cs.NIFeb 20, 2026
SaveRethinking Beam Management: Generalization Limits Under Hardware Heterogeneity
Organizations: Tampere University, Tampere, Finland · University of California San Diego, CA, USA
Abstract
Hardware heterogeneity across diverse user devices poses new challenges for beam-based communication in 5G and beyond. This heterogeneity limits the applicability of machine learning (ML)-based algorithms. This article highlights the critical need to treat hardware heterogeneity as a first-class design concern in ML-aided beam management. We analyze key failure modes in the presence of heterogeneity and present case studies demonstrating their performance impact. Finally, we discuss potential strategies to improve generalization in beam management.
Figures & tables
Fig. 1: Illustration of heterogeneity types in data-driven beam management.
Fig. 2: 3D layout of simulation scenario and SUMO snapshot.
Fig. 3: Demonstration of performance drop under different UE heterogeneity types. The bars indicate the sample mean of the SE drop relative to the genie-based SE (thus, in %), and the 90th percentile illustrates the variation in the distribution of the relative SE values. In the “train” setup, the DNN-based beam predictor (blue) always performs on par or better than two-level HS (orange) and ES (green). Under heterogeneous “test” conditions, however, the DNN-based predictor exhibits pronounced performance degradation and high variability, particularly under codebook and location heterogeneity.
| Nikita Zeulin is a Doctoral Researcher currently pursuing a doctoral degree with Tampere University, Finland. He received his B.Sc. degree from St. Petersburg State University of Aerospace Instrumentation, Russia, and his M.Sc. degree from Skolkovo Institute of Science and Technology, Russia. His research interests include federated learning and machine learning enhancements for wireless communications. |
| Olga Galinina is an Institute of Advanced Study Senior Research Fellow at Tampere University, Finland. She received her B.Sc. and M.Sc. degrees from the Department of Applied Mathematics, St. Petersburg State Polytechnical University of Peter the First, and her Ph.D. degree from Tampere University of Technology. Her research interests include applied mathematical modeling and analysis of wireless networks and statistical machine learning. |
| Ibrahim Kilinc is a Ph.D. student at the University of California, San Diego. He received his B.S. in Electrical and Electronics Engineering and his minor degree in Information Systems and Technologies from Bilkent University, Ankara, Turkey, in 2023. His research interests include sensor-aided beam alignment in heterogeneous devices in wireless systems and signal processing. |
| Sergey Andreev received the Ph.D. degree from Tampere University of Technology, and the Cand.Sc., and Dr.Habil. degrees from SUAI. He is a Professor of Wireless Communications and formerly an Academy Research Fellow with Tampere University, Finland. He is also a Research Specialist with Brno University of Technology, Czech Republic. He co-authored more than 300 published research works on intelligent IoT, mobile communications, and heterogeneous networking. |
| Robert W. Heath Jr. is the Charles Lee Powell Chair in Wireless Communication in the Department of ECE at the University of California, San Diego. He has received several awards including the 2019 IEEE Kiyo Tomiyasu Award, the 2020 North Carolina State University Innovator of the Year Award, and the 2021 IEEE Vehicular Technology Society James Evans Avant Garde Award. He is a licensed Amateur Radio Operator, a registered Professional Engineer in Texas, a Private Pilot, a Fellow of the National Academy of Inventors, a Fellow of the IEEE, and a Fellow of the AAAS. |
More in this topic
Wireless Communications
Wireless communications studies how information is sent over radio links and how networks manage signals, spectrum, and resources reliably.
See all 399 papers