BNSMAT

This project is an example implementation of Branch and Sandwich algorithm in Matlab for bi-level MILP problems. The usage is provided by main.m

The algorithm is implemented for scientific applications and, in fact, is not applicable to the real field cases, since MATLAB is not a good choice for it.

Currently I'm working on Julia implementation of the same algorithm.

Example

clear slv;
clear problemInstance;
clear BnSSolver;
clear propdict;


Co = [-1 -10];
Ao = [];
Bo = [];
ci = [0 1];
ai = [-25 20; 1 2; 2 -1; -2 -10];
bi = [30;10;15;-15];
lc = [0; 0];
uc = [8; 4];

problemInstance = bp_milp(Co, Ao, Bo, ci, ai, bi, [1 2], lc, uc, 1); 
slv = BnSSolver(problemInstance);
slv.solve();

Adding new classes of problems

BnS itself is a management strategy. Any kind of problem should be accompanied with a solver for five types of problems. In case of MILP it could be done in general, however some MINLP problems might require additional tuning.

The interface is described in @bilevelProblem class which should be inherited. Unfortunately, it is described partially at this moment (TODO)