D2D auxiliary equipment caching system and caching method based on reinforcement learning

Abstract

The invention discloses a D2D auxiliary equipment caching system and caching method based on reinforcement learning, and particularly relates to a D2D communication caching optimization system based on a DQN (Deep Q Network) reinforcement learning mode. The system is composed of a training data screening module, a service interaction module, a request processing module and a log recording sub-module of a server side, and a reinforcement learning module, a request processing module, a file caching module, a log recording module of auxiliary node equipment and all users in an auxiliary node D2Dcommunication coverage area where the users are located. The record of a user on a file request is statistically analyzed; the data is screened and then is used for reinforcement learning; a caching strategy adapting to popularity on auxiliary node equipment is obtained through learning by a neural network, so that the hit rate of system D2D auxiliary equipment unloading is improved, the load of abase station is effectively reduced, and the effects of reducing the user delay, saving the network bandwidth, increasing the resource reuse and the like of D2D communication are exerted.

Description

technical field [0001] The invention belongs to a D2D auxiliary device cache system, and relates to reinforcement learning and D2D communication in a next-generation communication network. Background technique [0002] According to Cisco's research, wireless and wired traffic is growing rapidly every day, especially video-on-demand and high-quality streaming media services occupy a large part of network traffic. On cable networks, video traffic already accounts for more than 50% of total Internet traffic. Device-to-Device (D2D) communication in future 5G wireless networks provides an effective way to solve this problem. Downloading base station files to devices capable of D2D communication not only greatly reduces the user's transmission delay, but also saves a lot of bandwidth for the network. The storage capacity of general user equipment is extremely limited, the size of the battery is also limited, and it involves personal privacy issues, making it difficult to directl...

AI Technical Summary

Problems solved by technology

[0003] Much recently, and building on the work of N. Golrezaei, K. Shanmugam, A.G. Dimakis, A.F. Molisch, and G. Caire et al., we found that the problem of how to cache files on secondary devices is NP-hard, but the enhanced Learning emerges, inspiring us

Images