Towards efficient support for massive Internet of Things over cellular networks
The usage of Internet of Things (IoT) devices over cellular networks is seeing tremendous growth in recent years, and that growth in only expected to increase in the near future. While existing 4G and 5G cellular networks offer several desirable features for this type of applications, their design has historically focused on accommodating traditional mobile devices (e.g. smartphones). As IoT devices have very different characteristics and use cases, they create a range of problems to current networks which often struggle to accommodate them at scale. Although newer cellular network technologies, such as Narrowband-IoT (NB-IoT), were designed to focus on the IoT characteristics, they were extensively based on 4G and 5G networks to preserve interoperability, and decrease their deployment cost. As such, several inefficiencies of 4G/5G were also carried over to the newer technologies. This thesis focuses on identifying the core issues that hinder the large scale deployment of IoT over cellular networks, and proposes novel protocols to largely alleviate them. We find that the most significant challenges arise mainly in three distinct areas: connection establishment, network resource utilisation and device energy efficiency. Specifically, we make the following contributions. First, we focus on the connection establishment process and argue that the current procedures, when used by IoT devices, result in increased numbers of collisions, network outages and a signalling overhead that is disproportionate to the size of the data transmitted, and the connection duration of IoT devices. Therefore, we propose two mechanisms to alleviate these inefficiencies. Our first mechanism, named ASPIS, focuses on both the number of collisions and the signalling overhead simultaneously, and provides enhancements to increase the number of successful IoT connections, without disrupting existing background traffic. Our second mechanism focuses specifically on the collisions at the connection establishment process, and used a novel approach with Reinforcement Learning, to decrease their number and allow a larger number of IoT devices to access the network with fewer attempts. Second, we propose a new multicasting mechanism to reduce network resource utilisation in NB-IoT networks, by delivering common content (e.g. firmware updates) to multiple similar devices simultaneously. Notably, our mechanism is both more efficient during multicast data transmission, but also frees up resources that would otherwise be perpetually reserved for multicast signalling under the existing scheme. Finally, we focus on energy efficiency and propose novel protocols that are designed for the unique usage characteristics of NB-IoT devices, in order to reduce the device power consumption. Towards this end, we perform a detailed energy consumption analysis, which we use as a basis to develop an energy consumption model for realistic energy consumption assessment. We then take the insights from our analysis, and propose optimisations to significantly reduce the energy consumption of IoT devices, and assess their performance.