Electronic Thesis and Dissertation Repository


Master of Science


Electrical and Computer Engineering


Wang, Xianbin

2nd Supervisor

Rahman, Quazi



Exponentially growing number of communicating devices and the need for faster, more reliable and secure communication are becoming major challenges for current mobile communication architecture. More number of connected devices means more bandwidth and a need for higher Quality of Service (QoS) requirements, which bring new challenges in terms of resource and traffic management. Traffic offload to the edge has been introduced to tackle this demand-explosion that let the core network offload some of the contents to the edge to reduce the traffic congestion. Device-to-Device (D2D) communication and edge caching, has been proposed as promising solutions for offloading data. D2D communication refers to the communication infrastructure where the users in proximity communicate with each other directly. D2D communication improves overall spectral efficiency, however, it introduces additional interference in the system. To enable D2D communication, efficient resource allocation must be introduced in order to minimize the interference in the system and this benefits the system in terms of bandwidth efficiency. In the first part of this thesis, low complexity resource allocation algorithm using stable matching is proposed to optimally assign appropriate uplink resources to the devices in order to minimize interference among D2D and cellular users.

Edge caching has recently been introduced as a modification of the caching scheme in the core network, which enables a cellular Base Station (BS) to keep copies of the contents in order to better serve users and enhance Quality of Experience (QoE). However, enabling BSs to cache data on the edge of the network brings new challenges especially on deciding on which and how the contents should be cached. Since users in the same cell may share similar content-needs, we can exploit this temporal-spatial correlation in the favor of caching system which is referred to local content popularity. Content popularity is the most important factor in the caching scheme which helps the BSs to cache appropriate data in order to serve the users more efficiently. In the edge caching scheme, the BS does not know the users request-pattern in advance. To overcome this bottleneck, a content popularity prediction using Markov Decision Process (MDP) is proposed in the second part of this thesis to let the BS know which data should be cached in each time-slot. By using the proposed scheme, core network access request can be significantly reduced and it works better than caching based on historical data in both stable and unstable content popularity.