Rules for generating the M* benchmarks were proposed in 2001. These benchmarks are designed
to be similar to real-life problems and have a large number of
near-optimal solutions. There are 6 classes of problems, and for each problem m = n.
| Class | Size (m = n) |
| MO | 100 |
| MP | 200 |
| MQ | 300 |
| MR | 500 |
| MS | 1000 |
| MT | 2000 |
Our package was originally generated by Laurent Michel. There are 5 instances of each class O, P, Q and R, and there is one instance for T and S, respectively. The package can be downloaded here. Furthermore, there is a generator available for MS Windows operating systems at the website of Jozef Kratica.
| [1] |
J. Kratica, D. Tosic, V. Filipovic and I. Ljubic Solving the simple plant location problem by genetic algorithm RAIRO Operations Research, 35:127-142. 2001. |