车联网中基于轨迹预测的无人机动态协同优化覆盖算法

针对城市车联网中出现的基站覆盖空洞及局部流量过载等问题，提出了一种基于车辆轨迹预测信息的动态预部署方案。首先，为了训练得到统一的seq2seq-GRU轨迹预测模型，多个携带边缘计算服务器的无人机在分布式联邦学习与区块链的架构下，去除中心聚合节点，采取改进的Raft算法，在每轮训练中根据贡献数据量的大小，选举得到节点来完成参数聚合及模型更新任务。其次，基于模型预测结果，提出了一种改进的虚拟力向导部署算法，通过各虚拟力来引导无人机进行动态地部署以提升车辆的接入率及通信质量。仿真结果表明，提出的训练架构能够加速模型的训练，部署算法在提升车辆接入率的同时提升了车辆与无人机之间的通信质量。

UAV dynamic collaborative optimization coverage algorithm based on trajectory prediction in Internet of Vehicles

Aiming at the problems of coverage voids of base stations and local traffic overload in urban vehicle networking, this paper proposed a dynamic pre-deployment scheme based on vehicle trajectory prediction information. Firstly, multiple UAVs equipped with edge computing servers remove the central aggregation node under the architecture of distributed federated learning and blockchain and adopt an improved Raft algorithm to train a unified seq2seq-GRU trajectory prediction model. In the round of training, according to the amount of contributed data, the scheme elected the nodes to complete the parameter aggregation and model updating tasks. Secondly, based on the prediction results of the model, this paper proposed an improved virtual force guide deployment algorithm, which guided the UAV to dynamically deploy through each virtual force to improve the access rate and communication quality of the vehicle. The simulation results show that the proposed training architecture can accelerate the training of the model, and the deployment algorithm improves the access rate of the vehicle while improving the communication quality between the vehicle and the UAV.