T3 – Continual Learning for Integrated Terrestrial and Non-Terrestrial Networks

Co-presenter: Muhammad Ali Jamshed, University of Glasgow, UK
Co-presenter: Muhammad Ahmed Mohsin, Stanford University, USA

Abstract: The integration of Non-Terrestrial Networks (NTN), including satellites, High-Altitude Platform Stations (HAPS), and Uncrewed Aerial Vehicles (UAVs), with Terrestrial Networks (TN) is a defining challenge for 5G-Advanced and 6G. TN-NTN integration promises global coverage and resilient connectivity for underserved regions and maritime operations. However, it introduces severe impairments: large Doppler shifts, rapid mobility, variable delays, high path loss, and spectrum coexistence issues. These challenges demand innovative signal processing solutions augmented with Artificial Intelligence (AI). Accurate and timely Channel State Information (CSI) prediction is a bottleneck for integrated TN-NTN systems. Channel aging due to mobility and 3GPP timing constraints drastically reduces spectral efficiency if not mitigated. While Deep Learning (DL)-based CSI predictors (RNNs, LSTMs, transformers) show promise, they suffer from poor generalization across configurations (e.g., antenna arrays, frequencies, mobility profiles), requiring costly retraining. Continual learning offers a path forward. By enabling models to adapt to new environments without catastrophic forgetting, continual learning can address domain shift, support online adaptation, and ensure robust CSI prediction in dynamic TN-NTN scenarios. Emerging methods, such as Learning without Forgetting (LwF), generative diffusion models for CSI augmentation, and foundation models adapted to wireless systems, open new avenues for reliable integration.

This tutorial provides a comprehensive overview of continual learning for integrated TN-NTN systems, bridging communication engineering, AI, and signal processing. Participants will gain insights into state-of-the-art methods, limitations, and future research directions to build intelligent, adaptive, and resilient 6G systems.

Co-presenter’s Bios:

Muhammad Ali Jamshed:

Muhammad Ali Jamshed has been with the University of Glasgow since 2021. He is currently affiliated as a Researcher with the Machine Learning and Communications Lab at Stanford University, USA and a Visiting Research Professor at KNU, South Korea. He also serves as Chief Technical Advisor at TEQ-IT Ltd. He was a Visiting Research Fellow at the University of Sussex from 2022 to 2024 and worked as a Wireless Technical Consultant with Briteyellow Ltd, between 2022 and 2023. He is endorsed by the Royal Academy of Engineering under the Exceptional Talent category and was nominated for the Departmental Prize for Excellence in Research in both 2019 and 2020 at the University of Surrey. He is a Senior Member of IEEE, a Fellow of the Royal Society of Arts (FRSA), a Fellow of the Higher Education Academy (FHEA), and a Member of the EURASIP Academy. He serves as an Editor for IEEE Wireless Communications Letters and as an Associate Editor for the IEEE Sensors Journal, IEEE Internet of Things Magazine, and IEEE Communications Standards Magazine. He is the Founding/Lead Chair of the IEEE ComSoc Special Interest Group on AI for Integrated Terrestrial and Non-Terrestrial Networks (AITNTN) and has served as General Chair for over 20 workshops at leading IEEE international conferences, including IEEE GLOBECOM, IEEE ICC, and IEEE INFOCOM. His research interests include energy-efficient networks, AI for wireless communications, intelligent signal processing for non-terrestrial networks (NTN), EMF exposure measurements, and backscatter communications. He has published over 90 research papers in these areas and edited six books.

Muhammad Ahmed Mohsin:

Muhammad Ahmed Mohsin received his undergraduate degree from National University of Sciences and Technology, Islamabad Pakistan where received the Rectors Gold Medal for his final year thesis. He is currently pursuing his Ph.D. under Dr. John M. Cioffi, and his research interests lie in reinforcement learning for optimization. He has broadly worked on non-linear optimization, resource allocation and reasoning models in LLMs and has published paper at top conferences like ICML, ICLR, ICC and ICASSP. He won the best paper award for ICC 2025 and has won several research grants for his Ph.D. thesis work and several travel grants to IEEE ComSoc conferences as well.