CN105071881A - Re-entry dynamic plasma sheath Markov channel modeling method - Google Patents

Re-entry dynamic plasma sheath Markov channel modeling method Download PDF

Info

Publication number
CN105071881A
CN105071881A CN201510513924.6A CN201510513924A CN105071881A CN 105071881 A CN105071881 A CN 105071881A CN 201510513924 A CN201510513924 A CN 201510513924A CN 105071881 A CN105071881 A CN 105071881A
Authority
CN
China
Prior art keywords
prime
omega
state
sigma
parameter
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN201510513924.6A
Other languages
Chinese (zh)
Other versions
CN105071881B (en
Inventor
石磊
方水汛
李小平
姚博
杨敏
刘彦明
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xidian University
Original Assignee
Xidian University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Xidian University filed Critical Xidian University
Priority to CN201510513924.6A priority Critical patent/CN105071881B/en
Publication of CN105071881A publication Critical patent/CN105071881A/en
Application granted granted Critical
Publication of CN105071881B publication Critical patent/CN105071881B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Landscapes

  • Other Investigation Or Analysis Of Materials By Electrical Means (AREA)
  • Complex Calculations (AREA)

Abstract

The invention discloses a re-entry dynamic plasma sheath Markov channel modeling method. The method comprises the following steps: firstly, inputting an amplitude attenuation sequence of a plasma sheath under a specific condition, establishing a multi-state Markov channel model, and obtaining model parameters by using the reversible hop Markov Monte Carlo algorithm; secondly, generating a state sequence according to a state transition matrix of the Markov channel model; and lastly, generating a signal random attenuation sequence according to a Gaussian distribution parameter and a state sequence in each state of the Markov channel model. The state number of an optimal Markov channel and the distribution parameter of a random process in each corresponding state are obtained through the reversible hop Markov Monte Carlo algorithm. Compared with an existing Markov plasma channel modeling method, the re-entry dynamic plasma sheath Markov channel modeling method has the advantages that a preset state number does not need to be preset, and errors caused by human factors can be reduced.

Description

One reenters dynamic plasma sheath cover Markov channel modeling method
Technical field
The invention belongs to TTC&T Technology field, be specifically related to one and reenter the cover channel Markov modeling of dynamic plasma sheath and analogy method.
Background technology
Reentry vehicle is in hypersonic flight process, and meeting coated one deck high temperature thermic plasmasphere around aircraft, is called as plasma sheath cover.Charged particle based on free electron in sheath cover will absorb, reflect and scattering electromagnetic wave, produces the effect of metalloid shielding, makes electromagnetic signal generation deep fades.These effects will cause communication quality to worsen, and signal of communication will be caused time serious to interrupt (black barrier phenomenon).Black barrier phenomenon by have a strong impact on ground station to the catching of aircraft, follow the tracks of and real-time telemetry transfer of data, cause unpredictable consequence.
From radio communication angle, plasma sheath cover channel circumstance and not mating of communication system, result in signal of communication interruption.Plasma sheath physical parameter has significantly dynamically time-varying characteristics, and it produces mainly due to factors such as attitude of flight vehicle adjustment, turbulent perturbation, non-homogeneous ablations.Become during physical parameter dynamic must cause electromagnetic dielectric parameter time become, and then cause the distortion of radio wave propagation and signal.Obviously can observe the width phase mudulation effect of signal at present by experiment.Therefore being adapted to communication system under this Special complex channel and method for finding, reentering setting up of dynamic plasma sheath cover channel model most important.
Because the factor affecting the dynamic change of plasma electric magnetic parameter is different, cause the fluctuation range of amplitude fading larger, describe not accurate enough according to single random process, therefore can consider to adopt multimode Markov Chain to be described, multimode markoff process is higher than single random process modeling method accuracy.Existing research adopts the multimode Markov approach of equiprobability criteria for classifying to carry out modeling to channel, but the method needs preset channel status number, and the adaptability of Channel Modeling is poor.
Summary of the invention
For the deficiencies in the prior art, the present invention is intended to propose a kind of new multimode Markov channel modeling method, the method obtains the distributed constant of random process under the status number of best Markov channel and each state by reversible saltus step Markov Monte carlo algorithm, advantage is not need preset state number compared with existing Markov plasma channel modeling method, can reduce the error that human factor is introduced.
To achieve these goals, the present invention adopts following technical scheme:
One reenters dynamic plasma sheath cover Markov channel modeling method, comprises following steps:
Under S1 input specified conditions, signal is through the decay sequence of plasma sheath cover;
S2 sets up the multimode Markov channel model of plasma sheath cover, according to the plasma sheath cover decay sequence of step S1 input, obtain channel model parameters by reversible saltus step Markov Monte carlo algorithm, comprise Gaussian Distribution Parameters under Markov channel model status number, channel model state-transition matrix and each state;
The state-transition matrix that S3 utilizes step S2 to obtain generates Markov state sequence z t, t=1,2 ... n, n are sequence length;
The Markov state sequence produced under Gaussian Distribution Parameters and step S3 under each state that S4 utilizes step S2 to obtain, produces corresponding random decay sequence simulation value y t, t=1,2 ... n, n are sequence length.
It should be noted that, step S2 is specifically implemented as follows:
2.1) each state of Markov channel plasma sheath overlapped is modeled as Gaussian process, and the first-order statistics characteristic of channel represents with mixed Gauss model:
Y = Σ j = 1 K ω j N ( μ j , σ j 2 ) ;
Wherein Y represents the probability density function of channel, and K is the number of Gaussian process in the status number of markoff process and mixed Gauss model, ω jrepresent state probability, namely label is the weight of Gaussian density function shared by Mixture Model Probability Density Function of j, and has represent that label is the Gaussian process of j, and represent that label is average and the variance of the Gaussian process of j, wherein j=1,2 ..., K;
2.2) Confirming model parameter (K, ω j, μ j, σ j) prior distribution: status number K obeys Poisson distribution, μ jgaussian distributed, i.e. μ j~ N (ξ, κ -1), and σ jobey inverse gamma distribution, namely wherein β obeys gamma distribution, i.e. β ~ G (g, f), mixed Gaussian process weights omega jform ω vector, and ω obeys Dirichlet distribution, i.e. ω ~ D (δ 1, δ 2...., δ k), wherein (ξ, κ, g, f, δ) is transcendent distributed constant, is definite value, and δ is (δ 1..., δ k) set;
2.3) each parameter of initialization model, comprising: init state number K (1), Gaussian Distribution Parameters under each state prior distribution hyper parameter (ξ, κ, g, f, δ) and birth probability b kwith POD d k=1-b k, k is the label of status number; Birth probability and POD obey binomial distribution, and set iterations N and suppose uneven iterations M, and M gets initialization iteration count h; In addition, increase the hidden status switch z of parameter, its initialization is determined by following formula:
z (1) i=argmax{Pr(z i=j)}i=1,2....nj=1,2,...K;
Pr ( z i = j ) = ω j 1 2 π σ j e - ( y i - μ j ) 2 2 σ j 2 ;
Wherein, y i, i=1,2 ... .n is signal attenuation sequence, and n is the length of sequence, z ifor in the hidden status switch z of parameter, label is the value of i, z i (1)it is then its initialization value;
2.4) the split degree process of the h time iteration is entered: produce (0, a 1) equally distributed random number U1, if U1 < is b k, then fission process is entered; Otherwise, enter merging process; Parameter space is updated to x 1 ( h + 1 ) = ( K ( h + 1 ) 1 , &omega; j ( h + 1 ) 1 , &mu; j ( h + 1 ) 1 , &sigma; j ( h + 1 ) 1 , z ( h + 1 ) 1 , &beta; ( h ) ) , Wherein be respectively status number K, state probability ω j, label is that the average of the Gaussian process of j and variance, the hidden status switch z of parameter complete the renewal after the split degree process of the h time iteration, hyper parameter β because do not upgrade in fission process, so retain the h-1 time iteration complete after updated value β (h);
2.5) the birth death process of the h time iteration is entered: produce (0, a 1) uniform random number U2, if U2 < is b k, then enter birth process, otherwise enter death process; Parameter space is updated to x 2 ( h + 1 ) = ( K ( h + 1 ) 2 , &omega; j ( h + 1 ) 2 , &mu; j ( h + 1 ) 2 , &sigma; j ( h + 1 ) 2 , z ( h + 1 ) 2 , &beta; ( h ) ) , Wherein be respectively status number K, state probability ω j, label is that the average of the Gaussian process of j and variance, the hidden status switch z of parameter complete the renewal after the birth death process of the h time iteration, hyper parameter β because do not upgrade in merging process, so retain the h-1 time iteration complete after updated value β (h);
2.6) Gibbs sampling undated parameter space is utilized obtain the h time iteration result x ( h + 1 ) = ( K ( h + 1 ) , &omega; j ( h + 1 ) , &mu; j ( h + 1 ) , &sigma; j ( h + 1 ) , z ( h + 1 ) , &beta; ( h + 1 ) ) , K (h+1), ω j (h+1), μ j (h+1), σ j (h+1), z (h+1)and β (h+1)be respectively status number K, state probability ω j, label is the final updated value that the average of the Gaussian process of j and variance, parameter hidden status switch z and hyper parameter β obtain after completing the h time iteration;
2.7) h=h+1, if h < is N, repeats 2.4)-2.6);
2.8) for parameter space iteration result x (1), x (2)...., x (N), M uneven iteration before removing, obtains x (M+1), x (M+2)...., x (N), to wherein parameter K (M+1), K (M+2)..., K (N)make histogram, obtain maximum k op, be optimum state number;
2.9) at parameter space x (M+1), x (M+2)...., x (N)in, for K (i)=k op, i=M+1, M+2 ..., the parameter (ω of N (i), μ (i), σ (i)), obtain mathematic expectaion (ω, μ, σ), be the Gaussian parameter under Markov model different conditions;
2.10) utilize 2.3) in the likelihood function and 2.9 of the hidden status switch z of parameter) in Gaussian parameter under the Markov model different conditions obtained, obtain hidden status switch t=1,2 ... n, and obtain state-transition matrix P further and be:
P(i,j)=N ij/N i
Wherein, P (i, j) represents the probability jumping to state j from state i, N ijfor the number of state j is transferred to, N from state i ifor be in the number of state i, i, j=1,2 ..., n.
It should be noted that, step 2.4) in, described fission process is specific as follows:
2.4.1.1) the parameter space x finally obtained after the h-1 time iteration being completed (h), from j=1,2 ..., K (h)stochastic choice state j s *, parameter is with j s1, j s2represent the label of the rear state of division, after division, status number becomes K '=K (h)+ 1, K (h)and z (h)be the final updated value of status number K and the hidden status switch z of parameter obtained after the h-1 time iteration completes, and be respectively the h-1 time iteration complete after state j s *ω j, μ jand σ jupdated value;
2.4.1.2) three random number u are produced 1, u 2, u 3, meet:
u 1~beta(2,2)u 2~beta(2,2)u 3~beta(1,1);
Beta is beta distribution, state weight and corresponding state Gaussian Distribution Parameters after division:
&omega; j s 1 &prime; = &omega; j s * ( h ) * u 1 &omega; j s 2 &prime; = &omega; j s * ( h ) * ( 1 - u 1 )
&mu; j s 1 &prime; = &mu; j s 1 ( h ) - u 2 &sigma; j s * ( h ) &omega; j s 1 &prime; &omega; j s 2 &prime; &mu; j s 2 &prime; = &mu; j s 2 ( h ) + u 2 &sigma; j s * ( h ) &omega; j s 2 &prime; &omega; j s 1 &prime; ;
&sigma; j s 1 &prime; = u 3 ( 1 - u 2 2 ) &sigma; j s * ( h ) &omega; j s * ( h ) &omega; j s 1 &prime; &sigma; j s 2 &prime; = ( 1 - u 3 ) ( 1 - u 2 2 ) &sigma; j s * ( h ) &omega; j s * ( h ) &omega; j s 2 &prime;
with represent state j respectively s1and j s2μ is obtained after the h-1 time corresponding iteration completes jupdated value;
2.4.1.3) the hidden status switch of undated parameter, will be originally sample point according to following probability assignments to j s1and j s2:
p a l l o c = &omega; j s 1 &prime; &sigma; j s 1 &prime; exp ( - ( y i - &mu; j s 1 &prime; ) 2 2 &sigma; j s 1 &prime; 2 ) &omega; j s 1 &prime; &sigma; j s 1 &prime; exp ( - ( y i - &mu; j s 1 &prime; ) 2 2 &sigma; j s 1 &prime; 2 ) + &omega; j s 2 &prime; &sigma; j s 2 &prime; exp ( - ( y i - &mu; j s 2 &prime; ) 2 2 &sigma; j s 2 &prime; 2 ) ;
P allocdistribute to j s1probability, q=1-P allocdistribute to j s2probability, the hidden status switch of the parameter after renewal is designated as z ';
2.4.1.4) acceptance probability split_accept=min (1, split_A) is calculated, wherein:
s p l i t _ A = ( l i k e l i h o o d r a t i o ) &times; p ( K ( h ) + 1 ) p ( K ( h ) ) &times; ( K ( h ) + 1 ) &times; &omega; j s 1 &prime; &delta; j s 1 - 1 + l 1 &omega; j s 2 &prime; &delta; j s 2 - 1 + l 2 &omega; j s * ( h ) &delta; j s * - 1 + l 1 + l 2 B ( &delta; j s * , K&delta; j s * ) &times; &kappa; 2 &pi; exp &lsqb; - 1 2 &kappa; { ( &mu; j s 1 &prime; - &xi; ) 2 + ( &mu; j s 2 &prime; - &xi; ) 2 - ( &mu; j s * ( h ) - &xi; ) 2 } &rsqb; &times; &beta; ( h ) &alpha; &Gamma; ( &alpha; ) ( &sigma; j s 1 &prime; 2 &sigma; j s 2 &prime; 2 &sigma; j s * ( h ) 2 ) - &alpha; - 1 exp ( - &beta; ( &sigma; j s 1 &prime; - 2 + &sigma; j s 2 &prime; - 2 - &sigma; j s * ( h ) - 2 ) ) &times; d K ( h ) + 1 b K ( h ) P a l l o c &times; { g 2 , 2 ( u 1 ) g 2 , 2 ( u 2 ) g 1 , 1 ( u 3 ) } - 1 &times; &omega; j s * ( h ) | &mu; j s 1 &prime; - &mu; j s 2 &prime; | &sigma; j s 1 &prime; 2 &sigma; j s 2 &prime; 2 u 2 ( 1 - u 2 2 ) u 3 ( 1 - u 3 ) &sigma; j s * ( h ) 2 ;
Wherein, (likehoodratio) represents the likelihood ratio of the front state value of the state value after division and division, p (K (h)) and p (K (h)+ 1) being respectively status number is K (h)and K (h)probability when+1, l 1, l 2for sample y ibelong to state j s1, j s2count, B (|) represents beta function, g p, qrepresent beta (p, q) probability density function, with represent that status number is K respectively (h)and K (h)the birth rate of+1 and the death rate, represent status number j respectively s1, j s2with corresponding δ;
2.4.1.5) (0, a 1) uniform random number U is produced sif, U s< split_accept, then model parameter space by raw parameter space x (h)in parameter be updated to otherwise parameter space initial value will be retained, namely x 1 ( h + 1 ) = x ( h ) .
It should be noted that, step 2.4) in, described merging process is as follows:
2.4.2.1) for parameter space x (h), from j=1,2 ..., K (h)middle Stochastic choice two states, are designated as j c1and j c2, state parameter is state after merging is designated as j c *, then status number becomes K '=K (h)-1;
2.4.2.2) originally state j will be belonged to c1and j c2sample point unified to incorporate into as state j c *, the status switch after merging is designated as z ', that is:
z i &prime; = j c * , i &Element; { i : z i ( h ) = j c 1 } &cup; { i : z i ( h ) = j c 2 } ;
2.4.2.3) by following equation group Renewal model parameter
&omega; j c * &prime; = &omega; j c 1 ( h ) + &omega; j c 2 ( h )
&omega; j c * &prime; &mu; j c * &prime; = &omega; j c 1 ( h ) &mu; j c 1 ( h ) + &omega; j c 2 ( h ) &mu; j c 2 ( h ) ;
&omega; j c * &prime; ( &mu; j c * &prime; 2 + &sigma; j c * &prime; 2 ) = &omega; j c 1 ( h ) ( &mu; j c 1 ( h ) 2 + &sigma; j c 1 ( h ) 2 ) + &omega; j c 2 ( h ) ( &mu; j c 2 ( h ) 2 + &sigma; j c 2 ( h ) 2 )
2.4.2.4) calculate acceptance probability combine_accept=min (1, combine_A), wherein combine_A is step 2.4.1.4) inverse of split_A in fission process;
2.4.2.5) (0, a 1) uniform random number U is produced cif, U c< combine_accept, then model parameter space by parameter space x (h)in parameter be updated to otherwise parameter space initial value will be retained, namely x 1 ( h + 1 ) = x ( h ) .
It should be noted that, step 2.5) in, described birth process is specific as follows:
2.5.1.1) utilize step 2.2) the prior distribution of hypothesis produce a new stochastic regime j b *, now status number becomes new state parameter is:
&omega; j b * &prime; ~ b e t a ( 1 , K &prime; ) &mu; j b * &prime; ~ N ( &xi; , &kappa; - 1 ) &sigma; j b * &prime; - 2 ~ &Gamma; ( &alpha; , &beta; ) ;
2.5.1.2) revise the weight of the previous status in parameter space:
&omega; j &prime; = &omega; j ( h + 1 ) 1 ( 1 - &omega; j b * ) , j = 1 , 2.... K ( h + 1 ) 1 ;
Thus make &omega; j b * + &Sigma; j = 1 K &omega; j &prime; = 1 ;
2.5.1.3) acceptance probability birth_accept=min (1, birth_A) is calculated, wherein
b i r t h _ A = p ( K ( h + 1 ) 1 + 1 ) p ( K ( h + 1 ) 1 ) 1 B ( K ( h + 1 ) 1 &delta; j b * , &delta; j b * ) &omega; j b * &prime; &delta; j b * - 1 ( 1 - &omega; j b * &prime; ) n + K ( h + 1 ) 1 &delta; j b * - K ( h + 1 ) 1 ( K ( h + 1 ) 1 + 1 ) &times; d K ( h + 1 ) 1 + 1 ( K 0 + 1 ) b K ( h + 1 ) 1 1 g 1 , K ( h + 1 ) 1 ( &omega; j b * &prime; ) ( 1 - &omega; j b * &prime; ) K ( h + 1 ) 1 ;
Wherein, k 0before representing newly-increased state status number in parameter space, with represent that status number is respectively with probability, with represent that status number is respectively with birth rate and the death rate;
2.5.1.4) (0, a 1) distribution random numbers U is produced bif, U b< birth_accept, then will increase state status number K ( h + 1 ) 2 = K &prime; ; Otherwise parameter space initial value will be retained, namely x 2 ( h + 1 ) = x ( h + 1 ) .
It should be noted that, step 2.5) in, described death process is specific as follows:
2.5.2.1) from raw parameter space selection state of dummy status equal probability, be designated as j d *, now state becomes
2.5.2.2) revise the weight of the previous status in parameter space:
&omega; j &prime; = &omega; j 1 - &omega; j d * , j &NotEqual; j d * ;
Thus ensure &Sigma; j &NotEqual; j d * &omega; j &prime; = 1 ;
2.5.2.3) calculate acceptance probability death_accept=min (1, death_A), wherein death_A is birth process step 2.5.1.3) in the inverse of birth_A;
2.5.2.4) (0, a 1) distribution random numbers U is produced dif, U d< death_accept, then by state delete, status number K ( h + 1 ) 2 = K &prime; ; Otherwise parameter space initial value will be retained, namely x 2 ( h + 1 ) = x 1 ( h + 1 ) .
It should be noted that, step 2.6) in, the h time iteration result x ( h + 1 ) = ( K ( h + 1 ) , &omega; j ( h + 1 ) , &mu; j ( h + 1 ) , &sigma; j ( h + 1 ) , z ( h + 1 ) , &beta; ( h + 1 ) ) In, K ( h + 1 ) = K ( h + 1 ) 2 , Parameter with hyper parameter β (h+1)upgrade as follows:
&omega; ( h + 1 ) ~ D i r i c h l e t ( &delta; 1 + n 1 , ... , &delta; K + n K ( h + 1 ) )
&mu; j ( h + 1 ) ~ N { &sigma; j ( h + 1 ) 2 - 2 &Sigma; i : z i ( h + 1 ) 2 = j y i + &kappa; &xi; &sigma; j ( h + 1 ) 2 - 2 n j + &kappa; , ( &sigma; j ( h + 1 ) 2 - 2 n j + &kappa; ) - 1 }
&sigma; j ( h + 1 ) - 2 ~ &Gamma; { &alpha; + 1 2 n j , &beta; ( h ) + 1 2 &Sigma; i : z i ( h + 1 ) 2 = j ( y i - &mu; j ( h + 1 ) ) 2 } ;
( z i ( h + 1 ) = j ) ~ &omega; j ( h + 1 ) &sigma; j ( h + 1 ) exp { - ( y i - &mu; j ( h + 1 ) ) 2 2 &sigma; j ( h + 1 ) 2 }
&beta; ( h + 1 ) ~ &Gamma; ( g + &kappa; &alpha; , f + &Sigma; j &sigma; j ( h + 1 ) - 2 )
Wherein Γ () represents gamma function, n jbe j=1,2 ..., K (h+1)number, for the hidden status switch z of parameter completes the renewal after the birth death process of the h time iteration middle label is the value of i.
It should be noted that, described step S3 detailed process is as follows:
3.1) generate (0, a 1) uniform random number u, be used for determining new state;
3.2) when setting models be in state k (k=1,2 ..., k op), cumulative probability is denoted as β k, provided by following formula:
&beta; k = &Sigma; j = 1 k p k j ;
Wherein p kjrepresent the probability forwarding state j from state k-hop to;
3.3) condition then jumping to state i is:
β i≤u≤β i+1
Determined next step state by this condition, and then produce the Markov state sequence z on emulation link t, t=1,2 ... n.
It should be noted that, in step S4, decay sequences y at random tsimulation value is specially:
y t ~ N ( &mu; z t , &sigma; z t ) ;
Wherein represent Markov state sequence z t, t=1,2 ... the average of state corresponding to n and variance.
Beneficial effect of the present invention is:
1, one provided by the invention reenters dynamic plasma Markov channel modeling method, utilize multimode Markov Chain can avoid single random process cannot the limitation of accurate description channel characteristics to Channel Modeling, and each for Markov state is modeled as Gaussian process, whole like this channel first-order statistics characteristic is equivalent to mixed Gaussian process, it is when different conditions number, can be used for approximate Independent Sources with Any Probability Density Function, ensure that the feasibility of Channel Modeling, simultaneously for the dynamic plasma under different condition, can be described by unified mathematical form,
2, in Markov Channel Modeling process, do not do default to status number, can avoid like this introducing human error, but calculate Gaussian Distribution Parameters under markovian optimum state number and each state by reversible saltus step Markov Monte carlo algorithm;
3, the present invention is applicable to reentry plasma sheath cover Channel Modeling, also be applicable to space flight near space vehicle plasma sheath cover Channel Modeling, the channel model set up can be used for algorithm design and the Performance Evaluation of the communication physical layer transmission technologys such as modulating/demodulating, chnnel coding, channel estimating and equilibrium.
Accompanying drawing explanation
Fig. 1 is plasma sheath of the present invention cover channel Markov Channel Modeling and modeling process chart;
Fig. 2 is that reversible saltus step Markov Monte carlo algorithm calculates optimum state number and each distributions parameter schematic diagram;
Fig. 3 is the plasma sheath cover decay sequence under certain condition of input;
Fig. 4 is the optimum state number histogram that reversible saltus step Markov Monte carlo algorithm is obtained;
Fig. 5 is that dynamic plasma Markov channel model probability density function simulation value and theoretical value contrast schematic diagram;
Fig. 6 is that dynamic plasma Markov channel model cumulative distribution function simulation value and theoretical value contrast schematic diagram.
Embodiment
Below with reference to accompanying drawing, the invention will be further described, it should be noted that, the present embodiment, premised on the technical program, give detailed execution mode and concrete operating process, but protection scope of the present invention is not limited to the present embodiment.
As shown in Figure 1, the present invention specifically comprises the steps:
Step S101, under input specified conditions, signal is through the decay sequences y of plasma sheath cover i, i=1,2 ... .n, n are the length of sequence, as shown in Figure 3.
Step S102, sets up the multimode Markov channel model of plasma sheath cover, according to the plasma sheath cover decay sequence of step S101 input, obtains channel model parameters by reversible saltus step Markov Monte carlo algorithm.As shown in Figure 2, concrete steps comprise flow process:
2.1) each state of Markov channel plasma sheath overlapped is modeled as Gaussian process, and the first-order statistics characteristic of channel represents with mixed Gauss model:
Y = &Sigma; j = 1 K &omega; j N ( &mu; j , &sigma; j 2 ) ;
Wherein Y represents the probability density function of channel, and K is the number of Gaussian process in the status number of markoff process and mixed Gauss model, ω jrepresent state probability, namely label is the weight of Gaussian density function shared by Mixture Model Probability Density Function of j, and has represent that label is the Gaussian process of j, and represent that label is average and the variance of the Gaussian process of j, wherein j=1,2 ..., K;
2.2) Confirming model parameter (K, ω j, μ j, σ j) prior distribution: status number K obeys Poisson distribution, μ jgaussian distributed, i.e. μ j~ N (ξ, κ -1), and σ jobey inverse gamma distribution, namely wherein β obeys gamma distribution, i.e. β ~ G (g, f), mixed Gaussian process weights omega jform ω vector, and ω obeys Dirichlet distribution, i.e. ω ~ D (δ 1, δ 2...., δ k), wherein (ξ, κ, g, f, δ) is transcendent distributed constant, is definite value, and δ is (δ 1..., δ k) set;
2.3) each parameter of initialization model, comprising: init state number K (1), Gaussian Distribution Parameters under each state prior distribution hyper parameter (ξ, κ, g, f, δ) and birth probability b kwith POD d k=1-b k, k is the label of status number; Birth probability and POD obey binomial distribution, and set iterations N and suppose uneven iterations M, and M gets initialization iteration count h.Specifically be initialized as N=3000, M=1000, h=1, K max=20, ξ=y min+ R/2, κ=1/R 2, g=0.2, f=10/R 2, δ=1, wherein R=y max-y min, y minand y maxrepresent y respectively i, i=1,2 ... the maximum in .n and minimum value; d 1=0, b k=d k=0.5, k=1,2 ..., K max-1. initialization by step 2.2) in hypothesis prior distribution determine, in addition, hidden status switch z initialization is determined by following formula:
z (1) i=argmax{Pr(z i=j)}i=1,2....nj=1,2,...K;
Pr ( z i = j ) = &omega; j 1 2 &pi; &sigma; j e - ( y i - &mu; j ) 2 2 &sigma; j 2 ;
Wherein, y i, i=1,2 ... .n is signal attenuation sequence, and n is the length of sequence, z ifor in the hidden status switch z of parameter, label is the value of i, z i (1)it is then its initialization value;
2.4) the split degree process of the h time iteration is entered: produce (0, a 1) equally distributed random number U1, if U1 < is b k, then enter fission process, namely enter step 2.4.1); Otherwise, enter merging process, namely enter step 2.4.2); Parameter space is updated to x 1 ( h + 1 ) = ( K ( h + 1 ) 1 , &omega; j ( h + 1 ) 1 , &mu; j ( h + 1 ) 1 , &sigma; j ( h + 1 ) 1 , z ( h + 1 ) 1 , &beta; ( h ) ) , Wherein be respectively status number K, state probability ω j, label is that the average of the Gaussian process of j and variance, the hidden status switch z of parameter complete the renewal after the split degree process of the h time iteration, hyper parameter β because do not upgrade in fission process, so retain the h-1 time iteration complete after updated value β (h), specific as follows:
2.4.1) fission process:
2.4.1.1) the parameter space x finally obtained after the h-1 time iteration being completed (h), from j=1,2 ..., K (h)stochastic choice state j s *, parameter is with j s1, j s2represent the label of the rear state of division, after division, status number becomes K '=K (h)+ 1, K (h)and z (h)be the final updated value of status number K and the hidden status switch z of parameter obtained after the h-1 time iteration completes, and be respectively the h-1 time iteration complete after state j s *ω j, μ jand σ jupdated value;
2.4.1.2) three random number u are produced 1, u 2, u 3, meet:
u 1~beta(2,2)u 2~beta(2,2)u 3~beta(1,1);
Beta is beta distribution, state weight and corresponding state Gaussian Distribution Parameters after division:
&omega; j s 1 &prime; = &omega; j s * ( h ) * u 1 &omega; j s 2 &prime; = &omega; j s * ( h ) * ( 1 - u 1 )
&mu; j s 1 &prime; = &mu; j s 1 ( h ) - u 2 &sigma; j s * ( h ) &omega; j s 1 &prime; &omega; j s 2 &prime; &mu; j s 2 &prime; = &mu; j s 2 ( h ) + u 2 &sigma; j s * ( h ) &omega; j s 2 &prime; &omega; j s 1 &prime; ;
&sigma; j s 1 &prime; = u 3 ( 1 - u 2 2 ) &sigma; j s * ( h ) &omega; j s * ( h ) &omega; j s 1 &prime; &sigma; j s 2 &prime; = ( 1 - u 3 ) ( 1 - u 2 2 ) &sigma; j s * ( h ) &omega; j s * ( h ) &omega; j s 2 &prime;
with represent state j respectively s1and j s2μ is obtained after the h-1 time corresponding iteration completes jupdated value;
2.4.1.3) the hidden status switch of undated parameter, will be originally sample point according to following probability assignments to j s1and j s2:
p a l l o c = &omega; j s 1 &prime; &sigma; j s 1 &prime; exp ( - ( y i - &mu; j s 1 &prime; ) 2 2 &sigma; j s 1 &prime; 2 ) &omega; j s 1 &prime; &sigma; j s 1 &prime; exp ( - ( y i - &mu; j s 1 &prime; ) 2 2 &sigma; j s 1 &prime; 2 ) + &omega; j s 2 &prime; &sigma; j s 2 &prime; exp ( - ( y i - &mu; j s 2 &prime; ) 2 2 &sigma; j s 2 &prime; 2 ) ;
P allocdistribute to j s1probability, q=1-P allocdistribute to j s2probability, the hidden status switch of the parameter after renewal is designated as z ';
2.4.1.4) acceptance probability split_accept=min (1, split_A) is calculated, wherein:
s p l i t _ A = ( l i k e l i h o o d r a t i o ) &times; p ( K ( h ) + 1 ) p ( K ( h ) ) &times; ( K ( h ) + 1 ) &times; &omega; j s 1 &prime; &delta; j s 1 - 1 + l 1 &omega; j s 2 &prime; &delta; j s 2 - 1 + l 2 &omega; j s * ( h ) &delta; j s * - 1 + l 1 + l 2 B ( &delta; j s * , K&delta; j s * ) &times; &kappa; 2 &pi; exp &lsqb; - 1 2 &kappa; { ( &mu; j s 1 &prime; - &xi; ) 2 + ( &mu; j s 2 &prime; - &xi; ) 2 - ( &mu; j s * ( h ) - &xi; ) 2 } &rsqb; &times; &beta; ( h ) &alpha; &Gamma; ( &alpha; ) ( &sigma; j s 1 &prime; 2 &sigma; j s 2 &prime; 2 &sigma; j s * ( h ) 2 ) - &alpha; - 1 exp ( - &beta; ( &sigma; j s 1 &prime; - 2 + &sigma; j s 2 &prime; - 2 - &sigma; j s * ( h ) - 2 ) ) &times; d K ( h ) + 1 b K ( h ) P a l l o c &times; { g 2 , 2 ( u 1 ) g 2 , 2 ( u 2 ) g 1 , 1 ( u 3 ) } - 1 &times; &omega; j s * ( h ) | &mu; j s 1 &prime; - &mu; j s 2 &prime; | &sigma; j s 1 &prime; 2 &sigma; j s 2 &prime; 2 u 2 ( 1 - u 2 2 ) u 3 ( 1 - u 3 ) &sigma; j s * ( h ) 2 ;
Wherein, (likehoodratio) represents the likelihood ratio of the front state value of the state value after division and division, p (K (h)) and p (K (h)+ 1) being respectively status number is K (h)and K (h)probability when+1, l 1, l 2for sample y ibelong to state j s1, j s2count, B (|) represents beta function, g p, qrepresent beta (p, q) probability density function, with represent that status number is K respectively (h)and K (h)the birth rate of+1 and the death rate, represent status number j respectively s1, j s2with corresponding δ;
2.4.1.5) (0, a 1) uniform random number U is produced sif, U s< split_accept, then model parameter space by raw parameter space x (h)in parameter be updated to otherwise parameter space initial value will be retained, namely x 1 ( h + 1 ) = x ( h ) .
2.4.2) merging process:
2.4.2.1) for parameter space x (h), from j=1,2 ..., K (h)middle Stochastic choice two states, are designated as j c1and j c2, state parameter is state after merging is designated as j c *, then status number becomes K '=K (h)-1;
2.4.2.2) originally state j will be belonged to c1and j c2sample point unified to incorporate into as state j c *, the status switch after merging is designated as z ', that is:
z i &prime; = j c * , i &Element; { i : z i ( h ) = j c 1 } &cup; { i : z i ( h ) = j c 2 } ;
2.4.2.3) by following equation group Renewal model parameter
&omega; j c * &prime; = &omega; j c 1 ( h ) + &omega; j c 2 ( h )
&omega; j c * &prime; &mu; j c * &prime; = &omega; j c 1 ( h ) &mu; j c 1 ( h ) + &omega; j c 2 ( h ) &mu; j c 2 ( h ) ;
&omega; j c * &prime; ( &mu; j c * &prime; 2 + &sigma; j c * &prime; 2 ) = &omega; j c 1 ( h ) ( &mu; j c 1 ( h ) 2 + &sigma; j c 1 ( h ) 2 ) + &omega; j c 2 ( h ) ( &mu; j c 2 ( h ) 2 + &sigma; j c 2 ( h ) 2 )
2.4.2.4) calculate acceptance probability combine_accept=min (1, combine_A), wherein combine_A is step 2.4.1.4) inverse of split_A in fission process;
2.4.2.5) (0, a 1) uniform random number U is produced cif, U c< combine_accept, then model parameter space by parameter space x (h)in parameter be updated to otherwise parameter space initial value will be retained, namely x 1 ( h + 1 ) = x ( h ) .
2.5) the birth death process of the h time iteration is entered: produce (0, a 1) uniform random number U2, if U2 < is b k, then enter birth process, namely enter step 2.5.1), otherwise enter death process, namely enter step 2.5.2); Parameter space is updated to x 2 ( h + 1 ) = ( K ( h + 1 ) 2 , &omega; j ( h + 1 ) 2 , &mu; j ( h + 1 ) 2 , &sigma; j ( h + 1 ) 2 , z ( h + 1 ) 2 , &beta; ( h ) ) , Specific as follows:
2.5.1) birth process:
2.5.1.1) utilize step 2.2) the prior distribution of hypothesis produce a new stochastic regime j b *, now status number becomes new state parameter is:
&omega; j b * &prime; ~ b e t a ( 1 , K &prime; ) &mu; j b * &prime; ~ N ( &xi; , &kappa; - 1 ) &sigma; j b * &prime; 2 ~ &Gamma; ( &alpha; , &beta; ) ;
2.5.1.2) revise the weight of the previous status in parameter space:
&omega; j &prime; = &omega; j ( h + 1 ) 1 ( 1 - &omega; j b * ) , j = 1 , 2 ... . K ( h + 1 ) 1 ;
Thus make &omega; j b * + &Sigma; j = 1 K &omega; j &prime; = 1 ;
2.5.1.3) acceptance probability birth_accept=min (1, birth_A) is calculated, wherein
b i r t h _ A = p ( K ( h + 1 ) 1 + 1 ) p ( K ( h + 1 ) 1 ) 1 B ( K ( h + 1 ) 1 &delta; j b * , &delta; j b * ) &omega; j b * &prime; &delta; j b * - 1 ( 1 - &omega; j b * &prime; ) n + K ( h + 1 ) 1 &delta; j b * - K ( h + 1 ) 1 ( K ( h + 1 ) 1 + 1 ) &times; d K ( h + 1 ) 1 + 1 ( K 0 + 1 ) b K ( h + 1 ) 1 1 g 1 , K ( h + 1 ) 1 ( &omega; j b * &prime; ) ( 1 - &omega; j b * &prime; ) K ( h + 1 ) 1 ;
Wherein, k 0before representing newly-increased state status number in parameter space, with represent that status number is respectively with probability, with represent that status number is respectively with birth rate and the death rate;
2.5.1.4) (0, a 1) distribution random numbers U is produced bif, U b< birth_accept, then will increase state status number K ( h + 1 ) 2 = K &prime; ; Otherwise parameter space initial value will be retained, namely x 2 ( h + 1 ) = x 1 ( h + 1 ) .
2.5.2) death process:
2.5.2.1) from raw parameter space selection state of dummy status equal probability, be designated as j d *, now state becomes
2.5.2.2) revise the weight of the previous status in parameter space:
&omega; j &prime; = &omega; j 1 - &omega; j d * , j &NotEqual; j d * ;
Thus ensure &Sigma; j &NotEqual; j d * &omega; j &prime; = 1 ;
2.5.2.3) calculate acceptance probability death_accept=min (1, death_A), wherein death_A is birth process step 2.5.1.3) in the inverse of birth_A;
2.5.2.4) (0, a 1) distribution random numbers U is produced dif, U d< death_accept, then by state delete, status number K ( h + 1 ) 2 = K &prime; ; Otherwise parameter space initial value will be retained, namely x 2 ( h + 1 ) = x 1 ( h + 1 ) .
2.6) Gibbs sampling undated parameter space is utilized obtain the h time iteration result x ( h + 1 ) = ( K ( h + 1 ) , &omega; j ( h + 1 ) , &mu; j ( h + 1 ) , &sigma; j ( h + 1 ) , z ( h + 1 ) , &beta; ( h + 1 ) ) , K (h+1), ω j (h+1), μ j (h+1), σ j (h+1), z (h+1)and β (h+1)be respectively status number K, state probability ω j, label is the final updated value that the average of the Gaussian process of j and variance, parameter hidden status switch z and hyper parameter β obtain after completing the h time iteration;
Wherein, K ( h + 1 ) = K ( h + 1 ) 2 , Parameter with hyper parameter β (h+1)upgrade as follows:
&omega; ( h + 1 ) ~ D i r i c h l e t ( &delta; 1 + n 1 , ... , &delta; K + n K ( h + 1 ) )
&mu; j ( h + 1 ) ~ N { &sigma; j ( h + 1 ) 2 - 2 &Sigma; i : z i ( h + 1 ) 2 = j y i + &kappa; &xi; &sigma; j ( h + 1 ) 2 - 2 n j + &kappa; , ( &sigma; j ( h + 1 ) 2 - 2 n j + &kappa; ) - 1 }
&sigma; j ( h + 1 ) - 2 ~ &Gamma; { &alpha; + 1 2 n j , &beta; ( h ) + 1 2 &Sigma; i : z i ( h + 1 ) 2 = j ( y i - &mu; j ( h + 1 ) ) 2 } ;
( z i ( h + 1 ) = j ) ~ &omega; j ( h + 1 ) &sigma; j ( h + 1 ) exp { - ( y i - &mu; j ( h + 1 ) ) 2 2 &sigma; j ( h + 1 ) 2 }
&beta; ( h + 1 ) ~ &Gamma; ( g + &kappa; &alpha; , f + &Sigma; j &sigma; j ( h + 1 ) - 2 )
Wherein Γ () represents gamma function, n jbe j=1,2 ..., K (h+1)number, for the hidden status switch z of parameter completes the renewal after the birth death process of the h time iteration middle label is the value of i.
2.7) h=h+1, if h < is N, repeats 2.4)-2.6);
2.8) for parameter space iteration result x (1), x (2)...., x (N), M uneven iteration before removing, obtains x (M+1), x (M+2)...., x (N), to wherein parameter K (M+1), K (M+2)..., K (N)make histogram, obtain maximum k op, be optimum state number, as shown in Figure 4;
2.9) at parameter space x (M+1), x (M+2)...., x (N)in, for K (i)=k op, i=M+1, M+2 ..., the parameter (ω of N (i), μ (i), σ (i)), obtain mathematic expectaion (ω, μ, σ), be the Gaussian parameter under Markov model different conditions;
2.10) utilize 2.3) in the likelihood function and 2.9 of the hidden status switch z of parameter) in Gaussian parameter under the Markov model different conditions obtained, obtain hidden status switch t=1,2 ... n, and obtain state-transition matrix P further and be:
P(i,j)=N ij/N i
Wherein, P (i, j) represents the probability jumping to state j from state i, N ijfor the number of state j is transferred to, N from state i ifor be in the number of state i, i, j=1,2 ..., n.
The state-transition matrix that S103 utilizes step S102 to obtain generates Markov state sequence z t, t=1,2 ... n, specifically comprises:
3.1) generate (0, a 1) uniform random number u, be used for determining new state;
3.2) when setting models be in state k (k=1,2 ...., k op), cumulative probability is denoted as β k, provided by following formula:
&beta; k = &Sigma; j = 1 k p k j ;
Wherein p kjrepresent the probability forwarding state j from state k-hop to;
3.3) condition then jumping to state i is:
β i≤u≤β i+1
Determined next step state by this condition, and then produce the Markov state sequence z on emulation link t, t=1,2 ... n.
Step S104: the status switch produced under Gaussian Distribution Parameters and step S103 under each state utilizing step S102 to obtain, produces the sequences y that decays at random t, t=1,2 ... n, that is:
y t ~ N ( &mu; z t , &sigma; z t ) ;
Wherein represent z t, t=1,2 ... the average of state corresponding to n and variance.
So far the simulation of article on plasma sheath cover channel model is completed.
As can be known from Fig. 5 and Fig. 6, the probability density function simulation value in first-order statistics characteristic and input value match, and illustrate that modeling result has higher accuracy.
For a person skilled in the art, according to above technical scheme and design, various corresponding change and distortion can be made, and all these change and distortion all should be included within the protection range of the claims in the present invention.

Claims (9)

1. reenter a dynamic plasma sheath cover Markov channel modeling method, it is characterized in that, described method comprises following steps:
Under S1 input specified conditions, signal is through the decay sequence of plasma sheath cover;
S2 sets up the multimode Markov channel model of plasma sheath cover, according to the plasma sheath cover decay sequence of step S1 input, obtain channel model parameters by reversible saltus step Markov Monte carlo algorithm, comprise Gaussian Distribution Parameters under Markov channel model status number, channel model state-transition matrix and each state;
The state-transition matrix that S3 utilizes step S2 to obtain generates Markov state sequence z t, t=1,2 ... n, n are sequence length;
The Markov state sequence produced under Gaussian Distribution Parameters and step S3 under each state that S4 utilizes step S2 to obtain, produces corresponding random decay sequence simulation value y t, t=1,2 ... n, n are sequence length.
2. according to claim 1ly reenter dynamic plasma sheath cover Markov channel modeling method, it is characterized in that, step S2 is specifically implemented as follows:
2.1) each state of Markov channel plasma sheath overlapped is modeled as Gaussian process, and the first-order statistics characteristic of channel represents with mixed Gauss model:
Y = &Sigma; j = 1 K &omega; j N ( &mu; j , &sigma; j 2 ) ;
Wherein Y represents the probability density function of channel, and K is the number of Gaussian process in the status number of markoff process and mixed Gauss model, ω jrepresent state probability, namely label is the weight of Gaussian density function shared by Mixture Model Probability Density Function of j, and has represent that label is the Gaussian process of j, and represent that label is average and the variance of the Gaussian process of j, wherein j=1,2 ..., K;
2.2) Confirming model parameter (K, ω j, μ j, σ j) prior distribution: status number K obeys Poisson distribution, μ jgaussian distributed, i.e. μ j~ N (ξ, κ -1), and σ jobey inverse gamma distribution, namely wherein β obeys gamma distribution, i.e. β ~ G (g, f), mixed Gaussian process weights omega jform ω vector, and ω obeys Dirichlet distribution, i.e. ω ~ D (δ 1, δ 2...., δ k), wherein (ξ, κ, g, f, δ) is transcendent distributed constant, is definite value, and δ is (δ 1..., δ k) set;
2.3) each parameter of initialization model, comprising: init state number K (1), Gaussian Distribution Parameters under each state prior distribution hyper parameter (ξ, κ, g, f, δ) and birth probability b kwith POD d k=1-b k, k is the label of status number; Birth probability and POD obey binomial distribution, and set iterations N and suppose uneven iterations M, and M gets initialization iteration count h; In addition, increase the hidden status switch z of parameter, its initialization is determined by following formula:
z (1) i=argmax{Pr(z i=j)}i=1,2....nj=1,2,...K;
Pr ( z i = j ) = &omega; j 1 2 &pi; &sigma; j e - ( y i - &mu; j ) 2 2 &sigma; j 2 ;
Wherein, y i, i=1,2 ... .n is signal attenuation sequence, and n is the length of sequence, z ifor in the hidden status switch z of parameter, label is the value of i, z i (1)it is then its initialization value;
2.4) the split degree process of the h time iteration is entered: produce (0, a 1) equally distributed random number U1, if U1 < is b k, then fission process is entered; Otherwise, enter merging process; Parameter space is updated to x 1 ( h + 1 ) = ( K ( h + 1 ) 1 , &omega; j ( h + 1 ) 1 , &mu; j ( h + 1 ) 1 , &sigma; j ( h + 1 ) 1 , z ( h + 1 ) 1 , &beta; ( h ) ) , Wherein be respectively status number K, state probability ω j, label is that the average of the Gaussian process of j and variance, the hidden status switch z of parameter complete the renewal after the split degree process of the h time iteration, hyper parameter β because do not upgrade in fission process, so retain the h-1 time iteration complete after updated value β (h);
2.5) the birth death process of the h time iteration is entered: produce (0, a 1) uniform random number U2, if U2 < is b k, then enter birth process, otherwise enter death process; Parameter space is updated to x 2 ( h + 1 ) = ( K ( h + 1 ) 2 , &omega; j ( h + 1 ) 2 , &mu; j ( h + 1 ) 2 , &sigma; j ( h + 1 ) 2 , z ( h + 1 ) 2 , &beta; ( h ) ) , Wherein be respectively status number K, state probability ω j, label is that the average of the Gaussian process of j and variance, the hidden status switch z of parameter complete the renewal after the birth death process of the h time iteration, hyper parameter β because do not upgrade in merging process, so retain the h-1 time iteration complete after updated value β (h);
2.6) Gibbs sampling undated parameter space is utilized obtain the h time iteration result x ( h + 1 ) = ( K ( h + 1 ) , &omega; j ( h + 1 ) , &mu; j ( h + 1 ) , &sigma; j ( h + 1 ) , z ( h + 1 ) , &beta; ( h + 1 ) ) , K (h+1), ω j (h+1), μ j (h+1), σ j (h+1), z (h+1)and β (h+1)be respectively status number K, state probability ω j, label is the final updated value that the average of the Gaussian process of j and variance, parameter hidden status switch z and hyper parameter β obtain after completing the h time iteration;
2.7) h=h+1, if h < is N, repeats 2.4)-2.6);
2.8) for parameter space iteration result x (1), x (2)...., x (N), M uneven iteration before removing, obtains x (M+1), x (M+2)...., x (N), to wherein parameter K (M+1), K (M+2)..., K (N)make histogram, obtain maximum k op, be optimum state number;
2.9) at parameter space x (M+1), x (M+2)...., x (N)in, for K (i)=k op, i=M+1, M+2 ..., the parameter (ω of N (i), μ (i), σ (i)), obtain mathematic expectaion (ω, μ, σ), be the Gaussian parameter under Markov model different conditions;
2.10) utilize 2.3) in the likelihood function and 2.9 of the hidden status switch z of parameter) in Gaussian parameter under the Markov model different conditions obtained, obtain hidden status switch t=1,2 ... n, and obtain state-transition matrix P further and be:
P(i,j)=N ij/N i
Wherein, P (i, j) represents the probability jumping to state j from state i, N ijfor the number of state j is transferred to, N from state i ifor be in the number of state i, i, j=1,2 ..., n.
3. according to claim 2ly reenter dynamic plasma sheath cover Markov channel modeling method, it is characterized in that, step 2.4) in, described fission process is specific as follows:
2.4.1.1) the parameter space x finally obtained after the h-1 time iteration being completed (h), from j=1,2 ..., K (h)stochastic choice state j s *, parameter is with j s1, j s2represent the label of the rear state of division, after division, status number becomes K '=K (h)+ 1, K (h)and z (h)be the final updated value of status number K and the hidden status switch z of parameter obtained after the h-1 time iteration completes, and be respectively the h-1 time iteration complete after state j s *ω j, μ jand σ jupdated value;
2.4.1.2) three random number u are produced 1, u 2, u 3, meet:
u 1~beta(2,2)u 2~beta(2,2)u 3~beta(1,1);
Beta is beta distribution, state weight and corresponding state Gaussian Distribution Parameters after division:
&omega; j s 1 &prime; = &omega; j s * ( h ) * u 1 &omega; j s 2 &prime; = &omega; j s * ( h ) * ( 1 - u 1 )
&mu; j s 1 &prime; = &mu; j s 1 ( h ) - u 2 &sigma; j s * ( h ) &omega; j s 1 &prime; &omega; j s 2 &prime; &mu; j s 2 &prime; = &mu; j s 2 ( h ) + u 2 &sigma; j s * ( h ) &omega; j s 2 &prime; &omega; j s 1 &prime; ;
&sigma; j s 1 &prime; = u 3 ( 1 - u 2 2 ) &sigma; j s * ( h ) &omega; j s * ( h ) &omega; j s 1 &prime; &sigma; j s 2 &prime; = ( 1 - u 3 ) ( 1 - u 2 2 ) &sigma; j s * ( h ) &omega; j s * ( h ) &omega; j s 2 &prime;
with represent state j respectively s1and j s2μ is obtained after the h-1 time corresponding iteration completes jupdated value;
2.4.1.3) the hidden status switch of undated parameter, will be originally sample point according to following probability assignments to j s1and j s2:
p a l l o c = &omega; j s 1 &prime; &sigma; j s 1 &prime; exp ( - ( y i - &mu; j s 1 &prime; ) 2 2 &sigma; j s 1 &prime; 2 ) &omega; j s 1 &prime; &sigma; j s 1 &prime; exp ( - ( y i - &mu; j s 1 &prime; ) 2 2 &sigma; j s 1 &prime; 2 ) + &omega; j s 2 &prime; &sigma; j s 2 &prime; exp ( - ( y i - &mu; j s 2 &prime; ) 2 2 &sigma; j s 2 &prime; 2 ) ;
P allocdistribute to j s1probability, q=1-P allocdistribute to j s2probability, the hidden status switch of the parameter after renewal is designated as z ';
2.4.1.4) acceptance probability split_accept=min (1, split_A) is calculated, wherein:
s p l i t _ A = ( l i k e l i h o o d r a t i o ) &times; p ( K ( h ) + 1 ) p ( K ( h ) ) &times; ( K ( h ) + 1 ) &times; &omega; j s 1 &prime; &delta; j s 1 - 1 + l 1 &omega; j s 2 &prime; &delta; j s 2 - 1 + l 2 &omega; j s * ( h ) &delta; j s * - 1 + l 1 + l 2 B ( &delta; j s * , K&delta; j s * ) &times; &kappa; 2 &pi; exp &lsqb; - 1 2 &kappa; { ( &mu; j s 1 &prime; - &xi; ) 2 + ( &mu; j s 2 &prime; - &xi; ) 2 - ( &mu; j s * ( h ) - &xi; ) 2 } &rsqb; &times; &beta; ( h ) &alpha; &Gamma; ( &alpha; ) ( &sigma; j s 1 &prime; 2 &sigma; j s 2 &prime; 2 &sigma; j s * &prime; 2 ) - &alpha; - 1 exp ( - &beta; ( &sigma; j s 1 &prime; - 2 + &sigma; j s 2 &prime; - 2 - &sigma; j s * ( h ) - 2 ) ) &times; d K ( h ) + 1 b K ( h ) P a l l o c &times; { g 2 , 2 ( u 1 ) g 2 , 2 ( u 2 ) g 1 , 1 ( u 3 ) } - 1 &times; &omega; j s * ( h ) | &mu; j s 1 &prime; - &mu; j s 2 &prime; | &sigma; j s 1 &prime; 2 &sigma; j s 2 &prime; 2 u 2 ( 1 - u 2 2 ) u 3 ( 1 - u 3 ) &sigma; j s * ( h ) 2 ;
Wherein, (likehoodratio) represents the likelihood ratio of the front state value of the state value after division and division, p (K (h)) and p (K (h)+ 1) being respectively status number is K (h)and K (h)probability when+1, l 1, l 2for sample y ibelong to state j s1, j s2count, B (|) represents beta function, g p, qrepresent beta (p, q) probability density function, with represent that status number is K respectively (h)and K (h)the birth rate of+1 and the death rate, represent status number j respectively s1, j s2with corresponding δ;
2.4.1.5) (0, a 1) uniform random number U is produced sif, U s< split_accept, then model parameter space by raw parameter space x (h)in parameter be updated to (K ', ω ' js1, ω ' js2, μ ' js1, μ ' js2, σ ' js1, σ ' js2, z '); Otherwise parameter space initial value will be retained, namely x 1 ( h + 1 ) = x ( h ) .
4. the dynamic plasma sheath that reenters according to Claims 2 or 3 overlaps Markov channel modeling method, it is characterized in that, step 2.4) in, described merging process is as follows:
2.4.2.1) for parameter space x (h), from j=1,2 ..., K (h)middle Stochastic choice two states, are designated as j c1and j c2, state parameter is state after merging is designated as j c *, then status number becomes K '=K (h)-1;
2.4.2.2) originally state j will be belonged to c1and j c2sample point unified to incorporate into as state j c *, the status switch after merging is designated as z ', that is:
z i &prime; = j c * , i &Element; { i : z i ( h ) = j c 1 } &cup; { i : z i ( h ) = j c 2 } ;
2.4.2.3) by following equation group Renewal model parameter
&omega; j c * &prime; &omega; j c 1 ( h ) + &omega; j c 2 ( h )
&omega; j c * &prime; &mu; j c * &prime; = &omega; j c 1 ( h ) &mu; j c 1 ( h ) + &omega; j c 2 ( h ) &mu; j c 2 ( h ) ;
&omega; j c * &prime; ( &mu; j c * &prime; 2 + &sigma; j c * &prime; 2 ) = &omega; j c 1 ( h ) ( &mu; j c 1 ( h ) 2 + &sigma; j c 1 ( h ) 2 ) + &omega; j c 2 ( h ) ( &mu; j c 2 ( h ) 2 + &sigma; j c 2 ( h ) 2 )
2.4.2.4) calculate acceptance probability combine_accept=min (1, combine_A), wherein combine_A is step 2.4.1.4) inverse of split_A in fission process;
2.4.2.5) (0, a 1) uniform random number U is produced cif, U c< combine_accept, then model parameter space by parameter space x (h)in parameter be updated to otherwise parameter space initial value will be retained, namely
5. according to claim 2ly reenter dynamic plasma sheath cover Markov channel modeling method, it is characterized in that, step 2.5) in, described birth process is specific as follows:
2.5.1.1) utilize step 2.2) the prior distribution of hypothesis produce a new stochastic regime j b *, now status number becomes new state parameter is:
&omega; j b * &prime; ~ b e t a ( 1 , K &prime; ) , &mu; j b * &prime; ~ N ( &xi; , &kappa; - 1 ) , &sigma; j b * &prime; - 2 ~ &Gamma; ( &alpha; , &beta; ) ;
2.5.1.2) revise the weight of the previous status in parameter space:
&omega; j &prime; = &omega; j ( h + 1 ) 1 ( 1 - &omega; j b * ) , j = 1 , 2 ... . K ( h + 1 ) 1 ;
Thus make &omega; j b * + &Sigma; j = 1 K &omega; j &prime; = 1 ;
2.5.1.3) acceptance probability birth_accept=min (1, birth_A) is calculated, wherein
b i r t h _ A = p ( K ( h + 1 ) 1 + 1 ) p ( K ( h + 1 ) 1 ) 1 B ( K ( h + 1 ) 1 &delta; j b * , &delta; j b * ) &omega; j b * &prime; &delta; j b * - 1 ( 1 - &omega; j b * &prime; ) n + K ( h + 1 ) 1 &delta; j b * - K ( h + 1 ) 1 ( K ( h + 1 ) 1 + 1 ) &times; d K ( h + 1 ) 1 + 1 ( K 0 + 1 ) b K ( h + 1 ) 1 1 g 1 , K ( h + 1 ) 1 ( &omega; j b * &prime; ) ( 1 - &omega; j b * &prime; ) K ( h + 1 ) 1 ;
Wherein, k 0before representing newly-increased state status number in parameter space, with represent that status number is respectively with probability, with represent that status number is respectively with birth rate and the death rate;
2.5.1.4) (0, a 1) distribution random numbers U is produced bif, U b< birth_accept, then will increase state status number otherwise parameter space initial value will be retained, namely x 2 ( h + 1 ) = x 1 ( h + 1 ) .
6. the dynamic plasma sheath that reenters according to claim 2 or 5 overlaps Markov channel modeling method, it is characterized in that, step 2.5) in, described death process is specific as follows:
2.5.2.1) from raw parameter space selection state of dummy status equal probability, be designated as j d *, now state becomes
2.5.2.2) revise the weight of the previous status in parameter space:
&omega; j &prime; = &omega; j 1 - &omega; j d * , j &NotEqual; j d * ;
Thus ensure &Sigma; j &NotEqual; j d * &omega; j &prime; = 1 ;
2.5.2.3) calculate acceptance probability death_accept=min (1, death_A), wherein death_A is birth process step 2.5.1.3) in the inverse of birth_A;
2.5.2.4) (0, a 1) distribution random numbers U is produced dif, U d< death_accept, then by state delete, status number otherwise parameter space initial value will be retained, namely x 2 ( h + 1 ) = x 1 ( h + 1 ) .
7. according to claim 2ly reenter dynamic plasma sheath cover Markov channel modeling method, it is characterized in that, step 2.6) in, the h time iteration result x ( h + 1 ) = ( K ( h + 1 ) , &omega; j ( h + 1 ) , &mu; j ( h + 1 ) , &sigma; j ( h + 1 ) , z ( h + 1 ) , &beta; ( h + 1 ) ) In, K ( h + 1 ) = K ( h + 1 ) 2 , Parameter with hyper parameter β (h+1)upgrade as follows:
&omega; ( h + 1 ) ~ D i r i c h l e t ( &delta; 1 + n 1 , ... , &delta; K + n K ( h + 1 ) )
&mu; j ( h + 1 ) ~ N { &sigma; j ( h + 1 ) 2 - 2 &Sigma; i : z i ( h + 1 ) 2 = j y i + &kappa; &xi; &sigma; j ( h + 1 ) 2 - 2 n j + &kappa; , ( &sigma; j ( h + 1 ) 2 - 2 n j + &kappa; ) - 1 }
&sigma; j ( h + 1 ) - 2 ~ &Gamma; { &alpha; + 1 2 n j , &beta; ( h ) + 1 2 &Sigma; i : z i ( h + 1 ) 2 = j ( y i - &mu; j ( h + 1 ) ) 2 } ;
( z i ( h + 1 ) = j ) ~ &omega; j ( h + 1 ) &sigma; j ( h + 1 ) exp { - ( y i - &mu; j ( h + 1 ) ) 2 2 &sigma; j ( h + 1 ) 2 }
&beta; ( h + 1 ) ~ &Gamma; ( g + &kappa; &alpha; , f + &Sigma; j &sigma; j ( h + 1 ) - 2 )
Wherein Γ () represents gamma function, n jbe number, for the hidden status switch z of parameter completes the renewal after the birth death process of the h time iteration middle label is the value of i.
8. according to claim 1 and 2ly reenter dynamic plasma sheath cover Markov channel modeling method, it is characterized in that, described step S3 detailed process is as follows:
3.1) generate (0, a 1) uniform random number u, be used for determining new state;
3.2) when setting models be in state k (k=1,2 ...., k op), cumulative probability is denoted as β k, provided by following formula:
&beta; k = &Sigma; j = 1 k p k j ;
Wherein p kjrepresent the probability forwarding state j from state k-hop to;
3.3) condition then jumping to state i is:
β i≤u≤β i+1
Determined next step state by this condition, and then produce the Markov state sequence z on emulation link t, t=1,2 ... n.
9. according to claim 1ly reenter dynamic plasma sheath cover Markov channel modeling method, it is characterized in that, in step S4, decay sequences y at random tsimulation value is specially:
y t ~ N ( &mu; z t , &sigma; z t ) ;
Wherein represent Markov state sequence z t, t=1,2 ... the average of state corresponding to n and variance.
CN201510513924.6A 2015-08-20 2015-08-20 One kind reenters dynamic plasma sheath set Markov channel modeling method Active CN105071881B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201510513924.6A CN105071881B (en) 2015-08-20 2015-08-20 One kind reenters dynamic plasma sheath set Markov channel modeling method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201510513924.6A CN105071881B (en) 2015-08-20 2015-08-20 One kind reenters dynamic plasma sheath set Markov channel modeling method

Publications (2)

Publication Number Publication Date
CN105071881A true CN105071881A (en) 2015-11-18
CN105071881B CN105071881B (en) 2017-09-12

Family

ID=54501179

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201510513924.6A Active CN105071881B (en) 2015-08-20 2015-08-20 One kind reenters dynamic plasma sheath set Markov channel modeling method

Country Status (1)

Country Link
CN (1) CN105071881B (en)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105846915A (en) * 2016-05-12 2016-08-10 清华大学 High dynamic information channel communication verifying system
CN107181565A (en) * 2017-05-04 2017-09-19 天津大学 A kind of bit-level mixed channel modeling method
CN107483131A (en) * 2017-06-30 2017-12-15 西安电子科技大学 The double satellite combined channel Markov state method for generating sequence of high-speed aircraft
CN110837688A (en) * 2019-09-30 2020-02-25 西安电子科技大学 Total field/scattered field plane wave source generation method in plasma sheath 3D-FDTD modeling
CN111835445A (en) * 2020-06-30 2020-10-27 西安理工大学 Dynamic plasma sheath channel amplitude and phase joint modeling method
CN112737659A (en) * 2020-12-03 2021-04-30 西安电子科技大学 Remote measuring channel incoherent large-scale SIMO processing method, system and application
CN113589191A (en) * 2021-07-07 2021-11-02 江苏毅星新能源科技有限公司 Power failure diagnosis system and method

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104378170A (en) * 2014-10-27 2015-02-25 西安电子科技大学 Near space dynamic plasma sheath channel modeling and simulating method

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104378170A (en) * 2014-10-27 2015-02-25 西安电子科技大学 Near space dynamic plasma sheath channel modeling and simulating method

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
D.L.POTTER: "Introduction of the PIRATE program for parametric reentry vehicle plasma effects studies", 《PRESENTED AT THE 37TH AIAA PLASMADYN. LASERS CONF.,SAN FRANCISCO,CA,USA》 *
M.YANG ETC.: "Density-modulation effects on the propagation of an electromagnetic wave in a plasma", 《PHYS. PLASMAS》 *
Y.MIN,ETC.: "propagation of electromagnetic signals in the time-varying plasma", 《ACTA PHYS. SINICA》 *

Cited By (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105846915A (en) * 2016-05-12 2016-08-10 清华大学 High dynamic information channel communication verifying system
CN107181565A (en) * 2017-05-04 2017-09-19 天津大学 A kind of bit-level mixed channel modeling method
CN107181565B (en) * 2017-05-04 2020-05-05 天津大学 Bit-level mixed channel modeling method
CN107483131A (en) * 2017-06-30 2017-12-15 西安电子科技大学 The double satellite combined channel Markov state method for generating sequence of high-speed aircraft
CN107483131B (en) * 2017-06-30 2021-06-15 西安电子科技大学 Method for generating high-speed aircraft double-satellite combined channel Markov state sequence
CN110837688A (en) * 2019-09-30 2020-02-25 西安电子科技大学 Total field/scattered field plane wave source generation method in plasma sheath 3D-FDTD modeling
CN111835445A (en) * 2020-06-30 2020-10-27 西安理工大学 Dynamic plasma sheath channel amplitude and phase joint modeling method
CN111835445B (en) * 2020-06-30 2022-06-07 西安理工大学 Dynamic plasma sheath channel amplitude and phase joint modeling method
CN112737659A (en) * 2020-12-03 2021-04-30 西安电子科技大学 Remote measuring channel incoherent large-scale SIMO processing method, system and application
CN112737659B (en) * 2020-12-03 2022-04-08 西安电子科技大学 Remote measuring channel incoherent large-scale SIMO processing method, system and application
CN113589191A (en) * 2021-07-07 2021-11-02 江苏毅星新能源科技有限公司 Power failure diagnosis system and method
CN113589191B (en) * 2021-07-07 2024-03-01 郴州雅晶源电子有限公司 Power failure diagnosis system and method

Also Published As

Publication number Publication date
CN105071881B (en) 2017-09-12

Similar Documents

Publication Publication Date Title
CN105071881A (en) Re-entry dynamic plasma sheath Markov channel modeling method
Gnedin et al. Multi-dimensional cosmological radiative transfer with a Variable Eddington Tensor formalism
Zheng et al. Primal mixed solution to unconfined seepage flow in porous media with numerical manifold method
Toro et al. ADER schemes for scalar non-linear hyperbolic conservation laws with source terms in three-space dimensions
CN104933245B (en) Naval reactor Shielding design optimization method based on artificial neural network and genetic algorithms
Langelage et al. Heavy dense QCD and nuclear matter from an effective lattice theory
CN104376581A (en) Gaussian mixture unscented particle filter algorithm employing adaptive resampling
Büyükçorak et al. Lognormal mixture shadowing
CN104052557A (en) Method for modeling Nakagami repeated fading channel
Dunbar et al. Ensemble inference methods for models with noisy and expensive likelihoods
CN105354860A (en) Box particle filtering based extension target CBMeMBer tracking method
CN105740204A (en) Low-frequency-band earth conductivity rapid inversion method under irregular terrain
Qaddouri et al. Optimized Schwarz methods with an overset grid for the shallow-water equations: preliminary results
Cao et al. Iterative learning control for multi-agent systems with impulsive consensus tracking
Bi et al. Three-dimensional cooperative positioning for Internet of Things provenance
Li et al. UAV trajectory optimization for spectrum cartography: A PPO approach
CN105095555A (en) Non-divergence smoothing processing method and apparatus for velocity field
CN104951617A (en) Antenna optimization design method based on area decomposition method
Jiao et al. Method for accurate and efficient eye diagram prediction of nonlinear high-speed links
CN102055694B (en) Non-linear system identification method based on particle swarm
Konrad Multifidelity monte carlo sampling in plasma microturbulence analysis
CN102263599B (en) Intelligent antenna array simulation method and apparatus thereof
Zhou et al. Time‐varying non‐line‐of‐sight (NLOS) satellite navigation channel modelling and simulation using multiple point‐scatterers
Attia Advanced Sampling Methods for Solving Large-Scale Inverse Problems
Qiang et al. Bayesian filter based on grid filtration and its application to Multi-UAV tracking

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant