Vehicular Ad-hoc Networks (VANETs) are based on vehicle to infrastructure communications in which the vehicles periodically broadcast information to update a Road Side Unit (RSU). The traffic data is forwarded from all RSUs to a cloud or a central server for global analysis and detection of congestion levels on the roads. However, communication costs may considerably increase when a large amount of data is transmitted to such cloud-like service providers. In this paper, we propose a data clustering framework to perform traffic information reduction at the edge of vehicular networks by exploiting fog computing. The proposed data clustering framework defines two methods for the reduction of the traffic data stream: (i) Baseline method, which is an ordinary traffic congestion detection approach, and (ii) two adapted clustering methods for a data stream; namely, the Ordering Points to Identify the Clustering Structure (OPTICS) and the Density-Based Spatial Clustering of Applications with Noise (DBSCAN). The results have shown that the proposed traffic data framework using clustering methods is accurate even when the vehicular traffic condition is highly congested, potentially reducing the communication costs and bringing significant results for the development of VANETs.

车辆自组织网络（VANET）基于车辆到基础设施的通信，其中车辆定期广播信息以更新路边单元（RSU）。交通数据从所有RSU转发到云或中央服务器，以进行全局分析和检测道路上的拥堵程度。但是，当将大量数据传输到类似云的服务提供商时，通信成本可能会大大增加。在本文中，我们提出了一种数据聚类框架，通过利用雾计算在车辆网络的边缘执行交通信息减少。拟议的数据聚类框架定义了两种减少交通数据流的方法：（i）基线方法，这是一种普通的交通拥堵检测方法，（ii）两种适用于数据流的聚类方法；即，识别聚类结构的排序点（OPTICS）和带有噪声的应用程序的基于密度的空间聚类（DBSCAN）。结果表明，即使在车辆交通状况高度拥挤的情况下，使用聚类方法提出的交通数据框架也是准确的，从而有可能降低通信成本，并为VANET的发展带来重大成果。