The input files are derived from six files of the TSPLIB library:
fnl4461, pr1002, nrw1379, d18512, d15112, brd14051. From each file we
created istances with different dimensions. For example, to generate
an input file with 15 vertices from "fnl4461.tsp" we create three
files: fnl4461_D15.tsp, matching_fnl4461_15.mtc,
starting_tour.str. The first file represents the complete graph. This
file is a copy of the original file "fnl4461.tsp" with the only
difference that the parameter dimension inside it is set to 15. This
means that the graph is represented by only the first 15 vertices.
The second file, "matching_fnl4461_15.mtc", contains the information
to state what are the pickup and delivery vertices of the graph and
how they are associated. More precisely, on each row of the .mtc file
there are 3 numbers: the first is the ID of the vertex, the second
states whether this vertex is a pickup or a delivery (0 = delivery, 1
= pickup, 3 = depot) and the last number is the ID of the vertex
associated to it. For instance, if in the row there is: 3 0 7 this
means that vertex 3 is a delivery vertex and its pickup is vertex 7
(the only exception is for the depot whose row is: 0 3 0). The
definition of pickup and delivery vertices and the generation of
matchings were executed in a random way. Finally, the last file
represents the starting tour ( .str files). We use eight starting
tours generated with the techniques specified by Cassani and Righini
(2004). These files are composed by a single column representing the
tour (a sequence of vertex IDs) and on the last row is shown the tour
cost. In order to better assemble the input files, the last two files
"matching_fnl4461_15.mtc" and "starting_tour.str", are saved into a
directory named "fnl4461_DIM_15". The same structure is repeated for
other dimensions and types of files.