| dc.description.abstract | The future Sixth-Generation (6G) network is anticipated to extend connectivity for millions of Unmanned Aerial Vehicles (UAVs) worldwide and support various innovative use cases, such as cargo transport, inspection, and intelligent agriculture. The terrestrial cellular networks provide real-time information exchange between UAVs and Ground Control Stations (GCS), which facilitates the evolution of UAV communication systems while bringing promising economic benefits to cellular network operators. However, the tremendous growth in the UAV data traffic, with diverse and stringent service requirements, would add another pressure on the already congested terrestrial cellular network that is facing a rigorous challenge to increase network capacity with the limited spectrum resources. Moreover, since Macro Base Station (MBS) antennas are typically downtilt, UAVs, which are served by the MBS antenna’s side lobes, suffer from sharp signal fluctuations causing throughput reduction and coverage drop. Besides, due to the Line-of-Sight (LoS) between UAVs and MBSs, UAVs experience higher uplink/downlink interference compared to ground Cellular Users (CUs). In this thesis, we propose two novel aerial network architectures in which we design efficient interference and resource management strategies to support the UAV Quality-of-Service (QoS) guarantee while considering different types of interference. Firstly, we propose a novel standalone aerial multi-cell network where multiple UAV Base Stations (UAV-BSs) provide cellular services to UAV Users by reusing the licensed and unlicensed spectrum. Our objective is to jointly optimize the subchannels and power allocations of UAV-Users in the licensed and unlicensed spectrum to maximize the network uplink sum rate, considering inter-cell interference, co-existence with terrestrial cellular and WiFi systems, and the QoS of UAV-Users. We prove mathematically that the formulated optimization problem is an NP-hard problem. Therefore, the original problem is decomposed into three subproblems to solve it efficiently. We first use convex optimization and the Hungarian algorithm to obtain the global optimal of power and subchannel allocations in the licensed spectrum, respectively. Then, we design a matching game with externalities and coalition game algorithms to obtain the Nash stable of the subchannel allocation in the unlicensed band. Local optimal power assignment in the unlicensed spectrum is obtained using the successive convex approximation method. Lastly, we develop an iterative algorithm to solve the three subproblems sequentially until convergence is reached. Simulation results demonstrate that the proposed algorithm achieves a significantly higher uplink sum rate compared with other resource allocation schemes. Moreover, the proposed algorithm improves the network throughput and capacity by nearly two times comparing to the Long Term Evolution-Advanced (LTE-A). Secondly, we propose a novel integrated aerial-terrestrial multi-operator network. In the network, each operator deploys a number of UAV-BSs besides the terrestrial MBS, where each BS reuses the operator’s licensed spectrum to provide downlink connectivity for UAV-Users. Moreover, the operators allow the UAV-Users, whose demand cannot be satisfied by the licensed band, to compete with others to obtain bandwidth from the unlicensed spectrum. Given the QoS requirements of UAV-Users, we aim to maximize the total sum rate by jointly optimizing user association, BSs transmit power, and dynamic spectrum allocation considering inter-cell interference in the licensed band and inter-operator interference in the unlicensed spectrum. In particular, we divide the resulting non-convex Mixed-Integer Non-Linear Programming (MINLP) optimization problem into two sequential subproblems: user association and power control in the licensed spectrum; and dynamic spectrum allocation and user association in the unlicensed spectrum. Furthermore, the former subproblem is decomposed into multiple subproblems for distributed and parallel problem-solving. Since the resulting former subproblem is still a non-convex MINLP problem, we propose a distributed iterative algorithm consisting of a matching game, coalition game, and successive convex approximation technique to solve it. Afterwards, in the latter subproblem, we first use a matching game to associate UAV-Users with the UAV-BSs for each operator in the unlicensed spectrum. Then, we propose a three-layers auction algorithm to allocate the unlicensed spectrum among operators dynamically. Extensive simulation results demonstrate that the proposed algorithm in the licensed spectrum significantly improves network throughput per operator than the conventional terrestrial network alone. Moreover, the achieved system throughput of the proposed algorithms in both licensed and unlicensed spectrum is 86.8% higher compared with that of using the licensed spectrum only. In summary, we have proposed integrated aerial-terrestrial network architectures that leverage the aerial network to complete the terrestrial network to serve cellular-connected UAVs by reusing licensed and unlicensed spectrum considering multi-cell and multi-operator scenarios. Under the proposed network architectures, we have investigated the subchannel allocation, UAV-Users’ transmit power, user association, BSs’ transmit power, and dynamic spectrum management to maximize the network throughput considering the QoS of UAV-User. The proposed architectures and algorithms should provide valuable guidelines for future research in designing resource and interference management schemes, improving network capacity, and enhancing spectrum utilization for complex interference environments in integrated UAV-cellular networks. | en |
