Model-based analysis and optimisation of Wireless Networks
In the last two decades, wireless communication has evolved to one of the most important techniques for exchanging information. Providing much more flexibility and possibilities compared to its wired alternative, applications utilising wireless communication require elaborated conceptions of planning the underlying network to achieve desired operational requirements and ensure reliability. However, realising these objectives requires to exploit network-specific knowledge, e.g., about the environment with its distinct characteristics or the used device hardware and applications, possibly entailing further efforts and costs to obtain, e.g., by performing time-consuming measurement campaigns. Thus, solutions are often required being practical and adapted to a given situation. Goal of this thesis is the development of universal approaches for analysing and improving wireless networks without constraining their applicability by underlying assumptions while usefully exploiting available network information. Breaking down the process of knowledge generation and exploitation into consecutive steps based on the amount of available information allows to provide granular approaches able to adapt to a wide range of scenarios considering their highly individual conditions. Distinguishing and addressing four leverage points, this thesis makes the following contributions to further understand and optimise wireless networks: (1) identifying the network topology and localise its devices, (2) modelling wireless communication in a fast and accurate way, (3) understanding the cause-effect relationship for changing network behaviour and exploit the knowledge for (4) improving the network with regard to specifiable performance metrics. The evaluation was conducted in several real-world scenarios, demonstrating the viability of automated general solutions for network analysis and optimisation at different levels of knowledge, managing to realise universal, network-independent applicability on one side and adaptability to specific systems and environments on the other side. The presented approaches are able to flexibly operate with variable amounts of information, ranging from the necessary minimum up to a large amount of further relevant details, enhancing the quality of results. Conclusively, each contribution of this thesis provides an approach to analyse a different aspect of the network and extract valuable knowledge to exploit. Altogether, they cover the complete range of knowledge acquisition to assist users in gaining insights about their network to ultimately improve the performance of their setup.