CN103401625B - Particle swarm optimization algorithm based cooperative spectrum sensing optimization method - Google Patents

Abstract

The invention discloses a particle swarm optimization algorithm based cooperative spectrum sensing optimization method. The method comprises the following implementation steps: firstly building a cooperative spectrum sensing optimization model and enabling to-be-optimized variables to correspond to relative parameters of a particle swarm optimization algorithm; initializing the iterative times as well as the position vector and the velocity vector of randomly generated particles; carrying out fitness assessment on the initialized particles and finding out the particle with minimum fitness value to serve as the global optimum particle; along with the increase of the iterative times, continuously updating the position vector and the velocity vector of the particles and updating the global optimum value; and when the iterative times reaches the maximum iterative times, outputting the global optimum particle and solving the corresponding detection probability. The method disclosed by the invention has the advantages of simplicity and convenience, little computation amount, easiness in realization, few adjustment parameter, strong search capability and the like when being used for the cooperative spectrum sensing optimization.

Description

Based on the collaborative spectrum sensing optimization method of particle swarm optimization algorithm

Technical field

The present invention relates to cognitive radio technology field, be specifically related to a kind of collaborative spectrum sensing optimization method based on particle swarm optimization algorithm, be applicable to the fields such as system, multiple-objection optimization, pattern recognition, scheduling, signal transacting, decision-making, robot application.

Background technology

As a kind of new intelligent wireless communication technology, cognitive radio (Cognitive Radio, CR) technology is by carrying out " secondary utilization " mandate frequency spectrum, and the contradiction lacked between growing wireless access demand for alleviating frequency spectrum resource provides feasible thinking.It can perception electromagnetic environment analyze wireless environment, understand and judge, the messaging parameter of adjustment System is with the change conformed adaptively, under the prerequisite not affecting authorized user communication, realizes the high-efficiency dynamic spectrum utilization of intelligence.

In cognition network, one of maximum focus of the current beyond doubt cognitive radio research of frequency spectrum perception, first, fast and reliable ground finds that frequency spectrum cavity-pocket is the prerequisite of carrying out frequency spectrum resource recycling; Secondly, unauthorized user needs higher receiver sensitivity with the spectrum activity of accurate measurements authorized user, thus avoids causing interference to authorized user, and this just has higher requirement to frequency spectrum perception.But by wireless channel interference, decline and the impact of time-varying characteristics, general single cognitive user is difficult to obtain reliable transient state perception information, can effectively tell is faint primary user's signal or primary user this two states idle on earth.And the testing result of single cognitive user of comparing, multiple cognitive user carries out the reliability that collaborative spectrum sensing effectively can improve detection, and therefore collaborative spectrum sensing technology has important theory value and realistic meaning.

Collaborative spectrum sensing is divided into perception and two stages of report usually, and in perception stage, each user completes local independently and detects; In the report stage, the local testing result of all users is sent to fusion center (Fusion Center, FC), FC carries out data fusion to the local statistic information received, and comprehensively makes the conclusive judgement whether primary user's signal exists.

Generally speaking, for the optimization of collaborative spectrum sensing, can by following two kinds of methods:

1, optimum linear combining method (Optimal Linear Cooperation, OPT.LIN), OPT.LIN method is divided into subproblem to former optimization problem, draws its infimum in theory, then solved by convex optimization method, thus the computing time of meeting at substantial.

2, offset coefficient method (Modified Deflection Coefficient, OPT.MDC), although OPT.MDC method is simple, its optimality is based upon the local SNR of cognitive user much larger than on the basis of 1, makes its range of application be subject to a definite limitation.

Summary of the invention

The object of the invention is to the deficiency for prior art, a kind of collaborative spectrum sensing optimization method based on particle swarm optimization algorithm is provided, while obtaining more excellent detection perform, reduce the algorithm time spent when frequency spectrum perception.

It should be noted that, the technical thought realizing the object of the invention is: first set up collaborative spectrum sensing Optimized model, and on this basis, the relevant parameter of initialization particle swarm optimization algorithm, by corresponding for the relevant parameter of variable to be optimized and particle swarm optimization algorithm; Initialization iterations, the position vector of the random particle produced and velocity vector; Fitness analysis is carried out to initialized each particle, and finds out the sequence number of the minimum particle of fitness value, as global optimum; Along with the increase of iterations, constantly update position vector and the velocity vector of particle, and upgrade weights optimal value; When iterations reaches maximum iteration time, export global optimum, obtain the detection probability corresponding to it.

To achieve these goals, technical scheme of the present invention is further described below:

Based on the collaborative spectrum sensing optimization method of particle swarm optimization algorithm, said method comprising the steps of:

(1) collaborative spectrum sensing Optimized model is set up;

(2) relevant parameter of initialization particle swarm optimization algorithm, comprising:

(2a) random generation P particle, note particle i is x _i=[x _i1, x _i2..., x _iD], i=1,2 ..., P, D=M, wherein D represents individual dimension, and wherein particle is defined as variable to be optimized, and namely control centre gives the weight w that each user's statistical information is distributed, and is also x _i=w;

(2b) iterations t=0 is put, the position vector of stochastic generation particle í $x_i^t={[x_{i1}^t,x_{i2}^t,\·\·\·,x_{\mathrm{iD}}^t]}^T$ And velocity vector $v_i^t={[v_{i1}^t,v_{i2}^t,\·\·\·,v_{\mathrm{iD}}^t]}^T$ , wherein $x_{\mathrm{id}}^t\∈\left[\mathrm{0,1}\right],$ and 1≤d≤M, 1≤i≤P; After each iteration, the location formula of each particle is:

$\stackrel{\‾}{x_{\mathrm{id}}^t}=\frac{\left|x_{\mathrm{id}}^t\right|}{\underset{d=1}{\overset{D}{\mathrm{\Σ}}}\left|x_{\mathrm{id}}^t\right|};$

(3) fitness value of described each particle is calculated wherein fitness function is defined as the testing result that the weights representated by each particle obtain.According to fitness value, determine local optimum particle $p_i^t=(p_{i1}^t,p_{i2}^t,\·\·\·,p_{\mathrm{iD}}^t)$ With global optimum's particle $p_b^t=(x_{b1}^t,x_{b2}^t,\·\·\·,x_{\mathrm{bD}}^t),$ Wherein b is the ordinal number that the minimum particle of fitness value is corresponding, for t=0,

(4) t=t+1 is made, and by formula $v_{\mathrm{id}}^t=\mathrm{\&omega;}{v}_{\mathrm{id}}^{t-1}+c_1\mathrm{\&xi;}(p_{\mathrm{id}}^{t-1}-x_{\mathrm{id}}^{t-1})+c_2\mathrm{\&eta;}(p_{\mathrm{bd}}^{t-1}-x_{\mathrm{id}}^{t-1})$ Renewal speed wherein c ₁and c ₂be normal number, ξ and η is distributed in the random number in [0,1], and ω is inertia weight, with the d of corresponding local optimum particle and global optimum's particle ties up element respectively, wherein, if $v_{\mathrm{id}}^t>V_{\max },$ Order $v_{\mathrm{id}}^t=V_{\max };$ If $v_{\mathrm{id}}^t<-V_{\max },$ Order $v_{\mathrm{id}}^t=-V_{\max };$

(5) by formula upgrade position be normalized according to step (2b);

(6) fitness value through each particle upgraded is calculated for particle if its current fitness value is less than the local optimum particle in last iteration fitness value, its local optimum particle is set to otherwise its local optimum particle remains unchanged if particle current fitness value is less than global optimum's particle in last iteration fitness value, by this particle be set to global optimum's particle if do not have the fitness value of particle to be less than fitness value, Ze Ling global optimum particle remains unchanged

(7) judge whether iterations t reaches maximum iteration time T _maxif reach, export global optimum, and according to formula obtain the detection probability P corresponding to it _d; Otherwise, repeat step (4) ~ (7).

It should be noted that, described collaborative spectrum sensing Optimized model is:

minf(w)

$s.t.{\left|\right|w\left|\right|}_2^2=1$

Wherein w is weight, $f\left(w\right)=\frac{Q^{-1}\left(p_f\right)\sqrt{w^T\mathrm{Aw}}-E_s{g}^{T}w}{\sqrt{w^T\mathrm{Bw}}},$ P _ffalse alarm probability, A=2Ndiag ²(σ)+diag (δ), B=2Ndiag ²(σ)+diag (δ)+4E _sdiag (g) diag (σ), wherein N is sampling number, σ is link noise, δ is control channel noise, g represents channel gain, f (w) utilizes the non-increasing abbreviation of Q function to obtain by detection probability, and maximize detection probability corresponding to minimizing f (w), wherein detection probability is

It should be noted that, the computing formula of described step (3) is:

According to formula fitness (x _i)=f (x _i), $f\left(w\right)=\frac{Q^{-1}\left(p_f\right)\sqrt{w^T\mathrm{Aw}}-E_s{g}^{T}w}{\sqrt{w^T\mathrm{Bw}}},$ Wherein x _i=w.

The present invention compared with prior art has the following advantages:

1, the present invention adopts Deterministic rules, but carries out random search according to the memory of history and society, and compared to prior art, the possibility that the present invention searches optimal solution increases greatly;

2, information flow of the present invention is unidirectional, can converge on optimal value faster;

3, the speed of service of the present invention is fast, effectively can improve detection perform;

4, convergence rate of the present invention is very fast, and algorithm is simple, easy programming realization.

Accompanying drawing explanation

Fig. 1 is the theory diagram that the present invention realizes;

Fig. 2 is the overview flow chart that the present invention realizes;

Fig. 3 is the comparison diagram of the inventive method and existing method detection perform;

Fig. 4 is the comparison diagram of the inventive method and existing method average time spent when collaborative spectrum sensing is optimized.

Embodiment

Below in conjunction with accompanying drawing, the invention will be further described.

It should be noted that, as shown in Figure 1, be the theory diagram of collaborative spectrum sensing, each user completes frequency spectrum perception independently, and then local testing result is sent to fusion center, fusion center is adjudicated final result.

As shown in Figure 2, concrete steps of the present invention are as follows:

Step 1, sets up collaborative spectrum sensing Optimized model;

Collaborative spectrum sensing Optimized model is:

minf(w)

$s.t.{\left|\right|w\left|\right|}_2^2=1;$

Wherein w is weight, $f\left(w\right)=\frac{Q^{-1}\left(p_f\right)\sqrt{w^T\mathrm{Aw}}-E_s{g}^{T}w}{\sqrt{w^T\mathrm{Bw}}},$ P _ffalse alarm probability, A=2Ndiag ²(σ)+diag (δ), B=2Ndiag ²(σ)+diag (δ)+4E _sdiag (g) diag (σ), wherein N is sampling number, σ is link noise, δ is control channel noise, g represents channel gain, f (w) utilizes the non-increasing abbreviation of Q function to obtain by detection probability, and maximize detection probability corresponding to minimizing f (w), wherein detection probability is

Step 2, the relevant parameter of initialization particle swarm optimization algorithm;

(2a) random generation P particle, note particle i is x _i=[x _i1, x _i2..., x _iD], i=1,2 ..., P, D=M, wherein D represents individual dimension;

It should be noted that, for collaborative spectrum sensing problem of the present invention, particle is defined as variable to be optimized, and namely control centre gives the weight w that each user's statistical information is distributed, and is also x _i=w;

(2b) iterations t=0 is put, the position vector of stochastic generation particle í $x_i^t={[x_{i1}^t,x_{i2}^t,\&CenterDot;\&CenterDot;\&CenterDot;,x_{\mathrm{iD}}^t]}^T$ And velocity vector $v_i^t={[v_{i1}^t,v_{i2}^t,\&CenterDot;\&CenterDot;\&CenterDot;,v_{\mathrm{iD}}^t]}^T,$ Wherein $x_{\mathrm{id}}^t\&Element;\left[\mathrm{0,1}\right],$ and 1≤d≤M, 1≤i≤P; The initial population of stochastic generation, the position of particle need meet constraints.After each iteration, formula is passed through in the position of each particle complete normalization operation.

Step 3, assessment; According to formula fitness (x _i)=f (x _i), wherein x _i=w, its calculation procedure is as follows:

The fitness value of each particle in step (2a) and establish b to be the ordinal number of the minimum particle of fitness value, then global optimum's particle is for local optimum particle as t=0, wherein fitness function is defined as the testing result that the weights representated by each particle obtain, i.e. fitness (x _i)=f (x _i), wherein x _i=w;

Need to further illustrate, individual fitness value is less, represents the detection probability P that the weights represented by this individuality obtain _dhigher, thus the quality of this individuality is better;

Step 4, renewal speed; Wherein, make t=t+1, and by formula $v_{\mathrm{id}}^t=\mathrm{\&omega;}{v}_{\mathrm{id}}^{t-1}+c_1\mathrm{\&xi;}(p_{\mathrm{id}}^{t-1}-x_{\mathrm{id}}^{t-1})+c_2\mathrm{\&eta;}(p_{\mathrm{bd}}^{t-1}-x_{\mathrm{id}}^{t-1})$ Renewal speed $v_{\mathrm{id}}^t,$ Wherein c ₁and c ₂be normal number, ξ and η is distributed in the random number in [0,1], and ω is inertia weight; with the d of corresponding local optimum particle and global optimum's particle ties up element respectively, if order $v_{\mathrm{id}}^t=V_{\max };$ If $v_{\mathrm{id}}^t<-V_{\max },$ Order $v_{\mathrm{id}}^t=-V_{\max }.$

Step 5, upgrades position; By formula upgrade position be normalized according to step (2b);

Step 6, upgrades optimal particle; Calculate the fitness value through each particle upgraded for particle if its current fitness value is less than the local optimum particle in last iteration fitness value, its local optimum particle is set to otherwise its local optimum particle remains unchanged if particle current fitness value is less than global optimum's particle in last iteration fitness value, by this particle be set to global optimum's particle if do not have the fitness value of particle to be less than fitness value, Ze Ling global optimum particle remains unchanged

Step 7, judges whether iterations t reaches maximum iteration time T _maxif reach, export global optimum, and according to formula obtain the detection probability P corresponding to it _d; Otherwise, repeat step 4 ~ 7.

In order to describe effect of the present invention better, prove by simulation example below further.

1, simulation parameter is arranged

Design parameter arranges as shown in table 1:

Simulation parameter	Number of particles	Iterations	Acceleration constant	Inertia weight
					Value	P=20	T max=50	c 1=c 2=2	w=0.4

2, emulation mode

The linear combining method of existing optimum, offset coefficient method and the inventive method.

3, content is emulated

Emulation 1

As shown in Figure 3, the linear combining method (OPT.LIN) of the inventive method and traditional optimum and the detection perform of offset coefficient method (OPT.MDC) compare.

In figure 3, curve 31 is that the inventive method emulates the detection perform curve obtained, curve 32 is that optimum linear combining method (OPT.LIN) emulates the detection perform curve obtained, and curve 33 is that offset coefficient method (OPT.MDC) emulates the detection perform curve obtained.

Correlation curve 31,32 and 33 can be found out, method of the present invention is close with optimum linear combining method (OPT.LIN) detection perform, and is better than the detection perform of offset coefficient method (OPT.MDC).But compare the inventive method, optimum linear combining method (OPT.LIN) is divided into subproblem to former optimization problem, draws its infimum in theory, is then solved by convex optimization method, the computing time of meeting at substantial.Offset coefficient method (OPT.MDC) is although simple, and its optimality is based upon the local SNR of cognitive user much larger than on the basis of 1.

Emulation 2

As shown in Figure 4, the linear combining method (OPT.LIN) of the inventive method and traditional optimum and offset coefficient method (OPT.MDC) are carried out emulation at average time spent when collaborative spectrum sensing is optimized to compare.

In the diagram, curve 41 is the inventive method average times spent when collaborative spectrum sensing is optimized, curve 42 is optimum linear combining method (OPT.LIN) average times spent when collaborative spectrum sensing is optimized, and curve 43 is offset coefficient method (OPT.MDC) average times spent when collaborative spectrum sensing is optimized.

Correlation curve 41,42 and 43 can be found out, the optimum time spent by linear combining method (OPT.LIN) is maximum, time spent by the inventive method takes second place, the time of offset coefficient method (OPT.MDC) spent by three kinds of methods is minimum, and computational efficiency and real-time so just undoubtedly for ensureing frequency spectrum perception provide effective optimum ideals.

The above-mentioned simulation result of comprehensive analysis, particle swarm optimization algorithm have when collaborative spectrum sensing is optimized simple and convenient, amount of calculation is little, be easy to realize, adjustment parameter few, search the advantages such as plain ability is strong.

For a person skilled in the art, according to technical scheme described above and design, other various corresponding change and distortion can be made, and all these change and distortion all should belong within the protection range of the claims in the present invention.

Claims (2)

1., based on the collaborative spectrum sensing optimization method of particle swarm optimization algorithm, it is characterized in that, said method comprising the steps of:

(1) collaborative spectrum sensing Optimized model is set up:

minf(w)

$s.t.{\left|\right|w\left|\right|}_2^2=1$

Wherein w is weight, $f\left(w\right)=\frac{Q^{-1}\left(P_f\right)\sqrt{w^T\mathrm{Aw}}-E_s{g}^{T}w}{\sqrt{w^T\mathrm{Bw}}},$ P _ffalse alarm probability, A=2Ndiag ²(σ)+diag (δ), B=2Ndiag ²(σ)+diag (δ)+4E _sdiag (g) diag (σ), wherein N is sampling number, and σ is link noise, and δ is control channel noise, and g represents channel gain, E _sfor primary user's transmitting power; F (w) utilizes the non-increasing abbreviation of Q function to obtain by detection probability, maximizes detection probability corresponding to minimizing f (w);

(2) relevant parameter of initialization particle swarm optimization algorithm, comprising:

(2a) random generation P particle, note particle i is x _i=[x _i1, x _i2..., x _iD], i=1,2 ..., P, D=M, wherein D represents individual dimension, and wherein particle is defined as variable to be optimized, x _i=w, w are the weights that control centre distributes to each user's statistical information;

(2b) iterations t=0 is put, the position vector of stochastic generation particle í $x_i^t={[x_{i1}^t,x_{i2}^t,\&CenterDot;\&CenterDot;\&CenterDot;,x_{\mathrm{iD}}^t]}^T$ And velocity vector $v_i^t={[v_{i1}^t,v_{i2}^t,\&CenterDot;\&CenterDot;\&CenterDot;,v_{\mathrm{iD}}^t]}^T,$ Wherein $x_{\mathrm{id}}^t\&Element;\left[\mathrm{0,1}\right],$ and 1#d M, 1#i P; After each iteration, the location formula of each particle is:

$\stackrel{\&OverBar;}{x_{\mathrm{id}}^t}=\frac{\left|x_{\mathrm{id}}^t\right|}{\underset{d=1}{\overset{D}{\mathrm{\&Sigma;}}}\left|x_{\mathrm{id}}^t\right|};$

(3) fitness value of described each particle is calculated wherein fitness function is defined as the testing result that the weights representated by each particle obtain, and according to fitness value, determines local optimum particle $p_i^t=(p_{i1}^t,p_{i2}^t,\&CenterDot;\&CenterDot;\&CenterDot;,p_{\mathrm{iD}}^t)$ With global optimum's particle $p_b^t=(x_{b1}^t,x_{b2}^t,\&CenterDot;\&CenterDot;\&CenterDot;,x_{\mathrm{bD}}^t),$ Wherein b is the ordinal number that the minimum particle of fitness value is corresponding, for t=0,

(4) t=t+1 is made, and by formula $v_{\mathrm{id}}^t={\mathrm{\&omega;v}}_{\mathrm{id}}^{t-1}+c_1\mathrm{\&xi;}(p_{\mathrm{id}}^{t-1}-x_{\mathrm{id}}^{t-1})+c_2\mathrm{\&eta;}(p_{\mathrm{bd}}^{t-1}-x_{\mathrm{id}}^{t-1})$ Renewal speed wherein c ₁and c ₂be normal number, ξ and η is distributed in the random number in [0,1], and ω is inertia weight, with the d of corresponding local optimum particle and global optimum's particle ties up element respectively, wherein, if $v_{\mathrm{id}}^t>V_{\max },$ Order $v_{\mathrm{id}}^t=V_{\max };$ If $v_{\mathrm{id}}^t<-V_{\max },$ Order $v_{\mathrm{id}}^t=-V_{\max };$

(5) by formula upgrade position be normalized according to step (2b);

(6) fitness value through each particle upgraded is calculated for particle if its current fitness value is less than the local optimum particle in last iteration fitness value, its local optimum particle is set to otherwise its local optimum particle remains unchanged if particle current fitness value is less than global optimum's particle in last iteration fitness value, by this particle be set to global optimum's particle if do not have the fitness value of particle to be less than fitness value, Ze Ling global optimum particle remains unchanged

(7) judge whether iterations t reaches maximum iteration time T _maxif reach, export global optimum, and according to formula $P_d=Q\left(\frac{Q^{-1}\left(P_f\right)\sqrt{w^T\mathrm{Aw}}-E_s{g}^{T}w}{\sqrt{w^T\mathrm{Bw}}}\right)$ Obtain the detection probability P corresponding to it _d; Otherwise, repeat step (4) ~ (7).

2. collaborative spectrum sensing optimization method according to claim 1, is characterized in that, the computing formula of described step (3) is:

According to formula fitness (x _i)=f (x _i), wherein x _i=w; Fitness (x _i) be the fitness value of corresponding particle