Handover decision-making schemes in UAV communication networks

In recent years, unmanned aerial vehicles (UAVs) have seen increasing application across various industries, including military operations, surveillance, telecommunications, delivery services, and rescue missions. Their ability to maneuver in three-dimensional space with high agility and their ease of deployment makes them highly adaptable for a wide range of tasks. In communication networks, UAVs can serve two primary roles: as aerial user equipment (UE) in cellular networks, known as cellular-connected UAVs, or as aerial base stations (BSs) to provide reliable, cost-effective, and on-demand wireless communication services. When acting as aerial BSs, UAVs enhance the existing terrestrial wireless network by increasing the likelihood of line-of-sight (LoS) communication links to ground UEs, thereby improving connectivity.

However, in both roles, handover (HO) is critical for maintaining connectivity as user associations change. While HO improves the quality of service (QoS) for mobile UEs, frequent HOs can lead to increased signaling overhead and a higher call drop rate. Given that modern terrestrial networks often consist of BSs with various access technologies—particularly in 5G ultra-dense small cell deployments—combined with the fast movement of UAVs and ground UEs, frequent HOs can easily occur. Therefore, reducing the number of HOs is essential to ensure efficient network performance. Additionally, in different scenarios, the purpose of HO goes beyond merely maintaining a high-quality link. For instance, in UAV-aided networks or UAV-aided multi-access edge computing (MEC) networks, other metrics like energy efficiency and computing resource availability play a crucial role in the HO process. By making appropriate HO decisions, these metrics can be optimised to achieve better overall service performance. This thesis proposes various HO decision-making schemes tailored to different UAV network scenarios.

Firstly, for cellular-connected UAVs, they rely on reliable communication links to complete tasks. As aerial UEs in communication network, UAVs suffer high interference and highly frequent unnecessary HO when they are moving and served by the ground communication network system. To address these issues, we proposed an improved Q-Learning (QL) based HO algorithm. The formation of the action space considers both the Signal-to-Interference-plus-noise ratio (SINR) and the distance between the BSs and UAV with different weights. We defined performance metrics to evaluate the algorithm, which are HO rate, throughput loss rate, and outage rate. The results show the proposed algorithm can further reduce the HO rate by 52.9% with slightly compromising the throughput and the outage rate. Furthermore, the optimal weight combination of ωsinr and ωd is found to be ωd = 4.75 and ωsinr = 5.25 are suggested to achieve the best trade-off of the three performance indicators.

Secondly, in the heterogeneous network (HetNet), the deployment of UAV can also cause problems. For example, in UAV-aided HetNet, the number of HO and the number of unnecessary HO will increase because of the dense distribution of small base stations (SBSs) and UAVs. Frequent HO and unnecessary HO can result in interruption, increased overhead and energy consumption, which is not desirable for battery powered UAVs. To solve the problem, a QL based HO decision-making algorithm is proposed with the aim to reduce HO number and improve energy efficiency (EE). In the algorithm, QL can be applied to address decision-making challenges in communication systems widely. However, a large volume of training data can pose challenges and complexities, therefore, in this algorithm, the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) is utilised to reduce the size of the action space in QL. The proposed hybrid TOPSIS-QL method enhances both the HO performance and the scalability. In this method, SINR, time of stay (ToS) and average EE are taken into account. The simulation results show that the number of HO and unnecessary HO is remarkably reduced by around 48.7% and 69.2% respectively, and the average EE is notably improved by around 46.1% in comparison with the baseline.

Finally, HO decision-making problems also happen in UAV-aided MEC systems. The rapid growth of smart devices and the increasing demand for data computational tasks in 5G-based dense cellular networks result in frequent HO and task migration. This leads to decreased quality of computing and increased energy consumption. A Fuzzy Inference System (FIS)-based HO decision-making scheme for UAV-aided MEC is proposed to address these challenges. Our objectives are to reduce the number of HOs for UEs, decrease energy consumption for BSs, and increase the efficiency of task processing. To achieve this, the FIS is divided into two stages: (i) HO decision and (ii) BS selection. In the HO decision stage, criteria such as the distance between the UE and the serving BS, receive signal strength (RSS), allocated CPU frequency, and task time threshold are considered to determine whether to execute an HO. In the BS selection stage, the system evaluates the distance between the UE and the target BS, the distance between the target BS and the serving BS, the number of connected UEs, the SINR, and the ToS to select the optimal BS for the UE to connect with. Simulation results demonstrate that the proposed HO scheme achieves better performance in reducing task completion time by 38.6%, increasing the number of successful tasks by 74.1%, and reducing the energy consumption by 95.6%, while also reducing the number of HOs by 84.8% in the comparison with baseline.