Model class
BOBILib contains bilevel optimization instances that are
all part of the class of MILP-MILP bilevel problems.
This means that both the upper- and the lower-level problem is a
mixed integer linear problem (MILP).
In particular, this includes the possibility that both the upper- and
the lower-level problem can be
- a continuous linear problem,
- a pure integer linear problem,
- a mixed integer linear problem.
Model statistics
The instances in the instance tables can be sorted with respect to the following statistics of the instances.
| Symbol | Explanation |
|---|---|
| \(n_x\) | Number of upper-level variables |
| \(|I_u|\) | Number of upper-level integer variables |
| \(|B_u|\) | Number of upper-level binary variables |
| \(n_y\) | Number of lower-level variables |
| \(|I_l|\) | Number of lower-level integer variables |
| \(|B_l|\) | Number of lower-level binary variables |
| \(n_\mathrm{Link}\) | Number of linking variables |
| \(n_\mathrm{Link}^I\) | Number of integer linking variables |
| \(n_\mathrm{Link}^B\) | Number of binary linking variables |
| \(n_\mathrm{Link}^C\) | Number of continuous linking variables |
| \(m_u\) | Number of upper-level constraints |
| \(m_l\) | Number of lower-level constraints |
| \(m_\mathrm{Coup}\) | Number of coupling constraints |
| \(m_\mathrm{Link}\) | Number of linking constraints |
| \(\alpha\) | Objective function alignment w.r.t. \(y\), i.e. \( d_u^\top e / (\|d_u\|_2 \|e\|_2) \) |
Input and solution file format
All instances in the BOBILib use
the (name-based) MibS input
file format and are thus given as pairs of mps and aux files.
For more information about the MPS file format, read the respective
MPS (format)
wikipedia page.
A detailed explanation with an example is given in Section 4 of the BOBILib report.
There, you also find the definition and an example for the solution format.
License
All instances are licensed under the Creative Commons CC BY-SA 4.0 license.