Advanced Machine Learning for 6G Networks
- Publication Type:
- Thesis
- Issue Date:
- 2024
Open Access
Copyright Clearance Process
- Recently Added
- In Progress
- Open Access
This item is open access.
Beyond 5G and 6G communications are foreseen to transform the world, connecting not only people but also vehicles, wearables, devices, sensors, and even physical and digital worlds. To achieve that, 6G systems are expected to employ various disruptive technologies (e.g., non-terrestrial networks (NTNs), mmWave communications, pervasive artificial intelligence, and ambient backscatter communications) to enable/support new use cases, e.g., autonomous cyber-physical systems and Metaverse/holographic teleportation. Thus, this thesis aims to leverage the latest advances in machine learning (ML) to address different problems facing 6G systems. We first envision that UAVs will play a critical role in 6G and NTNs, e.g., flying data collectors. To tackle the uncertainty in the data collection process and the UAV’s energy capacity limitation, we propose an innovative deep reinforcement transfer learning approach to control the UAV's speed and energy replenishment process and allow UAVs to ``share'' and ``transfer'' learning knowledge, thus reducing learning time and improving learning quality significantly.
6G is also envisioned as ubiquitous sensors thanks to the Integrated Communications and Sensing (ICAS) technology, e.g., for flood sensing/warning or in autonomous vehicles (Avs). Optimizing the waveform structure for ICAS applications to AVs is one of the most challenging tasks due to the strong influences between sensing and data communication functions under dynamic environments. Therefore, we develop a novel framework that intelligently and adaptively optimize its waveform structure to maximize sensing and data communication performance.
Another key application/service of 6G is to enable the seamless deployment and operation of Metaverse. Building and maintaining the Metaverse not only demand enormous resources but also need to address the dynamic, uncertain, and real-time resource demands. Thus, we develop a novel ML-based framework that offers a highly effective and comprehensive solution for managing various resource types for Metaverse by leveraging the similarities among applications.
Security is always one of the top concerns in wireless communications, especially for 6G connected by a massive number of heterogeneous devices. We design a lightweight framework leveraging ambient backscatter communications and deep meta-learning to counter eavesdropping attacks, effectively decode weak backscattered signals without requiring perfect information, and quickly adapt to new environments with very limited knowledge.
The above results demonstrate the great potential of advanced machine learning in addressing the emerging issues of 6G and enabling new applications/services. As future works, one may look into the applications of Generative AI to 6G and how to design 6G systems to enable Generative AI as a service.
Please use this identifier to cite or link to this item: