Deep Reinforcement Learning for Intelligent Computing and Content Edge Service in ICN-based IoV

Abstract

Driven by the development of communication and computing technologies, the intelligent Internet of Vehicles (IoV) has attracted much attention in recent years. Specifically, integration of communication, computing, caching, and AI at the network edge has become a key to realizing various exciting IoV applications. However, the dynamic nature of IoV imposes great challenges on the successful realization of integrated edge services. In this paper, we first propose an Information-Centric Networking (ICN)-based framework to accommodate both computing and content service requests in IoV. Next, considering the fact that making use of the popularity of the service requests and the caching of computing results may greatly improve the efficiency of the edge service, we propose an innovative algorithm based on deep Q-learning to learn the popularity of service requests and make joint computing and caching decisions accordingly. Simulation results show that the pro-posed algorithm can improve the satisfied request ratio by environment learning and data reuse.