BOBILib: Bilevel Optimization (Benchmark) Instance Library

BOBILib: Bilevel Optimization (Benchmark) Instance Library
Computational bilevel optimization for models including binary or general integer variables has been a highly active field of research over the last years and decades. To propel the development of new algorithms, it is very important for the community to have access to a large and well-curated set of test instances that can be used to test new algorithms and to compare different methods. There are very many success stories in computational optimization of such instance collections such as MIPLIB, MINLPLib, or QPLIB. The aim of BOBILib and this website is to provide such a test instance collection for (mixed) integer bilevel optimization including all best known feasible points. The entire collection consists of more than 2500 instances. Moreover, we compiled a benchmark instance set of 122 instances, which provides a meaningful basis for experimental comparisons of solution methods in a moderate time.

Contact us

The success of this instance collection also depends on the submissions from the community to let the collection grow and contain more and more challenging and realistic instances. If you have an instance that you want to contribute, please submit it via e-mail.

Latest news

July 2024: The report and the website are online in their first version!

Citation

If you use BOBILib in your paper or thesis, please cite us using the following BibTeX entry.
      @report{BOBILib:2024,
        author = {Johannes Thürauf and Thomas Kleinert and Ivana Ljubi\'c and Ted Ralphs and Martin Schmidt},
        title  = {BOBILib: Bilevel Optimization (Benchmark) Instance Library},
        year   = {2024},
        url    = {https://optimization-online.org/?p=27063}
      }
    
You can find the report here.

Disclaimer

The collection and processing of the instances that are available on this website is often intricate detail work. Thus, we cannot guarantee that there are no mistakes. If you spot any problems, please let us know via sending an e-mail.