With the ever continuing increase in the number of devices connecting to one another, forming the Internet of Things (IoT), understanding how the performance of these networks connect is of paramount importance.Modelling the location of devices/base stations by random sets, tools from stochastic geometry can be leveraged creating a framework which is far more tractable than previous lattice models. As such there has been significant progress into the fundamental capabilities of wireless networks, including connectivity and scalability. To date, the majority of literature has presumed a uniform Poisson Point Process in R2 to model the distribution of points which represent base stations or smart devices. Although very tractable, the model loses some important features that heavily influence real world net-works. For example, within this model there is a notion of a “typical user”,which is that the network performance (such as the probability you can send a signal) is independent of location. Naturally, this contradicts what we as consumers experience in reality, since connectivity varies due to base stations being far away, or buildings/hills etc causing signal degradation.
Furthermore, as the IoT continues to grow, the network becomes increasingly dynamic due to portable smart devices requiring connectivity. Motivated by this, we study the impact boundaries ( examples are geographical features like the body of water surrounding New York city) and human mobility have on the connectivity properties of wireless networks, both over single and multiple time slots. To achieve this, we use a Poisson Point Process with non-uniform measure in some finite domain and model the links between points as probabilistic connection function. Through this general model, referred to as a Soft Random Geometric Graph, we are able to study a range of different networks even within the wireless communication literature. These include the classical device to base station architecture, to device-to-device networks where information is relayed in a multihop fashion, to ultra dense5G networks where the base stations may themselves be mobile, for instance as drones. We later extend our work to spatio-temporal networks where we analyse how these graphs evolve over multiple time slots. Our results there-fore have a variety of applications ranging from the routing information and scalability of mobile ad hoc networks to how 5G networks should be deployed in urban environments to maximise user experience.
|Date of Award||25 Jun 2019|
|Supervisor||Carl P Dettmann (Supervisor), Woon Hau Chin (Supervisor) & Orestis Georgiou (Supervisor)|
- Wireless Networks
- Stochastic Geometry