CN101902747A - Spectrum allocation method based on fuzzy logic genetic algorithm - Google Patents

Abstract

The invention discloses a spectrum allocation method based on a fuzzy logic genetic algorithm, which mainly solves the problems of low allocation efficiency and high complexity of traditional cognitive network spectrum allocation. The spectrum allocation method based on the fuzzy logic genetic algorithm comprises a static network allocation method and a dynamic network allocation method. The static network allocation method comprises mapping a spectrum allocation matrix onto a chromosome, adaptively adjusting crossover rate and variation rate by fuzzy logics in an iterative process of a genetic algorithm, performing crossing-over and variation operation and inversely mapping the chromosome onto the chromosome so as to optimize the spectrum allocation. The dynamic network allocation method comprises establishing a discussion group for each mobile subscriber, mapping a spectrum allocation matrix in the discussion group onto a chromosome and optimizing spectrum resources in the group by the fuzzy logic genetic algorithm so that the spectrum resources of the whole network are optimized. The static network spectrum allocation method has the advantage of efficient spectrum allocation. The dynamic network allocation method has the advantages of low complexity and less time consumption. The spectrum allocation method based on the fuzzy logic genetic algorithm is useful for cognition of a static network and a dynamic network.

Description

Frequency spectrum distributing method based on fuzzy logic genetic algorithm

Technical field

The invention belongs to the cognitive radio technology field in the radio communication, relate to cognitive frequency spectrum distributing method, specifically utilize the genetic algorithm of fuzzy logic control, solve the spectrum allocation may problem, can be applicable in the cognition network from centralized and distributed two aspects.

Background technology

Cognitive radio technology in the radio communication is a kind of intelligent radio electrical communication technology, can reach the high reliability of communication system and the high efficiency of the availability of frequency spectrum whenever and wherever possible by perception surrounding environment, modulation operational factor.

Progressively perfect along with the development of cognitive radio technology and IEEE802.22 standard, the access of cognitive frequency spectrum resource and distribution technique also are subjected to scholar's extensive concern and research.In the cognition network of 802.22WRAN, N TV equipment user arranged, promptly main user, each main user uses an independently channel; M customer premises equipment, CPE CPE arranged, i.e. secondary user's is used main user's idle channel.

At present, from domestic and international existing technical situation, for the theoretical research of cognitive radio, frequency spectrum distributing method is divided three classes substantially:

The first kind is the Nash Equilibrium notion of utilizing in the theory of games, realize the fair effectively distribution of frequency spectrum, as G.Scutari, D.P.Palomar, S.Barbarossa, et al.MIMO Cognitive Radio:A Game Theoretical Approach[C], IEEE Workshop on Signal Processing Advances in Wireless Communications, Recife, Brazil, July 6-9,2008,426-430.

Second class is based on the auction theory in the economics, solve the frequency spectrum lease problem under the different auction mechanism, as Wu Yongle, Wang Beibei, Liu K.J.Ray, et al.A Scalable Collusion-resistant Multi-winner Cognitive Spectrum Auction Game[J], IEEE Transactions on Communications, 2009,57 (12): 3805-3816.

The 3rd class is the research that the thought of utilization graph theory is carried out Frequency Distribution and power control, the spectrum allocation may problem is converted into the map colouring problem of the color-sensitive in the graph theory, by a series of suboptimization algorithm approximate optimal solutions, as Peng Chunyi, Zheng Haitao, Zhao Ben Y. Utilization and Fairness in Spectrum Assignment for Opportunistic Spectrum Access[J] .ACM Mobile Networks and Applications, 2006,11 (4): 555-576.

The first kind and second class methods are because there are shortcomings such as computation complexity height, spectrum allocation may time length in himself theoretical characteristic; The 3rd class methods algorithm complex is low, but spectrum allocation may efficient is lower, and the local one to one negotiation algorithm that is used for greedy algorithm, maximum and the bandwidth algorithm of static network and is used for dynamic network all belongs to this type of.

Zhijin Zhao introduces genetic algorithm in the spectrum allocation may process in article Cognitive Radio Spectrum Allocation using Evolutionary Algorithms, quantum genetic algorithm (the Quantum Genetic Algorithm that proposes, QGA) has spectrum allocation may performance efficiently, but only considered static spectrum allocation may algorithm, during, wireless network environment dynamic change huge when network topology structure, static spectrum allocation may algorithm computation amount increases, and efficiency of algorithm reduces greatly.

Summary of the invention

The objective of the invention is to overcome the deficiency of above-mentioned prior art, a kind of static network frequency spectrum distributing method based on fuzzy logic genetic algorithm that is applicable to static network is proposed, with a kind of dynamic network frequency spectrum distributing method that is applicable to dynamic network based on fuzzy logic genetic algorithm, to improve spectrum allocation may efficient under the different network environments, strengthen network-adaptive.

For achieving the above object, the frequency spectrum distributing method based on fuzzy logic genetic algorithm that the present invention proposes comprises following two kinds of methods:

One, based on the static network frequency spectrum distributing method of fuzzy logic genetic algorithm, performing step comprises as follows:

(1) determines effective spectral matrix Γ={ u according to the cognition network structure _{M, n}| u _{M, n}∈ 0,1}} _{M * N}, benefit matrix E={e _{M, n}} _{M * N}With interference matrix C={c _{M, k}| c _{M, k}∈ 0,1}} _{M * M}, wherein M represents user's number, N represents channel bar number; u _{M, n}=1 expression user m can utilize channel n, u _{M, n}=0 expression user m can not utilize channel n; e _{M, n}Obtainable benefit when expression user m uses channel n; c _{M, k}=1 expression user m uses identical channel can produce interference, c with user k _{M, k}=0 expression can not produce interference;

(2) with spectrum allocation may matrix A={ a _{M, n}| a _{M, n}∈ 0,1}} _{M * N}Middle element is initialized as 0, wherein a _{M, n}=1 expression is distributed to user m, a with channel n _{M, n}The unallocated user m that gives of=0 expression channel n; And element in the spectrum allocation may matrix A is mapped as the chromosome code element one by one, chromosome length is len=MN;

(3) initialization genetic algorithm parameter comprises population size S, aberration rate P _mWith crossing-over rate P _cInitialization genetic algebra g=1 produces the population size at random and is the initial population of S, promptly produces S bar chromosome, and the chromosome matrix notation is

Wherein

K the code element of expression chromosome i;

(4) according to the chromosome matrix L ⁽ⁱ⁾, benefit matrix E and effective spectrum matrix Γ calculate every chromosomal fitness function

Calculate genetic algorithm converges speed according to fitness function

(5) according to above-mentioned convergence rate R (g) and genetic algebra g, with convergence rate and genetic algebra obfuscation, the obfuscation of convergence rate and genetic algebra output is respectively With

Wherein Low, and Med, High} are the rate of convergence fuzzy set, Low, Med, High represent the basic, normal, high degree of convergence rate respectively; Small, Med, Large} are the fuzzy set of genetic algebra, Small, Med, Large represent respectively genetic algebra little, in, big degree;

According to fuzzy logic ordination collection launch mode gelatinization crossing-over rate and aberration rate degree of membership

With

Wherein Smaller, Small, Med, Large, Larger represent respectively crossing-over rate and aberration rate degree be minimum, little, in, big, greatly, utilize following formula reverse gelatinization output P _cAnd P _m:

$P_c=\frac{D_{P_c}^{\mathrm{Smaller}}{x}_1+D_{P_c}^{\mathrm{Small}}{x}_2+D_{P_c}^{\mathrm{Med}}{x}_3+D_{P_c}^{L\arg e}{x}_4+D_{P_c}^{L\arg \mathrm{er}}{x}_5}{{\mathrm{\Σ}}_{i=1}^5{x}_i}$

$P_m=\frac{D_{P_m}^{\mathrm{Smaller}}{y}_1+D_{P_m}^{\mathrm{Small}}{y}_2+D_{P_m}^{\mathrm{Med}}{y}_3+D_{P_m}^{L\arg e}{y}_4+D_{P_m}^{L\arg \mathrm{er}}{y}_5}{{\mathrm{\Σ}}_{i=1}^5{y}_i}$

X wherein _iAnd y _iBe respectively crossing-over rate and aberration rate fuzzy set central point;

(6) according to adjusted crossing-over rate P _cWith aberration rate P _mAgain chromosome is intersected and mutation operation, thereby obtain S (P _c+ 1) bar chromosome is chosen S bar chromosome as population of future generation according to fitness function order from big to small;

(7) judge whether genetic algebra g reaches maximum evolutionary generation g _Max, if reach, the chromosome matrix inversion of fitness maximum shine upon back allocation matrix A, distribute corresponding frequency spectrum to each user; Otherwise genetic algebra g '=g+1 continues step (4)～(6).

Two, based on the dynamic network frequency spectrum distributing method of fuzzy logic genetic algorithm, performing step comprises as follows:

1) during the cognition network initial construction, utilize the static network frequency spectrum distributing method to obtain the spectrum allocation may matrix A, for each user distributes frequency spectrum;

2), mark the user's collection { U that moves in the network every a time interval Δ ₁, U ₂..., U _{N '}, N ' expression mobile subscriber number wherein is to mobile subscriber U _iFind out with user that it disturbs mutually and form discussion group τ _i

3) be each discussion group τ _iSet up a sub-conflict graph, obtain the effective spectrum matrix respectively

The benefit matrix

Interference matrix

With the spectrum allocation may matrix

M wherein _iExpression discussion group τ _iMiddle number of users, N _iExpression discussion group τ _iThe middle spendable number of channel of user; u _{I, m, n}=1 expression discussion group τ _iMiddle user m can use channel n, u _{I, m, n}=0 expression discussion group τ _iMiddle user m can not use channel n; e _{I, m, n}Expression discussion group τ _iObtainable income when middle user m uses channel n; c _{I, m, k}=1 expression discussion group τ _iMiddle user m uses identical channel can produce interference, c with user k _{I, m, k}=0 expression can not produce interference; a _{I, m, n}=1 expression discussion group τ _iIn channel n is distributed to user m, a _{I, m, n}=0 expression discussion group τ _iMiddle channel n is unallocated to give user m; Initialization i=1;

4) for discussion group τ ₁, with the spectrum allocation may matrix A _iIn be that 1 element is mapped as the chromosome code element one by one, obtain chromosome length

Produce S bar initial chromosome at random; Static network frequency spectrum distributing method according to proposing draws discussion group τ _iInner optimum spectrum allocation may A _i, upgrade corresponding frequency spectrum distribution in the overall spectrum allocation may matrix A;

5) make i '=i+1, repeating step 4) up to i '=N ', distribute corresponding frequency spectrum to the user according to the allocation matrix A of final updated.

The present invention has the following advantages:

1, static network frequency spectrum distributing method of the present invention owing to introduced the fuzzy logic control genetic algorithm parameter, makes crossing-over rate and the aberration rate can adaptive change, has therefore improved spectrum allocation may efficient, has the allocative efficiency comparable with the method for exhaustion.

2, dynamic network frequency spectrum distributing method of the present invention, because only needing that mobile subscriber in the network is carried out frequency spectrum redistributes, and do not need overall network is carried out spectrum allocation may, low with respect to static network frequency spectrum distributing method complexity, greatly reduce the algorithm time consumption, be applicable in the demanding dynamic network of real-time.

Description of drawings

Fig. 1 is the static network frequency spectrum distributing method flow process that the present invention is based on fuzzy logic genetic algorithm;

Fig. 2 is the dynamic network frequency spectrum distributing method flow process that the present invention is based on fuzzy logic genetic algorithm;

Fig. 3 is existing cognition network structure chart;

Fig. 4 is convergence rate used herein, genetic algebra fuzzy set membership function figure;

Fig. 5 is crossing-over rate used herein, aberration rate fuzzy set membership function figure;

Static network frequency spectrum distributing method and existing greedy algorithm, maximum and bandwidth algorithm that Fig. 6 proposes for the present invention, and the spectrum allocation may benefit comparison diagram of the method for exhaustion;

Dynamic network frequency spectrum distributing method and local machinery of consultation one to one that Fig. 7 proposes for the present invention, and the spectrum allocation may benefit comparison diagram of static network frequency spectrum distributing method;

The dynamic network frequency spectrum distributing method that Fig. 8 proposes for the present invention, with local machinery of consultation one to one, and the static network frequency spectrum distributing method time consumption comparison diagram of the present invention's proposition.

Embodiment

With reference to Fig. 1, the static network frequency spectrum distributing method that the present invention is based on fuzzy logic genetic algorithm comprises the steps:

Step 1 is determined effective spectral matrix Γ, interference matrix C and benefit matrix E according to the cognition network structure.

With reference to Fig. 3, the cognition network structure comprises M secondary user's and N main user, independent channel of each main CU, main user's interference range as shown in phantom in FIG., the secondary user's interference range is shown in chain-dotted line among the figure.

A. according to the cognition network structure, spectral matrix is Γ={ u _{M, n}| u _{M, n}∈ 0,1}} _{M * N}, when secondary user's m is not within the interference range of main user n, then use this main user's channel, at this moment u _{M, n}=1, otherwise u _{M, n}=0, u _{M, n}=1 expression user m can utilize channel n, u _{M, n}=0 expression user m can not utilize channel n;

B. according to the cognition network structure, interference matrix is C={c _{M, k}| c _{M, k}∈ 0,1}} _{M * M}, as two secondary user's m, the distance of k is during greater than the secondary user's interference distance, c _{M, k}=0, otherwise c _{M, k}=1; c _{M, k}=1 expression user m uses identical channel can produce interference, c with user k _{M, k}=0 expression can not produce interference;

C. according to the cognition network structure, in the time of can obtaining secondary user's n and on channel m, transmit and transmitter apart from d _{M, n}, the benefit matrix can by Try to achieve, wherein e _{M, n}Obtainable benefit when expression user m uses channel n, P _tBe transmitting power, B is a channel width, N ₀Be noise power spectral density.

Step 2 is to the genetic algorithm initialization.

(2.1) with spectrum allocation may matrix A={ a _{M, n}| a _{M, n}∈ 0,1}} _{M * N}Middle element is initialized as 0, wherein a _{M, n}=1 expression is distributed to user m, a with channel n _{M, n}The unallocated user m that gives of=0 expression channel n; And element in the spectrum allocation may matrix A is mapped in the binary digit of chromosome correspondence one by one, chromosome length is len=MN;

(2.2) initialization genetic algebra g=1 according to interference matrix C, produces the population size that meets noiseless condition at random and is the initial population of S, promptly produces S bar chromosome, with the chromosome matrix notation is:

Wherein

K the code element of expression chromosome i.

Step 3 is determined chromosome fitness function and genetic algorithm converges speed.

With crossing-over rate P _cWith aberration rate P _mChromosome is intersected and mutation operation, thereby obtain S (Pm+1) bar chromosome,, calculate the chromosomal fitness of i bar according to chromosome matrix L, benefit matrix E and effective spectrum matrix Γ

Calculate genetic algorithm converges speed according to fitness function

Step 4 is adjusted crossing-over rate and aberration rate by fuzzy logic.

Convergence rate used herein, genetic algebra fuzzy set membership function are as shown in Figure 4, wherein Fig. 4 (a) is a rate of convergence fuzzy set degree of membership, Fig. 4 (b) is a genetic algebra fuzzy set degree of membership, can obtain convergence rate and genetic algebra fuzzy set degree of membership and fuzzy set interval by this Fig. 4.

Crossing-over rate used herein, aberration rate fuzzy set membership function are as shown in Figure 5, wherein Fig. 5 (a) is a crossing-over rate fuzzy set degree of membership, Fig. 5 (b) is an aberration rate fuzzy set degree of membership, can obtain crossing-over rate and aberration rate fuzzy set interval and fuzzy set central point by this figure.

(4.1) with reference to convergence rate R shown in Figure 4 and genetic algebra g membership function figure, obtain the convergence rate fuzzy set Low, Med, High} and genetic algebra fuzzy set the membership function of Large} is respectively for Small, Med:

$D_R^{\mathrm{Low}}\left(x\right)=1-50x,0\&le;x<0.02$

$D_R^{\mathrm{Med}}\left(x\right)=\left\{\begin{array}{cc}50x,& 0\&le;x<0.02\\ 2-50x,& 0.02\&le;x<0.04\end{array}\right.---\left(1\right)$

$D_R^{\mathrm{High}}\left(x\right)=\left\{\begin{array}{cc}-1+50x,& 0.02\&le;x<0.04\\ 1,& x\&GreaterEqual;0.04\end{array}\right.$

$D_g^{\mathrm{Small}}\left(y\right)=\left\{\begin{array}{cc}1,& 0\&le;y<50\\ 2-0.02y,& 50\&le;y<100\end{array}\right.$

$D_g^{\mathrm{Med}}\left(y\right)=\left\{\begin{array}{cc}0.02y,& 50\&le;y<100\\ 3-0.02y,& 100\&le;y<150\end{array}\right.---\left(2\right)$

$D_g^{\mathrm{High}}\left(y\right)=\left\{\begin{array}{cc}-1+50y,& 100\&le;y<150\\ 1,& 150\&le;y\&le;200\end{array}\right.$

X, y represent the value of convergence rate R and genetic algebra g respectively, and with value difference substitution x, the y of convergence rate and genetic algebra, the obfuscation degree of membership that obtains convergence rate and genetic algebra is

With

Wherein Low, Med, High characterize the basic, normal, high of convergence rate respectively, and its scope is respectively [0,0.02], [0,0.04], [0.02,1]; Small, Med, Large respectively representing genetic algebraically little, in, big, its scope is respectively [0,100], [50,150], [100,200];

(4.2) according to convergence rate shown in Figure 4 and genetic algebra fuzzy set interval, and crossing-over rate shown in Figure 5 and aberration rate fuzzy set interval, provide following 9 fuzzy inference rule collection:

Rule 1: if convergence rate R in interval [0,0.02], genetic algebra g releases crossing-over rate P in interval [0,100] _cAt interval [0.8,1], aberration rate P _mIn interval [0.011,1];

Rule 2: if convergence rate R in interval [0,0.02], genetic algebra g releases crossing-over rate P in interval [50,150] _cAt interval [0.7,0.9], aberration rate P _mIn interval [0.008,0.012];

Rule 3: if convergence rate R in interval [0,0.02], genetic algebra g releases crossing-over rate P in interval [100,200] _cAt interval [0.6,0.8], aberration rate P _mIn interval [0.005,0.009];

Rule 4: if convergence rate R in interval [0,0.04], genetic algebra g releases crossing-over rate P in interval [0,100] _cAt interval [0.8,1], aberration rate P _mIn interval [0.008,0.012];

Rule 5: if between the convergence rate Zone R [0,0.04], genetic algebra g releases crossing-over rate P in interval [50,150] _cAt interval [0,0.6], aberration rate P _mIn interval [0.005,0.009];

Rule 6: if between the convergence rate Zone R [0,0.04], genetic algebra g releases crossing-over rate P in interval [100,200] _cAt interval [0.5,0.7], aberration rate P _mIn interval [0.002,0.006];

Rule 7: if convergence rate R in interval [0.02,1], genetic algebra g releases crossing-over rate p in interval [0,100] _cAt interval [0.6,0.8], aberration rate P _mIn interval [0.005,0.009];

Rule 8: if convergence rate R in interval [0.02,1], genetic algebra is in interval g[50,150], release crossing-over rate p _cAt interval [0.5,0.7], aberration rate P _mIn interval [0.002,0.006];

Rule 9: if convergence rate R in interval [0.02,1], genetic algebra g releases crossing-over rate P in interval [100,200] _cAt interval [0,0.6], aberration rate P _mIn interval [0,0.003].

According to above obfuscation rule set, and, obtain crossing-over rate and aberration rate obfuscation degree of membership by the obfuscation degree of membership of convergence rate and genetic algebra

With

Wherein Smaller, Small, Med, Large, Larger represent respectively crossing-over rate and aberration rate degree be minimum, little, in, big, greatly, its scope is to be respectively [0,0.6], [0.5,0.7], [0.6,0.8], [0.7,0.9], [0.8,1] and [0,0.003], [0.002,0.006], [0.005,0.009], [0.008,0.012], [0.011,1], solution formula is as follows:

$\left\{\begin{array}{c}{D}_{P_c}^{\mathrm{Smaller}}=\min \{D_R^{\mathrm{High}},D_g^{L\arg e}\}\\ D_{P_c}^{\mathrm{Small}}=\min \{D_R^{\mathrm{High}},D_g^{\mathrm{Med}}\}+\min \{D_R^{\mathrm{Med}},D_g^{L\arg e}\}\\ D_{P_c}^{\mathrm{Med}}=\min \{D_R^{\mathrm{High}},D_g^{\mathrm{Small}}\}+\min \{D_R^{\mathrm{Med}},D_g^{\mathrm{Med}}\}+\min \{D_R^{\mathrm{Low}},D_g^{L\arg e}\}\\ D_{P_c}^{L\arg e}=\min \{D_R^{\mathrm{Med}},D_g^{\mathrm{Small}}\}+\min \{D_R^{\mathrm{Low}},D_g^{\mathrm{Med}}\}\\ D_{P_c}^{L\arg \mathrm{er}}=\min \{D_R^{\mathrm{Low}},D_g^{\mathrm{Small}}\}\end{array}\right.---\left(3\right)$

$\left\{\begin{array}{c}{D}_{P_m}^{\mathrm{Smaller}}=\min \{D_R^{\mathrm{High}},D_g^{L\arg e}\}\\ D_{P_m}^{\mathrm{Small}}=\min \{D_R^{\mathrm{High}},D_g^{\mathrm{Med}}\}+\min \{D_R^{\mathrm{Med}},D_g^{L\arg e}\}\\ D_{P_m}^{\mathrm{Med}}=\min \{D_R^{\mathrm{High}},D_g^{\mathrm{Small}}\}+\min \{D_R^{\mathrm{Med}},D_g^{\mathrm{Med}}\}+\min \{D_R^{\mathrm{Low}},D_g^{L\arg e}\}\\ D_{P_m}^{L\arg e}=\min \{D_R^{\mathrm{Med}},D_g^{\mathrm{Small}}\}+\min \{D_R^{\mathrm{Low}},D_g^{\mathrm{Med}}\}\\ D_{P_m}^{L\arg \mathrm{er}}=\min \{D_R^{\mathrm{Low}},D_g^{\mathrm{Small}}\}\end{array}\right.---\left(4\right)$

(4.3) central point of crossing-over rate and aberration rate fuzzy set is as shown in Figure 5 used x respectively _iAnd y _i(i=1,2 ..., 5) expression, obtain the crossing-over rate P that the reverse gelatinization is exported according to following formula _cWith aberration rate P _m, be respectively:

$P_c=\frac{D_{P_c}^{\mathrm{Smaller}}{x}_1+D_{P_c}^{\mathrm{Small}}{x}_2+D_{P_c}^{\mathrm{Med}}{x}_3+D_{P_c}^{L\arg e}{x}_4+D_{P_c}^{L\arg \mathrm{er}}{x}_5}{{\mathrm{\&Sigma;}}_{i=1}^5{x}_i}---\left(5\right)$

$P_m=\frac{D_{P_m}^{\mathrm{<figure-callout\; id="1"\; label="Smaller\; y"\; filenames="HSA00000186147100011.png,HSA00000186147100021.png"\; state="\{\{state\}\}">Smaller</figure-callout>}}{y}_{}{}_1+D_{P_m}^{\mathrm{<figure-callout\; id="2"\; label="Small\; y"\; filenames="HSA00000186147100041.png,HSA00000186147100042.png"\; state="\{\{state\}\}">Small</figure-callout>}}{y}_{}{}_2+D_{P_m}^{\mathrm{Med}}{y}_3+D_{P_m}^{L\arg e}{y}_4+D_{P_m}^{L\arg \mathrm{er}}{y}_5}{{\mathrm{\&Sigma;}}_{i=1}^5{y}_i}.---\left(6\right)$

Step 5 is carried out genetic algorithm and is intersected and mutation operation, selects excellent chromosome, and judges whether to reach maximum genetic algebra.

(5.1) according to adjusted crossing-over rate P _cWith aberration rate P _mAgain chromosome is intersected and mutation operation, thereby obtain S (P _c+ 1) bar chromosome is chosen S bar chromosome as population of future generation according to fitness function order from big to small;

(5.2) obtain crossing-over rate and aberration rate after, judge whether genetic algebra g reaches maximum evolutionary generation g _Max, if reach, the chromosome matrix inversion of fitness maximum shine upon back allocation matrix A, distribute corresponding frequency spectrum to each user; Otherwise genetic algebra g '=g+1 continues step 3～5.

With reference to Fig. 2, the dynamic frequency spectrum deployment method step based on fuzzy logic genetic algorithm of the present invention is as follows:

Step 1 during the cognition network initial construction, utilizes the static network frequency spectrum distributing method to obtain the spectrum allocation may matrix A, for each user distributes frequency spectrum.

Step 2, every a time interval Δ, the mobile subscriber who finds out in the network secondary user's forms mobile subscriber's collection { U ₁, U ₂..., U _{N '}N ' ∈ 1,2 ..., and N}, N ' expression mobile subscriber number wherein is to mobile subscriber U _iFind out the secondary user's of disturbing with its generation, form discussion group τ _i

Step 3 is obtained each discussion group τ respectively _iThe effective spectrum matrix

The benefit matrix

Interference matrix

With the spectrum allocation may matrix

M wherein _iExpression discussion group τ _iMiddle number of users, N _iExpression discussion group τ _iThe middle spendable number of channel of user; u _{I, m, n}=1 expression discussion group τ _iMiddle user m can use channel n, u _{I, m, n}=0 expression discussion group τ _iMiddle user m can not use channel n; e _{I, m, n}Expression discussion group τ _iObtainable income when middle user m uses channel n; c _{I, m, k}=1 expression discussion group τ _iMiddle user m uses identical channel can produce interference, c with user k _{I, m, k}=0 expression can not produce interference; a _{I, m, n}=1 expression discussion group τ _iIn channel n is distributed to user m, a _{I, m, n}=0 expression discussion group τ _iMiddle channel n is unallocated to give user m; Initialization i=1.

Step 4 is to discussion group τ _iFrequency spectrum redistribute.

(4.1) for discussion group τ _i, with the spectrum allocation may matrix A _iIn be that 1 element is mapped as the chromosome code element one by one, obtain chromosome length

Produce S bar initial chromosome at random, suppose a _{I, m, n}Be mapped as chromosome k position, this position is that 1 expression nth user transfers mobile subscriber U to m bar channel _iUse, this position is that 0 expression is not transferred the possession of;

(4.2), draw discussion group τ according to the static network frequency spectrum distributing method _iFinal spectrum allocation may matrix A _i, upgrading the corresponding spectrum allocation may in the spectrum allocation may matrix A, other users' spectrum allocation may remains unchanged in the network;

(4.3) make i '=i+1, repeating step 4 is up to i '=N ', and the allocation matrix A of final updated is the current frequency spectrum optimum allocation, and distributes corresponding frequency spectrum to the user.

Effect of the present invention can further prove by following simulation example:

One, simulated conditions

At a 10 * 10km ²In the cognition network sub-district, 5 the main user transmitters that distributing take 5 separate channels respectively, every channel width is 6MHz, and it disturbs radius is 5km, and K secondary user's is randomly dispersed in the network, the scope of K is 10～40, and it disturbs radius is 1km, main user transmitter transmitting power P _t=1W, the power spectral density N of interference noise ₀=10 ^-6W/Hz.

Two, emulation content

Emulation 1 a: example that is static network frequency spectrum distributing method of the present invention, its content is static network distribution method and existing greedy algorithm, maximum and the bandwidth algorithm (Max-Sum-Bandwidth that proposes with the present invention, MSB), the method for exhaustion is compared, wherein the secondary user's number increases to 40 from 10, and simulation result as shown in Figure 6.

Emulation 2 a: example that is dynamic network frequency spectrum distributing method of the present invention, its content is to compare with existing local one to one machinery of consultation, static network frequency spectrum distributing method of the present invention with dynamic network frequency spectrum distributing method of the present invention, wherein the secondary user's number increases to 40 from 10, the ratio that number of mobile users accounts for total number of users is 50%, the mobile subscriber moves with the speed of 50m/s, and simulation result as shown in Figure 7.

Emulation 3: the time loss that is static network of the present invention and dynamic network frequency spectrum distributing method, its content is that the normalization time consumption to dynamic network frequency spectrum distributing method and existing local one to one machinery of consultation, static network frequency spectrum distributing method compares, wherein the secondary user's number is 20, the mobile subscriber moves with the speed of 50m/s, the ratio that number of mobile users accounts for total number of users changes to 80% from 10%, and simulation result as shown in Figure 8.

Three, simulation result

As seen from Figure 6, the spectrum allocation may performance of static network frequency spectrum distributing method under different secondary user's numbers that the present invention proposes is all extremely near frequency spectrum optimum allocation performance, compare with the bandwidth algorithm with maximum with greedy algorithm, average frequency spectrum distributes performance to improve 4.3% and 3.0% respectively.

As seen from Figure 7, the local optimum characteristic of dynamic network frequency spectrum distributing method makes its spectrum allocation may benefit be lower than the static network frequency spectrum distributing method.But static network spectrum allocation may algorithm of the present invention has on average improved 3.7% than existing local one to one machinery of consultation spectrum allocation may benefit.

As seen from Figure 8, the dynamic network frequency spectrum distributing method is short slightly and fluctuate little than local machinery of consultation spended time one to one, the dynamic network frequency spectrum distributing method only needs to reconfigure the frequency spectrum resource of mobile subscriber and interference user thereof, so the cost of the time of algorithm is much smaller than the static network frequency spectrum distributing method.

The above-mentioned simulation result of analysis-by-synthesis, the present invention introduces fuzzy logic genetic algorithm and carries out spectrum allocation may, discusses and analyzes at static network and two kinds of situations of dynamic network, has proved the superiority of frequency spectrum distributing method that the present invention puies forward.

Specific implementation of the present invention can be applied in a flexible way according to the variation of network environment.Before changing, topological structure utilize the static network frequency spectrum distributing method to reach the optimized distribution of frequency spectrum; When secondary user's is moved, main user inserts with certain probability or discharge channel, utilizes the dynamic network frequency spectrum distributing method that the spectrum allocation may result is finely tuned, thereby realize the combination of static network and dynamic network frequency spectrum distributing method.

Claims (4)

1. the static network frequency spectrum distributing method based on fuzzy logic genetic algorithm comprises the steps:

(1) determines effective spectral matrix Γ={ u according to the cognition network structure _{M, n}| u _{M, n}∈ 0,1}} _{M * N}, benefit matrix E={e _{M, n}} _{M * N}With interference matrix C={c _{M, k}| c _{M, k}∈ 0,1}} _{M * N}, wherein M represents user's number, N represents channel bar number; u _{M, n}=1 expression user m can utilize channel n, u _{M, n}=0 expression user m can not utilize channel n; e _{M, n}Obtainable benefit when expression user m uses channel n; c _{M, k}=1 expression user m uses identical channel can produce interference, c with user k _{M, k}=0 expression can not produce interference;

(2) with spectrum allocation may matrix A={ a _{M, n}| a _{M, n}∈ 0,1}} _{M * N}Middle element is initialized as 0, wherein a _{M, n}=1 expression is distributed to user m, a with channel n _{M, n}The unallocated user m that gives of=0 expression channel n; And element in the spectrum allocation may matrix A is mapped as the chromosome code element one by one, chromosome length is len=MN;

(3) initialization genetic algorithm parameter comprises population size S, aberration rate P _mWith crossing-over rate P _cInitialization genetic algebra g=1 produces the population size at random and is the initial population of S, promptly produces S bar chromosome, and the chromosome matrix notation is

Its by

K the code element of expression chromosome i;

(4) according to the chromosome matrix L ⁽ⁱ⁾, benefit matrix E and effective spectrum matrix Γ calculate every chromosomal fitness function

Calculate genetic algorithm converges speed according to fitness function

(5) according to above-mentioned convergence rate R (g) and genetic algebra g, with convergence rate and genetic algebra obfuscation, the obfuscation of convergence rate and genetic algebra output is respectively

With

Wherein Low, and Med, High} are the rate of convergence fuzzy set, Low, Med, High represent the basic, normal, high degree of convergence rate respectively; Small, Med, Large} are the fuzzy set of genetic algebra, Small, Med, Large represent respectively genetic algebra little, in, big degree;

According to fuzzy logic ordination collection launch mode gelatinization crossing-over rate and aberration rate degree of membership With

Wherein Smaller, Small, Med, Large, Larger represent respectively crossing-over rate and aberration rate degree be minimum, little, in, big, greatly, utilize following formula reverse gelatinization output P _cAnd P _m:

$P_c=\frac{D_{P_c}^{\mathrm{Smaller}}{x}_1+D_{P_c}^{\mathrm{Small}}{x}_2+D_{P_c}^{\mathrm{Med}}{x}_3+D_{P_c}^{L\arg e}{x}_4+D_{P_c}^{L\arg \mathrm{er}}{x}_5}{{\mathrm{\&Sigma;}}_{i=1}^5{x}_i}$

$P_m=\frac{D_{P_m}^{\mathrm{Smaller}}{y}_1+D_{P_m}^{\mathrm{Small}}{y}_2+D_{P_m}^{\mathrm{Med}}{y}_3+D_{P_m}^{L\arg e}{y}_4+D_{P_m}^{L\arg \mathrm{er}}{y}_5}{{\mathrm{\&Sigma;}}_{i=1}^5{y}_i}$

X wherein _iAnd y _iBe respectively crossing-over rate and aberration rate fuzzy set central point;

(6) according to adjusted crossing-over rate p _cWith aberration rate P _mAgain chromosome is intersected and mutation operation, thereby obtain S (P _c+ 1) bar chromosome is chosen S bar chromosome as population of future generation according to fitness function order from big to small;

(7) judge whether genetic algebra g reaches maximum evolutionary generation g _Max, if reach, the chromosome matrix inversion of fitness maximum shine upon back allocation matrix A, distribute corresponding frequency spectrum to each user; Otherwise genetic algebra g '=g+1 continues step (4)～(6).

2. according to claims 1 described static network frequency spectrum distributing method, it is characterized in that the described fuzzy logic ordination of step (5), comprise following 9 based on fuzzy logic genetic algorithm:

Rule 1: if convergence rate R in interval [0,0.02], genetic algebra g releases crossing-over rate p in interval [0,100] _cAt interval [0.8,1], aberration rate P _mIn interval [0.011,1];

Rule 2: if convergence rate R in interval [0,0.02], genetic algebra g releases crossing-over rate P in interval [50,150] _cAt interval [0.7,0.9], aberration rate P _mIn interval [0.008,0.012];

Rule 3: if convergence rate R in interval [0,0.02], genetic algebra g releases crossing-over rate P in interval [100,200] _cAt interval [0.6,0.8], aberration rate P _mIn interval [0.005,0.009];

Rule 4: if convergence rate R in interval [0,0.04], genetic algebra g releases crossing-over rate P in interval [0,100] _cAt interval [0.8,1], aberration rate P _mIn interval [0.008,0.012];

Rule 5: if between the convergence rate Zone R [0,0.04], genetic algebra g releases crossing-over rate P in interval [50,150] _cAt interval [0,0.6], aberration rate P _mIn interval [0.005,0.009];

Rule 6: if between the convergence rate Zone R [0,0.04], genetic algebra g releases crossing-over rate P in interval [100,200] _cAt interval [0.5,0.7], aberration rate P _mIn interval [0.002,0.006];

Rule 7: if convergence rate R in interval [0.02,1], genetic algebra g releases crossing-over rate p in interval [0,100] _cAt interval [0.6,0.8], aberration rate P _mIn interval [0.005,0.009];

Rule 8: if convergence rate R in interval [0.02,1], genetic algebra is in interval g[50,150], release crossing-over rate p _cAt interval [0.5,0.7], aberration rate P _mIn interval [0.002,0.006];

Rule 9: if convergence rate R in interval [0.02,1], genetic algebra g releases crossing-over rate P in interval [100,200] _cAt interval [0,0.6], aberration rate P _mIn interval [0,0.003].

3. the dynamic network frequency spectrum distributing method based on fuzzy logic genetic algorithm comprises the steps:

1) during the cognition network initial construction, utilize the static network frequency spectrum distributing method to obtain the spectrum allocation may matrix A, for each user distributes frequency spectrum;

2), mark the user's collection { U that moves in the network every a time interval Δ ₁, U ₂..., U _{N '}, N ' expression mobile subscriber number wherein is to mobile subscriber U _iFind out with user that it disturbs mutually and form discussion group τ _i

3) to each discussion group τ _iObtain the effective spectrum matrix respectively

The benefit matrix Interference matrix With the spectrum allocation may matrix

M wherein _iExpression discussion group τ _iMiddle number of users, N _iExpression discussion group τ _iThe middle spendable number of channel of user; u _{I, m, n}=1 expression discussion group τ _iMiddle user m can use channel n, u _{I, m, n}=0 expression discussion group τ _iMiddle user m can not use channel n; e _{I, m, n}Expression discussion group τ _iObtainable income when middle user m uses channel n; c _{I, m, k}=1 expression discussion group τ _iMiddle user m uses identical channel can produce interference, c with user k _{I, m, k}=0 expression can not produce interference; a _{I, m, n}=1 expression discussion group τ _iIn channel n is distributed to user m, a _{I, m, n}=0 expression discussion group τ _iMiddle channel n is unallocated to give user m; Initialization i=1;

4) for discussion group τ _i, with the spectrum allocation may matrix A _iIn be that 1 element is mapped as the chromosome code element one by one, obtain chromosome length

Produce S bar initial chromosome at random; According to the static network frequency spectrum distributing method, draw discussion group τ _iInner optimum spectrum allocation may A _i, upgrade corresponding frequency spectrum distribution in the overall spectrum allocation may matrix A;

5) make i '=i+1, repeating step 4) up to i '=N ', distribute corresponding frequency spectrum to the user according to the allocation matrix A of final updated.

4. according to claims 3 described dynamic network frequency spectrum distributing methods, it is characterized in that the 4th based on fuzzy logic genetic algorithm) described chromosome code element of step, comprise two kinds of situations: code element is that the channel of 1 this use of expression transfers mobile subscriber U _i, code element is that the channel of 0 this use of expression is not transferred the possession of.

Images