CN105611574A

CN105611574A - Method for combining dynamic access and subcarrier allocation under cache-based ultra-dense network

Info

Publication number: CN105611574A
Application number: CN201510994459.2A
Authority: CN
Inventors: 李曦; 刘宜明; 纪红; 张鹤立; 王珂
Original assignee: Beijing University of Posts and Telecommunications
Current assignee: Beijing University of Posts and Telecommunications
Priority date: 2015-12-25
Filing date: 2015-12-25
Publication date: 2016-05-25
Anticipated expiration: 2035-12-25
Also published as: CN105611574B

Abstract

The invention discloses a method for joint dynamic access and sub-carrier allocation based on a cache-based ultra-dense network. The specific steps are as follows: first, multiple users simultaneously send request information to all access points to find cache content; then each access point Judging whether there is the cached content requested by the current user K, all access points satisfying the user K will transmit their attribute parameters to the local control, and the local control will assign the best access point to the user K; otherwise, the user K directly sends the remote The server sends a request to obtain content; the remote server uses popularity analysis to update the cache according to the user request information; finally, after each user is matched with its own access point, subcarrier allocation is performed to make the connection between the user and the access point to communicate. The advantage lies in that the access selection is completed by integrating multiple factors, the improvement of resource management efficiency and the dynamic allocation of sub-carriers are realized, and the spectrum utilization rate is significantly improved.

Description

A Method of Joint Dynamic Access and Subcarrier Allocation Based on Cache in Ultra-dense Network

技术领域technical field

本发明属于组网及资源分配技术领域，具体是指一种基于缓存的超密集网络下联合动态接入和子载波分配的方法。The invention belongs to the technical field of networking and resource allocation, and specifically refers to a method for joint dynamic access and subcarrier allocation in a buffer-based ultra-dense network.

背景技术Background technique

超密集网络是5G的有力候选技术，超密集组网技术通过增加基站部署密度，可实现频率复用效率的巨大提升，极大地提高系统容量，满足5G千倍容量增长需求。然而，愈发密集的网络部署使得网络拓扑更加复杂，现有的内容分发机制在实现图片、音频、视频等海量信息传送的同时，存在大量的内容重复传输，对频谱资源等造成了极大地浪费。针对这个问题，将缓存技术引入超密集网络，通过在接入点或者核心网进行内容缓存，可以减少冗余数据传输，有效降低回程链路消耗和网络时延，从而提高了频谱利用效率和能效利用率。The ultra-dense network is a strong candidate technology for 5G. The ultra-dense networking technology can greatly improve the frequency reuse efficiency by increasing the deployment density of base stations, greatly improve the system capacity, and meet the demand for 5G capacity growth of a thousand times. However, the increasingly dense network deployment makes the network topology more complicated. While the existing content distribution mechanism realizes the transmission of massive information such as pictures, audio, and video, there is a large amount of repeated transmission of content, which causes a great waste of spectrum resources. . In response to this problem, the caching technology is introduced into the ultra-dense network. By caching content at the access point or the core network, redundant data transmission can be reduced, and backhaul link consumption and network delay can be effectively reduced, thereby improving spectrum utilization efficiency and energy efficiency. utilization rate.

在基于缓存的密集网络下，文献1：在回程链路受限的密集无线网络中基于物理层缓存的吞吐量增益优化方法，提出了一种基于回程链路受限的新型缓存无线网络架构，并且在该架构下提出了物理层缓存方案以提高系统吞吐量，但是该方案仅考虑回程链路等因素，并显示结果与基站缓存容量大小是相关的。Under the cache-based dense network, Literature 1: A throughput gain optimization method based on physical layer cache in a dense wireless network with limited backhaul links, proposes a new cache wireless network architecture based on backhaul links, And under this architecture, a physical layer caching scheme is proposed to improve system throughput, but this scheme only considers factors such as the backhaul link, and shows that the result is related to the cache capacity of the base station.

文献2：在异构网络中联合路由和内容缓存优化的方法，采用联合路由选择和缓存分配的方案进行资源分配优化，该方案下所考虑的问题单一，没有考虑频效、能效等资源利用率问题。Document 2: A joint routing and content cache optimization method in a heterogeneous network, using a joint routing selection and cache allocation scheme to optimize resource allocation. The problem considered in this scheme is single, and resource utilization such as frequency efficiency and energy efficiency are not considered. question.

在超密集的场景下，现有技术没有考虑接入点的负载均衡，进行接入点的动态选择。此外，没有考虑频谱效率的优化，对子载波等资源进行分配优化。In an ultra-dense scene, the prior art does not consider load balancing of access points, and dynamically selects access points. In addition, the optimization of the spectrum efficiency is not considered, and the resource allocation such as subcarriers is optimized.

发明内容Contents of the invention

本发明针对现有技术中不能高效地利用接入点的存储资源为用户提供服务，频谱和能效资源有效利用率不能实现最大化，提出了一种基于缓存的超密集网络下联合动态接入和子载波分配的方法，Aiming at the inability to efficiently utilize the storage resources of access points in the prior art to provide services for users, and the effective utilization of frequency spectrum and energy efficiency resources cannot be maximized, the present invention proposes a cache-based joint dynamic access and sub-network under ultra-dense networks method of carrier assignment,

具体步骤如下：Specific steps are as follows:

步骤一、多个用户用广播方式发送请求信息给所有接入点，寻找缓存内容；Step 1. Multiple users broadcast request information to all access points to search for cached content;

请求信息是指缓存内容；用户数量为O个；The request information refers to the cache content; the number of users is O;

步骤二、将用户K作为当前用户，各个接入点判断是否存在当前用户请求的缓存内容，如果某个接入点空闲且存在该缓存内容，该接入点反馈1给用户K，进入步骤三；否则反馈0；进入步骤四；Step 2. Taking user K as the current user, each access point judges whether there is cached content requested by the current user. If an access point is idle and has the cached content, the access point feeds back 1 to user K, and proceeds to step 3. ;Otherwise feedback 0; go to step 4;

1≤K≤O；1≤K≤O;

步骤三、满足用户K的所有接入点将各自的属性参量传送给本地控制，本地控制将最佳接入点分配给用户K；Step 3, all access points satisfying user K transmit their attribute parameters to the local control, and the local control assigns the best access point to user K;

满足用户K的所有接入点为m个；每个接入点的属性参量均包括缓存容量、时延和信噪比等，共n个属性；There are m access points that satisfy user K; the attribute parameters of each access point include buffer capacity, delay, and signal-to-noise ratio, etc., and there are n attributes in total;

具体步骤如下：Specific steps are as follows:

步骤301、根据用户K的请求信息，针对每个候选接入点，本地控制对n个属性参量中每两个属性参量之间的相对重要性进行一一比较，得到决策矩阵M：Step 301, according to the request information of user K, for each candidate access point, the local control compares the relative importance of every two attribute parameters among the n attribute parameters one by one, and obtains the decision matrix M:

本地控制将每个候选接入点的各个属性参量之间的相对重要性进行比较，得到该候选接入点的决策矩阵M：The local control compares the relative importance of each attribute parameter of each candidate access point to obtain the decision matrix M of the candidate access point:

其中a_ij代表接入点中属性参量i与属性参量j的相对重要性比较值；Wherein a _ij represents the relative importance comparison value of attribute parameter i and attribute parameter j in the access point;

${a a}_{i i j j} > > 00;; {a a}_{i i j j} = = \frac{11}{{a a}_{j j i i}};; {a a}_{i i i i} = = 11;;$

步骤302、对决策矩阵M进行归一化，得到标准化后的决策矩阵B：Step 302. Normalize the decision matrix M to obtain a standardized decision matrix B:

其中b_ij代表接入点中对比较值a_ij归一化后的值；Where b _ij represents the normalized value of the comparison value a _ij in the access point;

步骤303、对决策矩阵B的一致性进行校验，判断决策矩阵是否有效，如果有效，进行步骤304，否则返回步骤302；Step 303, check the consistency of decision matrix B, judge whether the decision matrix is valid, if valid, go to step 304, otherwise return to step 302;

一致性比率CR定义如下： The consistency ratio CR is defined as follows:

其中，CI表示不一致性指标：λ_max是决策矩阵B的最大特征值，n是决策矩阵B中属性参量的个数，RI是已知的平均随机一致性指标；Among them, CI represents the inconsistency index: λ _max is the maximum eigenvalue of the decision matrix B, n is the number of attribute parameters in the decision matrix B, and RI is the known average random consistency index;

当CR＜0.1时，认为决策矩阵B具有可接受的一致性，否则重新构造决策矩阵B。When CR<0.1, the decision matrix B is considered to have acceptable consistency, otherwise, the decision matrix B is reconstructed.

步骤304、获取决策矩阵B中n个属性参量综合产生的的权重向量ω；Step 304, obtaining the weight vector ω generated by the synthesis of n attribute parameters in the decision matrix B;

ω＝(ω₁,ω₂,...ω_j,...,ω_n)ω＝(ω ₁ ,ω ₂ ,...ω _j ,...,ω _n )

ω_j为第j个属性参量的权重；ω _j is the weight of the jth attribute parameter;

步骤305、针对m个候选接入点，生成所有属性参量的状态矩阵S；Step 305, generating a state matrix S of all attribute parameters for the m candidate access points;

状态矩阵S为m行n列，每一行代表每个接入点的n个属性参量；The state matrix S has m rows and n columns, and each row represents n attribute parameters of each access point;

其中，s_mn表示第m个接入点对应第n个属性参量的值。Wherein, s _mn represents the value of the nth attribute parameter corresponding to the mth access point.

步骤306、将权重向量ω与状态矩阵S相乘得到加权决策矩阵Q：Step 306, multiply the weight vector ω by the state matrix S to obtain the weighted decision matrix Q:

步骤307、根据加权决策矩阵Q，确定最佳接入方案Q_best和最差接入方案Q_worst，Step 307, according to the weighted decision matrix Q, determine the best access scheme Q _best and the worst access scheme Q _worst ,

Q_best＝(ω₁·s_1best,ω₂·s_2best,...ω_j·s_jbest,...,ω_n·s_nbest)Q _best ＝(ω ₁ ·s _1best ,ω ₂ ·s _2best ,...ω _j ·s _jbest ,...,ω _n ·s _nbest )

Q_worst＝(ω₁·s_1worst,ω₂·s_2worst,...ω_j·s_jworst,...,ω_n·s_nworst)Q _worst ＝(ω ₁ ·s _1worst ,ω ₂ ·s _2worst ,...ω _j ·s _jworst ,...,ω _n ·s _nworst )

s_jbest表示所有m个接入点的第j个属性参量中最佳值；s_jworst表示所有m个接入点的第j个属性参量中最差值；s _jbest represents the best value of the jth attribute parameter of all m access points; s _jworst represents the worst value of the jth attribute parameter of all m access points;

步骤308、针对某个接入点l，分别计算候选接入方案x_lj与最佳接入方案Q_best的欧氏距离，和候选接入方案x_lj与最差接入方案Q_worst的欧氏距离；Step 308. For a certain access point l, calculate the Euclidean distance between the candidate access scheme x _lj and the best access scheme Q _best , and the Euclidean distance between the candidate access scheme x _lj and the worst access scheme Q _worst distance;

候选接入方案x_lj与最佳接入方案Q_best的欧氏距离为Q_lbest，具体是指候选接入方案x_lj的每个属性参量与该属性最优值s_jbest的欧式距离，如下：The Euclidean distance between the candidate access scheme x _lj and the best access scheme Q _best is Q _lbest , which specifically refers to the Euclidean distance between each attribute parameter of the candidate access scheme x _lj and the optimal value s _jbest of the attribute, as follows:

${Q Q}_{l l b b e e s the s t t} = = \sqrt{{Σ Σ}_{j j = = 11}^{n no} {ω ω}_{j j} {(({x x}_{l l j j} - - {s the s}_{j j b b e e s the s t t}))}^{22}}$

候选接入方案x_lj与最差接入方案Q_worst的欧氏距离为Q_lworst，具体是指候选接入方案x_lj的每个属性参量与该属性最差值s_jworst的欧式距离，如下：The Euclidean distance between the candidate access scheme x _lj and the worst access scheme Q _worst is Q _lworst , which specifically refers to the Euclidean distance between each attribute parameter of the candidate access scheme x _lj and the worst value s _jworst of the attribute, as follows:

${Q Q}_{l l w w o o r r s the s t t} = = \sqrt{{Σ Σ}_{j j = = 11}^{n no} {ω ω}_{j j} {(({x x}_{l l j j} - - {s the s}_{j j w w o o r r s the s t t}))}^{22}}$

步骤309、针对某个接入点l，计算候选方案x_lj与最佳方案之间的偏好值P_l。Step 309 , for a certain access point l, calculate the preference value P _l between the candidate solution x _lj and the best solution.

公式如下：The formula is as follows:

${P P}_{l l} = = \frac{{Q Q}_{l l w w o o r r s the s t t}}{{Q Q}_{l l b b e e s the s t t} + + {Q Q}_{l l w w o o r r s the s t t}}$

其中P_l代表的是用户K对于第l个接入点的偏好值。Among them, P _l represents user K's preference value for the lth access point.

步骤310、计算所有用户分别对于每个接入点的偏好值，并按序排列，生成用户选择矩阵R；Step 310, calculating the preference values of all users for each access point, and arranging them in order to generate a user selection matrix R;

如下所示：As follows:

其中，P_mO代表第O个用户选择第m接入点的偏好值。每一行代表的所有O个用户选择每个相同接入点的偏好值；用户选择矩阵R的列表示每个用户选择各个不同接入点的偏好值；Among them, P _mO represents the preference value of the Oth user choosing the mth access point. All O users represented by each row select the preference value of each same access point; the columns of the user selection matrix R represent each user's preference value of selecting each different access point;

步骤311、针对每个用户，本地控制将最大偏好值对应的接入点分配给该用户，并删除该用户和匹配的接入点，依次分配，直至所有用户都完成网络接入。Step 311 , for each user, the local control assigns the access point corresponding to the maximum preference value to the user, deletes the user and the matching access point, and assigns them in sequence until all users complete network access.

步骤四、用户K没有得到任何接入点的响应，则用户K直接向远端服务器发送请求，获取内容；远端服务器根据用户请求信息，利用流行度分析，完成缓存更新；Step 4: User K does not get a response from any access point, then user K directly sends a request to the remote server to obtain the content; the remote server completes the cache update according to the user request information by using popularity analysis;

步骤五、每个用户与各自的接入点匹配后，进行子载波分配，使用户与接入点之间进行通信。Step 5. After each user is matched with its own access point, subcarrier allocation is performed to enable communication between the user and the access point.

步骤501、初始化子载波集合和用户集合；Step 501, initialize subcarrier set and user set;

子载波集合为N＝{n'|n'＝1,2,...,N}，用户集合I＝{i'|i'＝1,2,...,I}，分配给用户i'的子载波索引X_i'＝φ。The set of subcarriers is N={n'|n'=1,2,...,N}, the user set I={i'|i'=1,2,...,I}, allocated to user i 'The subcarrier index Xi _' = φ.

步骤502、根据注水算法算出每个用户所对应每个子载波的发射功率和信道容量。Step 502: Calculate the transmit power and channel capacity of each subcarrier corresponding to each user according to the water filling algorithm.

发射功率p_i',n'：Transmit power p _i',n' :

${p p}_{{i i}^{' '},, {n no}^{' '}} = = \frac{{P P}_{t t o o t t} + + {Σ Σ}_{{n no}^{' '} = = 11}^{N N} 11 / / {γ γ}_{{i i}^{' '},, {n no}^{' '}}}{N N} - - 11 / / {γ γ}_{{i i}^{' '},, {n no}^{' '}}$

p_i',n'表示对用户i'分配子载波n'的功率，P_tot表示最大发射功率；γ_i',n'表示对用户i'分配子载波n'的信噪比。p _i',n' represents the power of subcarrier n' assigned to user i', P _tot represents the maximum transmission power; γ _i',n' represents the signal-to-noise ratio of subcarrier n' assigned to user i'.

信道容量C_i',n'：Channel capacity C _i',n' :

${C C}_{{i i}^{' '},, {n no}^{' '}} = = \frac{B B}{N N} {log log}_{22} ((11 + + {p p}_{{i i}^{' '},, {n no}^{' '}} {γ γ}_{{i i}^{' '},, {n no}^{' '}}))$

B代表该系统总带宽；B represents the total bandwidth of the system;

步骤503、针对子载波n'，分别计算每个用户下该子载波的信道容量，并进行降序排列选出最大值 Step 503, for the subcarrier n', calculate the channel capacity of the subcarrier for each user, and arrange in descending order to select the maximum value

子载波n'初始值为1；The initial value of subcarrier n' is 1;

表示为：Expressed as:

${C C}_{{n no}^{' '}}^{max max} = = {max max}_{{i i}^{' '} &Element; &Element; I I} {{{C C}_{{i i}^{' '},, {n no}^{' '}}}}$

步骤504、将子载波n'分配给最大信道容量值对应的用户，并将子载波n'从子载波集合N中移除，返回步骤503继续按序选取下一个子载波，直到将所有的子载波分配完。Step 504, assign subcarrier n' to the maximum channel capacity value The corresponding user removes the subcarrier n' from the subcarrier set N, returns to step 503 and continues to select the next subcarrier in sequence until all subcarriers are allocated.

步骤505、对分配后的子载波进行注水，根据子载波的发射功率p_i',n'计算系统频谱利用效率；Step 505, filling the allocated subcarriers with water, and calculating the system spectrum utilization efficiency according to the transmit power p _i',n' of the subcarriers;

系统频谱利用效率最大化目标函数如下：The objective function of system spectrum utilization efficiency maximization is as follows:

${max max}_{{{{a a}_{{i i}^{' '},, {n no}^{' '}},, {p p}_{{i i}^{' '},, {n no}^{' '}}}}} \frac{C C}{B B} = = {max max}_{{{{a a}_{{i i}^{' '},, {n no}^{' '}},, {p p}_{{i i}^{' '},, {n no}^{' '}}}}} {Σ Σ}_{{i i}^{' '} = = 11}^{I I} {Σ Σ}_{{n no}^{' '} = = 11}^{N N} {a a}_{{i i}^{' '},, {n no}^{' '}} {log log}_{22} ((11 + + {p p}_{{i i}^{' '},, {n no}^{' '}} {γ γ}_{{i i}^{' '},, {n no}^{' '}}))$

要满足的条件为：The conditions to be met are:

$\{\begin{matrix} {Σ Σ}_{{i i}^{' '} = = 11}^{I I} {Σ Σ}_{{n no}^{' '} = = 11}^{N N} {a a}_{{i i}^{' '},, {n no}^{' '}} {p p}_{{i i}^{' '},, {n no}^{' '}} \leq \leq {P P}_{t t o o t t} \\ {p p}_{{i i}^{' '},, {n no}^{' '}} &GreaterEqual; &Greater Equal; 00 \\ \underset{{i i}^{' '}}{Σ Σ} {a a}_{{i i}^{' '},, {n no}^{' '}} = = 11 \end{matrix}$

{a_i',n'}表示子载波分配集合，值为0或1，1表示将子载波n'分配给用户i'，0表示子载波n'没有分配给用户i'；{a _i',n' } indicates the subcarrier allocation set, the value is 0 or 1, 1 indicates that subcarrier n' is allocated to user i', and 0 indicates that subcarrier n' is not allocated to user i';

第一个约束条件表示总功率的限制条件，其中P_tot表示最大发射功率；第二个约束条件表示用户i'分配子载波n'的功率大于等于0；最后一个约束条件表示每一个子载波只能分配一次。The first constraint condition represents the constraint condition of the total power, where P _tot represents the maximum transmission power; the second constraint condition indicates that the power of user i'allocating subcarrier n' is greater than or equal to 0; the last constraint condition indicates that each subcarrier only can be assigned once.

本发明的优点在于：The advantages of the present invention are:

1)、一种基于缓存的超密集网络下联合动态接入和子载波分配的方法，根据仿真结果可以看出，该方法有效地提高了频谱利用效率，这一结果证明了该机制在密集场景下满足多种业务需求的可行性和适用性。1) A cache-based joint dynamic access and subcarrier allocation method in an ultra-dense network. According to the simulation results, it can be seen that this method effectively improves the spectrum utilization efficiency. This result proves that the mechanism is effective in dense scenarios. Feasibility and applicability to meet various business needs.

2)、一种基于缓存的超密集网络下联合动态接入和子载波分配的方法，可以综合多个因素完成接入选择，实现资源管理效率的提升。2) A method for joint dynamic access and subcarrier allocation based on a cache-based ultra-dense network, which can integrate multiple factors to complete access selection and improve resource management efficiency.

3)、一种基于缓存的超密集网络下联合动态接入和子载波分配的方法，可以实现子载波的动态分配，使得频谱利用效率显著提升。。3) A cache-based method for joint dynamic access and subcarrier allocation in an ultra-dense network, which can realize dynamic allocation of subcarriers and significantly improve spectrum utilization efficiency. .

附图说明Description of drawings

图1是本发明的系统模型示意图；Fig. 1 is a schematic diagram of a system model of the present invention;

图2是本发明的系统模型架构框图；Fig. 2 is a system model architecture block diagram of the present invention;

图3是本发明基于缓存的超密集网络下联合动态接入和子载波分配的方法流程图；Fig. 3 is a flow chart of the method for joint dynamic access and subcarrier allocation under the cache-based ultra-dense network of the present invention;

图3a是本发明本地控制将最佳接入点分配给用户的流程图；Fig. 3a is a flow chart of the present invention for local control to assign the best access point to users;

图3b是本发明每个用户与各自的接入点进行子载波分配的流程图；FIG. 3b is a flow chart of subcarrier allocation between each user and their respective access points in the present invention;

图4是本发明多属性决策算法下4个用户的权重因子仿真示意图；Fig. 4 is the emulation schematic diagram of the weight factor of 4 users under the multi-attribute decision-making algorithm of the present invention;

图5是本发明网络接入选择偏好排序仿真示意图；Fig. 5 is a schematic diagram of network access selection preference sorting simulation in the present invention;

图6是本发明频谱利用效率与用户个数关系图；Fig. 6 is a graph showing the relationship between spectrum utilization efficiency and the number of users in the present invention;

图7是本发明系统频谱效率与信噪比关系图；Fig. 7 is a diagram of the relationship between spectral efficiency and signal-to-noise ratio of the system of the present invention;

具体实施方式detailed description

下面将结合附图对本发明作进一步的详细说明。The present invention will be further described in detail below in conjunction with the accompanying drawings.

如图1所示，超密集网络中采用远端服务器作为核心网，连接多个接入点，接入点根据内容流行度分析缓存一部分内容，每个用户分别连接一个接入点，如图2所示，接入点将各自的属性参量传送给本地控制，本地控制根据收到的用户请求信息和接入点属性参量，利用多属性决策矩阵为用户动态选择接入点，选择过程中，需要综合考虑接入点的各个属性，如缓存容量、时延、信噪比等进行多属性接入选择。完成动态接入过程后，在贪婪算法的基础上完成子载波分配，基于发送功率、频谱带宽以及缓存大小限制对AP和子载波进行选取，以最大化频谱效率为目标，完成子载波分配。As shown in Figure 1, a remote server is used as the core network in an ultra-dense network, and multiple access points are connected. The access point analyzes and caches part of the content according to the popularity of the content, and each user connects to an access point, as shown in Figure 2. As shown in , the access points transmit their respective attribute parameters to the local control, and the local control uses the multi-attribute decision matrix to dynamically select the access point for the user according to the received user request information and the attribute parameters of the access point. During the selection process, it is necessary to Comprehensively consider various attributes of the access point, such as buffer capacity, delay, signal-to-noise ratio, etc., for multi-attribute access selection. After the dynamic access process is completed, the subcarrier allocation is completed on the basis of the greedy algorithm, and APs and subcarriers are selected based on the transmission power, spectrum bandwidth, and buffer size limitations, and the subcarrier allocation is completed with the goal of maximizing spectrum efficiency.

本发明首先根据用户请求信息，采用广播方式寻找缓存内容，反馈结果用S表示，若某个接入点空闲且存在该缓存内容，则记为S＝1，否则记为S＝0。接入点将自己的属性信息如缓存容量、时延、信噪比等，传送给本地控制。若用户没有得到任何接入点的响应，则直接向服务器端发送请求，获取内容，服务器端根据用户请求信息，利用流行度分析，完成缓存更新；若用户得到接入点响应，根据用户的内容和业务类型请求，进行层次性分析，选择最佳的接入点，由本地控制通知被选择的接入点提供服务，完成接入选择，更新接入选择集合Q。The present invention first uses broadcasting to search for cached content according to user request information, and the feedback result is denoted by S. If a certain access point is idle and has the cached content, it is recorded as S=1, otherwise it is recorded as S=0. The access point transmits its own attribute information, such as buffer capacity, delay, signal-to-noise ratio, etc., to the local control. If the user does not get a response from any access point, it will directly send a request to the server to obtain the content, and the server will use the popularity analysis to complete the cache update according to the user request information; if the user gets a response from the access point, it will According to the service type request, perform hierarchical analysis, select the best access point, and the local control notifies the selected access point to provide services, completes the access selection, and updates the access selection set Q.

根据接入选择结果，在已知信道状态信息的前提下，通过注水算法算出每个用户对应子载波的信道容量，在计算出的信道容量的基础上，对具有最好信道容量的子载波优先分配，然后进行迭代运算，最终完成子载波分配。According to the access selection result, on the premise of knowing the channel state information, the channel capacity of each user’s corresponding subcarrier is calculated through the water filling algorithm, and on the basis of the calculated channel capacity, the subcarrier with the best channel capacity is given priority Allocation, and then iterative operation, and finally complete subcarrier allocation.

一种基于缓存的超密集网络下联合动态接入和子载波分配的方法，如图3所示，具体步骤如下：A method for joint dynamic access and subcarrier allocation based on a cache-based ultra-dense network, as shown in Figure 3, the specific steps are as follows:

步骤一、多个用户同时用广播方式发送请求信息给所有接入点，寻找缓存内容；Step 1. Multiple users broadcast request information to all access points at the same time, looking for cached content;

请求信息包括：缓存内容；用户数量为O个；The request information includes: cache content; the number of users is O;

步骤二、将用户K作为当前用户，根据用户K的请求信息，各个接入点判断是否存在当前用户请求的缓存内容，如果某个接入点空闲且存在该缓存内容，该接入点反馈1给用户K，进入步骤三；否则反馈0；进入步骤四；Step 2. User K is taken as the current user. According to the request information of user K, each access point judges whether there is cached content requested by the current user. If an access point is idle and has the cached content, the access point will feedback 1 Give user K, go to step 3; otherwise feedback 0; go to step 4;

1≤K≤O；1≤K≤O;

本地控制根据用户的请求内容和业务类型，进行层次性分析，选择最佳的接入点，通知被选择的接入点提供服务，完成接入选择，然后更新接入选择集合Q。The local control performs hierarchical analysis according to the user's request content and service type, selects the best access point, notifies the selected access point to provide services, completes the access selection, and then updates the access selection set Q.

如图3a所示，具体步骤如下：As shown in Figure 3a, the specific steps are as follows:

步骤301、根据用户K的请求信息，针对每个候选接入点的n个属性参量，本地控制对每两个属性参量之间的相对重要性进行一一比较，得到决策矩阵M：Step 301. According to the request information of user K, for n attribute parameters of each candidate access point, the local control compares the relative importance of each two attribute parameters one by one to obtain a decision matrix M:

缓存内容不同，对于接入点选择的属性要求不同，构造判决矩阵来表示每个候选接入点各个属性参量之间的重要性关系，通过两两比较不同属性之间的重要性，以决定某个属性的重要程度，通常按1～9的比例标度对重要性程度进行赋值，表1中给出了1～9标度的含义：The cache content is different, and the attribute requirements for access point selection are different. A decision matrix is constructed to represent the importance relationship between each attribute parameter of each candidate access point, and the importance of different attributes is compared in pairs to determine a certain The importance of each attribute is usually assigned on a scale of 1 to 9. The meaning of the scale of 1 to 9 is given in Table 1:

表1Table 1

其中a_ij代表接入点中参量i与参量j的的相对重要性比较值；Wherein a _ij represents the relative importance comparison value of parameter i and parameter j in the access point;

步骤303、对标准化后的决策矩阵B进行一致性校验；判断决策矩阵是否有效，如果有效，进行步骤304，否则返回步骤302；Step 303, check the consistency of the standardized decision matrix B; judge whether the decision matrix is valid, if it is valid, go to step 304, otherwise return to step 302;

在计算各属性权值之前，需要对决策矩阵进行一致性校验。因为如果决策矩阵过于偏离一致性，计算得到的权重向量将不具可信性，因此有必要对决策矩阵B的一致性进行校验。定义如下：Before calculating the weight of each attribute, it is necessary to check the consistency of the decision matrix. Because if the decision matrix deviates too much from the consistency, the calculated weight vector will not be credible, so it is necessary to check the consistency of the decision matrix B. It is defined as follows:

b_ik×b_kj＝b_ij,i,j,k＝1,2,…,n(3)b _ik ×b _kj ＝b _ij ,i,j,k=1,2,...,n(3)

b_ik为标准化后接入点中参量i与参量k的的相对重要性比较值；b _ik is the relative importance comparison value of parameter i and parameter k in the access point after standardization;

如果等式成立的话，则表示判断矩阵具有一致性。即表示决策者在进行属性的两两比较时，思维具有一致性。但是由于人的思维具有一定的主观性，难以保持绝对的一致性，所以为了衡量矩阵的一致性，引入不一致性指标CI、一致性比率CR等概念：If the equation holds true, it means that the judgment matrix is consistent. That is to say, decision makers have consistency in their thinking when making pairwise comparisons of attributes. However, due to the subjectivity of human thinking, it is difficult to maintain absolute consistency, so in order to measure the consistency of the matrix, concepts such as inconsistency index CI and consistency ratio CR are introduced:

$C C I I = = \frac{{λ λ}_{m m a a x x} - - n no}{n no - - 11} - - - - - - ((44))$

$C C R R = = \frac{C C I I}{R R I I} - - - - - - ((55))$

其中，λ_max是决策矩阵B的最大特征值，n是决策矩阵B中属性参量的个数，RI是已知的平均随机一致性指标，如下表所示：Among them, λ _max is the maximum eigenvalue of the decision matrix B, n is the number of attribute parameters in the decision matrix B, and RI is the known average random consistency index, as shown in the following table:

表2Table 2

nno 11 22 33 44 55 66 77 88 99 RIRI 00 00 0.580.58 0.900.90 1.121.12 1.241.24 1.321.32 1.411.41 1.451.45

步骤304、计算决策矩阵B中n个属性参量综合产生的的权重向量ω；Step 304, calculating the weight vector ω generated by the synthesis of n attribute parameters in the decision matrix B;

ω＝(ω₁,ω₂,...ω_j,...,ω_n)ω＝(ω ₁ ,ω ₂ ,...ω _j ,...,ω _n )

ω_j为第j个属性参量的权重；本发明采用归一化后的特征向量作为网络接入选择的权重向量。ω _j is the weight of the jth attribute parameter; the present invention uses the normalized feature vector as the weight vector for network access selection.

步骤305、生成m个候选接入节点的所有属性参量的状态矩阵S；Step 305, generating a state matrix S of all attribute parameters of m candidate access nodes;

步骤307、根据加权决策矩阵Q，确定最佳接入方案Q_best和最差接入方案Q_worst。Step 307 , according to the weighted decision matrix Q, determine the best access scheme Q _best and the worst access scheme Q _worst .

最佳接入方案即是选择每一个参量的最好的情况，比如说对于缓存容量来讲，就选择最大值，对于功耗来讲，就选择最小值。而最差接入方案则与之相反，计算方式如下：The best access solution is to select the best situation for each parameter. For example, for the cache capacity, the maximum value is selected, and for the power consumption, the minimum value is selected. The worst access scheme is the opposite, and the calculation method is as follows:

Q_best＝(ω₁·s_1best,ω₂·s_2best,...ω_j·s_jbest,...,ω_n·s_nbest)(8)Q _best ＝(ω ₁ ·s _1best ,ω ₂ ·s _2best ,...ω _j ·s _jbest ,...,ω _n ·s _nbest )(8)

Q_worst＝(ω₁·s_1worst,ω₂·s_2worst,...ω_j·s_jworst,...,ω_n·s_nworst)(9)Q _worst ＝(ω ₁ ·s _1worst ,ω ₂ ·s _2worst ,...ω _j ·s _jworst ,...,ω _n ·s _nworst )(9)

${Q Q}_{l l b b e e s the s t t} = = \sqrt{{Σ Σ}_{j j = = 11}^{n no} {ω ω}_{j j} {(({x x}_{l l j j} - - {s the s}_{j j b b e e s the s t t}))}^{22}} - - - - - - ((1010))$

${Q Q}_{l l w w o o r r s the s t t} = = \sqrt{{Σ Σ}_{j j = = 11}^{n no} {ω ω}_{j j} {(({x x}_{l l j j} - - {s the s}_{j j w w o o r r s the s t t}))}^{22}} - - - - - - ((1111))$

通过计算偏好值，得出候选网络与最优网络和最差网络之间距离的比值。计算的公式如下：By calculating the preference value, the ratio of the distance between the candidate network and the optimal network and the worst network is obtained. The calculation formula is as follows:

${P P}_{l l} = = \frac{{Q Q}_{l l w w o o r r s the s t t}}{{Q Q}_{l l b b e e s the s t t} + + {Q Q}_{l l w w o o r r s the s t t}} - - - - - - ((1212))$

用户K对每个接入点的最大偏好值：P_k＝(P_1k,P_2k,...P_lk,...P_mk)；User K's maximum preference value for each access point: P _k = (P _1k , P _2k ,...P _lk ,...P _mk );

步骤310、计算所有用户分别对于每个接入点的偏好值，并按序排列，生成用户选择矩阵R；如下所示：Step 310, calculate the preference values of all users for each access point, and arrange them in order to generate a user selection matrix R; as shown below:

其中，P_mO代表第O个用户选择第m接入点的偏好值。每一行代表的O个用户选择各个接入点的偏好值；Among them, P _mO represents the preference value of the Oth user choosing the mth access point. The O users represented by each row select the preference value of each access point;

本地控制根据用户选择矩阵R的列，选择最大的偏好值，将该偏好值对应的接入点分配给对应用户，分配后删除匹配的用户和接入点，继续依次分配，直至所有用户都完成网络接入。The local control selects the largest preference value according to the column of the user selection matrix R, assigns the access point corresponding to the preference value to the corresponding user, deletes the matching user and access point after assignment, and continues to assign in sequence until all users complete Internet access.

用户选择矩阵R每一列表示，每一个用户对m个接入点的偏好值；Each column of the user selection matrix R represents each user's preference value for m access points;

表3接入点的仿真参数Table 3 Simulation parameters of the access point

信噪比(/dBm)SNR(/dBm) 时延(/ms)Latency (/ms) 覆盖半径(km)Coverage radius (km) 缓存容量(TB)Cache capacity (TB) 功耗(W)power consumption(W) 接入点1access point 1 7070 2525 5050 100100 1/1001/100 接入点2access point 2 6060 5050 4040 200200 1/1001/100 接入点3access point 3 7373 8080 7070 400400 1/501/50 接入点4access point 4 5050 3030 3030 200200 1/701/70

根据内容流行度分析法，内容命中率与Zipf指数有关。在本次仿真中，设定Zipf参数为0.8，缓存命中率为0.7，不失一般性，设定得到接入点响应的用户个数为4，其余2个用户需要从后端服务器获取内容。这4个用户处于不同的区域，有着不同的业务。用户1属于中心用户，请求网络浏览业务；用户2属于中心用户，请求流媒体业务；用户3属于边缘用户，请求流媒体业务；用户4属于边缘用户，请求网络浏览业务。根据本节提出的多属性决策算法，仿真结果如图4所示，横坐标表示了5个参量：信噪比，时延，覆盖半径，缓存容量，功耗。纵坐标表示对应每个参量的权重值，可以看出针对不同的用户请求，权重因子是有很大差异的。对于流媒体业务来说，主要是单向传输，不需要双向的实时通信，在时延上要求不高，但是文件一般较大，需要较高的缓存容量，用户2和用户3都是请求流媒体业务，所以时延因子权重较低，缓存容量因子权重较高。对于网络浏览业务来说，缓存容量也低，但需要较高的服务质量和较低的时延，所以对信噪比要求较高，如用户1和用户4请求网络浏览业务，时延因子的权重较高，缓存容量因子的权重较低。中心用户对覆盖半径要求较低，所以用户1和用户2的覆盖半径因子的权重较低；而用户3和用户4覆盖半径因子的权重较高。According to the content popularity analysis method, the content hit rate is related to the Zipf index. In this simulation, the Zipf parameter is set to 0.8, the cache hit rate is 0.7, without loss of generality, the number of users who get the response from the access point is set to 4, and the remaining 2 users need to obtain content from the back-end server. These four users are in different regions and have different services. User 1 belongs to the central user and requests the web browsing service; user 2 belongs to the central user and requests the streaming media service; user 3 belongs to the edge user and requests the streaming media service; user 4 belongs to the edge user and requests the web browsing service. According to the multi-attribute decision-making algorithm proposed in this section, the simulation results are shown in Figure 4, and the abscissa indicates five parameters: signal-to-noise ratio, delay, coverage radius, cache capacity, and power consumption. The vertical axis represents the weight value corresponding to each parameter. It can be seen that the weight factor is very different for different user requests. For streaming media business, it is mainly one-way transmission, no two-way real-time communication is required, and the requirement for time delay is not high, but the file is generally large and requires a high cache capacity. Both user 2 and user 3 are request streams For media services, the weight of the delay factor is low, and the weight of the cache capacity factor is high. For web browsing services, the cache capacity is also low, but higher quality of service and lower delay are required, so the signal-to-noise ratio is relatively high. For example, when users 1 and 4 request web browsing services, the delay factor The higher the weight, the lower the cache capacity factor. The central users have lower requirements on the coverage radius, so the weights of the coverage radius factors of users 1 and 2 are lower; while the weights of the coverage radius factors of users 3 and 4 are higher.

根据权重因子和各个接入点的属性信息，得到了每个用户对于接入点的选择偏好值，仿真结果如图5所示：According to the weight factor and the attribute information of each access point, the selection preference value of each user for the access point is obtained, and the simulation results are shown in Figure 5:

由图5可以看出，对于同一个接入点来说，其参量信息是不变的，但是不同用户计算得到的偏好值却是不同的，说明本文提出的方法充分考虑了用户的需求信息，对于同一用户来讲，其网络需求也是不变的，但是在不同的接入点下计算得到的偏好值也是不同的，说明算法也充分考虑了接入点的参量信息。因而，本文提出的网络接入选择算法同时兼顾用户需求和接入点性能，是有效进行子载波分配的基础。最终用户的网络接入选择结果如表4所示：It can be seen from Figure 5 that for the same access point, its parameter information remains unchanged, but the preference values calculated by different users are different, indicating that the method proposed in this paper fully considers the user's demand information, For the same user, the network requirements are also unchanged, but the preference values calculated under different access points are also different, indicating that the algorithm also fully considers the parameter information of the access point. Therefore, the network access selection algorithm proposed in this paper takes into account both user requirements and access point performance, which is the basis for effective subcarrier allocation. The network access selection results of end users are shown in Table 4:

表4用户接入选择结果Table 4 User access selection results

用户user 接入点\后端服务器Access Point\Backend Server 用户1user 1 接入点3access point 3 用户2user 2 接入点2access point 2 用户3user 3 接入点1access point 1 用户4user 4 接入点4access point 4 用户5user 5 后端服务器backend server 用户6user 6 后端服务器backend server

完成用户接入选择之后，在贪婪算法的基础上，提出了一种基于系统容量最大化子载波优先分配算法。在已知信道状态信息的前提下，通过注水算法算出每个用户对应子载波的信道容量，然后在计算的信道容量的基础上，对具有最好信道容量的子载波优先分配。After the user access selection is completed, on the basis of the greedy algorithm, a subcarrier priority allocation algorithm based on system capacity maximization is proposed. On the premise of knowing the channel state information, the channel capacity of the subcarriers corresponding to each user is calculated through the water filling algorithm, and then on the basis of the calculated channel capacity, the subcarriers with the best channel capacity are allocated preferentially.

如图3b所示，具体步骤如下：As shown in Figure 3b, the specific steps are as follows:

发射功率p_i',n'：Transmit power p _i',n' :

${p p}_{{i i}^{' '},, {n no}^{' '}} = = \frac{{P P}_{t t o o t t} + + {Σ Σ}_{{n no}^{' '} = = 11}^{N N} 11 / / {γ γ}_{{i i}^{' '},, {n no}^{' '}}}{N N} - - 11 / / {γ γ}_{{i i}^{' '},, {n no}^{' '}} - - - - - - ((1414))$

信道容量C_i'，n'：Channel capacity C _i',n' :

${C C}_{{i i}^{' '},, {n no}^{' '}} = = \frac{B B}{N N} {log log}_{22} ((11 + + {p p}_{{i i}^{' '},, {n no}^{' '}} {γ γ}_{{i i}^{' '},, {n no}^{' '}})) - - - - - - ((1515))$

B代表总系统架构模型的带宽；B represents the bandwidth of the total system architecture model;

子载波n'初始值为1；The initial value of subcarrier n' is 1;

表示为：Expressed as:

${C C}_{{n no}^{' '}}^{max max} = = {max max}_{{i i}^{' '} &Element; &Element; I I} {{{C C}_{{i i}^{' '},, {n no}^{' '}}}} - - - - - - ((1616))$

如子载波信道容量按降序排列为：For example, the subcarrier channel capacities are arranged in descending order as follows:

${C C}_{11,, n no} &GreaterEqual; &Greater Equal; {C C}_{22,, n no} &GreaterEqual; &Greater Equal; {C C}_{33,, n no} &GreaterEqual; &Greater Equal; ... ... &GreaterEqual; &Greater Equal; {C C}_{I I,, n no},, {C C}_{n no}^{max max} = = {C C}_{11,, n no}$

${max max}_{{{{a a}_{{i i}^{' '},, {n no}^{' '}},, {p p}_{{i i}^{' '},, {n no}^{' '}}}}} \frac{C C}{B B} = = {max max}_{{{{a a}_{{i i}^{' '},, {n no}^{' '}},, {p p}_{{i i}^{' '},, {n no}^{' '}}}}} {Σ Σ}_{{i i}^{' '} = = 11}^{I I} {Σ Σ}_{{n no}^{' '} = = 11}^{N N} {a a}_{{i i}^{' '},, {n no}^{' '}} {log log}_{22} ((11 + + {p p}_{{i i}^{' '},, {n no}^{' '}} {γ γ}_{{i i}^{' '},, {n no}^{' '}})) - - - - - - ((1717))$

要满足的条件为：The conditions to be met are:

$\{\begin{matrix} {Σ Σ}_{{i i}^{' '} = = 11}^{I I} {Σ Σ}_{{n no}^{' '} = = 11}^{N N} {a a}_{{i i}^{' '},, {n no}^{' '}} {p p}_{{i i}^{' '},, {n no}^{' '}} \leq \leq {P P}_{t t o o t t} \\ {p p}_{{i i}^{' '},, {n no}^{' '}} &GreaterEqual; &Greater Equal; 00 \\ \underset{{i i}^{' '}}{Σ Σ} {a a}_{{i i}^{' '},, {n no}^{' '}} = = 11 \end{matrix} - - - - - - ((1818))$

B代表总系统架构模型的带宽，{a_i',n'}表示子载波分配集合，值为0或1，1表示将子载波n'分配给用户i'，0表示子载波n'没有分配给用户i'；B represents the bandwidth of the total system architecture model, {a _i',n' } represents the subcarrier allocation set, the value is 0 or 1, 1 indicates that subcarrier n' is allocated to user i', 0 indicates that subcarrier n' is not allocated to user i';

本实施例中仿真采用瑞利衰落信道模型，子载波数为64个，用户数为16个，总功率P_tot为1W，噪声功率谱密度N₀为10e-8W/Hz，总带宽B为1MHz。并且与最小容量最大法(MAX-MIN)和公平比例法(FPS)这两种经典的子载波分配方法作对比，仿真结果如图6和图7所示：In this embodiment, the simulation adopts the Rayleigh fading channel model, the number of subcarriers is 64, the _number of users is 16, the total power P _tot is 1W, the noise power spectral density N is 10e-8W/Hz, and the total bandwidth B is 1MHz . And compared with the two classic subcarrier allocation methods, the minimum capacity maximum method (MAX-MIN) and the fair proportional method (FPS), the simulation results are shown in Figure 6 and Figure 7:

由图6可以看出，在用户个数相同的情况下，本发明提出的算法所得到的频谱利用效率大大高于比例公平算法和最小容量最大化法，因为本发明的算法整体考虑系统容量，根据信道状态信息来分配子载波和功率，从而大大提高了系统的资源利用率。同时比例公平算法的频谱利用率略高于最小容量最大化法，因为虽然都考虑了公平性，但比例公平算法同时考虑了系统容量最大化，而最小容量最大化法只考虑了子载波在用户之间的分配，没有考虑功率在子载波之间的自适应分配，从而频谱利用效率率较低。It can be seen from Figure 6 that when the number of users is the same, the spectrum utilization efficiency obtained by the algorithm proposed by the present invention is much higher than that of the proportional fairness algorithm and the minimum capacity maximization method, because the algorithm of the present invention considers the system capacity as a whole, Subcarriers and power are allocated according to channel state information, thereby greatly improving system resource utilization. At the same time, the spectrum utilization rate of the proportional fairness algorithm is slightly higher than that of the minimum capacity maximization method, because although fairness is considered, the proportional fairness algorithm also considers the maximum capacity of the system, while the minimum capacity maximization method only considers the subcarriers in the user The allocation between subcarriers does not consider the adaptive allocation of power between subcarriers, so the spectrum utilization efficiency is low.

图7是频谱利用效率随信噪比变化的曲线，可以看出在相同信噪比的情况下，本文提出的算法所得到的频谱利用效率大大高于比例公平算法和最小容量最大化法，这是因为本算法通过不断进行迭代注水计算信道容量来分配子载波，可以更好地避免将具有较差信道质量的子载波分配给用户，从而可以提高系统频谱利用效率。Figure 7 is a curve of spectrum utilization efficiency changing with SNR. It can be seen that under the same SNR condition, the spectrum utilization efficiency obtained by the algorithm proposed in this paper is much higher than that of the proportional fairness algorithm and the minimum capacity maximization method. This is because the algorithm allocates subcarriers through continuous iterative water injection to calculate channel capacity, which can better avoid allocating subcarriers with poor channel quality to users, thereby improving the efficiency of system spectrum utilization.

Claims

1. A method for joint dynamic access and sub-carrier allocation under a buffer-based ultra-dense network, characterized in that, comprising the following steps:

Step 1. Multiple users broadcast request information to all access points to search for cached content;

The number of users is O;

Step 2. Taking user K as the current user, each access point judges whether there is cached content requested by the current user. If a certain access point is idle and has the cached content, the access point feeds back 1 to user K, and proceeds to step 3. ;Otherwise feedback 0; go to step 4;

Step 3, all access points satisfying user K transmit their attribute parameters to the local control, and the local control assigns the best access point to user K;

All access points satisfying user K are m;

Step 4: User K does not get a response from any access point, then user K directly sends a request to the remote server to obtain the content; the remote server completes the cache update according to the user request information by using popularity analysis;

Step 5. After each user is matched with their own access point, subcarrier allocation is performed to enable communication between the user and the access point;

Specifically include:

Step 501, initialize subcarrier set and user set;

The set of subcarriers is N={n'|n'=1,2,...,N}, the user set I={i'|i'=1,2,...,I}, allocated to user i 'The subcarrier index Xi _' = φ;

Step 502, calculate the transmit power and channel capacity of each subcarrier corresponding to each user according to the water filling algorithm;

Transmit power p _i',n' :

{p p}_{{i i}^{' '},, {n no}^{' '}} = = \frac{{P P}_{t t o o t t} + + {Σ Σ}_{{n no}^{' '} = = 11}^{N N} 11 / / {γ γ}_{{i i}^{' '},, {n no}^{' '}}}{N N} - - 11 / / {γ γ}_{{i i}^{' '},, {n no}^{' '}}

p _{i', n'} represents the power of subcarrier n' assigned to user i', P _tot represents the maximum transmit power; γ _{i', n'} represents the signal-to-noise ratio of subcarrier n' assigned to user i';

Channel capacity C _i',n' :

{C C}_{{i i}^{' '},, {n no}^{' '}} = = \frac{B B}{N N} {log log}_{22} ((11 + + {p p}_{{i i}^{' '},, {n no}^{' '}} {γ γ}_{{i i}^{' '},, {n no}^{' '}}))

B represents the total bandwidth of the system;

Step 503, for the subcarrier n', calculate the channel capacity of the subcarrier for each user, and arrange in descending order to select the maximum value

The initial value of subcarrier n' is 1;

{C C}_{{n no}^{' '}}^{max max} = = {max max}_{{i i}^{' '} &Element; &Element; I I} {{{C C}_{{i i}^{' '},, {n no}^{' '}}}}

Step 504, assign subcarrier n' to the maximum channel capacity value corresponding user, and remove the subcarrier n' from the subcarrier set N, return to step 503 and continue to select the next subcarrier in sequence until all subcarriers are allocated;

Step 505, filling the allocated subcarriers with water, and calculating the system spectrum utilization efficiency according to the transmit power p _i',n' of the subcarriers;

The objective function of system spectrum utilization efficiency maximization is as follows:

{max max}_{{{{a a}_{{i i}^{' '},, {n no}^{' '}},, {p p}_{{i i}^{' '},, {n no}^{' '}}}}} \frac{C C}{B B} = = {max max}_{{{{a a}_{{i i}^{' '},, {n no}^{' '}},, {p p}_{{i i}^{' '},, {n no}^{' '}}}}} {Σ Σ}_{{i i}^{' '} = = 11}^{I I} {Σ Σ}_{{n no}^{' '} = = 11}^{N N} {a a}_{{i i}^{' '},, {n no}^{' '}} {log log}_{22} ((11 + + {p p}_{{i i}^{' '},, {n no}^{' '}} {γ γ}_{{i i}^{' '},, {n no}^{' '}}))

The conditions to be met are:

\{\begin{matrix} {Σ Σ}_{{i i}^{' '} = = 11}^{I I} {Σ Σ}_{{n no}^{' '} = = 11}^{N N} {a a}_{{i i}^{' '},, {n no}^{' '}} {p p}_{{i i}^{' '},, {n no}^{' '}} \leq \leq {P P}_{t t o o t t} \\ {p p}_{{i i}^{' '},, {n no}^{' '}} &GreaterEqual; &Greater Equal; 00 \\ \underset{{i i}^{' '}}{Σ Σ} {a a}_{{i i}^{' '},, {n no}^{' '}} = = 11 \end{matrix}

{a _i',n' } indicates the subcarrier allocation set, the value is 0 or 1, 1 indicates that subcarrier n' is allocated to user i', and 0 indicates that subcarrier n' is not allocated to user i';

The first constraint condition represents the constraint condition of the total power, where P _tot represents the maximum transmission power; the second constraint condition indicates that the power of user i'allocating subcarrier n' is greater than or equal to 0; the last constraint condition indicates that each subcarrier only can be assigned once.

2. The method for joint dynamic access and subcarrier allocation under a cache-based ultra-dense network according to claim 1, wherein the attribute parameters of each access point include cache capacity, time delay and Signal-to-noise ratio, etc., a total of n attributes.

3. The method for joint dynamic access and subcarrier allocation under a cache-based ultra-dense network according to claim 1, wherein said step three is specifically:

Step 301, according to the request information of user K, for each candidate access point, the local control compares the relative importance of every two attribute parameters among the n attribute parameters one by one, and obtains the decision matrix M:

a _ij represents the relative importance comparison value of attribute parameter i and attribute parameter j in the access point; a _ii =1;

Step 302. Normalize the decision matrix M to obtain a standardized decision matrix B:

Where b _ij represents the normalized value of the comparison value a _ij in the access point;

Step 303, check the consistency of decision matrix B, judge whether the decision matrix is valid, if valid, go to step 304, otherwise return to step 302;

The consistency ratio CR is defined as follows:

Among them, CI represents the inconsistency index: λ _max is the maximum eigenvalue of the decision matrix B, n is the number of attribute parameters in the decision matrix B, and RI is the known average random consistency index;

When CR<0.1, the decision matrix B is considered to have acceptable consistency, otherwise, the decision matrix B is reconstructed;

Step 304, obtaining the weight vector ω generated by the synthesis of n attribute parameters in the decision matrix B;

ω＝(ω ₁ ,ω ₂ ,...ω _j ,...,ω _n )

ω _j is the weight of the jth attribute parameter;

Step 305, generating a state matrix S of all attribute parameters for the m candidate access points;

The state matrix S has m rows and n columns, and each row represents n attribute parameters of each access point;

Among them, s _mn represents the value of the mth access point corresponding to the nth attribute parameter;

Step 306, multiply the weight vector ω by the state matrix S to obtain the weighted decision matrix Q:

Step 307, according to the weighted decision matrix Q, determine the best access scheme Q _best and the worst access scheme Q _worst ,

Q _best ＝(ω ₁ ·s _1best ,ω ₂ ·s _2best ,...ω _j ·s _jbest ,...,ω _n ·s _nbest )

Q _worst ＝(ω ₁ ·s _1worst ,ω ₂ ·s _2worst ,...ω _j ·s _jworst ,...,ω _n ·s _nworst )

s _jbest represents the best value of the jth attribute parameter of all m access points; s _jworst represents the worst value of the jth attribute parameter of all m access points;

Step 308. For a certain access point l, calculate the Euclidean distance between the candidate access scheme x _lj and the best access scheme Q _best , and the Euclidean distance between the candidate access scheme x _lj and the worst access scheme Q _worst distance;

The Euclidean distance between the candidate access scheme x _lj and the best access scheme Q _best is Q _lbest , which specifically refers to the Euclidean distance between each attribute parameter of the candidate access scheme x _lj and the optimal value s _jbest of the attribute, as follows:

{Q Q}_{l l b b e e s the s t t} = = \sqrt{{Σ Σ}_{j j = = 11}^{n no} {ω ω}_{j j} {(({x x}_{l l j j} - - {s the s}_{j j b b e e s the s i i}))}^{22}}

The Euclidean distance between the candidate access scheme x _lj and the worst access scheme Q _worst is Q _lworst , which specifically refers to the Euclidean distance between each attribute parameter of the candidate access scheme x _lj and the worst value s _jworst of the attribute, as follows:

{Q Q}_{l l w w o o r r s the s t t} = = \sqrt{{Σ Σ}_{j j = = 11}^{n no} {ω ω}_{j j} {(({x x}_{l l j j} - - {s the s}_{j j w w o o r r s the s i i}))}^{22}}

Step 309. For a certain access point l, calculate the preference value P _l between the candidate solution x _lj and the best solution;

The formula is as follows:

{P P}_{l l} = = \frac{{Q Q}_{l l w w o o r r s the s t t}}{{Q Q}_{l l b b e e s the s t t} + + {Q Q}_{l l w w o o r r s the s t t}}

Where P _l represents user K's preference value for the lth access point;

Step 310, calculating the preference values of all users for each access point, and arranging them in order to generate a user selection matrix R;

Among them, P _m0 represents the preference value of the Oth user choosing the mth access point; each row represents the preference value of all O users choosing each same access point; the column of the user selection matrix R represents each user’s choice of each Preference values for different access points;

Step 311 , for each user, the local control assigns the access point corresponding to the maximum preference value to the user, deletes the user and the matching access point, and assigns them in sequence until all users complete network access.