The paper considers the bi-objective extension of the classical (symmetric) TSP problem, i.e. it is a TSP problem with two cost matrices.
The instance set is created by taking 5 TSP problems with 100 cities (kroa100.tsp, krob100.tsp, kroc100.tsp, krod100.tsp, kroe100.tsp) from the TSPLIB and combining them into 10 BITSP testsets, (kro100ab.raw, kro100ac.raw, kro100ad.raw, kro100ae.raw, kro100bc.raw, kro100bd.raw, kro100be.raw, kro100cd.raw, kro100ce.raw, kro100de.raw).
All instance files are given solely in the raw format.
We use the following parameter names:
-
$n$ = dimension/size -
$c_{i,j}$ = the distance from city$i$ to city$j$ . Notice that all distances have integer value (as in TSPLIB)
The instances have the following format, one dimension number and two full, symmetric two-dimensional matrices, where each line correspond to one entry to a matrix. The distance between the city and itself is naturally 0.
n
0 0 c_{0,0}
0 1 c_{0,1}
0 2 c_{0,2}
....
n-1 n-3 c_{n-1,n-3}
n-1 n-2 c_{n-1,n-2}
n-1 n-1 c_{n-1,n-1}
0 0 c_{0,0}
0 1 c_{0,1}
0 2 c_{0,2}
....
n-1 n-3 c_{n-1,n-3}
n-1 n-2 c_{n-1,n-2}
n-1 n-1 c_{n-1,n-1}
That is, first the dimension, then the distance for each pair of cities, first in objective 1 matrix, then a blank line then in objective 2 matrix.