Delay-Aware Task Offloading for Mobile Fog Nodes in Smart Cities: A Fuzzy and Q-Learning Approach

Abstract

Vehicular services, such as remote intelligent control, ve hicular video streaming and virtual reality-based driving assis tance, are delay-sensitive tasks [1]. These tasks request more computational powers and consume more energy for process ing. Thus, these tasks are offloaded to the fog node (FN) in the fog computing (FC) infrastructure. Therefore, integrating vehicular networks and FC leads to fog computing-based vehicular ad-hoc networks (FCVANETs) [2]. In FCVANETs, vehicular services are disrupted when the FN’s battery is depleted, particularly in remote areas lacking consistent power, such as forests and mountainous terrain. Reviving the FNs necessitates either solar energy or human intervention, and providing continuous services until the next recharge cycle without exhausting energy is challenging. As a result, we consider that FNs are equipped with powerful batteries that are periodically recharged but lack a permanent power source. Further, we consider the FNs coverage regions overlapping with neighbouring FNs and delivering services to the re quested vehicles in the overlapping regions. However, vehicles without service requests and ample computational capabilities are designated as mobile fog nodes (MFNs). We introduce a task-offloading approach called fuzzy and reinforcement learning task offloading (FRLTO) to increase the lifetime of FCVANETs. The existing work [2] addresses FN factors, such as energy consumption, latency, delay constraints, and relia bility, but overlooks the challenges that arise in FCVANETs when the coverage areas of FNs intersect with neighbouring FNs. Conversely, the task offloading approach presented in this work considers factors such as task data size, delay constraints, and MFNs’ sojourn time within the overlapping region.