Imitation-Learning-Enabled Vehicular Edge Computing: Toward Online Task Scheduling

Abstract

Vehicular edge computing (VEC) is a promising paradigm to enable information sharing and acquisition among vehicles. Since information freshness has significantly influence on task scheduling, an emerging metric, named the age of information (AoI), has been utilized to evaluate it. Recent research generally focuses on AoI minimization but pays little attention to information personality. However, different impacts caused by distinct information may be posed on user decisions. This article first briefly introduces the state of the art of imitation learning in wireless networks. After that, an imitation-learning-based online task scheduling scheme is designed with the support of VEC. It intends to minimize the average age of critical information (AoCI), referring to the age of information that has significant impacts on vehicle decisions. Performance evaluations show that the proposed scheme outperforms other algorithms from several aspects. At last, we discuss several potential research challenges and open issues for artificial intelligence in the Internet of Vehicles.