A cooperative caching scheme utilizing regional feature and dynamic vehicle clustering in vehicular edge networks

Abstract

Empowered with edge caching technology, the near-end storage resources of edge nodes in the Internet of Vehicles (IoV) can be fully utilized to accelerate the process of responding to content requests. In this paper, we propose a cooperative edge caching scheme utilizing regional feature and dynamic vehicle clustering (CRFDC). We take into consideration the fact that the change frequency of content popularity and regional feature is much lower than that of network topology and channel changes affected by vehicle mobility. To address this, we establish a double time-scale model. On the larger time-scale, we consider changes in content popularity and regional feature. On the smaller time-scale, we use the Prediction by Partial Matching (PPM) algorithm to predict vehicle's position. Additionally, we implement a dynamic cluster management approach, where vehicles with similar paths are grouped together, and use a consistent hashing algorithm to distribute contents among cooperative nodes. Finally, we employ deep reinforcement learning (DRL) approach to optimize our cooperative caching strategy for achieving lower content delivery latency. Simulation experiments demonstrate that our CRFDC scheme outperforms other cooperative caching schemes and benchmark algorithms in terms of reducing content transmission delay, improving cache hit ratio and decreasing communication overhead.