Deep Reinforcement Learning for Cooperative Edge Caching in Vehicular Networks

In order to enable more and more multimedia content to be shared in the vehicular network, edge caching is a promising approach to cache content near the vehicles to reduce the burden of communication link and improve quality of service. However, the high mobility of vehicles and change in content popularity bring new challenges to edge caching in dynamic environment. Under the limitation of cache capacity, we propose a collaborative caching strategy in vehicular network to maximize the data throughput obtained from edge devices. Specifically, we first use Hawkes process to adapt to the dynamic change of contents’ popularity. Then, a cooperative content caching scheme based on deep reinforcement learning (DRL) is proposed. Finally, the performance of the scheme is evaluated by simulation experiments.