Exploiting Fog Computing With An Adapted Dbscan For Traffic Congestion Detection System

In order to feed a Traffic Congestion Detection System (TCDS), road safety messages (beacons) are continuously exchanged on Vehicular Ad hoc Networks (VANETs) through the IEEE 802.11p control channel. In VANET, the number of beacons in the communication network increases as the number of vehicles on the roads increases, raising communication costs. For a TCDS, clustering algorithms have been used to detect source and level of the traffic congestion based on vehicular density, as well as group similar traffic data that may lead to a reduction in the amount of data on the network. However, these clustering approaches have been employed to work only in a static dataset. Therefore, we propose a Fog Computing Framework that employs an adapted DBSCAN to reduce the amount of data produced in an online traffic data stream environment. The aim is to offer a more suitable approach for reducing the online traffic data stream, which is sent from Fog to the Cloud without losing accuracy of information related to road congestion. The evaluation results have shown that there is a dependence relationship between the size of the DBSCAN's radius, the amount of reduced data, and the congestion level accuracy.