Hardware realization of neuromorphic computing with a 4–port photonic reservoir for modulation format identification
Congratulations to Enes Şeker and the entire team on our latest paper published in Neuromorphic Computing and Engineering:
🧠💡 “Hardware realization of neuromorphic computing with a 4–port photonic reservoir for modulation format identification”
Co-authored with Rijil Thomas, Guillermo von Hünefeld, Stephan Suckow, Mahdi Kaveh, Gregor Ronniger, Pooyan Safari, Isaac Sackey, David Stahl, Colja Schubert, Johannes Karl Fischer, Ronald Freund, and Max Lemme.
📡 In this work, we present NeuroPIC — a hybrid photonic–electronic reservoir computing system designed for modulation format identification in C-band telecom networks. Built on a silicon-on-insulator platform, the NeuroPIC features a 4-port, 16-node reservoir architecture and achieves ~100% accuracy in identifying 4QAM, 16QAM, 32QAM, and 64QAM formats over 20 km fiber links at 32 Gbaud.
✨ What makes this special?
• Combines energy-efficient photonic analog processing with digital readout
• Demonstrates robust performance despite fabrication imperfections
• Outperforms simulations thanks to richer signal interference in the physical system
This work showcases the potential of neuromorphic photonic hardware to revolutionize high-speed temporal data processing in real-world applications.
Read the full paper: https://lnkd.in/g84tmDbR
AMO GmbH, Lehrstuhl für Elektronische Bauelemente (ELD) at RWTH Aachen University, Fraunhofer Heinrich Hertz Institute HHI
This research was funded by the Federal Ministry of Research, Technology and Space of Germany (BMFTR) in the framework of CELTIC-NEXT AI-NET PROTECT (AMO GmbH), FKZ 16KIS1281 (Fraunhofer-Institut für Nachrichtentechnik, Heinrich Hertz Institute), and FKZ 16KIS1301 (ID Photonics GmbH)], under the 6G-RIC project with Grant 16KISK020K and in the Cluster4Future NeuroSys (FKZ 03ZU1106BB).
Special thanks to Piotr Cegielski (now with Infineon Technologies, Munich), Anna Lena Schall-Giesecke (now with the University of Duisburg-Essen and Fraunhofer IMS, Duisburg), and Peter Bienstman (Ghent University) for their invaluable insights and stimulating discussions on the subject matter of this research.
