基于强化学习的车联网路由协议设计与性能分析研究

The Internet of Vehicles (IOV) is a dynamic network with large numbers of vehicles and infrastructures which are connected by the wired or wireless way. One of the fundamental problems of IOV is how to provide a better data transmission service and enhance the business supporting capability. By using the hybrid communication architecture (the infrastructure support + the ad hoc way), the project carries on studies of the high efficient routing protocol design and data transmission performance analysis for IOV. Through seizing the high dynamic property of IOV and mining the characteristics of the vehicle trajectory, the logic relationship formed by vehicle contacts and the traffic flow, the project uses the latent regularity of IOV and the idea of the reinforcement learning to construct the reinforcement learning model with the characteristics of IOV, and proposes the high efficient routing protocol under the hybrid communication architecture. The project also verifies the performance of data transmission protocol from the theoretical and practical aspects. First, the project constructs a more realistic data transmission system model for IOV, and then gives the scaling laws of the throughput and delay of IOV. They provide the performance evaluation indicant of data transmission in IOV. Second, combining the real vehicular data, the project builds the simulation platform to evaluate the performance of the proposed protocol. The project motivates the optimization of the data transmission protocol design and provides the feasible and effective solution for the data communication of IOV.

车联网主要是由大量车辆和基础设施以有线、无线方式互连构成的动态网络，是当前蓬勃发展的应用领域。其中重要的基础问题之一便是如何提供更好的数据传输服务质量，增强业务保障能力。本项目以基础设施支持模式和车辆自组织模式混合为车联网通信架构，展开对车联网高效路由协议及数据传输性能的研究。项目抓住车联网的高动态性，深入挖掘车辆的行驶轨迹、车辆相遇构成的逻辑关系、交通流等特征，利用车联网中隐藏的规律性，以强化学习类方法为思维突破，建立具有车联网特性的强化学习模型，提出混合架构下基于强化学习的高效路由协议。本项目从理论和实践两方面验证数据传输协议的性能，首先，建立更贴合现实的车联网数据传输系统模型，给出车联网吞吐量和延迟的网络标度律结果，提供车联网数据传输性能的评价指标；其次，结合实际数据，构建仿真平台，评价协议的性能。本项目为车联网数据传输协议设计的优化提供动力，为车联网数据通信提供行之有效的解决方案。

车联网主要是由大量车辆和基础设施以有线、无线方式互连构成的动态网络，是当前蓬勃发展的应用领域。其中重要的基础问题之一便是如何提供更好的数据传输服务质量，增强业务保障能力。本项目以基础设施支持模式和车辆自组织模式混合为车联网通信架构，展开对车联网高效路由协议及数据传输性能的研究。项目深入挖掘车辆的行驶轨迹、车辆相遇构成的逻辑关系、交通流等特征，利用车联网中隐藏的规律性，以强化学习方法为思维突破，建立具有车联网特性的强化学习模型。针对车辆传输的接入过程，给出了基于强化学习的车联网传输负载均衡优化策略，同时基于建立的强化学习模型，给出了面向车联网网络场景的认知高效路由协议。从理论方面建立更贴合现实的基于中心点的车联网数据传输系统模型，给出车联网最常见的通信会话模式-广播模式的吞吐量界的结果。最后，结合实际收集的车辆行驶数据，构建仿真平台，验证所提出的数据传输协议性能要优于当前广泛使用的车联网数据传输协议。