| Zugriffsnummer | 53641 |
| Dokumenttyp | Zeitschriftenartikel |
| Peer Review | mit Peer Review |
| Sprache | Englisch |
| Titel | Hybrid AI-PID control for precision metrology |
| Autor(in); Institution |
Degenhardt, Johannes; 5.2, Dimensionelle Nanometrologie, PTB-Braunschweig
Bounaim, Mohamed Wassim; 5.2, Dimensionelle Nanometrologie, PTB-Braunschweig
Tutsch, Rainer; Institute of Production Metrology IPROM TU Braunschweig, Braunschweig, GERMANY
|
| Quelle/Jahr | Measurement: Sensors: 38 (2025), Suppl., e1 - e7 |
| Artikelnummer | 101786 |
| Availability | [online only] |
| ISSN | 2665-9174 (online) |
| DOI | |
| URL | |
| Verlag | Amsterdam: Elsevier |
| Freie Schlagworte | Artificial intelligence ; Deep reinforcement learning ; Optimal control ; Atomic force microscopy ; Metrology |
| Zusammenfassung | Artificial intelligence (AI) offers great potential for implementing high-performance feedback controllers for precision measurement instruments. However, since AI is critical from a reliability standpoint, the concept of hybrid control, which combines AI with proportional-integral-differential control (PID), is investigated here. Using an atomic force microscope as an example, the approach of backup hybrid control and cooperative hybrid control is compared. In the backup hybrid control approach, AI primarily controls the system, and PID only intervenes if necessary. In the cooperative approach, both controllers contribute continuously to the control signal. While the backup hybrid approach can theoretically deliver better control performance, evidence suggests that the cooperative approach is superior, especially for metrological applications where reliability and trustworthiness are paramount. |
| Kostenfreier Zugang | Open Access Gold |