WO2021232848A1 - 一种异构网络下基于支持向量机的资源分配方法 - Google Patents
一种异构网络下基于支持向量机的资源分配方法 Download PDFInfo
- Publication number
- WO2021232848A1 WO2021232848A1 PCT/CN2021/074165 CN2021074165W WO2021232848A1 WO 2021232848 A1 WO2021232848 A1 WO 2021232848A1 CN 2021074165 W CN2021074165 W CN 2021074165W WO 2021232848 A1 WO2021232848 A1 WO 2021232848A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- user
- wifi
- support vector
- network
- vector machine
- Prior art date
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W24/00—Supervisory, monitoring or testing arrangements
- H04W24/06—Testing, supervising or monitoring using simulated traffic
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W72/00—Local resource management
- H04W72/50—Allocation or scheduling criteria for wireless resources
- H04W72/53—Allocation or scheduling criteria for wireless resources based on regulatory allocation policies
Definitions
- the technical field of wireless communication of the present invention specifically relates to a resource allocation method based on a support vector machine in a heterogeneous network.
- the traffic offloading between cellular network and WiFi network is the most common. Its implementation form is to offload user traffic in cellular network to WiFi network to alleviate the problem of tight cellular network bandwidth.
- WiFi network and cellular network can be offloaded. Traffic offloading is achieved due to: 1) The frequency band used by the WiFi network is an unlicensed frequency band, which can overlap with the cellular network without interfering with each other; 2) The WiFi network has sufficient bandwidth to provide high-throughput and high-reliability wireless communication; 3) The WiFi network technology is mature, the access point is cheap, and it is widely used. It has a high penetration rate in public or indoor environments. Because it can greatly increase the network system capacity, alleviate network congestion, and achieve low cost, cellular networks and WiFi networks Inter-flow offloading has caused widespread concern in the industry.
- the purpose of the invention In order to overcome the shortcomings in the prior art, provide a support vector machine-based resource allocation method under a heterogeneous network, and put more computing resources on the generation of data sets and training, so that in the actual application process
- the corresponding model can be directly called to solve the problem, which reduces the delay time of making a decision, and the decision of whether to uninstall a user is only related to the current user’s state, without considering the situation of other users.
- the status of the network user determines whether to perform uninstallation, so as to maximize the overall user satisfaction of the heterogeneous network.
- the present invention provides a support vector machine-based resource allocation method in a heterogeneous network, which includes the following steps:
- N RB N RB
- N WiFi the number of WiFi users of the access point corresponding to the user
- D BS the distance between the user and the base station
- D AP the distance between the user and the WiFi access point
- Step S4 is continuously executed until the judgment of all cellular network users is completed.
- the test set is used to test the accuracy of the support vector machine model in training, and the training model is saved after the requirements are met.
- the training process of the support vector machine model in the step S3 is:
- the final decision function is expressed by the following formula (8).
- the present invention constructs a user-centered support vector machine offloading model to improve user satisfaction in heterogeneous networks.
- the concept of user satisfaction is introduced to better meet the different QoS requirements of users; then the generated data set is used to train the support vector machine model, Turn it into a classification problem to reallocate network resources, judge whether to perform uninstall based on the status of WiFi network and cellular network users in each state, and find the best WiFi network access point to achieve overall user satisfaction on heterogeneous networks
- this method greatly reduces the complexity of the algorithm, increases the user rate, and finally achieves an increase in satisfaction.
- the present invention has the following advantages:
- the present invention takes user satisfaction as the measurement value, and makes offloading decisions based on the user's own rate requirements, which can effectively improve the satisfaction of cellular network users, avoiding the situation that the traditional resource allocation method allocates too much or too little resources, and is better It meets the needs of users.
- the present invention puts more computing resources on generating data sets and training, so that in the actual application process, the corresponding model can be directly called to solve the problem, and the delay time of making a decision is reduced.
- the present invention's decision on whether to uninstall a certain user is only related to the status of the current user, and there is no need to consider the situation of other users.
- Figure 1 is a schematic diagram of a system model of the method of the present invention
- Figure 2 is a schematic flow diagram of the method of the present invention.
- the present invention provides a support vector machine-based resource allocation method in a heterogeneous network, which is applied to the system model shown in Figure 1.
- the macro base station BS
- the four WiFi access points APs
- Cellular network users are randomly distributed on a circle with the base station BS as the center and a radius of R.
- the user positions of the WiFi network are randomly distributed on a circle with the AP access point as the center and r as the radius.
- N RB is used to represent the number of resource blocks (RB)
- P max is used to represent the maximum power of the base station
- B c is the bandwidth occupied by each resource block.
- N 0 is the thermal noise power density
- L k is the transmission path loss between the access point and user k
- L k is calculated using the Friis formula in the large-scale path loss model
- ⁇ is the wavelength of the signal
- G t is the gain of the transmitting antenna
- G r is the gain of the receiving antenna
- d is the distance between the user and the base station in m
- L is the loss factor independent of the propagation path
- ⁇ is the distance Attenuation factor. Since a user can transmit data through multiple resource blocks at the same time, the total rate of user k is:
- the WiFi network uses the CSMA/CA (Carrier Sense Multiple Access with Collision detection) protocol.
- the rate of user l in the WiFi network is, where B w represents the bandwidth of the WiFi network, T represents the total time slot of the WiFi, and Pm represents the power of the WiFi ,
- the time slot occupied by user l is t l :
- a model calculation scheme is used to simulate user satisfaction, which is a function of user data rate.
- R k of user k is higher than R required (data rate required by user k)
- user k’s satisfaction will slowly rise; when R k is lower than R required , user k’s satisfaction will drop sharply . Therefore, the definition of user satisfaction k is as follows:
- the offloading of heterogeneous network resources can be regarded as a dynamic game process.
- This embodiment defines this game as Where M is the set of all players in the game, Is the strategy set of player m, and the utility function U m of player m is defined as:
- S m and S- m are the strategies of the m-th AP and other APs except m
- ⁇ n is the user satisfaction of the m-th AP
- ⁇ k is the user satisfaction of the cellular network using the current strategy.
- Formula (8) shows that if any player changes strategy, the change in its own utility function is equal to the change in the potential function.
- Theorem 1 The game discussed in this embodiment is a precise potential game.
- the most important property of the precise potential game is that it has at least one pure strategy Nash equilibrium.
- the pure strategy Nash equilibrium of the game model proposed in this paper can maximize the overall user satisfaction, because the potential function set here is the total user satisfaction of the system.
- the BR algorithm is a classic algorithm for finding Nash equilibrium in game theory. Therefore, it can be used to find the optimal solution.
- an access point is randomly selected without repetition, the strategy of other access points is fixed, and its value is calculated.
- the total user satisfaction of all different strategies choose the best strategy among them to maximize Um in the current iteration; if the value of the total user satisfaction reaches convergence, stop the iteration.
- the above strategy is the best strategy, and the detailed process is as method 2.
- this embodiment applies a support vector machine-based resource allocation method in a heterogeneous network provided by the present invention to a system model. Referring to FIG. 2, the specific steps are as follows:
- N RB N RB
- N WiFi the number of WiFi users of the access point corresponding to the user
- D BS the distance between the user and the base station
- D AP the distance between the user and the WiFi access point
- Step S4 is continuously executed until the judgment of all cellular network users is completed.
- X i can be divided into two categories. Assuming that these two classes are linearly separable, at least one hyperplane defined by a vector W ⁇ R 4 and an offset b ⁇ R 4 can be found, which can separate the two classes without error.
- the decision is executed according to the value of the function sgn[f(X)], where f(X) is the decision function associated with the hyperplane, defined as:
- the distance between the nearest training sample and the separating hyperplane should be maximized, and the distance can be expressed as 1/
- the hyperplane that can maximize the distance between itself and the nearest vector is the best hyperplane. Therefore, the problem of solving the optimal hyperplane can be transformed into the following optimization problem:
- equation (12) itself is a convex quadratic programming problem, and the "dual problem" can be obtained by using the Lagrange multiplier method. Specifically, adding the Lagrangian multiplier a i ⁇ 0 to the constraint in equation (12), the Lagrangian function of the above optimization problem can be written as:
- X can be replaced by a kernel function K(X i ,X), which mainly includes linear kernels and Gaussian kernels. So the final decision function can be expressed by the following formula.
- Step this embodiment S4 is the trained SVM model for the actual network determines whether the user needs to unload, the specific process is: the input used Support Vector Machine as a vector X i (N RB, N WiFi , D BS , D WiFi ), where N RB is the number of resource blocks occupied by users currently judged by the SVM model, N WiFi is the number of WiFi users at the access point corresponding to the user, D BS is the distance between the user and the base station, and D AP For the distance between the user and the WiFi access point, the maximum coverage range of the WiFi access point is defined as R.
- the possible values of the output y are 1 and -1, which respectively indicate to perform uninstallation and not to perform uninstallation.
- Randomly generate N cellular network users use the resource block allocation method to evenly allocate resource blocks to these users, and generate a resource block table B.
- N ti N t +1, using the resource block allocation method introduced above to update the user parameters N RB and N WiFi ;
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Signal Processing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
本发明公开了一种异构网络下基于支持向量机的资源分配方法,包括如下步骤:获得已经进行过网络流量卸载区域的用户及其网络状态数据;将获得的用户和网络状态数据整理后作为数据集,且将数据集分为训练集和测试集;将训练集中每个数据的参数作为初始支持向量机模型输入向量,网络状态数据作为每个数据的标记值,训练优化支持向量机模型的参数;获得当前异构网络下蜂窝网络用户的相关参数,将其作为训练后支持向量机模型的输入向量,得到输出值用以判断是否需要卸载。本发明把更多的计算资源放在生成数据集和训练上,降低了做出决策的延迟时间,对异构网络中蜂窝网络用户的状态判断是否执行卸载,以达到异构网络整体用户满意度最大化。
Description
本发明无线通信技术领域,具体涉及一种异构网络下基于支持向量机的资源分配方法。
随着第五代移动通信的迅猛发展,手机用户对于在社交,购物,出行,娱乐等方方面面都有更高的要求,尤其是对于较快网速的要求。然而,由于蜂窝网络的用户众多,且离基站的距离较远,故通信损耗大,用户难以达到最满意的速率要求,因此,越来越多研究关注通过LTE(Long Term Evolution)网络与WiFi网络之间的协作,使得部分蜂窝流量通过WiFi网络卸载,以使得蜂窝网络用户获得更快的速率。从网络运营商的角度来看,蜂窝网络过载将得到缓解。从用户的角度来看,WiFi网络将会带来更好的用户体验,他们也愿意连接到WiFi。因此,无线网络卸载提供了一种非常经济高效的方式来帮助LTE网络扩展容量,并帮助用户获得更好的服务质量(QoS)和体验质量(QoE)。
异构网络中,蜂窝网络与WiFi网络间的流量卸载最为常见,其实现形式是将蜂窝网络中的用户流量卸载到WiFi网络中,以缓解蜂窝网络带宽紧张的问题.WiFi网络和蜂窝网络间可实现流量卸载是由于:1)WiFi网络使用的频段是免许可频段,可以与蜂窝网络重叠,网络间互相没有干扰;2)WiFi网络带宽充足,可提供高吞吐量、高可靠性的无线通信;3)WiFi网络技术成熟,接入点价格低廉,并且应用广泛,在公众场合或室内环境普及率高.由于可大幅度提升网络***容量,缓解网络拥塞,且实现成本低廉,蜂窝网络与WiFi网络间流量卸载引起业界广泛的关注。
现有的方法往往需要进行多次迭代,需要花费较长的算法执行时间,增加了网络计算资源开销。
发明内容
发明目的:为了克服现有技术中存在的不足,提供一种异构网络下基于支持向量机的资源分配方法,把更多的计算资源放在生成数据集和训练上,这样在实际应用的过程中可以直接调用相应的模型来解决问题,降低了做出决策的延迟时间,而且对于某个用户是否卸载的决策仅与当前用户的状态有关,不必考虑其他用户的情况,对异构网络中蜂窝网络用户的状态判断是否执行卸载,以达到异构网络整体用户满意度最大化。
技术方案:为实现上述目的,本发明提供一种异构网络下基于支持向量机的资源分配方法,包括如下步骤:
S1:获得已经进行过网络流量卸载区域的用户X
i(N
RB,N
WiFi,D
BS,D
WiFi)及其网络状态数据yi∈{-1,+1},其中N
RB为用户所占用的资源块数量,N
WiFi为用户所对应接入点的WiFi用户数量,D
BS为用户离基站的距离,D
AP为用户到WiFi接入点之间的距离,y
i=1和y
i=-1分别表示用户执行卸载和不执行卸载;
S2:将获得的用户X
i(N
RB,N
WiFi,D
BS,D
WiFi)和网络状态数据yi∈{-1,+1}整理后作为数据集,且将数据集分为训练集和测试集;
S3:将训练集中每个数据的参数X
i(N
RB,N
WiFi,D
BS,D
WiFi)作为初始支持向量机模型输入向量,yi∈{-1,+1}作为每个数据的标记值,训练优化支持向量机模型的参数,生成模型对应的函数f(x);
S4:获得当前异构网络下蜂窝网络用户的相关参数,将其作为训练后支持向量机模型的输入向量,根据函数f(x)得到输出值y,并判断y是否大于0;若y小于0,则无须执行卸载,继续用训练后的支持向量机模型判断下一个用户;若y大于0,则对该用户执行卸载,并将其占用的资源块分配给其他用户,同时更新被分配资源用户的参数N
RB和N
WiFi;
S5:不断执行步骤S4直至完成所有蜂窝网络用户的判断。
进一步的,所述步骤S3中利用测试集对训练中的支持向量机模型进行精度测试,达到要求后保存训练模型。
进一步的,所述步骤S3中支持向量机模型的训练过程为:
将函数f(X)定义为:
f(X)=W·X+b (1)
通过下式来估计函数f(X)的系数W和b:
y
i(W·X
i+b)>0 (2)
将参数W和B重新定义为:
y
i(W·X
i+b)≥1 (3)
将求解最优超平面的问题转化为下面的优化问题:
使用拉格朗日乘子法得到上面优化问题的拉格朗日函数:
可将式(3)中的问题转化为它的对偶问题:
通过二次规划方法解出相关参数的值:
最终的决策函数通过下式(8)表达。
本发明构造了一个以用户为中心的支持向量机卸载模型,用来提升异构网络中的用户满意度。首先在有一个宏基站和多个无线接入点的异构网络场景下,引入用户满意度的概念以更好的满足用户不同的QoS需求;之后利用生成的数据集进行训练支持向量机模型,将其转化为分类问题来对网络资源进行重新分配,基于每一状态的WiFi网络及蜂窝网络用户的状态判断是否执行卸载,并寻找最佳的WiFi网络接入点达到异构网络整体用户满意度最大化,该方法大幅度缩小了算法复杂度,提高用户速率,最后实现满意度提升。
有益效果:本发明与现有技术相比,具备如下优点:
1、本发明以用户满意度为度量值,根据用户自身速率需求作出卸载决策,能够有效提高蜂窝网络用户的满意度,避免了传统的资源分配方法分配过多或过少资源的情况,更好的满足了用户需求。
2、本发明把更多的计算资源放在生成数据集和训练上,这样在实际应用的过程中可以直接调用相应的模型来解决问题,降低了做出决策的延迟时间。
3、本发明对于某个用户是否卸载的决策仅与当前用户的状态有关,不必考虑其他用户的情况。
图1为本发明方法的***模型示意图;
图2为本发明方法的流程示意图。
下面结合附图和具体实施例,进一步阐明本发明,应理解这些实施例仅用于说明本发明而不用于限制本发明的范围,在阅读了本发明之后,本领域技术人员对本发明的各种等价形式的修改均落于本申请所附权利要求所限定的范围。
本发明提供一种异构网络下基于支持向量机的资源分配方法,其应用在如图1所示的***模型,该模型在一个异构网络中,宏基站(BS)位于中心位置,周围有四个WiFi的接入点(AP)随机部署,彼此之间互相不干扰。蜂窝网络用户在以基站BS为中心,半径为R的圆上随机分布,WiFi网络的用户位置是在以AP接入点为圆心,r为半径的圆上随机分布。用N
RB来表示资源块(RB)的数量,用P
max来表示基站的最大功率,B
c表示每个资源块所占用的带宽。
假定所有的资源块(RB)以相同的功率进行传输,则单个资源块的传输功率P
tr=P
max/N
RB,那么用户k占用第n个资源块的速率为:
其中N
0是热噪声功率密度,L
k是接入点到用户k之间传输的路径损耗,L
k采用大尺度路径损耗模型中的Friis公式进行计算
其中λ为信号的波长,G
t是发射天线的增益;G
r是接收天线的增益;d是用户与基站之间的距离,单位为m;L是与传播路径无关的损耗因子;α为距离衰减因子。由于一个用户可以同时通过多个资源块进行数据的传输,用户k的总速率为:
WiFi网络采用了CSMA/CA(Carrier Sense Multiple Access with Collision detection)协议,用户l在WiFi网络中的速率为,其中B
w代表WiFi网络的带宽,T表示WiFi的总时隙,Pm代表WiFi的功率,用户l占用的时隙为t
l:
本实施例中基于上述***模型,通过一个模型计算方案来模拟用户满意度,它是用户数据速率的函数。当用户k的实际传输速率R
k高于R
required(用户k所需的数据速率)时, 用户k的满意度将缓慢上升;当R
k低于R
required时,用户k的满意度将急剧下降。因此,用户满意度k的定义如下:
在蜂窝网络过载且WiFi网络能够提供更高数据速率的情况下,考虑将重叠覆盖范围内的一些蜂窝用户卸载到WiFi网络。由于一部分蜂窝用户已经转移到WiFi,剩余的蜂窝用户将占用更多的资源。因此,蜂窝网络用户的速率将随着用户满意度的提高而提高。WiFi网络的结果与蜂窝网络完全相反,WiFi用户传输速率会随着卸载用户满意度速率的提高而下降。因此,最终的目标是最大限度地提高整个***的总体用户满意度,同时将考虑一个约束:如果被卸载的用户在新网络中获得的数据速率低于原始网络,用户将失去卸载的动机。同时,这种情况不能通过卸载带来总用户满意度的提升,因此确保卸载的用户实现更高的数据速率至关重要,所以优化目标数学建模如下:
要求解此优化问题,首先要设计一个资源块分配算法,用来重新分配那些已经卸载到WiFi网络的LTE用户所占用的资源块。根据之前在上一部分中对用户满意度的定义,网络速率越是低的蜂窝网络用户,获得更多资源块后所带来的满意度的提升也就比那些速率快的用户获得后带来的提升更大,因此这一分配算法总是将剩余的资源块依次分给当前速率最低的用户,这样就可以确保用户卸载后总体满意度的最大化。表格B用来存储资源块的索引,它的行表示蜂窝网络用户的序号,列表示资源块的序号,表格中的数字“1”代表当前资源块已经被对应的用户所占用。用S和U分别表示卸载用户集和剩余的蜂窝网络用户集。详细求解过程为方法1:
方法1:资源块分配过程
1确定L=size(S),生成一个1*L的列表来存储集合S中用户占用的资源块的索引,并重置相关的位置。
2 for i=1:L
3计算Nr=sum(B(S(i),:)
4 for j=1:Nr
6令k=argmin(Г
k)
7将B(S(k),I{1,i}(1,j))的值置为1
8 end for
9 end for
其中S
m和S
-m分别是第m个AP和除了m以外其他AP的策略,Γn是第m个AP中用户的满意度,Γk是采用当前策略的蜂窝网络用户满意度。下面提供了一些关于纳什均衡和精确势博弈的定义。
定义1:如果一个策略集合(S1,S2,...,SM)满足
则可以认为这个博弈存在一个纳什均衡。
上面的等式意味着没有任何一个人可以通过单方面改变策略来提高它的收益。
U
m(S'
m,S
-m)-U
m(S
m,S
-m)=Φ(S'
m,S
-m)-Φ(S
m,S
-m) (9)
公式(8)表明,如果任何玩家改变策略,其自身效用函数的改变等于势函数的改变。
定理1:本实施例所讨论的博弈是一个精确势博弈。
将(7)中的γ定义为势函数。那么可以证明本文提出的博弈模型满足定义2中的条件,所以这个博弈是一个精确势博弈。
精确势博弈最重要的性质是它至少存在一个纯策略纳什均衡。本文中所提出的博弈模型的纯策略纳什均衡,可以最大化总体的用户满意度,因为这里设置的势函数是***的总用户满意度。
BR算法是博弈论中寻找纳什均衡的经典算法,因此可以采用它来寻找最优解在每次迭代中,随机选择一个接入点而不重复,固定其他接入点的策略,并计算其在所有不同策略上的总用户满意度;选择其中的最佳战略,使Um在当前迭代中最大化;如果总的用户满意度的值达到收敛,则停止迭代。以上策略即为最佳策略,详细过程如方法2。
方法2:
2:生成随机打乱的集合M,记为N
3:for i=1:M
4:选择AP m=N(i)
5:令Nm=Sm的大小
6:for j=1:Nm
7:选择S
m,j,
8:根据上文资源块分配算法更新b,
9:根据Um计算公式计算Um:
10:end for
11:选择Sm=argmax(Um),作为AP m的策略,
12:更新策略集S
13::end for
14:S是最佳策略,Γ=Γ(S)
基于上述资源块分配方法以及BR算法等,本实施例将本发明提供的一种异构网络下基于支持向量机的资源分配方法应用***模型,参照图2,其具体步骤如下:
S1:获得已经进行过网络流量卸载区域的用户X
i(N
RB,N
WiFi,D
BS,D
WiFi)及其网络状态数据yi∈{-1,+1},其中N
RB为用户所占用的资源块数量,N
WiFi为用户所对应接入点的WiFi用户数量,D
BS为用户离基站的距离,D
AP为用户到WiFi接入点之间的距离,y
i=1和y
i=-1分别表示用户执行卸载和不执行卸载;
S2:将获得的用户X
i(N
RB,N
WiFi,D
BS,D
WiFi)和网络状态数据yi∈{-1,+1}整理后作为数据集,且将数据集分为训练集和测试集;
S3:将训练集中每个数据的参数X
i(N
RB,N
WiFi,D
BS,D
WiFi)作为初始支持向量机模型输入向量,yi∈{-1,+1}作为每个数据的标记值,训练优化支持向量机模型的参数,生成模型对应的函数f(x);利用测试集对训练中的支持向量机模型进行精度测试,达到要求后保存训练模型;
S4:获得当前异构网络下蜂窝网络用户的相关参数,将其作为训练后支持向量机模型的输入向量,根据函数f(x)得到输出值y,并判断y是否大于0;若y小于0,则无须执行卸载,继续用训练后的支持向量机模型判断下一个用户;若y大于0,则对该用户执行卸载,并将其占用的资源块分配给其他用户,同时更新被分配资源用户的参数N
RB和N
WiFi;
S5:不断执行步骤S4直至完成所有蜂窝网络用户的判断。
本实施例的步骤S3中具体为:如何通过产生的数据获得支持向量机模型,根据支持向量机理论,输入向量的维度为4,训练样本来自4维特征空间X
i∈R
4(i=1,2,...m),映射值yi∈{-1,+1}与每个矢量X
i相关联。如前所述,X
i可以分为两类。假设这两类是 线性可分的,则可找到至少一个由向量W∈R
4和一个偏置b∈R
4定义的超平面,它可以无误差地分离这两类。决策根据函数sgn[f(X)]的值来执行,其中f(X)是与超平面相关联的决策函数,定义为:
f(X)=W·X+b (10)
将通过下式来估计函数f(X)的系数W和b:
y
i(W·X
i+b)>0 (11)
最近的训练样本和分离超平面之间的距离应该最大化,并且该距离可以表示为1/||W||。
因此,超平面参数W和B应该被重新定义为:
y
i(W·X
i+b)≥1 (12)
在所有满足要求的超平面中,能够让自身和最邻近的向量之间的距离最大的超平面就是最佳的超平面,因此可以将求解最优超平面的问题转化为下面的优化问题:
注意到式(12)本身是一个凸二次规划问题,可以使用拉格朗日乘子法得到其“对偶问题”。具体来说,对式(12)中的约束添加拉格朗日乘子a
i≥0,上面优化问题的拉格朗日函数可写为:
接下来可将(12)中的问题转化为它的对偶问题:
接下来可以通过二次规划方法解出相关参数的值:
通常情况下,X可以被核函数K(X
i,X)代替,核函数主要包括线性核和高斯核。所以最终的决策函数可以通过下式表达。
因此任务可以通过f(X)的取值来分类,若f(X)>0,则对应y
i=1,若f(X)<0,则对应y
i=-1。
本实施例中步骤S4中训练后的支持向量机模型用于实际网络中判断用户是否需要卸载,其具体过程为:由于所用的支持向量机的输入为向量X
i(N
RB,N
WiFi,D
BS,D
WiFi),其中N
RB为当前用SVM模型判断用户所占用的资源块数量,N
WiFi为该用户所对应接入点的WiFi用户数量,D
BS为该用户离基站的距离,D
AP为用户到WiFi接入点之间的距离,定义WiFi接入点最大覆盖范围为R。输出y的可能取值为1和-1,它们分别表示执行卸载和不进行卸载。
用支持向量机模型进行判断的详细流程如下:
1、首先产生网络分布信息列表L=(N1,N2,N3,N4),其中N
t表示第t个接入点wifi用户人数;
2、随机生成N个蜂窝网络用户,采用资源块分配方法平均分配资源块给这些用户,生成资源块表格B。
3、判断i<=N是否成立,如果成立则执行资源块分配方法中的4到9的步骤,否则结束:
4、分别计算用户i到最近接入点的距离D
AP和到基站的距离D
BS,如果DAP<=R,读取列表L中对应的WiFi接入点的人数N
t和当前用户所占用的资源块数目N
RB;
5、将X
i(N
RB,N
WiFi,D
BS,D
WiFi)作为支持向量机模型f(X)的输入数据;
6、如果f(Xi)>0,则令yi=1,此时将执行卸载。N
ti=N
t+1,采用上文介绍的资源块分配方法,更新用户参数N
RB和N
WiFi;
7、如果f(Xi)<=0,则令yi=-1,此时将不进行卸载,并继续判断下一个用户的最优决定。
Claims (5)
- 一种异构网络下基于支持向量机的资源分配方法,其特征在于:包括如下步骤:S1:获得已经进行过网络流量卸载区域的用户X i(N RB,N WiFi,D BS,D WiFi)及其网络状态数据yi∈{-1,+1},其中N RB为用户所占用的资源块数量,N WiFi为用户所对应接入点的WiFi用户数量,D BS为用户离基站的距离,D AP为用户到WiFi接入点之间的距离,y i=1和y i=-1分别表示用户执行卸载和不执行卸载;S2:将获得的用户X i(N RB,N WiFi,D BS,D WiFi)和网络状态数据yi∈{-1,+1}整理后作为数据集,且将数据集分为训练集和测试集;S3:将训练集中每个数据的参数X i(N RB,N WiFi,D BS,D WiFi)作为初始支持向量机模型输入向量,yi∈{-1,+1}作为每个数据的标记值,训练优化支持向量机模型的参数,生成模型对应的函数f(x);S4:获得当前异构网络下蜂窝网络用户的相关参数,将其作为训练后支持向量机模型的输入向量,根据函数f(x)得到输出值y,并判断y是否大于0;若y小于0,则无须执行卸载,继续用训练后的支持向量机模型判断下一个用户;若y大于0,则对该用户执行卸载,并将其占用的资源块分配给其他用户,同时更新被分配资源用户的参数N RB和N WiFi;S5:不断执行步骤S4直至完成所有蜂窝网络用户的判断。
- 根据权利要求1所述的一种异构网络下基于支持向量机的资源分配方法,其特征在于:所述步骤S3中利用测试集对训练中的支持向量机模型进行精度测试,达到要求后保存训练模型。
- 根据权利要求1所述的一种异构网络下基于支持向量机的资源分配方法,其特征在于:所述步骤S4中利用训练后的支持向量机模型判断用户是否进行卸载的具体过程为:S4-1:产生网络分布信息列表L=(N1,N2,N3,N4),其中N t表示第t个接入点wifi用户人数;S4-2:随机生成N个蜂窝网络用户,采用资源块分配方法平均分配资源块给蜂窝网络用户,生成资源块表格B;S4-3:判断i<=N是否成立,如果成立则执行分配步骤,否则结束:S4-4:分别计算用户i到最近接入点的距离D AP和到基站的距离D BS,如果DAP<=R,读取列表L中对应的WiFi接入点的人数N t和当前用户所占用的资源块数目N RB;S4-5:将X i(N RB,N WiFi,D BS,D WiFi)作为支持向量机模型f(X)的输入数据;S4-6:如果f(Xi)>0,则令yi=1,执行卸载,N ti=N t+1,采用资源块分配方法, 更新用户参数N RB和N WiFi;如果f(Xi)<=0,则令yi=-1,不进行卸载,并继续判断下一个用户的最优决定。
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010417039.9A CN111669775B (zh) | 2020-05-18 | 2020-05-18 | 一种异构网络下基于支持向量机的资源分配方法 |
CN202010417039.9 | 2020-05-18 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2021232848A1 true WO2021232848A1 (zh) | 2021-11-25 |
Family
ID=72383896
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2021/074165 WO2021232848A1 (zh) | 2020-05-18 | 2021-01-28 | 一种异构网络下基于支持向量机的资源分配方法 |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN111669775B (zh) |
WO (1) | WO2021232848A1 (zh) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114302497A (zh) * | 2022-01-24 | 2022-04-08 | 厦门大学 | 一种应用于非授权毫米波段异构网络共存的调度方法 |
CN115002721A (zh) * | 2022-06-06 | 2022-09-02 | 南京大学 | 一种面向b5g/6g全解耦蜂窝车联网的随机优化资源分配方法 |
CN116939668A (zh) * | 2023-09-15 | 2023-10-24 | 清华大学 | 车载WiFi-蜂窝异构网络通信资源分配方法、装置 |
CN117032936A (zh) * | 2023-09-28 | 2023-11-10 | 之江实验室 | 一种数据调度方法、装置和计算机设备 |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111669775B (zh) * | 2020-05-18 | 2022-07-29 | 南京邮电大学 | 一种异构网络下基于支持向量机的资源分配方法 |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103036802A (zh) * | 2013-01-08 | 2013-04-10 | 中国科学院计算技术研究所 | 一种流量卸载的方法和*** |
CN110113190A (zh) * | 2019-04-24 | 2019-08-09 | 西北工业大学 | 一种移动边缘计算场景中卸载时延优化方法 |
CN110798842A (zh) * | 2019-01-31 | 2020-02-14 | 湖北工业大学 | 一种基于多用户深度强化学习的异构蜂窝网络流量卸载方法 |
CN111669775A (zh) * | 2020-05-18 | 2020-09-15 | 南京邮电大学 | 一种异构网络下基于支持向量机的资源分配方法 |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109558240A (zh) * | 2018-10-31 | 2019-04-02 | 东南大学 | 一种移动终端应用下基于支持向量机的任务卸载方法 |
CN109600774B (zh) * | 2019-01-17 | 2021-09-28 | 南京邮电大学 | 一种LTE网络中基于联盟博弈的WiFi卸载方法 |
CN110941667B (zh) * | 2019-11-07 | 2022-10-14 | 北京科技大学 | 一种移动边缘计算网络中的计算卸载方法及*** |
-
2020
- 2020-05-18 CN CN202010417039.9A patent/CN111669775B/zh active Active
-
2021
- 2021-01-28 WO PCT/CN2021/074165 patent/WO2021232848A1/zh active Application Filing
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103036802A (zh) * | 2013-01-08 | 2013-04-10 | 中国科学院计算技术研究所 | 一种流量卸载的方法和*** |
CN110798842A (zh) * | 2019-01-31 | 2020-02-14 | 湖北工业大学 | 一种基于多用户深度强化学习的异构蜂窝网络流量卸载方法 |
CN110113190A (zh) * | 2019-04-24 | 2019-08-09 | 西北工业大学 | 一种移动边缘计算场景中卸载时延优化方法 |
CN111669775A (zh) * | 2020-05-18 | 2020-09-15 | 南京邮电大学 | 一种异构网络下基于支持向量机的资源分配方法 |
Non-Patent Citations (1)
Title |
---|
WU SIYUN; XIA WEIWEI; CUI WENQING; CHAO QIAN; LAN ZHUORUI; YAN FENG; SHEN LIANFENG: "An Efficient Offloading Algorithm Based on Support Vector Machine for Mobile Edge Computing in Vehicular Networks", 2018 10TH INTERNATIONAL CONFERENCE ON WIRELESS COMMUNICATIONS AND SIGNAL PROCESSING (WCSP), 18 October 2018 (2018-10-18), pages 1 - 6, XP033460210, DOI: 10.1109/WCSP.2018.8555695 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114302497A (zh) * | 2022-01-24 | 2022-04-08 | 厦门大学 | 一种应用于非授权毫米波段异构网络共存的调度方法 |
CN115002721A (zh) * | 2022-06-06 | 2022-09-02 | 南京大学 | 一种面向b5g/6g全解耦蜂窝车联网的随机优化资源分配方法 |
CN116939668A (zh) * | 2023-09-15 | 2023-10-24 | 清华大学 | 车载WiFi-蜂窝异构网络通信资源分配方法、装置 |
CN116939668B (zh) * | 2023-09-15 | 2023-12-12 | 清华大学 | 车载WiFi-蜂窝异构网络通信资源分配方法、装置 |
CN117032936A (zh) * | 2023-09-28 | 2023-11-10 | 之江实验室 | 一种数据调度方法、装置和计算机设备 |
CN117032936B (zh) * | 2023-09-28 | 2024-02-06 | 之江实验室 | 一种数据调度方法、装置和计算机设备 |
Also Published As
Publication number | Publication date |
---|---|
CN111669775B (zh) | 2022-07-29 |
CN111669775A (zh) | 2020-09-15 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2021232848A1 (zh) | 一种异构网络下基于支持向量机的资源分配方法 | |
CN110035559B (zh) | 一种基于混沌q-学习算法的竞争窗口大小智能选择方法 | |
TW201931912A (zh) | 用於無線通訊的電子裝置和方法以及電腦可讀儲存媒體 | |
Zheng et al. | 5G network-oriented hierarchical distributed cloud computing system resource optimization scheduling and allocation | |
Gaur et al. | Application specific thresholding scheme for handover reduction in 5G Ultra Dense Networks | |
Li et al. | Deep reinforcement learning-based resource allocation and seamless handover in multi-access edge computing based on SDN | |
WO2022134854A1 (zh) | 信息共享方法及通信装置 | |
KR101568081B1 (ko) | 셀룰러 시스템에서의 d2d 통신을 위한 자원 할당 방법 및 그 장치 | |
Liu et al. | Deep reinforcement learning-based MEC offloading and resource allocation in uplink NOMA heterogeneous network | |
CN106060876B (zh) | 一种异构无线网络均衡负载的方法 | |
JP2022514716A (ja) | 無線通信に用いられる電子装置、方法、及びコンピュータ読み取り可能な記憶媒体 | |
CN111212405B (zh) | 基于多流行度的分组d2d多阈值缓存放置方法 | |
CN104618934B (zh) | 一种基于吞吐量预测的整体优化中继节点选择方法 | |
Abdulshakoor et al. | Outage-aware matching game approach for cell selection in LTE/WLAN multi-RAT HetNets | |
Ye et al. | Hybrid-clustering game Algorithm for resource allocation in macro-femto hetnet | |
Ibrahimi et al. | Prediction of the content popularity in the 5G network: Auto-regressive, moving-average and exponential smoothing approaches | |
Hua et al. | Content caching policy with edge caching user classification in fog radio access networks | |
Nakazato et al. | Revenue model with multi-access edge computing for cellular network architecture | |
Vasquez-Toledo et al. | Mathematical analysis of highly scalable cognitive radio systems using hybrid game and queuing theory | |
Bulti et al. | Clustering-based adaptive low-power subframe configuration with load-aware offsetting in dense heterogeneous networks | |
Dong et al. | Deep Reinforcement Learning-based Adaptive Clustering Approach in Short Video Sharing through D2D Communication | |
Gao et al. | Analysis of acquired indoor LTE-A data from an actual HetNet cellular deployment | |
Sun et al. | WiFi Offloading Algorithm Based on Q‐Learning and MADM in Heterogeneous Networks | |
Li et al. | Adaptive handover based on traffic balancing and multi-dimensional collaborative resource management in MEC environment | |
Kikuchi et al. | Vm migration in mobile edge computing for qos improvement with wireless multi-hop access networks |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 21808569 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 21808569 Country of ref document: EP Kind code of ref document: A1 |