Vehicle to Vehicle (V2V) communication has recently been considered in 4G and 5G cellular networks. One of the challenging issues in cellular V2V is allocating radio resources to the vehicles. Although previous work have addressed this issue, the fast varying nature of vehicular traffic and its regularities implies that the mobility of the vehicles should be more attended. To this goal, we propose an autonomous geo-based resource selection algorithm that uses deep learning to predict vehicle locations in the future and alleviate the computation and signalling overhead of the cellular infrastructure in contrast to previous geo-based resource allocation algorithms. We utilize the current and the future of vehicle densities in a formulated matching problem to find the optimum assignment of sub-resource pools to geographic areas. Simulation results of a highway with diverse density scenarios and different number of available resources show that the proposed method guarantees a considerable reduction in computation and signalling overhead while in low awareness ranges, it provides up to 10% improvement in Packet Reception Ratio (PRR) and the error rate of vehicles compared to the previous Dynamic Geo-based Resource Selection Algorithm (DGRSA). The proposed method also provides up to 15% improvement in PRR and error rate compared to the modified DGRSA, which we have changed to run with an overhead equal to the overhead of our proposed method. Furthermore, our results demonstrate up to 67% and 76% improvement in blocking rate compared to DGRSA and modified DGRSA, respectively.

最近在 4G 和 5G 蜂窝网络中考虑了车对车 (V2V) 通信。蜂窝 V2V 中具有挑战性的问题之一是为车辆分配无线电资源。虽然以前的工作已经解决了这个问题，但车辆交通的快速变化性质及其规律意味着应该更多地关注车辆的机动性。为此，我们提出了一种基于地理的自主资源选择算法，与以前的基于地理的资源分配算法相比，该算法使用深度学习来预测未来的车辆位置，并减轻蜂窝基础设施的计算和信令开销。我们在制定的匹配问题中利用车辆密度的当前和未来来找到子资源池到地理区域的最佳分配。具有不同密度场景和不同可用资源数量的高速公路的仿真结果表明，所提出的方法在低感知范围内保证了计算和信令开销的显着减少，它提供了高达 10% 的数据包接收率 (PRR) 和与之前的基于地理的动态资源选择算法 (DGRSA) 相比，车辆的错误率。与修改后的 DGRSA 相比，所提出的方法还提供了高达 15% 的 PRR 和错误率改进，我们已将其更改为以与我们提出的方法的开销相等的开销运行。此外，我们的结果表明，与 DGRSA 和改良的 DGRSA 相比，阻断率分别提高了 67% 和 76%。