Design and Implementation of a Fog Computing Based Collision Warning System in VANETs

Abstract

Intelligent transportation system (ITS) is becoming an important part of our society. Vehicular ad hoc networks (VANETs), as one of the most fundamental technologies in ITS, have attracted great interests in both industry and academia. Due to ever-increasing data demands in VANETs, it is challenging to meet the requirements of ITS, especially for safety-critical applications. Fog computing is an emerging paradigm with the vision of supporting low communication latency, enhancing quality of service, and improving efficiency of network. Meanwhile, collision warning (CW) is critical for enhancing driving safety, which has stringent timing requirement. This work is dedicated to addressing safety issues in VANETs by designing and implementing a fog computing based collision warning system. Distinguishing from conventional cloud-based solutions, where vehicles have to communicate with servers resided in the backbone network via cellular network, the fog computing based system offloads communication and computation workloads to the infrastructures which are closer to end users (i.e., vehicles). In particular, the road-side unit (RSU) with dedicated short range communication (DSRC) interface and local computation unit is acted as the fog node, and it is installed at the road intersection. Then, vehicles approaching to the intersection will periodically report their status including positions, speed, driving direction and acceleration, to the RSU via DSRC. Based on the collected real-time status of surrounding vehicles, a collision prediction algorithm is designed and implemented in the computation unit of the fog node. The warning message will be triggered if potential collision is detected, and it will be transmitted to corresponding vehicles. For performance evaluation, we have implemented the system in realistic VANETs scenario. In addition, we have also implemented a cloud computing based CW system for performance comparison, where the LTE interface is adopted for transmitting messages to the cloud server, where the collision prediction algorithm is executed. Based on the field testing, we have demonstrated the superiority of the fog computing based CW system in terms of enhancing driving safety by providing real-time communication and distribute computation capabilities.