Study On Caching Placement And Resource Allocation Algorithm In UAV Cellular Network

Due to its advantages of low cost,easy deployment and reliable communication channels,unmanned aerial vehicle(UAV)has become one of the effective technologies to realize emergency communication and hot spot coverage.Due to the user’s centralized request for content in these scenarios,the contradiction between a large number of content requests and limited communication resources leads to the fact that transmission capacity of UAV assisted cellular network is difficult to meet the user’s needs.For the repeated requests of the same content,UAV caches the popular content,so as to effectively offload the traffic of wireless backhaul link and meet the user’s demand for low content delivery delay.At the same time,the mobile characteristics of UAV lead to the complex channel characteristics of communication channel.Considering the dynamic environment factors in the actual scenario,the joint optimization algorithm of caching placement and resource allocation in dynamic scenario is required.This paper is based on National Natural Science Foundation of China,"Research on Multi-level Cooperative Caching Method in UAV-Assisted Cellular Network".The caching placement and resource allocation algorithm in UAV aided cellular network is studied.Two scenarios of single UAV aided cellular network and multi-UAV aided cellular network are considered respectively.The joint optimization algorithm of caching placement and resource allocation in single UAV aided cellular network and the joint optimization algorithm of caching placement and resource allocation in multi UAV aided cellular network are proposed.The main work of this paper is as follows:1)In this paper,the research background and literature review of caching placement and resource allocation algorithm in UAV assisted cellular network are reviewed.Firstly,the research background of UAV assisted cellular network is summarized,and the research situation of emergency communication and hot spot area coverage is introduced.Then the cache placement in UAV aided cellular network is sorted out.Finally,the resource allocation algorithms in UAV aided cellular networks are summarized.2)In order to deal with the dynamic UAV location and content request in the actual scenario,the caching placement and resource allocation optimization problem is formulated as a Markov decision process to minimize the long-term content delivery delay.It is assumed that the UAV acts as an agent to perform caching placement and resource allocation,including scheduling of user requests for content and power allocation among users.In order to solve this problem,a caching placement and resource allocation algorithm based on reinforcement learning is proposed,where a soft greedy strategy is used to learn and select actions to find the best match between actions and states.In order to solve the problem that the Q-table of the joint optimization algorithm based on reinforcement learning in large-scale dynamic networks increases with the increase of the state number,an function approximation based algorithm,which consists of stochastic gradient descent and deep neural network.Numerical results show that,compared with the benchmark algorithms,the proposed algorithm has good performance and achieves a tradeoff between network performance and computational complexity.3)The joint optimization of user association,power allocation between users,UAV deployment and UAV caching placement is formulated to minimize content delivery delay.A joint optimization algorithm based on branch and bound is proposed to optimize the content delivery delay in each time slot.To deal with the dynamic content request、mobility of users and the mobility of UAV in the actual scenario,the original optimization problem is transformed into a Stackelberg game.Specifically,the game is decomposed into a leader user association sub-problem and multiple followers power allocation,UAV deployment and UAV caching placement sub problems.The joint optimization algorithm based on deep reinforcement learning is further proposed to solve the problem of long-term content delivery delay minimization.The simulation results show that the joint optimization algorithm based on branch and bound can get the optimal solution in each time slot,and its content delivery delay is much lower than that of the benchmark algorithms.Besides,the joint optimization algorithm based on deep reinforcement learning achieves relatively low long-term content delivery delay in dynamic environment,and its computational complexity is lower than branch and bound based algorithm.