|
|
 |
|
Mardi 22 Mars
| Heure: |
12:30 - 13:30 |
| Lieu: |
Salle B107, bâtiment B, Université de Villetaneuse |
| Résumé: |
Combinatorial Optimization subject to PDE constraints |
| Description: |
Christoph Buchheim We investigate a class of optimal control problems where the control parameters are binary variables, which are supposed to be static and probably subject to combinatorial constraints. Additionally, the problem contains state constraints involving semi-linear elliptic partial differential equations (PDEs). As an example, the binary variables may correspond to the activation of given heat sources attached to a metal sheet, the optimization problem then consists in switching on the smallest number of heat sources such that the point-wise temperature of the metal sheet is in a specified range. Our main result is that each state is a concave function in the binary control vector. This allows us to define an outer approximation scheme in which we alternate between the solution of a non-linear PDE and an integer linear program. Using this approach, we can solve instances on up to 2000 binary variables to global optimality. |
|
|
