Considering the infrastructure deployment cost and energy consumption, it is unrealistic to provide seamless coverage of the vehicular network. The presence of uncovered areas tends to hinder the prevalence of the in-vehicle services with large data volume. To this end, we propose a predictive cooperative multi-relay transmission strategy (PreCMTS) for the intermittently connected vehicular networks, fulfilling the 6G vision of semantic and green communications. Specifically, we introduce a task-driven knowledge graph (KG)-assisted semantic communication system, and model the KG into a weighted directed graph from the viewpoint of transmission. Meanwhile, we identify three predictable parameters about the individual vehicles to perform the following anticipatory analysis. Firstly, to facilitate semantic extraction, we derive the closed-form expression of the achievable throughput within the delay requirement. Then, for the extracted semantic representation, we formulate the mutually coupled problems of semantic unit assignment and predictive relay selection as a combinatorial optimization problem, to jointly optimize the energy efficiency and semantic transmission reliability. To find a favorable solution within limited time, we proposed a low-complexity algorithm based on Markov approximation. The promising performance gains of the PreCMTS are demonstrated by the simulations with realistic vehicle traces generated by the SUMO traffic simulator.

翻译：考虑到基础设施部署成本和能耗，为车载网络提供无缝覆盖是不现实的。未覆盖区域的存在往往阻碍了大容量车载服务的普及。为此，我们提出了一种预测性协同多中继传输策略（PreCMTS），用于间歇性连接的车载网络，以实现6G愿景中的语义和绿色通信。具体而言，我们引入了任务驱动的知识图谱（KG）辅助语义通信系统，并从传输角度将KG建模为加权有向图。同时，我们识别了关于单个车辆的三个可预测参数，以执行后续的预测分析。首先，为促进语义提取，我们推导了在延迟要求内可达吞吐量的闭式表达式。然后，针对提取的语义表示，我们将语义单元分配和预测性中继选择的相互耦合问题构建为组合优化问题，以联合优化能效和语义传输可靠性。为了在有限时间内找到可行解，我们提出了一种基于马尔可夫近似的低复杂度算法。通过使用SUMO交通模拟器生成的真实车辆轨迹进行仿真，证明了PreCMTS的显著性能增益。