Design and Optimization of Ultra-Reliable Low-Latency Communications in Beyond 5G Networks

The fifth generation (5G) and beyond wireless networks mark a pivotal shift in the realm of telecommunications. These advanced networks aim to provide an array of different services, fulfilling the diverse needs of modern-day connectivity. They are expected to provide services with high data rates, large connection density, ultra-low latency, and extraordinary reliability. To achieve these goals, there are three primary service categories in 5G and beyond networks: enhanced mobile broadband (eMBB), massive machine-type communications (mMTC), and ultra-reliable low-latency communications (URLLC). With eMBB, users can communicate with a substantial increase in data rates, enabling swift and high-bandwidth content consumption. On the other hand, mMTC sets the stage for the seamless integration of billions of devices into the network. However, it's URLLC that stands out as the linchpin of these networks, providing unprecedented levels of reliability and ultra-low latency, considering mission-critical applications and real-time responsiveness as the norm. This service is expected to open groundbreaking changes in fields such as healthcare, autonomous vehicles, industrial automation, and beyond. Given the above context, this dissertation focuses on designing effective communication protocols for different URLLC-related systems. In particular, the study delves into three key aspects: (1) Average block error rate (BLER) and minimum blocklength analysis for short-packet communications, a promising transmission method for URLLC; (2) Deep reinforcement learning (DRL)-based resource management strategy for uplink URLLC within the context of grant-free access, an advanced access technology for latency-sensitive dense networks; and (3) Joint optimization and DRL-based resource allocation for harmonious coexistence of diverse services such as eMBB, mMTC, and URLLC. Firstly, we study a promising transmission method for URLLC, namely short packet communications (SPC), to fulfill its stringent requirements. Specifically, we investigate SPC in downlink non-orthogonal multiple access (NOMA) systems using multiple-input multiple-output (MIMO) schemes. The main focus of this work is a comprehensive evaluation of system performance by analyzing the average block error rate (BLER) and minimizing the blocklength to reduce transmission latency. Our findings reveal that MIMO NOMA exhibits the capability to efficiently serve multiple users in a concurrent fashion while employing a lower blocklength in comparison with MIMO Orthogonal Multiple Access (OMA). These results effectively highlight the advantages of MIMO NOMA-based SPC, primarily in its ability to significantly reduce transmission latency. Secondly, we investigate the application of DRL techniques for designing highly efficient resource management solutions in grant-free NOMA (GF-NOMA) systems tailored to meet the stringent demands of URLLC. Our focus centers on maximizing network energy efficiency (EE) and ensuring the fulfillment of URLLC users' specific requirements. The outcomes of our simulations demonstrate that the methods we propose achieve better convergence properties, smaller signaling overhead, and larger network EE than other benchmark methods. Finally, our focus turns to the seamless combination of diverse services including eMBB, mMTC, and URLLC in NOMA-based systems. In this context, we develop an innovative resource management solution applying a joint optimization and cooperative multi-agent DRL approach. The primary goal of this strategy is to maximize network EE for the considered system while adhering to users' diverse demands. Our extensive simulations indicate that our proposed method provides superior performance regarding convergence property and system EE over other considered benchmark methods.