| dc.description.abstract | Equipped with communication modules, drones can perform as drone-cells (DCs) that provide on-demand communication services to users in various scenarios, such as traffic monitoring, Internet of things (IoT) data collections, and temporal communication provisioning. As the aerial relay nodes between terrestrial users and base stations (BSs), DCs are leveraged to extend wireless connections for uncovered users of radio access networks (RAN), which forms the drone-assisted RAN (DA-RAN). In DA-RAN, the communication coverage, quality-of-service (QoS) performance and deployment flexibility can be improved due to the line-of-sight DC-to-ground (D2G) wireless links and the dynamic deployment capabilities of DCs. Considering the special mobility pattern, channel model, energy consumption, and other features of DCs, it is essential yet challenging to design the flying trajectories and resource allocation schemes for DA-RAN. In specific, given the emerging D2G communication models and dynamic deployment capability of DCs, new DC deployment strategies are required by DA-RAN. Moreover, to exploit the fully controlled mobility of DCs and promote the user fairness, the flying trajectories of DCs and the D2G communications must be jointly optimized. Further, to serve the high-mobility users (e.g. vehicular users) whose mobility patterns are hard to be modeled, both the trajectory planning and resource allocation schemes for DA-RAN should be re-designed to adapt to the variations of terrestrial traffic. To address the above challenges, in this thesis, we propose a DA-RAN architecture in which multiple DCs are leveraged to relay data between BSs and terrestrial users. Based on the theoretical analyses of the D2G communication, DC energy consumption, and DC mobility features, the deployment, trajectory planning and communication resource allocation of multiple DCs are jointly investigated for both quasi-static and high-mobility users.
We first analyze the communication coverage, drone-to-BS (D2B) backhaul link quality, and optimal flying height of the DC according to the state-of-the-art drone-to-user (D2U) and D2B channel models. We then formulate the multi-DC three-dimensional (3D) deployment problem with the objective of maximizing the ratio of effectively covered users while guaranteeing D2B link qualities. To solve the problem, a per-drone iterated particle swarm optimization (DI-PSO) algorithm is proposed, which prevents the large particle searching space and the high violating probability of constraints existing in the pure PSO based algorithm. Simulations show that the DI-PSO algorithm can achieve higher coverage ratio with less complexity comparing to the pure PSO based algorithm.
Secondly, to improve overall network performance and the fairness among edge and central users, we design 3D trajectories for multiple DCs in DA-RAN. The multi-DC 3D trajectory planning and scheduling is formulated as a mixed integer non-linear programming (MINLP) problem with the objective of maximizing the average D2U throughput. To address the non-convexity and NP-hardness of the MINLP problem due to the 3D trajectory, we first decouple the MINLP problem into multiple integer linear programming and quasi-convex sub-problems in which user association, D2U communication scheduling, horizontal trajectories and flying heights of DBSs are respectively optimized. Then, we design a multi-DC 3D trajectory planning and scheduling algorithm to solve the sub-problems iteratively based on the block coordinate descent (BCD) method. A k-means-based initial trajectory generation scheme and a search-based start slot scheduling scheme are also designed to improve network performance and control mutual interference between DCs, respectively. Compared with the static DBS deployment, the proposed trajectory planning scheme can achieve much lower average value and standard deviation of D2U pathloss, which indicate the improvements of network throughput and user fairness.
Thirdly, considering the highly dynamic and uncertain environment composed by high-mobility users, we propose a hierarchical deep reinforcement learning (DRL) based multi-DC trajectory planning and resource allocation (HDRLTPRA) scheme for high-mobility users. The objective is to maximize the accumulative network throughput while satisfying user fairness, DC power consumption, and DC-to-ground link quality constraints. To address the high uncertainties of environment, we decouple the multi-DC TPRA problem into two hierarchical sub-problems, i.e., the higher-level global trajectory planning sub-problem and the lower-level local TPRA sub-problem. First, the global trajectory planning sub-problem is to address trajectory planning for multiple DCs in the RAN over a long time period. To solve the sub-problem, we propose a multi-agent DRL based global trajectory planning (MARL-GTP) algorithm in which the non-stationary state space caused by multi-DC environment is addressed by the multi-agent fingerprint technique. Second, based on the global trajectory planning results, the local TPRA (LTPRA) sub-problem is investigated independently for each DC to control the movement and transmit power allocation based on the real-time user traffic variations. A deep deterministic policy gradient based LTPRA (DDPG-LTPRA) algorithm is then proposed to solve the LTPRA sub-problem. With the two algorithms addressing both sub-problems at different decision granularities, the multi-DC TPRA problem can be resolved by the HDRLTPRA scheme. Simulation results show that 40% network throughput improvement can be achieved by the proposed HDRLTPRA scheme over the non-learning-based TPRA scheme.
In summary, we have investigated the multi-DC 3D deployment, trajectory planning and communication resource allocation in DA-RAN considering different user mobility patterns in this thesis. The proposed schemes and theoretical results should provide useful guidelines for future research in DC trajectory planning, resource allocation, as well as the real deployment of DCs in complex environments with diversified users. | en |
