Context-Aware Proactive Optimisation in Cellular Networks

5G and beyond networks are expected to meet the exponential traffic growth and fast-changing environments. Time-efficiency decides whether the 5G and beyond network optimisation can absorb the traffic growth and ensure a low latency simultaneously. These requirements bring challenges for operators to remain profitable while reducing operational expenditure. This thesis aims to improve the time efficiency by designing a more intelligent and user-oriented network-optimisation framework which is denoted as Context-Aware Proactive Optimisation (CAPO). This thesis quantifies the improved time-efficiency in three research lines: 1) aerial base stations (BSs) deployment and user association, 2) aerial BSs aided network off-loading, and 3) proactive load balancing. All of them share common characteristics of limited serving time and high computational complexity, so their performance becomes sensitive to the complexity. This thesis manages to keep the real-time complexity at a low level.
Firstly, Unmanned Aerial Vehicles (UAVs) are ideal carriers to substitute the terrestrial BSs and associate the ground users temporally. However, the UAV assisted BSs (UAV-BSs) deployment is a non-deterministic polynomial-time hard (NP-hard) problem that is difficult to be solved with time-efficiency. This thesis proposes a CAPO-based deployment strategy to solve this problem by reserving time-efficiency and energy-efficiency. The results indicate that problem-solving efficiency is improved at least ten times.
Secondly, fast deploying UAV-BSs will off-load the terrestrial overloaded BSs. Nevertheless, it still faces the time-efficiency problem because of jointly optimising multiple objectives, such as UAV-BSs' amount, locations, and allocating resource blocks. This thesis transforms the above joint optimisation problem into a combinatorial problem and uses a CAPO-aided heuristic algorithm to solve it with both time-efficiency and robustness. In this result, under a time constraint, my design could finish 30% more optimisation compared with non-CAPO ones.
Lastly, the terrestrial nodes should own the ability to balance their overload to neighbouring idle cells. However, existing load balancing algorithms need more time to react to the intense traffic changes. This shortage leads to cold-start problems, which cause slower convergence speed and lower time-efficiency. This thesis employs CAPO to enable event detection from social networks and prepare the capacity to absorb an upcoming demand peak. The results indicated CAPO's ability to make the load balancing converge with eliminated overshoot.
In 5G and beyond networks, the optimisation will be proactive, service-oriented, and user-oriented. The CAPO approach presented in this thesis becomes an indispensable path to increase the quality of experience and reduce OPEX. This thesis's wider impact includes better cross-fertilising the academic fields of data analytics, mobile edge computing, artificial intelligence, and wireless communications, as well as informing the industry of the promising potentials in this area.