Deep Reinforcement Learning for Content Caching Optimization in the Internet of Vehicles

Abstract

Modern progress in edge caching has contributed significantly to the advancements of the Internet of Vehicles(IoVs). However, high mobility constraints, time-variant dynamics, and heterogeneous content requirements lead to challenging problematics for edge servers that strive to use their resources optimally. Deep Reinforcement Learning allows incorporating cognition and intelligence into the Internet of Vehicles and permits the intelligent allocation of resources to resolve such challenging problems with various features. We propose an inter-vehicle-enabled caching for IoVs based on Deep Reinforcement Learning to define the optimal caching resource allocation strategies. The framework objective is to encourage sharing caching resources while reducing the probability of under/over-estimation of caching requirements. Extensive simulation results with different system parameters demonstrate the efficiency of the proposed solution.