Vehicular Fog Computing for Video Crowdsourcing: Applications, Feasibility, and Challenges

Abstract

With the growing adoption of dash cameras, we are seeing great potential for innovations by analyzing the video collected from vehicles. On the other hand, transmitting and analyzing a large amount of video, especially high-resolution video in real time, requires a lot of communications and computing resources. In this work, we investigate the feasibility and challenges of applying vehicular fog computing for real- time analytics of crowdsourced dash camera video. Instead of forwarding all the video to the cloud, we propose to turn commercial fleets (e.g., buses and taxis) into vehicular fog nodes, and to utilize these nodes to gather and process the video from the vehicles within communication ranges. We assess the feasibility of our proposal in two steps. First, we analyze the availability of vehicular fog nodes based on a real-world traffic dataset. Second, we explore the serviceability of vehicular fog nodes by evaluating the networking performance of fog-enabled video crowdsourcing over two mainstream access technologies, DSRC and LTE. Based on our findings, we also summarize the challenges to largescale real-time analytics of crowdsourced videos over vehicular networks.