UAV assisted cooperative caching on network edge using multi agent Actor critic reinforcement learning

Abstract

In recent times, Multi-access edge computing (MEC) has been introduced to assist cloud servers by bringing the computation closer to the edge. This is a well-known replacement to deal with the strict latency faced by users while retrieving contents from long-distance data centers. To cope up with this latency while simultaneously improving users’ QOS poses a limitation which can be handled through caching at edge nodes. However, where to cache and what to cache so that a higher cache hit rate is achieved also poses another significant issue which is addressed in this research. In this paper we have approached the problem of dynamic caching along with the selection of edge node that leads to better cache hit rate. We have also proposed the use of UAVs as aerial Base Station(BS) to assist in peak hours where a ground base station is not enough to support the surge in user requests.It also elaborates the optimal relocation of UAVs to e↵ectively support user mobility, which then caters a cluster of users by the K-means clustering algorithm. In addition, to maximize the cache hit ratio we have proposed a cooperative deep reinforcement learning algorithm which ensured a global increase in cache hit ratio and also an ecient allocation of storage. We have shown simulations on UAV reallocation based on user mobility patterns and also achieved higher global cache hit ratio using our proposed multi-agent actor-critic algorithm. In this paper, emphasis was given on how to cache and where to cache based on the cooperation of UAV and GBS which open doors for further research.