Stuttgart 2012 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
Q: Fachverband Quantenoptik und Photonik
Q 57: Kalte Atome
Q 57.7: Vortrag
Freitag, 16. März 2012, 12:15–12:30, V7.02
A self-optimizing experimental apparatus — •Ilka Geisel1, Stefan Jöllenbeck1, Jan Mahnke1, Kai Cordes2, Wolfgang Ertmer1, and Carsten Klempt1 — 1Institut für Quantenoptik, Leibniz Universität Hannover — 2Institut für Informationsverabeitung, Leibniz Universität Hannover
Even though most parameters in a typical cold atom experimental setup are controlled by one computer program, optimization is still usually done by hand. One has to find the correlations between unknown parameters in the experiment in order to reach the optimum. Systematically scanning the whole parameter space will quickly become impossible as one goes to higher dimensions.
The logical step is to use an automated optimization procedure. The demands on such a program include finding the global optimum and being robust against experimental noise while reaching a sensible solution within a small number of experimental cycles. We present a genetic algorithm based on Differential Evolution, which quickly finds the optimum even with strong experimental noise. Relying only on basic mathematics it requires little computing power and is easy to implement.
Using the algorithm we improved our magneto-optical trap in a nine dimensional partly correlated parameter space by over 20%. A simulation allows for studying the behavior of the algorithm under different noise levels and parameters and thus reaching the optimal configuration for optimizing a wide range of experimental tasks.