Improving the travel efficiency of citizens and the operation efficiency of urban has always been the goal of Intelligent Transportation System. Due to the neglect of strong traffic demand and driving behaviour preferences, coupled with the insufficient of communication and computing power, the existing measures based on the centralized control of vehicles or mandatory traffic restrictions lead to the traffic efficiency dramatically deviates from the system optimum. Which puts forward an urgent demand for multi-vehicle collaborative apportionment, but also brings challenges. In this paper, a double-layer collaborative apportionment method for connected vehicles, 2L-CoV for short, is proposed under the assistance of Space-air-ground integrated networks. 2L-CoV includes the traffic flow scheduling in global-layer and the vehicle routing planning in local-layer. Firstly, a distributed collaborative framework based SAGIN is presented to make a large-scale of virtual vehicles can interact with each other. Then, at the global-layer, traffic flow is guided speedily by an improved back-pressure algorithm to complete traffic flow scheduling; at the local-layer, considering the driving behaviour preferences, a game evolution online learning approach based on dominant strategy is proposed to plan the vehicle routing. Finally, the simulation results show that 2L-CoV can effectively balance the traffic network, improves the network throughput, and reduces the total travel time.

提高市民的出行效率和城市的运营效率一直是智能交通系统的目标。由于忽视了强大的交通需求和驾驶行为偏好，再加上通信和计算能力不足，基于车辆集中控制或强制性交通限制的现有措施导致交通效率大大偏离了系统最佳状态。这就提出了对多车辆协同分配的迫切需求，但同时也带来了挑战。本文提出了一种在空空地一体化网络的辅助下，对互联车辆进行双层协同分配的方法，简称为2L-CoV。2L-CoV包括全局层中的交通流调度和本地层中的车辆路由规划。首先，提出了一种基于分布式协作框架的SAGIN，以使大型虚拟车辆可以相互交互。然后，在全局层，通过改进的背压算法快速引导交通流，以完成交通流调度；在本地层，考虑驾驶行为偏好，提出一种基于优势策略的游戏进化在线学习方法来规划车辆路径。最后，仿真结果表明2L-CoV可以有效地平衡交通网络，提高网络吞吐量，并减少总旅行时间。提出了一种基于分布式协作框架的SAGIN，以使大型虚拟车辆可以彼此交互。然后，在全局层，通过改进的反压算法快速引导交通流，以完成交通流调度；在本地层，考虑驾驶行为偏好，提出一种基于优势策略的游戏进化在线学习方法来规划车辆路径。最后，仿真结果表明2L-CoV可以有效地平衡交通网络，提高网络吞吐量，并减少总旅行时间。提出了一种基于分布式协作框架的SAGIN，以使大型虚拟车辆可以彼此交互。然后，在全局层，通过改进的反压算法快速引导交通流，以完成交通流调度；在本地层，考虑驾驶行为偏好，提出一种基于优势策略的游戏进化在线学习方法来规划车辆路径。最后，仿真结果表明2L-CoV可以有效地平衡交通网络，提高网络吞吐量，并减少总旅行时间。提出了一种基于优势策略的游戏进化在线学习方法来规划车辆路径。最后，仿真结果表明2L-CoV可以有效地平衡交通网络，提高网络吞吐量，并减少总旅行时间。提出了一种基于优势策略的游戏进化在线学习方法来规划车辆路径。最后，仿真结果表明2L-CoV可以有效地平衡交通网络，提高网络吞吐量，并减少总旅行时间。