Multi-Agent Deep Reinforcement Learning for content caching within the Internet of Vehicles

Abstract

In the Internet of Vehicles (IoV), multiple applications and multimedia services are deployed to enhance the quality of traveling experienced by passengers. These software components ingest heterogeneous multimedia content (e.g., streams, safety information, etc.) acquired through the Internet. However, the high mobility constraints and time-variant dynamics of the IoV may delay the acquisition process. Caching lowers acquisition delays of the content by storing requested content in the vicinity of its consumers.We propose a Multi-Agent Deep Reinforcement Learning (MADRL) based framework to design an optimal caching strategy for IoVs. The framework objective is to incentivize sharing caching resources while minimizing errors in estimating content needs. Extensive simulation results with different system parameters demonstrate the efficiency of the proposed solution.