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

Abstract

The invention discloses a method for combining dynamic access and subcarrier allocation under a cache-based ultra-dense network. The method comprises the following specific steps that firstly, multiple users simultaneously send request information to all access points to search a caching content; secondly, each access point judges whether a caching content requested by the current user K exists, all access points meeting the user K transmit respective property parameters to local control, and the local control allocates the optimal access point to the user K, otherwise the user K directly sends a request to a remote server to obtain the content; thirdly, the remote server utilizes popularity analysis to complete cache updating according to the user request information; and lastly, after each user is matched with the respective access point, subcarrier allocation is carried out, so that each user is communicated with the respective access point. The method has the following advantages that multiple factors are synthesized to complete access selection, and promotion of the resource management efficiency and dynamic allocation of the subcarrier are realized, so that the spectral efficiency is remarkably improved.

Description

Under a kind of super-intensive network based on buffer memory, combine the method that dynamic access and subcarrier distribute

Technical field

The invention belongs to networking and resource allocation techniques field, specifically refer to a kind of super-intensive network second line of a couplet based on buffer memoryClose the method that dynamic access and subcarrier distribute.

Background technology

Super-intensive network is the strong candidate technologies of 5G, and super-intensive networking technology, can by increasing base station deployment densityRealize the tremendous increase of channeling efficiency, greatly improve power system capacity, meet thousand times of capacity increase demands of 5G.But more intensive network design makes network topology more complicated, existing content distribution mechanism realize picture,When the magnanimity information such as audio frequency, video transmits, exist a large amount of contents to repeat transmission, frequency spectrum resource etc. has been causedGreatly waste. For this problem, caching technology is introduced to super-intensive network, by access point or core netCarry out content caching, can reduce redundant data transmissions, effectively reduce back haul link consumption and network delay, thereby carryHigh spectrum utilization efficiency and efficiency utilization rate.

Under the dense network based on buffer memory, document 1: in the limited intensive wireless network of back haul link based on physicsThe throughput gain optimization method of layer buffer memory, has proposed a kind of based on the limited novel buffer memory wireless network frame of back haul linkStructure, and physical layer buffering scheme has been proposed under this framework to improve throughput of system, but this scheme is only considered backThe factors such as journey link, and show that result is relevant to base station buffer memory capacity size.

Document 2: the method that joint route and content caching are optimized in heterogeneous network, adopts joint route to select gentleDeposit the scheme of distribution and carry out resource allocation optimization, the problem of considering under this scheme is single, does not consider effect, efficiency frequentlyEtc. resource utilization problem.

Under the scene of super-intensive, prior art is not considered the load balancing of access point, carries out the dynamic choosing of access pointSelect. In addition, do not consider the optimization of spectrum efficiency, the resources such as subcarrier are carried out to allocation optimized.

Summary of the invention

The present invention is directed to the storage resources that can not utilize efficiently access point in prior art and provide service, frequency spectrum for userCan not realize maximization with efficiency utilization of resources rate, propose to combine under a kind of super-intensive network based on buffer memory movingThe method that state access and subcarrier distribute,

Concrete steps are as follows:

Step 1, multiple user send request information to all access points with broadcast mode, find cache contents;

Solicited message refers to cache contents; Number of users is O;

Step 2, using user K as active user, each access point judges whether the buffer memory that exists active user to askContent, if certain access point is idle and have this cache contents, this access point feedback 1 is given user K, enters stepThree; Otherwise feedback 0; Enter step 4;

1≤K≤O；

Step 3, all access points that meet user K send attribute parameter separately to local control, local controlBest access point is distributed to user K;

The all access points that meet user K are m; The attribute parameter of each access point include buffer memory capacity, timeProlong with signal to noise ratio etc., altogether n attribute;

Concrete steps are as follows:

Step 301, according to the solicited message of user K, for each candidate's access point, local control n attributeRelative importance in parameter between every two attribute parameters compares one by one, obtains decision matrix M:

Local control compares the relative importance between each attribute parameter of each candidate's access point, obtains thisThe decision matrix M of candidate's access point:

Wherein a_ijRepresent the relative importance fiducial value of attribute parameter i and attribute parameter j in access point;

$a_{ij}>0;a_{ij}=\frac{1}{a_{ji}};a_{ii}=1;$

Step 302, decision-making matrix M is normalized, obtains the decision matrix B after standardization:

Wherein b_ijRepresent in access point fiducial value a_ijValue after normalization;

Step 303, the uniformity of decision-making matrix B is carried out to verification, judge that whether decision matrix is effective, if effectively,Carry out step 304, otherwise return to step 302;

Consistency Ratio CR is defined as follows:

Wherein, CI represents inconsistency index:λ_maxThe eigenvalue of maximum of decision matrix B,N is the number of attribute parameter in decision matrix B, and RI is known mean random coincident indicator;

In the time of CR < 0.1, think that decision matrix B has acceptable uniformity, otherwise re-construct decision matrixB。

Step 304, obtain in decision matrix B n attribute parameter comprehensive produce weight vectors ω;

ω＝(ω₁,ω₂,...ω_j,...,ω_n)

ω_jIt is the weight of j attribute parameter;

Step 305, for m candidate's access point, generate the state matrix S of all properties parameter;

State matrix S is the capable n row of m, and every a line represents n attribute parameter of each access point;

Wherein, s_mnRepresent the value of corresponding n the attribute parameter of m access point.

Step 306, weight vectors ω and state matrix S are multiplied each other and obtain weighting decision matrix Q:

Step 307, according to weighting decision matrix Q, determine best access scheme Q_bestThe poorest access scheme Q_worst，

Q_best＝(ω₁·s_1best,ω₂·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)

s_jbestRepresent optimum value in j attribute parameter of all m access point; s_jworstRepresent all m access pointJ attribute parameter in worst-case value;

Step 308, for certain access point l, respectively calculated candidate access scheme x_ljWith best access scheme Q_bestEuropeFamily name's distance, and candidate's access scheme x_ljWith the poorest access scheme Q_worstEuclidean distance;

Candidate's access scheme x_ljWith best access scheme Q_bestEuclidean distance be Q_lbest, specifically refer to the candidate side of accessCase x_ljEach attribute parameter and this attribute optimal value s_jbestEuclidean distance, as follows:

$Q_{lbest}=\sqrt{\underset{j=1}{\overset{n}{\&Sigma;}}{\mathrm{\&omega;}}_j{(x_{lj}-s_{jbest})}^2}$

Candidate's access scheme x_ljWith the poorest access scheme Q_worstEuclidean distance be Q_lworst, specifically refer to that candidate accessesScheme x_ljEach attribute parameter and this attribute worst-case value s_jworstEuclidean distance, as follows:

$Q_{lworst}=\sqrt{\underset{j=1}{\overset{n}{\&Sigma;}}{\mathrm{\&omega;}}_j{(x_{lj}-s_{jworst})}^2}$

Step 309, for certain access point l, calculated candidate scheme x_ljAnd the preference value P between preferred plan_l。

Formula is as follows:

$P_l=\frac{Q_{lworst}}{Q_{lbest}+Q_{lworst}}$

Wherein P_lWhat represent is the preference value of user K for l access point.

Step 310, calculate all users respectively for the preference value of each access point, and arranged in sequence, user generatedSelection matrix R;

As follows:

Wherein, P_mORepresent that O user selects the preference value of m access point. All O use of every a line representativeThe preference value of each identical access point is selected at family; The list of user's selection matrix R shows that each user selects each difference to connectEnter preference value a little;

Step 311, for each user, access point corresponding maximum preference value is distributed to this user by local control,And delete the access point of this user and coupling, distribute successively, until all users complete network insertion.

Step 4, user K do not obtain the response of any access point, and user K directly sends to far-end serverRequest, obtains content; Far-end server, according to user request information, utilizes popularity analysis, completes buffer update;

Step 5, each user carry out subcarrier distribution after mating with access point separately, make user and access point itBetween communicate.

Step 501, initialization subcarrier set and user's set;

Subcarrier set is N={n'|n'=1,2 ..., N}, user gathers I={i'|i'=1, and 2 ..., I}, distributes to userThe sub-carrier indices X of i'_i'＝φ。

Step 502, according to water-filling algorithm calculate each user the transmitting power of corresponding each subcarrier and channel holdAmount.

Transmitting power p_i',n'：

$p_{i^{\&prime;},n^{\&prime;}}=\frac{P_{tot}+\underset{n^{\&prime;}=1}{\overset{N}{\&Sigma;}}1/{\mathrm{\&gamma;}}_{i^{\&prime;},n^{\&prime;}}}{N}-1/{\mathrm{\&gamma;}}_{i^{\&prime;},n^{\&prime;}}$

p_i',n'Represent the power to user i' allocation of subcarriers n', P_totRepresent maximum transmission power; γ_i',n'Represent userThe signal to noise ratio of i' allocation of subcarriers n'.

Channel capacity C_i',n'：

$C_{i^{\&prime;},n^{\&prime;}}=\frac{B}{N}{\log }_2(1+p_{i^{\&prime;},n^{\&prime;}}{\mathrm{\&gamma;}}_{i^{\&prime;},n^{\&prime;}})$

B represents this overall system bandwidth;

Step 503, for subcarrier n', calculate respectively the channel capacity of this subcarrier under each user, and fallOrder is arranged and is selected maximum

Subcarrier n' initial value is 1;

Be expressed as:

$C_{n^{\&prime;}}^{\max }={\max }_{i^{\&prime;}\&Element;I}\left\{C_{i^{\&prime;},n^{\&prime;}}\right\}$

Step 504, subcarrier n' is distributed to maximum channel capacity valueCorresponding user, and by subcarrier n' fromIn subcarrier set N, remove, return to step 503 and continue to choose according to the order of sequence next subcarrier, until all sons are carriedRipple distributes.

Step 505, to distribute after subcarrier carry out water filling, according to the transmitting power p of subcarrier_i',n'Computing system frequentlySpectrum utilization ratio;

It is as follows that system spectrum utilization ratio maximizes object function:

${\max }_{\{a_{i^{\&prime;},n^{\&prime;}},p_{i^{\&prime;},n^{\&prime;}}\}}\frac{C}{B}={\max }_{\left\{a_{i^{\&prime;},n^{\&prime;}},p_{i^{\&prime;},n^{\&prime;}}\right\}}\underset{i^{\&prime;}=1}{\overset{I}{\&Sigma;}}\underset{n^{\&prime;}=1}{\overset{N}{\&Sigma;}}{a}_{i^{\&prime;},n^{\&prime;}}{\log }_2(1+p_{i^{\&prime;},n^{\&prime;}}{\mathrm{\&gamma;}}_{i^{\&prime;},n^{\&prime;}})$

Condition that will be satisfied is:

$\left\{\begin{array}{c}\underset{i^{\&prime;}=1}{\overset{I}{\&Sigma;}}\underset{n^{\&prime;}=1}{\overset{N}{\&Sigma;}}{a}_{i^{\&prime;},n^{\&prime;}}{p}_{i^{\&prime;},n^{\&prime;}}\&le;P_{tot}\\ p_{i^{\&prime;},n^{\&prime;}}\&GreaterEqual;0\\ \underset{i^{\&prime;}}{\&Sigma;}{a}_{i^{\&prime;},n^{\&prime;}}=1\end{array}\right.$

{a_i',n'Represent that subcarrier distributes set, and value is that subcarrier n' is distributed to user i' by 0 or 1,1 expression, 0 representsSubcarrier n' does not distribute to user i';

First constraints represents the restrictive condition of general power, wherein P_totRepresent maximum transmission power; Second approximatelyBundle condition represents that the power of user i' allocation of subcarriers n' is more than or equal to 0; Last constraints represents each heightCarrier wave can only distribute once.

The invention has the advantages that:

1), combine the method that dynamic access and subcarrier distribute under a kind of super-intensive network based on buffer memory, according to emulationResult can find out, the method has improved spectrum utilization efficiency effectively, and this result has proved that this mechanism is intensiveUnder scene, meet feasibility and the applicability of multiple business demand.

2), combine the method that dynamic access and subcarrier distribute under a kind of super-intensive network based on buffer memory, can be comprehensiveMultiple factors complete access and select, and realize the lifting of resource management efficiency.

3), combine the method that dynamic access and subcarrier distribute under a kind of super-intensive network based on buffer memory, can realizeThe dynamic assignment of subcarrier, significantly promotes spectrum utilization efficiency. .

Brief description of the drawings

Fig. 1 is system model schematic diagram of the present invention;

Fig. 2 is system model block architecture diagram of the present invention;

Fig. 3 the present invention is based under the super-intensive network of buffer memory, to combine the method flow that dynamic access and subcarrier distributeFigure;

Fig. 3 a is the local flow chart of best access point being distributed to user of controlling of the present invention;

Fig. 3 b is the flow chart that the each user of the present invention and access point separately carry out subcarrier distribution;

Fig. 4 is lower 4 users' of multiple attribute decision making (MADM) algorithm of the present invention weight factor emulation schematic diagram;

Fig. 5 is that ordering of optimization preference emulation schematic diagram is selected in network insertion of the present invention;

Fig. 6 is spectrum utilization efficiency of the present invention and user's number graph of a relation;

Fig. 7 is system spectral efficiency of the present invention and Between Signal To Noise Ratio figure;

Detailed description of the invention

Below in conjunction with accompanying drawing, the present invention is described in further detail.

As shown in Figure 1, in super-intensive network, adopt far-end server as core net, connect multiple access points, accessPoint is analyzed buffer memory part content according to content popularit, and each user connects respectively an access point, as shown in Figure 2,Access point sends attribute parameter separately to local control, local control according to user request information and the access of receivingPoint attribute parameter, utilizes multiple attribute decision making (MADM) matrix for user's Dynamic Selection access point, in selection course, need to comprehensively examineConsider each attribute of access point, carry out multiattribute access as buffer memory capacity, time delay, signal to noise ratio etc. and select. Complete movingAfter state access procedure, on the basis of greedy algorithm, complete subcarrier and distribute, based on transmitted power, spectral bandwidth andCache size restriction is chosen AP and subcarrier, taking maximum spectral efficiency as target, completes subcarrier and distributes.

The present invention is first according to user request information, adopts broadcast mode to find cache contents, and feedback result represents with S,If certain access point is idle and have this cache contents, is designated as S=1, otherwise is designated as S=0. Access point by oneselfAttribute information, as buffer memory capacity, time delay, signal to noise ratio etc., sends local control to. If user does not obtain any accessThe response of point, directly sends request to server end, obtains content, and server end is according to user request information, profitUse popularity analysis, complete buffer update; If user obtains access point response, according to user's content and type of serviceRequest, carries out level analysis, selects best access point, is controlled and is notified selecteed access point that clothes are provided by this localityBusiness, completes access and selects, and upgrades access and selects set Q.

According to access selection result, under the prerequisite of known channel state information, calculate each user by water-filling algorithmThe channel capacity of corresponding subcarrier, on the basis of the channel capacity calculating, carries the son with optimum channel capacityRipple priority allocation, then carries out interative computation, finally completes subcarrier and distributes.

Under super-intensive network based on buffer memory, combine the method that dynamic access and subcarrier distribute, as shown in Figure 3,Concrete steps are as follows:

Step 1, multiple user send request information to all access points with broadcast mode simultaneously, find cache contents;

Solicited message comprises: cache contents; Number of users is O;

Step 2, using user K as active user, according to the solicited message of user K, each access point judges whetherThe cache contents that exists active user to ask, if certain access point is idle and have this cache contents, this access point is anti-User K is given in feedback 1, enters step 3; Otherwise feedback 0; Enter step 4;

1≤K≤O；

Step 3, all access points that meet user K send attribute parameter separately to local control, local controlBest access point is distributed to user K;

The all access points that meet user K are m; The attribute parameter of each access point include buffer memory capacity, timeProlong with signal to noise ratio etc., altogether n attribute;

Local control according to user's request content and type of service, carries out level analysis, selects best access point,Notify selecteed access point that service is provided, complete access and select, then upgrade access and select set Q.

As shown in Figure 3 a, concrete steps are as follows:

Step 301, according to the solicited message of user K, for n attribute parameter of each candidate's access point, this localityControl the relative importance between every two attribute parameters compared one by one, obtain decision matrix M:

Cache contents difference, the attribute specification difference of selecting for access point, structure judgement matrix represents each candidateImportance relation between each attribute parameter of access point, by comparing between two the importance between different attribute, withDetermine the significance level of certain attribute, conventionally in 1～9 ratio scale, importance degree is carried out to assignment, in table 1, giveGo out the implication of 1～9 scale:

Table 1

Local control compares the relative importance between each attribute parameter of each candidate's access point, obtains thisThe decision matrix M of candidate's access point:

Wherein a_ijRepresent parameter i and parameter j in access point relative importance fiducial value;

$a_{ij}>0;a_{ij}=\frac{1}{a_{ji}};a_{ii}=1;$

Step 302, decision-making matrix M is normalized, obtains the decision matrix B after standardization:

Wherein b_ijRepresent in access point fiducial value a_ijValue after normalization;

Step 303, the decision matrix B after standardization is carried out to consistency desired result; Judge that whether decision matrix is effective,If effectively, carry out step 304, otherwise returns to step 302;

Before calculating each attribute weights, need to carry out consistency desired result to decision matrix. Because if decision matrixToo depart from uniformity, the weight vectors calculating will not had a credibility, is therefore necessary decision-making matrix BUniformity is carried out verification. Be defined as follows:

b_ik×b_kj＝b_ij,i,j,k＝1,2,…,n(3)

b_ikFor parameter i in access point after standardization and parameter k relative importance fiducial value;

If equation is set up, represent that judgment matrix has uniformity. Represent that policymaker is carrying out attributeWhile comparing between two, thinking has uniformity. But because people's thinking has certain subjectivity, be difficult to keep absolutelyRight uniformity, so in order to weigh the uniformity of matrix, introduce inconsistency index CI, Consistency Ratio CREtc. concept:

$CI=\frac{{\mathrm{\&lambda;}}_{max}-n}{n-1}---\left(4\right)$

$CR=\frac{CI}{RI}---\left(5\right)$

Wherein, λ_maxBe the eigenvalue of maximum of decision matrix B, n is the number of attribute parameter in decision matrix B,RI is known mean random coincident indicator, as shown in the table:

Table 2

n	1	2	3	4	5	6	7	8	9
										RI	0	0	0.58	0.90	1.12	1.24	1.32	1.41	1.45

In the time of CR < 0.1, think that decision matrix B has acceptable uniformity, otherwise re-construct decision matrixB。

Step 304, calculate n the comprehensive generation of attribute parameter in decision matrix B weight vectors ω;

ω＝(ω₁,ω₂,...ω_j,...,ω_n)

ω_jIt is the weight of j attribute parameter; The present invention adopts the characteristic vector after normalization to select as network insertionWeight vectors.

The state matrix S of all properties parameter of step 305, m candidate's access node of generation;

State matrix S is the capable n row of m, and every a line represents n attribute parameter of each access point;

Wherein, s_mnRepresent the value of corresponding n the attribute parameter of m access point.

Step 306, weight vectors ω and state matrix S are multiplied each other and obtain weighting decision matrix Q:

Step 307, according to weighting decision matrix Q, determine best access scheme Q_bestThe poorest access scheme Q_worst。

Best access scheme is the best situation of selecting each parameter, such as for buffer memory capacity,Just select maximum, for power consumption, just select minimum of a value. The poorest access scheme in contrast, calculatesMode is as follows:

Q_best＝(ω₁·s_1best,ω₂·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)

s_jbestRepresent optimum value in j attribute parameter of all m access point; s_jworstRepresent all m access pointJ attribute parameter in worst-case value;

Step 308, for certain access point l, respectively calculated candidate access scheme x_ljWith best access scheme Q_bestEuropeFamily name's distance, and candidate's access scheme x_ljWith the poorest access scheme Q_worstEuclidean distance;

Candidate's access scheme x_ljWith best access scheme Q_bestEuclidean distance be Q_lbest, specifically refer to the candidate side of accessCase x_ljEach attribute parameter and this attribute optimal value s_jbestEuclidean distance, as follows:

$Q_{lbest}=\sqrt{\underset{j=1}{\overset{n}{\&Sigma;}}{\mathrm{\&omega;}}_j{(x_{lj}-s_{jbest})}^2}---\left(10\right)$

Candidate's access scheme x_ljWith the poorest access scheme Q_worstEuclidean distance be Q_lworst, specifically refer to candidateAccess scheme x_ljEach attribute parameter and this attribute worst-case value s_jworstEuclidean distance, as follows:

$Q_{lworst}=\sqrt{\underset{j=1}{\overset{n}{\&Sigma;}}{\mathrm{\&omega;}}_j{(x_{lj}-s_{jworst})}^2}---\left(11\right)$

Step 309, for certain access point l, calculated candidate scheme x_ljAnd the preference value P between preferred plan_l。

By calculating preference value, draw the ratio of distance between candidate network and optimal network and the poorest network. CalculateFormula as follows:

$P_l=\frac{Q_{lworst}}{Q_{lbest}+Q_{lworst}}---\left(12\right)$

Wherein P_lWhat represent is the preference value of user K for l access point.

The maximum preference value of user K to each access point: P_k＝(P_1k,P_2k,...P_lk,...P_mk)；

Step 310, calculate all users respectively for the preference value of each access point, and arranged in sequence, user generatedSelection matrix R; As follows:

Wherein, P_mORepresent that O user selects the preference value of m access point. O user's choosing of every a line representativeSelect the preference value of each access point;

Step 311, for each user, access point corresponding maximum preference value is distributed to this user by local control,And delete the access point of this user and coupling, distribute successively, until all users complete network insertion.

Local control according to the row of user's selection matrix R, selects maximum preference value, by access corresponding this preference valuePoint is distributed to respective user, deletes user and the access point of coupling after distributing, and continues to distribute successively, until institute is usefulFamily all completes network insertion.

Each list of user's selection matrix R is shown, the preference value of each user to m access point;

The simulation parameter of table 3 access point

	Signal to noise ratio (/dBm)	Time delay (/ms)	Covering radius (km)	Buffer memory capacity (TB)	Power consumption (W)
						Access point 1	70	25	50	100	1/100
Access point 2	60	50	40	200	1/100
						Access point 3	73	80	70	400	1/50
Access point 4	50	30	30	200	1/70

According to content popularit analytic approach, content hit rate is relevant with Zipf index. In this emulation, set ZipfParameter is 0.8, and cache hit rate is 0.7, and without loss of generality, user's number that setting obtains access point response is 4,All the other 2 users need to obtain content from back-end server. These 4 users, in different regions, have differentBusiness. User 1 belongs to central user, request network browsing business; User 2 belongs to central user, request Streaming MediaBusiness; User 3 belongs to edge customer, request streaming media service; User 4 belongs to edge customer, request network browsingBusiness. The multiple attribute decision making (MADM) algorithm proposing according to this section, as shown in Figure 4, abscissa has represented 5 ginsengs to simulation resultAmount: signal to noise ratio, time delay, covering radius, buffer memory capacity, power consumption. Ordinate represents the weight of corresponding each parameterValue, can find out that weight factor has very big-difference for different user's requests. Come for streaming media serviceSaying, is mainly one-way transmission, does not need two-way real-time Communication for Power, less demanding in time delay, but file is generalLarger, need higher buffer memory capacity, user 2 and user 3 are request streaming media service, so time delay factor powerHeavy lower, buffer memory capacity factor weight is higher. For network browsing business, buffer memory capacity is also low, but needsHigher service quality and lower time delay, thus signal to noise ratio is had relatively high expectations, as user 1 and user 4 ask networkBrowse service, the weight of time delay factor is higher, and the weight of the buffer memory capacity factor is lower. Central user is to covering radiusRequire lower, so the weight of user 1 and user's 2 the covering radius factor is lower; And user 3 and user 4 cover halfThe weight of the footpath factor is higher.

According to the attribute information of weight factor and each access point, obtain each user inclined to one side for the selection of access pointGood value, simulation result as shown in Figure 5:

As seen from Figure 5, for same access point, its parameter information is constant, but different userThe preference value calculating is but different, illustrates that method in this paper has taken into full account user's demand information,For same user, its network demand is also constant, and what still under different access points, calculate is inclined to one sideGood value is also different, illustrates that algorithm has also taken into full account the parameter information of access point. Thereby, net in this paperNetwork access selection algorithm is taken into account user's request and access point performance simultaneously, is the basis of effectively carrying out subcarrier distribution.End user's network insertion selection result is as shown in table 4:

Table 4 user accesses selection result

User	Access point back-end server
		User 1	Access point 3
User 2	Access point 2
		User 3	Access point 1
User 4	Access point 4
		User 5	Back-end server
User 6	Back-end server

Step 4, user K do not obtain the response of any access point, and user K directly sends to far-end serverRequest, obtains content; Far-end server, according to user request information, utilizes popularity analysis, completes buffer update;

Step 5, each user carry out subcarrier distribution after mating with access point separately, make user and access point itBetween communicate.

According to access selection result, under the prerequisite of known channel state information, calculate each user by water-filling algorithmThe channel capacity of corresponding subcarrier, on the basis of the channel capacity calculating, carries the son with optimum channel capacityRipple priority allocation, then carries out interative computation, finally completes subcarrier and distributes.

After completing user access is selected, on the basis of greedy algorithm, propose a kind of based on power system capacity maximumBeggar's carrier wave priority allocation algorithm. Under the prerequisite of known channel state information, calculate each use by water-filling algorithmThe channel capacity of the corresponding subcarrier in family, then on the basis of the channel capacity of calculating, to having optimum channel capacitySubcarrier priority allocation.

As shown in Figure 3 b, concrete steps are as follows:

Step 501, initialization subcarrier set and user's set;

Subcarrier set is N={n'|n'=1,2 ..., N}, user gathers I={i'|i'=1, and 2 ..., I}, distributes to userThe sub-carrier indices X of i'_i'＝φ。

Step 502, according to water-filling algorithm calculate each user the transmitting power of corresponding each subcarrier and channel holdAmount.

Transmitting power p_i',n'：

$p_{i^{\&prime;},n^{\&prime;}}=\frac{P_{tot}+\underset{n^{\&prime;}=1}{\overset{N}{\&Sigma;}}1/{\mathrm{\&gamma;}}_{i^{\&prime;},n^{\&prime;}}}{N}-1/{\mathrm{\&gamma;}}_{i^{\&prime;},n^{\&prime;}}---\left(14\right)$

p_i',n'Represent the power to user i' allocation of subcarriers n', P_totRepresent maximum transmission power; γ_i',n'Represent userThe signal to noise ratio of i' allocation of subcarriers n'.

Channel capacity C_i'，n'：

$C_{i^{\&prime;},n^{\&prime;}}=\frac{B}{N}{\log }_2(1+p_{i^{\&prime;},n^{\&prime;}}{\mathrm{\&gamma;}}_{i^{\&prime;},n^{\&prime;}})---\left(15\right)$

B represents the bandwidth of total system framework model;

Step 503, for subcarrier n', calculate respectively the channel capacity of this subcarrier under each user, and fallOrder is arranged and is selected maximum

Subcarrier n' initial value is 1;

Be expressed as:

$C_{n^{\&prime;}}^{\max }={\max }_{i^{\&prime;}\&Element;I}\left\{C_{i^{\&prime;},n^{\&prime;}}\right\}---\left(16\right)$

As sub-carrier channels capacity by descending is:

$C_{1,n}\&GreaterEqual;C_{2,n}\&GreaterEqual;C_{3,n}\&GreaterEqual;...\&GreaterEqual;C_{I,n},C_n^{\max }=C_{1,n}$

Step 504, subcarrier n' is distributed to maximum channel capacity valueCorresponding user, and by subcarrier n' fromIn subcarrier set N, remove, return to step 503 and continue to choose according to the order of sequence next subcarrier, until all sons are carriedRipple distributes.

Step 505, to distribute after subcarrier carry out water filling, according to the transmitting power p of subcarrier_i',n'Computing system frequentlySpectrum utilization ratio;

It is as follows that system spectrum utilization ratio maximizes object function:

${\max }_{\{a_{i^{\&prime;},n^{\&prime;}},p_{i^{\&prime;},n^{\&prime;}}\}}\frac{C}{B}={\max }_{\left\{a_{i^{\&prime;},n^{\&prime;}},p_{i^{\&prime;},n^{\&prime;}}\right\}}\underset{i^{\&prime;}=1}{\overset{I}{\&Sigma;}}\underset{n^{\&prime;}=1}{\overset{N}{\&Sigma;}}{a}_{i^{\&prime;},n^{\&prime;}}{\log }_2(1+p_{i^{\&prime;},n^{\&prime;}}{\mathrm{\&gamma;}}_{i^{\&prime;},n^{\&prime;}})---\left(17\right)$

Condition that will be satisfied is:

$\left\{\begin{array}{c}\underset{i^{\&prime;}=1}{\overset{I}{\&Sigma;}}\underset{n^{\&prime;}=1}{\overset{N}{\&Sigma;}}{a}_{i^{\&prime;},n^{\&prime;}}{p}_{i^{\&prime;},n^{\&prime;}}\&le;P_{tot}\\ p_{i^{\&prime;},n^{\&prime;}}\&GreaterEqual;0\\ \underset{i^{\&prime;}}{\&Sigma;}{a}_{i^{\&prime;},n^{\&prime;}}=1\end{array}\right.---\left(18\right)$

B represents the bandwidth of total system framework model, { a_i',n'Representing that subcarrier distributes set, value is that 0 or 1,1 expression willSubcarrier n' distributes to user i', and 0 represents that subcarrier n' does not distribute to user i';

First constraints represents the restrictive condition of general power, wherein P_totRepresent maximum transmission power; Second approximatelyBundle condition represents that the power of user i' allocation of subcarriers n' is more than or equal to 0; Last constraints represents each heightCarrier wave can only distribute once.

In the present embodiment, emulation adopts rayleigh fading channel model, and sub-carrier number is 64, and number of users is 16, totalPower P_totFor 1W, noise power spectral density N₀For 10e-8W/Hz, total bandwidth B is 1MHz. And hold with minimumMeasure these two kinds of classical sub-carrier wave distribution methods of maximum method (MAX-MIN) and fair rule of three (FPS) and compare, emulationResult is as shown in Figure 6 and Figure 7:

As seen from Figure 6, in the situation that user's number is identical, the frequency spectrum profit that the algorithm that the present invention proposes obtainsBe much higher than equitable proportion algorithm and minimum capacity by efficiency and maximize method, because algorithm of the present invention entirety is considered systemSystem capacity, comes allocation of subcarriers and power according to channel condition information, thereby has greatly improved the utilization of resources of systemRate. The availability of frequency spectrum of equitable proportion algorithm maximizes method a little more than minimum capacity simultaneously, although because all consideredFairness, but equitable proportion algorithm has been considered power system capacity maximization simultaneously, and minimum capacity maximization method is only consideredThe distribution of subcarrier between user, do not consider that the self adaptation of power between subcarrier distribute, thus frequency spectrumUtilization ratio rate is lower.

Fig. 7 is the curve that spectrum utilization efficiency changes with signal to noise ratio, can find out the in the situation that of identical signal to noise ratio, thisThe spectrum utilization efficiency that the algorithm that literary composition proposes obtains is much higher than equitable proportion algorithm and minimum capacity maximizes method,This is because this algorithm carrys out allocation of subcarriers by constantly carrying out iteration water filling calculating channel capacity, can keep away betterExempt from the subcarrier with poor channel quality to distribute to user, thereby can improve system spectrum utilization ratio.

Claims (3)

1. under the super-intensive network based on buffer memory, combine the method that dynamic access and subcarrier distribute, its feature existsIn, comprise the following steps:

Step 1, multiple user send request information to all access points with broadcast mode, find cache contents;

Number of users is O;

Step 2, using user K as active user, each access point judges whether the buffer memory that exists active user to askContent, if certain access point is idle and have this cache contents, this access point feedback 1 is given user K, enters stepThree; Otherwise feedback 0; Enter step 4;

Step 3, all access points that meet user K send attribute parameter separately to local control, local controlBest access point is distributed to user K;

The all access points that meet user K are m;

Step 4, user K do not obtain the response of any access point, and user K directly sends to far-end serverRequest, obtains content; Far-end server, according to user request information, utilizes popularity analysis, completes buffer update;

Step 5, each user carry out subcarrier distribution after mating with access point separately, make user and access point itBetween communicate;

Specifically comprise:

Step 501, initialization subcarrier set and user's set;

Subcarrier set is N={n'|n'=1,2 ..., N}, user gathers I={i'|i'=1, and 2 ..., I}, distributes to userThe sub-carrier indices X of i'_i'＝φ；

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

Transmitting power p_i',n'：

$p_{i^{\&prime;},n^{\&prime;}}=\frac{P_{tot}+\underset{n^{\&prime;}=1}{\overset{N}{\&Sigma;}}1/{\mathrm{\&gamma;}}_{i^{\&prime;},n^{\&prime;}}}{N}-1/{\mathrm{\&gamma;}}_{i^{\&prime;},n^{\&prime;}}$

p_i',n'Represent the power to user i' allocation of subcarriers n', P_totRepresent maximum transmission power; γ_i',n'Represent userThe signal to noise ratio of i' allocation of subcarriers n';

Channel capacity C_i',n'：

$C_{i^{\&prime;},n^{\&prime;}}=\frac{B}{N}{\log }_2(1+p_{i^{\&prime;},n^{\&prime;}}{\mathrm{\&gamma;}}_{i^{\&prime;},n^{\&prime;}})$

B represents this overall system bandwidth;

Step 503, for subcarrier n', calculate respectively the channel capacity of this subcarrier under each user, and fallOrder is arranged and is selected maximum

Subcarrier n' initial value is 1;

$C_{n^{\&prime;}}^{\max }={\max }_{i^{\&prime;}\&Element;I}\left\{C_{i^{\&prime;},n^{\&prime;}}\right\}$

Step 504, subcarrier n' is distributed to maximum channel capacity valueCorresponding user, and by subcarrier n' fromIn subcarrier set N, remove, return to step 503 and continue to choose according to the order of sequence next subcarrier, until all sons are carriedRipple distributes;

Step 505, to distribute after subcarrier carry out water filling, according to the transmitting power p of subcarrier_i',n'Computing system frequentlySpectrum utilization ratio;

It is as follows that system spectrum utilization ratio maximizes object function:

${\max }_{\{a_{i^{\&prime;},n^{\&prime;}},p_{i^{\&prime;},n^{\&prime;}}\}}\frac{C}{B}={\max }_{\{a_{i^{\&prime;},n^{\&prime;}},p_{i^{\&prime;},n^{\&prime;}}\}}\underset{i^{\&prime;}=1}{\overset{I}{\&Sigma;}}\underset{n^{\&prime;}=1}{\overset{N}{\&Sigma;}}{a}_{i^{\&prime;},n^{\&prime;}}{\log }_2(1+p_{i^{\&prime;},n^{\&prime;}}{\mathrm{\&gamma;}}_{i^{\&prime;},n^{\&prime;}})$

Condition that will be satisfied is:

$\left\{\begin{array}{c}\underset{i^{\&prime;}=1}{\overset{I}{\&Sigma;}}\underset{n^{\&prime;}=1}{\overset{N}{\&Sigma;}}{a}_{i^{\&prime;},n^{\&prime;}}{p}_{i^{\&prime;},n^{\&prime;}}\&le;P_{tot}\\ p_{i^{\&prime;},n^{\&prime;}}\&GreaterEqual;0\\ \underset{i^{\&prime;}}{\&Sigma;}{a}_{i^{\&prime;},n^{\&prime;}}=1\end{array}\right.$

{a_i',n'Represent that subcarrier distributes set, and value is that subcarrier n' is distributed to user i' by 0 or 1,1 expression, 0 representsSubcarrier n' does not distribute to user i';

First constraints represents the restrictive condition of general power, wherein P_totRepresent maximum transmission power; Second approximatelyBundle condition represents that the power of user i' allocation of subcarriers n' is more than or equal to 0; Last constraints represents each heightCarrier wave can only distribute once.

2. under a kind of super-intensive network based on buffer memory as claimed in claim 1, combine dynamic access and subcarrier dividesThe method of joining, is characterized in that, the attribute parameter of described each access point includes buffer memory capacity, time delay and signal to noise ratioDeng, n attribute altogether.

3. under a kind of super-intensive network based on buffer memory as claimed in claim 1, combine dynamic access and subcarrier dividesThe method of joining, is characterized in that, described step 3 is specially:

Step 301, according to the solicited message of user K, for each candidate's access point, local control joined n attributeRelative importance in amount between every two attribute parameters compares one by one, obtains decision matrix M:

a_ijRepresent the relative importance fiducial value of attribute parameter i and attribute parameter j in access point;a_ii＝1；

Step 302, decision-making matrix M is normalized, obtains the decision matrix B after standardization:

Wherein b_ijRepresent in access point fiducial value a_ijValue after normalization;

Step 303, the uniformity of decision-making matrix B is carried out to verification, judge that whether decision matrix is effective, if hadEffect, carry out step 304, otherwise returns to step 302;

Consistency Ratio CR is defined as follows:

Wherein, CI represents inconsistency index:λ_maxThe eigenvalue of maximum of decision matrix B,N is the number of attribute parameter in decision matrix B, and RI is known mean random coincident indicator;

In the time of CR < 0.1, think that decision matrix B has acceptable uniformity, otherwise re-construct decision matrixB；

Step 304, obtain in decision matrix B n attribute parameter comprehensive produce weight vectors ω;

ω＝(ω₁,ω₂,...ω_j,...,ω_n)

ω_jIt is the weight of j attribute parameter;

Step 305, for m candidate's access point, generate the state matrix S of all properties parameter;

State matrix S is the capable n row of m, and every a line represents n attribute parameter of each access point;

Wherein, s_mnRepresent the value of corresponding n the attribute parameter of m access point;

Step 306, weight vectors ω and state matrix S are multiplied each other and obtain weighting decision matrix Q:

Step 307, according to weighting decision matrix Q, determine best access scheme Q_bestThe poorest access scheme Q_worst，

Q_best＝(ω₁·s_1best,ω₂·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)

s_jbestRepresent optimum value in j attribute parameter of all m access point; s_jworstRepresent all m access pointJ attribute parameter in worst-case value;

Step 308, for certain access point l, respectively calculated candidate access scheme x_ljWith best access scheme Q_bestEuropeFamily name's distance, and candidate's access scheme x_ljWith the poorest access scheme Q_worstEuclidean distance;

Candidate's access scheme x_ljWith best access scheme Q_bestEuclidean distance be Q_lbest, specifically refer to the candidate side of accessCase x_ljEach attribute parameter and this attribute optimal value s_jbestEuclidean distance, as follows:

$Q_{lbest}=\sqrt{\underset{j=1}{\overset{n}{\&Sigma;}}{\mathrm{\&omega;}}_j{(x_{lj}-s_{jbesi})}^2}$

Candidate's access scheme x_ljWith the poorest access scheme Q_worstEuclidean distance be Q_lworst, specifically refer to that candidate accessesScheme x_ljEach attribute parameter and this attribute worst-case value s_jworstEuclidean distance, as follows:

$Q_{lworst}=\sqrt{\underset{j=1}{\overset{n}{\&Sigma;}}{\mathrm{\&omega;}}_j{(x_{lj}-s_{jworsi})}^2}$

Step 309, for certain access point l, calculated candidate scheme x_ljAnd the preference value P between preferred plan_l；

Formula is as follows:

$P_l=\frac{Q_{lworst}}{Q_{lbest}+Q_{lworst}}$

Wherein P_lWhat represent is the preference value of user K for l access point;

Step 310, calculate all users respectively for the preference value of each access point, and arranged in sequence, user generatedSelection matrix R;

Wherein, P_mORepresent that O user selects the preference value of m access point; All O use of every a line representativeThe preference value of each identical access point is selected at family; The list of user's selection matrix R shows that each user selects each difference to connectEnter preference value a little;

Step 311, for each user, access point corresponding maximum preference value is distributed to this user by local control,And delete the access point of this user and coupling, distribute successively, until all users complete network insertion.