Distributed machine learning for 6G intelligent vehicular communication networks

The rapid advancements in 6G technologies are transforming Vehicular Networks (VNs) into smarter, more connected systems. As autonomous vehicles and intelligent transportation systems become central to modern society, ensuring efficient, scalable, and secure communication is paramount. This thesis addresses these challenges by integrating Distributed Learning (DL) techniques with network slicing and integrated Terrestrial and Non‑Terrestrial Networks (T/NTNs) to satisfy the diverse and dynamic requirements of vehicular applications, such as autonomous driving and real-time traffic management, within the 6G ecosystem. The primary aim is to develop an adaptive, scalable, and secure framework, DL-as-a-Service (DLaaS), that enables real-time data processing and ultra-low latency in 6G VNs. DLaaS unifies various DL techniques to optimize both communication and computation while preserving user privacy. It is designed to handle heterogeneous vehicular devices and the complexity of multilayer networks, ensuring seamless operation across integrated T/NTN layers. Methodologically, the thesis introduces Federated Split Transfer Learning and its generalized version to address challenges related to resource-constrained devices and heterogeneous network environments. These DL techniques are tailored to meet the needs of 6G VNs, offering scalable solutions to real-time learning, model synchronization, and privacy protection. The practical viability of these approaches is validated through real‑world simulations and hardware implementations on edge computing platforms. The key findings of this research demonstrate that the proposed frameworks significantly enhance the accuracy, scalability, and latency of VNs. Furthermore, the application of deep reinforcement learning for dynamic adaptive streaming optimizes bitrate allocation, caching, and transcoding, improving quality of experience for users in real-time. These results confirm the feasibility of integrating advanced machine learning techniques and T/NTNs into the design of 6G internet of vehicle systems. In the end, proper conclusive remarks and several future directions are provided for the proposed solutions, offering valuable insights into the future of smart cities and intelligent mobility.