Self-organised multi-objective network clustering for coordinated communications in future wireless networks

The fifth generation (5G) cellular system is being developed with a vision of 1000 times more capacity than the fourth generation (4G) systems to cope with ever increasing mobile data traffic. Interference mitigation plays an important role in improving the much needed overall capacity especially in highly interference-limited dense deployment scenarios envisioned for 5G. Coordinated multi-point (CoMP) is identified as a promising interference mitigation technique where multiple base stations (BS) can cooperate for joint transmission/reception by exchanging user/control data and perform joint signal processing to mitigate inter-cell interference and even exploit it as a useful signal. CoMP is already a key feature of long term evolution-advanced (LTE-A) and envisioned as an essential function for 5G. However, CoMP cannot be realized for the whole network due to its computational complexity, synchronization requirement between coordinating BSs and high backhaul capacity requirement. BSs need to be clustered into smaller groups and CoMP can be activated within these smaller clusters. This PhD thesis aims to investigate optimum dynamic CoMP clustering solutions in 5G and beyond wireless networks with massive small cell (SC) deployment. Truly self-organised CoMP clustering algorithms are investigated, aiming to improve much needed spectral efficiency and other network objectives especially load balancing in future wireless networks. Low complexity, scalable, stable and efficient CoMP clustering algorithms are designed to jointly optimize spectral efficiency, load balancing and limited backhaul availability. Firstly, we provide a self organizing, load aware, user-centric CoMP clustering algorithm in a control and data plane separation architecture (CDSA) proposed for 5G to maximize spectral efficiency and improve load balancing. We introduce a novel re-clustering algorithm for user equipment (UE) served by highly loaded cells and show that unsatisfied UEs due to high load can be significantly reduced with minimal impact on spectral efficiency. Clustering with load balancing algorithm exploits the capacity gain from increase in cluster size and also the traffic shift from highly loaded cells to lightly loaded neighbours. Secondly, we develop a novel, low complexity, stable, network-centric clustering model to jointly optimize load balancing and spectral efficiency objectives and tackle the complexity and scalability issues of user-centric clustering. We show that our clustering model provide high spectral efficiency in low-load scenario and better load distribution in high-load scenario resulting in lower number of unsatisfied users while keeping spectral efficiency at comparably high levels. Unsatisfied UEs due to high load are reduced by $68.5\%$ with our algorithm when compared to greedy clustering model. In this context, the unique contribution of this work that it is the first attempt to fill the gap in literature for multi-objective, network-centric CoMP clustering, jointly optimizing load balancing and spectral efficiency. Thirdly, we design a novel multi-objective CoMP clustering algorithm to include backhaul-load awareness and tackle one of the biggest challenges for the realization of CoMP in future networks i.e. the demand for high backhaul bandwidth and very low latency. We fill the gap in literature as the first attempt to design a clustering algorithm to jointly optimize backhaul/radio access load and spectral efficiency and analyze the trade-off between them. We employ 2 novel coalitional game theoretic clustering methods, 1-a novel merge/split/transfer coalitional game theoretic clustering algorithm to form backhaul and load aware BS clusters where spectral efficiency is still kept at high level, 2-a novel user transfer game model to move users between clusters to improve load balancing further. Stability and complexity analysis is provided and simulation results are presented to show the performance of the proposed method under different backhaul availability scenarios. We show that average system throughout is increased by 49.9% with our backhaul-load aware model in high load scenario when compared to a greedy model. Finally, we provide an operator's perspective on deployment of CoMP. Firstly, we present the main motivation and benefits of CoMP from an operator's viewpoint. Next, we present operational requirements for CoMP implementation and discuss practical considerations and challenges of such deployment. Possible solutions for these experienced challenges are reviewed. We then present initial results from a UL CoMP trial and discuss changes in key network performance indicators (KPI) during the trial. Additionally, we propose further improvements to the trialed CoMP scheme for better potential gains and give our perspective on how CoMP will fit into the future wireless networks.