Multimedia Internet of Things: Game Theoretic Approaches for Distributed Resource Allocation

Abstract

Multimedia Internet of Things (IoT) is an emerging paradigm enabling devices to form an interconnected network of all types of communication. Multimedia IoT is integral to future wireless networks, but there are many challenges that must be addressed, including: network modelling; resource management; massive scale; interference management; clustering; and user quality-of-experience. This thesis focuses on three key enabling technologies of multimedia IoT: (i) device-to-device (D2D) communications; (ii) machine-to-machine (M2M) communications; and (iii) unmanned aerial vehicles (UAVs) as low flying base stations (BSs). Using game theory, we develop novel solutions for decision making processes required to address the key challenges and technologies of multimedia IoT, where interactions between users competing for resources are modelled in a fully distributed and autonomous manner. In the first half of this thesis, we study game theoretic approaches for resource allocation in underlaid D2D communications. First, we propose a flexible application-driven resource allocation scheme, to enhance and improve D2D user quality-of-experience and reliability. We propose a multiple objective Stackelberg game using a non-scalarised approach to enable flexible resource allocation, by coordinating and reducing the effects of intra-cell interference, while ensuring D2D user quality-of-experience. We demonstrate that best social welfare is guaranteed across all D2D users. Next, to further reduce intra-cell interference in underlaid D2D networks, we jointly optimise D2D mode selection, resource block allocation, and interference management. In existing D2D mode selection techniques, the BS assists D2D pairs in selecting a transmission mode, requiring complete channel state information (CSI), i.e., a network-assisted scenario. We propose both a non-cooperative game and a coalitional game to solve this joint optimisation problem in a network-assisted scenario. Next, we investigate a user-assisted approach as a viable D2D mode selection solution, where the BS has partial CSI. Hence, we extend the coalitional game to consider the user-assisted approach, while also coordinating intra-cell interference, scheduling D2D pairs, and allocating resource blocks efficiently. Extensive simulations validate the effectiveness of the proposed coalitional game in a user-assisted scenario. In the second half of this thesis, we explore two emerging topics in M2M communications: (i) correlation-aware clustering in dense M2M networks; and (ii) UAVs as low flying BSs. We first propose a clustering algorithm to cluster a massive number of machine-type devices based on data correlation, which reduces the number of redundant bits being sent to the BS. To solve this, we propose an evolutionary game, and derive a novel utility function that captures the average machine-type device transmission power per cluster, using stochastic geometry. The proposed algorithm converges to a stable cluster formation, which is robust to stochastic changes in the M2M network environment. Finally, we investigate an energy-aware scheduling scheme for mission critical M2M communications, utilising UAVs as low flying BSs. To solve this, we combine Lyapunov optimisation and a one-to-many matching game, where UAVs schedule themselves to collect sensed data from aggregators, to satisfy the ultra-reliable and low-latency constraint, whereas aggregators aim to decrease their energy consumption. Two-sided stable matching and stable power allocation is demonstrated for the proposed algorithm.