Abstract
This PhD thesis focuses on developing novel self organising functionalities in wireless cellular systems. Since the root cause of suboptimal performance in wireless cellular system is the mismatch between its semi-static design and its dynamically changing environment, the main objective of self organising functionalities is to counter measure the effect of this mismatch. Towards this end, we classify wireless cellular system dynamics based on their time scale into three main classes - short, medium and long term dynamics and develop self organising solution suitable for each class. Through investigation of case studies of self organising systems in nature, we first identify the desirable characteristics of self organising systems. By building on these case studies we propose a general framework called Biomimmetic Self Organising Framework (BSOF), for designing adaptive solution in engineering system that bear characteristics of self organisation. First major contribution of this thesis consists of a novel solution to cope with short term dynamics e.g. pop up hot spots. This solution optimizes antenna tilts in distributed manner for system wide spectral efficiency optimization in face of heterogeneous user geographical distributions. The solution is developed analytically by applying BSOF and its performance is evaluated against centralised fixed tilting benchmarks through system level simulations. Results show a 30% improvement in average spectral efficiency along with advantages of a self organising distributed solution i.e. very low signalling overhead, agility and scalability. Second major contribution in this thesis provides a novel solution to cope with medium term dynamics e.g. uneven traffic load among cells. This solution optimises cell load through cell coverage adaptation in distributed manner in order to minimise system wide average call blocking. The analytical framework behind this solution is developed by following the steps of BSOF. The Numerical results show 280% reduction in average blocking probability compared to no load balancing in place. The performance of this distributed solution is also compared against a bench mark of centrafised control based load balancing algorithm. Results show that a performance very close to the centralised solution can be obtained with proposed distributed solution, with added advantages of a distributed solution. Our final major contribution aims for providing a self organising functionality for long term dynamics e.g. demographical and socio economical changes. First we develop a novel framework for long term performance characterisation of wireless system in terms of three key performance indicators i.e. capacity, quality of service and energy efficiency. Then, by following the steps of BSOF, we develop a novel solution for self organisation of frequency reuse and deployment architecture for joint optimization of spectral efficiency, fairness and energy efficiency.