Abstract
In the 5G network, dense deployment of small cells and utilisation of millimetre wave (mmWave) band are some of the key approaches to boost network capacity. In such scenario, however, V2X (Vehicle to Vehicle and Vehicle to Infrastructure) communication systems are required to apply strict mobility management proceedings to reduce delay due to frequent handovers during attachment of different points. Moreover, dense deployment of mmWave small cells using narrow directional beams will escalate the cell and beam related handovers for high mobility of vehicles, which may in turn limits the performance gain promised by 5G-mmWave based vehicle-to infrastructure (V2I) communication. Thus, it is vitally important to suppress the frequent handovers in such networks. Handover reduction mechanisms are proposed in this thesis by identifying long-lasting connections. This thesis has three main contributions.
End-to-end system-level simulations are conducted to evaluate the impact of mobility on performance metrics, such as communication latency and packet loss ratio, in densely deployed small cells Heterogeneous Network (HetNet). To analyse the impact, we develop a windowing mechanism to distinguish the packet transmission performance around the handover period from the whole connection duration to capture the actual impact. Thus, the findings state that the impact of mobility becomes more significant in dense networks due to frequent exposure to cell borders and handovers.
As one of the key promising aspects of 5G for V2I is directional mmWave networks, frequent handovers even become more dominant due to small cell sizes and directional beam coverage in such networks. In this regard, an analytical model is proposed to find the theoretical upper-bound for vehicle sojourn time in a dense mmWave network. The theoretical upper-bound is developed to be employed as a benchmark for the performance of any practical design. Furthermore, we propose a Fuzzy Logic (FL) based beam-centric distributed algorithm to determine the beam among all visible beams where a vehicle can achieve the longest displacement within it. In this regard, we consider a densely deployed mmWave network and propose a handover decision to minimise the chance of handover events by selecting the beam offering the longest sojourn time when a vehicle travels across the network. State-of-the-art schemes, such as the strongest signal-based handover and Sticky Handover schemes are used to show the effectiveness of the proposed systems.
With the emerging popularity of Machine Learning (ML) and Artificial Intelligence (AI) in cellular networks, a Deep Q-Network based beam-centric handover decision (DQN-BD) method is proposed. The DQN-BD adaptively finds optimal cell and beam to maximize effective beam connection time while minimizing handover related cost and beam misalignment time. Different from the previous work, the analytical model is revisited for a specific vehicular environment on the road section of highway with dynamic beam direction and extended to consider blockage probability.