基于Q-learning的分布式基站缓存替换策略

Distributed Cache Replacement for Base Stations Based on Q-Learning

针对传统替换策略的不足,提出一种基于Q-learning的缓存替换策略。该方法使用马尔科夫博弈模型描述多基站协作替换问题,以降低网络服务延迟为目标,利用分布式Q-learning算法获得Nash均衡点作为最优策略。实验表明,与其他缓存替换策略相比,该方法能够有效降低网路延迟,提升服务质量。

To improve the weakness of traditional replacement policy,this paper proposes a cache replacement policy based on Q-learning.This method uses Markov Game model to describe multiple base station cooperation replacement,for the purpose of reducing the delay of network services,and uses the distributed Q-learning algorithm to obtain Nash equilibrium as the optimal strategy.The experiment shows that this method can reduce network delay effectively and improve service quality compared with other cache replacement policies.