Abstract: This PhD thesis focuses on enhancing Intelligent Transportation Systems (ITS) by integrating Vehicular Edge Computing (VEC) into 5G networks to overcome the hardware limitations of individual vehicles. Since critical V2X services are highly sensitive to latency, the research leverages SDN and NFV technologies to process data at the network edge rather than in a distant cloud, thereby ensuring high bandwidth and rapid response times. The primary objective is to develop a joint resource allocation framework that efficiently manages both radio and computational resources to minimize the cumulative delay caused by transmission and processing.

Abstract: While Artificial Intelligence (AI) has significantly advanced earth observation and remote sensing, deep learning models often act as opaque "black boxes," limiting their trustworthiness and practical adoption. Although Explainable AI (XAI) improves transparency, its application has largely focused on image classification rather than segmentation. This project bridges that gap by adapting recent XAI algorithms for multi-class image segmentation, focusing specifically on building extraction from high-resolution satellite imagery. Furthermore, we introduce a novel, robust XAI evaluation methodology based on "Entropy" to measure model uncertainty, demonstrating its superiority over conventional evaluation metrics. Ultimately, this research paves the way for more reliable and interpretable AI deployments in remote sensing..

Abstract: This PhD thesis investigates the enhancement of Physical Layer Security (PLS) and energy efficiency in next-generation wireless networks by combining NOMA with Ambient Backscatter Communication (AmBC). The study develops novel resource allocation strategies to maximize Secrecy Energy-Efficiency (SEE), providing exact mathematical solutions that simplify complex optimization problems into efficient, low-complexity searches. To address real-time dynamic environments, the project integrates Explainable Artificial Intelligence (XAI) to predict optimal allocations with high accuracy while ensuring model interpretability and significantly reduced computational overhead.

Abstract: Reliable and fast channel estimation is crucial for next-generation wireless networks supporting a wide range of vehicular and low-latency services. Recently, deep learning (DL)-based channel estimation has been explored as an efficient alternative to conventional least-square (LS) and linear minimum mean square error (LMMSE) approaches. Most of these DL approaches have not been realized on system on chip (SoC), and preliminary study shows that their complexity exceeds the complexity of the entire physical layer (PHY). The high latency of DL is another concern. This article considers the design and implementation of deep neural network (DNN) augmented LS (LSDNN)-based channel estimation for preamble-based orthogonal frequency-division multiplexing (OFDM) PHY on SoC. We demonstrate the gain in performance compared with the conventional LS and LMMSE approaches. Via software–hardware codesign, word-length optimization, and reconfigurable architectures, we demonstrate the superiority of the LSDNN over LS and LMMSE for a wide range of signal-to-noise ratio (SNR), number of pilots, preamble types, and wireless channels. Furthermore, we evaluate the performance, power, and area (PPA) of the LS and LSDNN application-specific integrated circuit (ASIC) implementations in 45-nm technology. We demonstrate that word-length optimization can substantially improve PPA for the proposed architecture in ASIC implementations.

Abstract: In Vehicle-to-Everything (V2X) communications, highly dynamic environments and vehicle velocities cause severe fading in both the time and frequency domains. This project evaluates and compares the channel estimation performance across three physical layer (PHY) frame structures: IEEE 802.11p, IEEE 802.11bd-draft, and a Unique-Word (UW)-based PHY. Simulation results demonstrate that the UW-based approach achieves robust, interference-free channel estimation using low-complexity techniques. In contrast, reaching a comparable estimation performance using the 802.11bd standard requires significantly higher computational complexity, making the UW-based PHY a highly efficient alternative for vehicular networks.

Abstract: Next-generation vehicular networks are key enablers for intelligent transportation systems, but highly dynamic environments with time-varying channels and Doppler shifts present significant reliability challenges. Conventional channel estimation methods often struggle in these scenarios, suffering from either impractical computational complexity or a lack of robustness. This research investigates the integration of Deep Learning (DL) to overcome these limitations. By adapting optimized DL architectures for high-mobility vehicular networks, the study demonstrates that data-driven estimators can provide highly accurate, robust, and low-complexity solutions. Ultimately, these DL-based models achieve an optimal trade-off between system performance and computational efficiency, ensuring reliable real-time communication in highly dynamic vehicular environments.