Time‐Critical Fog Computing for Vehicular Networks

Moving applications from the local infrastructure to a central data center using cloud computing has been the highlight of the current distributed computing era. This paradigm shift has enabled various new use cases such as computing and storage everywhere and on‐demand. A prominent example for cloud computing is off‐loading computation and data from smartphones to data centers. The Internet of Things (IoT) aims to connect people and objects such as vehicles, machines, and products with people through the Internet and the cloud. The number of IoT devices significantly has grown in recent years resulting in massive volume of data transferred to the cloud for analysis. Connecting vehicles as a smart thing is required for enhancing the vehicle's perception of its surroundings and increasing road safety as well as traffic efficiency. Some applications such as navigation may be delay‐tolerant and can be still further distributed among the vehicles and the cloud. Other applications such as collision warnings are delay‐critical and accordingly short‐lived. For example, detecting and identifying an immediate obstacle on the road requires a quick processing of onboard sensor data (e.g. camera, radar, and LIDAR) in a few tens of milliseconds. With this unprecedented evolution, fog computing has been proposed to support delay‐critical computational demand, security issues, communication latency and improve quality of service of vehicular applications. Fog computing is complementary to cloud computing. The interplay of both concepts is now considered as the real enabling architecture for virtually all vehicular networked applications. In this chapter, we first present various scenarios of time‐critical applications and their timeliness requirement. We define our application model based on a set of middleware building blocks in order to reach the timeliness guarantees. The key building blocks are resource monitoring, task scheduling, real‐time computation, and real‐time communication. Then we show how the perturbations can interrupt the communication and the computation between vehicles and/or infrastructures, which is a serious problem in such critical scenarios. Next, we critically review existing research efforts to cope with failures, threats, and constraints of the network, computation, and data management so as to efficiently meet the timeliness requirements despite the perturbation. In particular, we provide a taxonomy of fog computing according to the research topics. Throughout the chapter, we identify research gaps and sketch future research directions.