Performance modeling and enhancement for next generation wireless IoT networks

Remote connection of devices to the Internet was once a distant dream, however with the current ubiquity of wireless access networks, remote connection became an integral part of modern intelligent economy. Reliable remote applications and cloud services are becoming more crucial to our daily lives. The overwhelming need for a better-connected world is urging technologists, regulators and researchers to continuously devise better solutions for coping with this accelerating demand. The Internet of Things includes the framework for sensors to communicate over wireless networks with minimal or no human intervention. To cater for these demands, studies have projected that vast numbers of IoT devices are being deployed in a ubiquitous manner connecting billions of small sensors to the Internet. These projections serve many emerging business cases relying on sensors such as utility metering, precision agriculture, and hazard prevention, public health, logistic management, and environmental monitoring. Many of these applications have stringent requirements in terms of low energy-consumption, long distance communications, mobility, and affordability. When the utilized devices rely on batteries as their sole energy source, then energy-efficiency becomes of utmost importance in order to prolong their lifespan. As a result, next generation IoT networks rely on simple spectrum access methods to maintain simplicity and compatibility with limited-capability sensors. With these many benefits of IoT, some fundamental challenges emerge. The first challenge is the scarcity of the radio spectrum where IoT services are likely to underlay the current wireless network with the risk of congesting the already busy sub-GHz band. Thus, spectrum authorities continuously develop certain regulations on the devices operating in such shared spectra in order to harmonize the coexistence of different wireless services. These measures, however, do not completely resolve the coexistence challenge and IoT technologies still need to implement their own intelligent interference mitigation techniques. The second challenge is that the current cellular network technologies are designed for high speed and high-quality communications utilizing complex spectrum access methods, which are not suitable for low power applications. As a result, broadband cellular networks are not wildly tailored for extended battery life and limited throughput requirements. In this research, we address these challenges through analytical modeling supported with field and lab measurements to assess the readiness of the spectrum resources, and explore different techniques to enable next generation IoT cellular networks. To address the first challenge, we build a temporal mathematical model that captures the time correlation present in the shared IoT sub-GHz spectrum. The model is built based on machine learning analysis of actual measurements conducted in campaigns at different urban environments in Melbourne, Australia. Moreover, we derive a spectrum occupancy model that estimates the effective spectrum utilization of common random access methods that are adopted by IoT technologies. The radio spectrum models are utilized to draw future paradigms on the available spectral resources for additional IoT services. To address the second challenge, we derive closed-form analytic models that capture the energy-consumption and outage probability profiles for open access cellular IoT networks. This is achieved by presenting a practical energy-consumption model based on extensive measurements on typical IoT devices. Then, we derive the device energy minimum bounds and assess the contribution of network regularity on the bound by utilizing tools from stochastic geometry. The derived models are applied to real-world Telco deployments to understand their readiness for IoT technologies in terms of energy consumption. Furthermore, we propose an adaptive method to enhance the energy consumption profile of typical IoT devices in a cellular network by varying the different transmission parameters. Some of the developed models in this thesis have already been implemented in the largest open-access IoT network in Victoria, Australia, as part of the Smart Cities and Suburbs program funded by the Australian Government, the network is consisted of 48 gateways and 297 sensors. In this implementation, we tailored the proposed frameworks to suit the practical needs of the network.