Multi-Agent Caching Strategy for Spatial-Temporal Popularity in IoV
Peng He, Cao Li, Yaping Cui, Ruyan Wang, Dapeng Wu
Abstract
Motivated by connected vehicles, the internet of vehicles (IoV) has a prosperous development, thus a variety of IoV applications have emerged, which causes the dramatic growth of data traffic and content request, thus affecting the quality of service (QoS) of IoV. Edge caching can effectively improve the QoS by caching the contents in the road side unit (RSU) near the requester. Moreover, most existing IoV caching strategies only consider that the content popularity has temporal correlation or assume that the content popularity is given, which will impact the accuracy of content popularity. In this article, we propose a spatial-temporal correlation approach to predict the content popularity, which can improve the accuracy of the content popularity by the spatial-temporal correlation analyzing of the RSU historical content requests. Then, we introduce a multi-agent reinforcement learning (MARL) caching strategy based on the result of prediction, where each RSU agent individually chooses the action with the constraint of caching resources and updates the action to maximize the cumulative reward. Simulation results demonstrate that the proposed caching strategy has better performance. Take the Zipf parameter is 0.8 as an example, compared with the probability caching scheme (PCS), the proposed caching strategy improves the caching hit ratio by 15% and reduces the delay by 34%.