Resource allocation and incentive mechanism design in next generation wireless communication and mobile computing networks

Abstract

In light of the broader vision for digitalization and smart-X vertical industries (e.g., smart city, healthcare, manufacturing), an unprecedented increase of data-hungry and compute-intensive applications takes place, imposing stringent communication and computing requirements on the network. As a means of extending connectivity, increasing spectral and energy efficiency, and reducing latency, several prominent technologies and architectural paradigms emerge as key enablers of the Next Generation (NextG) 5G wireless networks. In this context, the simplistic adoption of existing solutions to traditional resource management problems is inefficient in exploring and exploiting the network’s capabilities. On the contrary, several degrees of freedom should be concurrently determined about the nature and type of the resources to be allocated while accounting for the multifaceted competition between different stakeholders. In this thesis, we tackle the problem of efficient resource allocation in heterogeneous and multi-tier wireless communication and mobile computing networks in a holistic, though distributed, manner. In the direction of self-organizing networks, where each network entity makes autonomous decisions regarding its resource utilization and allocation, we develop novel frameworks that capture the interdependencies between different network entities’ behaviors, interactions, and decisions by accounting for their different and/or conflicting objectives and the existence of incompleteness of information between them. To provide such a real-life spirit in modeling the resource management problems, we resort to Game Theory and Contract Theory, which result in low-complexity methodologies and algorithms for solving the formulated problems. First, multi-variable resource allocation problems are studied in heterogeneous and multitier wireless communication networks. Targeting to address the problem of incomplete/partial Channel State Information (CSI) from the Base Stations’ (BSs’) behalf, as well as the user-BS conflict regarding the users’ uplink transmission power investment, a synergistic approach based on the principles of Contract Theory is initially introduced. The BSs design a menu of contracts based on their statistical knowledge about the users’ CSI, which comprise indicative power levels along with a corresponding reward, such that the designed power levels enable the decoding of the users’ transmitted signals at the BSs’ receivers via the combination of Non-Orthogonal Multiple Access (NOMA) and Successive Interference Cancellation (SIC). The users autonomously select the one contract out of the menu that best fits their experienced channel conditions while employing a Reinforcement Learning (RL) algorithm to distributively determine the most beneficial BS association. Taking one step further to the underlying network architecture and employed technologies, we aim to concurrently account for and properly configure the resources across the wireless access and backhaul network parts of an Unmanned Aerial Vehicle (UAV)-assisted and Reconfigurable Intelligent Surface (RIS)-aided communication network. By capitalizing on the structural hierarchy between the UAV-mounted BS and the users, a Stackelberg game is formulated to distributively deal with a highly combinatorial resource management problem. The UAV, acting as leader, determines the RIS’s phase shifts such that the sum users’ signal strength is maximized in the uplink and, in a second phase, jointly calculates the bandwidth and uplink transmission power allocation in the backhaul link to the core network. In the third phase, the users, acting as the followers, optimize their uplink transmission powers to the UAV in a distributed manner. The second and third phases are iteratively repeated to conclude with the game’s Stackelberg equilibrium point, at which the end-to-end energy efficiency is maximized. Respecting the need for converged radio and computing resource allocation frameworks within the context of NextG 5G networks, a multi-tier mobile computing topology is considered, and the joint problem of computation task offloading and uplink transmission/offloading power allocation is studied. In contrast to the existing literature, where single-tier computing networks are modeled, consisting of an edge, a fog, or a cloud service layer, concurrently utilizing a wide range of computing capabilities and options is pursued. Given the users’ selfish behavior towards offloading to the edge due to the latter’s proximity, an incentivization mechanism based on Contract Theory is designed to motivate them to allow a percentage of their initially offloaded tasks to the edge to be further forwarded and processed at the fog. Having determined this percentage, the ultimate users’ computation task offloading and uplink transmission power to edge are jointly derived in a distributed manner by playing a game among them. By utilizing multiple computing service layers, an energy-efficient solution is derived while extending the edge service layer’s computing capacity. Finally, aiming to further study the distribution of computation tasks horizontally, i.e., within the same computing tier, while considering multiple servers, a multi-server Multi-Access Edge Computing (MEC) network is modeled. The users leverage the different available Radio Access Networks (RANs) nearby to offload their compute-intensive and latency-critical applications to multiple MEC servers simultaneously. To address the critical interference management problem under the resulting multi-user multi-server network topology while being motivated by the advancements in the non-orthogonal multiple access techniques, the application of Rate-Splitting Multiple Access (RSMA) is scrutinized. In this context, the minimization of the sum of the users’ maximum experienced delay among the different MEC servers is pursued by jointly optimizing their computation task offloading ratios, rate, uplink transmission power, and computing resource allocation across the offloading MEC servers. The formulated joint optimization problem is analytically solved using conventional optimization techniques based on the Karush-Kuhn-Tucker (KKT) conditions. This complements the ultimate purpose of this thesis to provide a holistic approach and view to the converged wireless communication and mobile computing networks. The considered resource management problems are thoroughly evaluated via modeling and simulation. The proposed frameworks’ pure operational characteristics and the designed algorithms’ convergence behavior are examined under different scenarios and values, while comparative results against other benchmarking and state-of-the-art solutions are provided to demonstrate the proposed frameworks’ effectiveness and efficiency.