Abstract

Next-generation wireless networks (Beyond 5G, 6G) aim to provide tremendous improvements over previous generations by promising a massive connectivity, ultra-reliable and low-latency communications, and soaring broadband speeds. Such transformation will give rise to a wide range of propitious Internet-of-Things (IoT) applications such as intelligent transportation systems (ITS), tactile internet, augmented/virtual reality, industry 4.0, etc. These applications possess stringent requirements of fresh and timely information updates to make critical decisions. Out-dated or stale information updates are highly undesirable for these applications as they may call forth unreliable or erroneous decisions. The conventional performance metrics such as delay and latency may not fully characterize the freshness of information for time-critical IoT applications. Recently, information freshness has been investigated through defining a new performance metric termed as Age of Information (AoI). AoI offers a rigorous way to quantify the information freshness as compared to other performance metrics and is deemed suitable for real-time IoT applications.
In reality, the limited energy and computing resources of IoT devices (IoTDs) is a significant challenge towards realizing the timely delivery of information updates. To address this challenge, the first aim of this dissertation is to examine the capability of multi-access edge computing (MEC) towards minimizing the AoI. In fact, MEC offers an expedited computation of resource-intensive tasks, which, if processed locally at the IoTDs, may experience excessive computational latency. In this context, an optimization problem is setup to determine the optimal scheduling policy with the goal of minimizing the expected sum AoI of multiple IoTDs, while considering the combined impact of unreliable channel conditions and random packet arrivals.
Another acute challenge is the high randomness and uncontrollable behaviour of wireless communication environments, which may severely impede the timely and reliable delivery of information updates. Towards addressing this challenge, reconfigurable intelligent surface (RIS) is leveraged to mitigate the propagation-induced impairments of the wireless environment and enhance the quality of wireless links to preserve the information freshness. First, a wireless network consisting of a base station (BS) that is forwarding information updates of multiple real-time traffic streams to their destinations is studied. The considered multiple access technique is frequency division multiple access (FDMA), which is an orthogonal multiple access (OMA) technique. A joint user scheduling and phase-shift matrix (passive beamforming) optimization problem is formulated with the objective of minimizing the expected sum AoI of the coexisting multiple traffic streams. The resulting problem is a mixed integer non-convex optimization problem. To evade the high coupling of the invoked optimization variables, the bi-level optimization technique is utilized, where the original problem is decomposed into an outer traffic stream scheduling problem and an inner RIS phase-shift matrix problem. Owing to the stochastic nature of packet arrivals, a deep reinforcement learning (DRL) solution is employed to solve the outer problem. To do so, the traffic stream scheduling is modeled as a Markov Decision Process (MDP) and Proximal Policy Optimization (PPO) is invoked to solve it. On the other hand, the inner problem that determines the RIS configuration is solved through semi-definite relaxation (SDR).
Due to the limitations of OMA techniques in terms of the number of served IoTDs and the spectral efficiency, the focus of this dissertation shifts to explore non-orthogonal multiple access (NOMA) scheme towards achieving the goal of minimizing the AoI in an uplink setting. In this context, an optimization problem is formulated to optimize the RIS configuration, the transmit power of IoTDs and their clustering policy. To solve this mixed-integer non-convex problem, the RIS configuration is obtained first by resorting to difference-of-convex (DC) along with successive convex approximation (SCA). On the other hand, the bi-level optimization is used to solve the power allocation and the clustering problems. Optimal closed-form expressions are derived for the power control scheme and the one-to-one matching is employed to solve the clustering problem. Aiming to further improve the information freshness in time-critical IoT applications, an extended version of NOMA, termed as Cooperative-NOMA (C-NOMA), is adopted. In C-NOMA, the cooperation between IoTDs through device-to-device (D2D) communication and full-duplex (FD) relaying is invoked within the NOMA scheme. In this context, the integration of RIS and C-NOMA is investigated towards achieving the goal of minimizing the average sum AoI. Precisely, it is investigated how much performance gain in terms of AoI reduction can be brought by the RIS-enabled uplink C-NOMA system compared to the conventional C-NOMA and NOMA schemes, both with and without RIS. Results elucidate the superiority of our proposed approaches against other baseline schemes. The findings in this dissertation shed light on the choice of effective design of wireless communication networks leveraging the core future enabling technologies.