A growing collection of bilevel problems
mb_2007_03 : Linear-Quadratic bilevel problem from [Mitsos & Barton, 2007]
Objective values | Solution point(s) |
---|---|
F* = -1.000 | y* = -1.000 |
f* = 1.000 |
The feasible set for the inner problem is the union of [-10,-1] and [1,10] and its optimal solutions are y = +/-1. Thus unique optimal solution of the bilevel problem is y* = -1 with F* = -1 and f* = 1.
Outer Problem | Inner Problem |
---|---|
![]() |
![]() |
Original source:
Other sources:
AMPL
formatvar y >= -10, <= 10; # Inner variable
var l{1..3} >= 0, <= 2; # KKT Multipliers
minimize outer_obj: y; # Outer objective
subject to
# Inner objective:
inner_obj: y^2 = 0;
# Inner constraints
inner_con: 1 - y^2 <= 0;
# KKT conditions
stationarity: 2*y - 2*y*l[1] - l[2] + l[3] = 0;
complementarity_1: l[1]*(1 - y^2) = 0;
complementarity_2: l[2]*(-10 - y) = 0;
complementarity_3: l[3]*(y - 10) = 0;