Reinhard Wiesmayr will present at the NVIDIA 6G Developer Day
Sign up for the first NVIDIA 6G Developer Day to attend Reinhard’s talk about Machine-Learning-Assisted Physical Layer Processing For 6G and many other interesting talks from academia and industry.
Machine learning (ML)-based physical layer (PHY) processing has the potential to outperform traditional wireless transceiver. Significant performance improvements are typically observed in real-world scenarios involving transmission over wireless channels with actual hardware. While most state-of-the-art research on ML-based PHY algorithms relies on synthetic data, NVIDIA’s ARC-OTA platform enables us to develop and validate new algorithms using over-the-air measurements in a real-time 5G system. Our current research explores two key ML-based PHY paradigms: (i) model-based methods and (ii) neural network (NN)-based approaches. The first paradigm unfolds traditional PHY algorithms and utilizes ML techniques to automatically tune algorithm parameters. In contrast, the second paradigm replaces entire PHY algorithms with NNs. In this talk, we will present results from both paradigms and compare their respective pros and cons. Furthermore, we will outline our next steps toward deploying these paradigms using the NVIDIA ARC-OTA system at ETH Zurich—this is a critical step for comprehensively evaluating candidate algorithms for next-generation wireless systems, including 6G.
Registration and further information on the external page event homepage!