertex>
35 <edge cost="1.000000000000000e+00">0</edge>
36 <edge cost="2.000000000000000e+00">1</edge>
37 <edge cost="1.000000000000000e+00">2</edge>
38 <edge cost="2.000000000000000e+00">3</edge>
39 </vertex>
40 </graph>
41 </travellingSalesmanProblemInstance>
Listing 1: An instance of the Symmetric travelling salesman problem.

All costs are written as real numbers (also if they were given as integers)
and the costs were not rounded. For comparing results with the classical
TSPLIB data, costs should be rounded first (following the TSPLIB rules).
Only for the 12 cases of geographical distances (ali535, bayg29, bays29,
burma14, gr137, gr202, gr229, gr431, gr666, gr96, ulysses16,
ulysses22) the costs were already rounded (but are still given as real num-
bers) according to the sophisticated rounding rules required by the TSPLIB
for coordinates on the globe. Thereby, the instances guarantee the same
optimal solution values as listed in the classical TSPLIB.

Extensions The data format can be easily extended for many problems
related to the TSP – e.g. for the Hamiltonian cycle problem (HCP), the
Capacitated vehicle routing problem (CVRP) or for the Orienteering problem
(OP). This can be done e.g. by introducing new general tags (as for the
knapsack constraint in the OP) or by defining new vertex or edge attributes
(e.g. edge capacities, vertex profits etc.). The format can also be used for
non-complete graphs or extended for multigraphs.

January, 2013

Ulrich Pferschy
Rostislav Staněk
Department of Statistics and Operations Research
University of Graz
Universitätsstraße 15
8010 Graz, Austria.
{pferschy, rostislav.stanek}@uni-graz.at

You might also like