Research On Edge Caching Strategy For Internet Of Things

Actively caching data in the edge communication equipment can effectively reduce the data traffic of backhaul links,bring down the average delay of data transmission and avoid network congestion during the traffic peaks.In the cache scenarios,developing an efficient cache placement strategy to improve the cache hit rate of files by predicting the interests of users is one of the key technologies to be considered.In addition,when the caching strategy is introduced to the dynamic scenarios of Internet of things(Io T),the freshness and the dynamic popularity of Io T data should also be considered for the timeliness of Io T data.Therefore,how to develop an appropriate cache replacement strategy is also one of the problems should be solved.This thesis mainly studies the edge caching technology in Io T.By predicting user interest,considering the freshness of data and mobility of users,using the machine learning approach,the cache placement and replacement strategy is proposed to optimize the cache hit rate and the total cost of the system.The details are as follows:(1)Aiming at the minimization problem of total cost for small base stations caching data in Io T scenario,an Online Collaborative with Freshness Caching Algorithm(OCWFCA)is proposed from the economic point of view.On the premise of limited storage capacity of small base stations and unknown content popularity,two types of costs are considered.The first type of cost is the caching cost,which is paid to the network service providers when SBSs request files from the Internet.The second type of cost is the download cost corresponding to the traffic on the transmission links.The total cost of the base stations is defined as a cost function,which is composed of the total caching cost and the total download cost.On the premise of ensuring the age of files satisfies the freshness constraint,the optimization problem of minimizing the total cost of BSs is constructed,and its NP completeness is proved by the mapping method of the set cover problem,finally the optimal caching strategy for each time slot can be obtained through an iterative process.The simulation results manifest that,the proposed algorithm minimizes the total cost on the premise of ensuring that the age of the file meets the freshness constraint.(2)Aiming at the maximization problem of cache hit ratio for D2 D content sharing scenario in Io T,we propose an optimization algorithm for solving the joint problem of file caching and updating.In the algorithm,a user mobility model based on Markov chain and a user interest prediction model based on the social proximity,user preference and freshness are constructed.Under the constraints of user capacity and file freshness,we formulate the mobility-aware,freshness-based and user interest-predicted optimization problem as a 0-1 multiple knapsack problem to maximize the caching hit probability and the extent to which a pair of D2 D users are interested in the files they’re going to obtain or cache,which is decomposed into two sub-problems,i.e.,cache problem and update problem,to be solved respectively.The simulation results manifest that,compared with existing schemes,the optimization algorithm proposed by us can predict the interests of users more accurately,improve the caching hit probability of files effectively,and maximize the whole utility of users in the Io T network.(3)Aiming at the minimization problem of total cost for users obtaining their desired files,a cache replacement strategy based on deep reinforcement learning is proposed.First of all,the cost of obtaining the Io T data is defined as a cost function,which is composed of the delay cost and the freshness loss cost.Then,an optimization model is constructed to minimize the total cost of all users in the network.To adapt to the dynamic environment of Io T scenarios,the cache replacement problem is modeled by Markov decision process and obtained by the deep reinforcement learning method.Because the action space in this scenario is discrete,the A3 C algorithm which fits the discrete space is used to optimize the cache replacement strategy.Simulation results show that,the proposed cache replacement strategy can effectively reduce the long-term cost of obtaining Io T data,achieve a balance between transmission delay and data freshness,and obtain higher cahce hit rate.