CN103428770A - Flow distribution method in multi-connection parallel-transmission of heterogeneous wireless network - Google Patents

Abstract

A flow distribution method in multi-connection parallel-transmission of a heterogeneous wireless network includes: according to transmission performance parameters of different connections in the heterogeneous wireless network, estimating transmission delays of data packets passing through the different connections, calculating the optimized flow distribution rate by a genetic simulated annealing algorithm, and meanwhile, taking the maximum transmission rate, the average transmission delay, the bit error rate and the maximum retransmission frequency of each network resource into consideration comprehensively to provide the available flow distribution method so as to realize the flow distribution with minimized maximum transmission delay difference between the connections. According to the implinit parallelism of the genetic algorithm and the climbing property of the annealing algorithm, the calculation complexity is reduced effectively, solving speed and convergence speed are increased greatly, properties of the algorithms and consumption of time are balanced further by setting the convergence threshold. Compared with the other flow distribution methods, the flow distribution method is more adaptable to specific application scenes of the heterogeneous wireless network, utilizes the network connection resources fully and increases the throughput.

Description

Flow allocation method in the multi-link parallel transmission of heterogeneous wireless network

Technical field

The present invention relates to a kind of assignment of traffic technology of the parallel transmission for heterogeneous wireless network, definite says, relates to the flow rate ratio distribution method of a plurality of connection parallel transmissions in a kind of heterogeneous wireless network, belongs to the mobile communication technology field in radio communication.

Background technology

Following cordless communication network system will be one kind of multiple heterogeneous networks (such as comprising: the standards such as 2G/2.5G/3G, LTE, WLAN, WPAN) and deposit, complex system collaborative and that constantly merge.These heterogeneous networks in overlay area, the characteristic of the aspect such as bandwidth, reliability, cost and fail safe is different.And they will meet terminal use's QoS demand jointly in following one period long term in the mode complemented each other.

The development of multi-mode terminals, made use simultaneously multiple heterogeneous network parallel transmission data become may with reality.Multi-link parallel transmission technology under heterogeneous network can take full advantage of Internet resources, meets the future communications business demand.Yet, because the various connected modes of heterogeneous wireless network all there are differences aspect technical characterstic and load capacity, therefore, multi-link parallel transmission certainly will bring problems.

Referring to Fig. 1, introduce the problem that the multi-link parallel transmission of heterogeneous wireless network exists: because the time delay that packet connects (being three in figure) arrival receiving terminal by difference is different, caused the data packet disorder phenomenon in multi-link parallel transmission system very serious, receiving terminal must be set up the buffer area that reorders, and for packet, at this, resequences.The packet of a certain particular number is waited for than it and is numbered the time that little packet all arrives, is exactly the time delay that reorders of this packet in the buffer area that reorders.The time delay that reorders has increased the end-to-end time delay of Business Stream, has had a strong impact on the performance of upper strata host-host protocol.

In multi-link parallel transmission transmission, a major issue is exactly how to reduce the extra time delay that reorders.Each assignment of traffic ratio between connecting, can reduce the time delay that reorders effectively when optimizing multi-link parallel transmission, is a kind of desirable approach that solves the delay problem that reorders.

In existing multi-link parallel transmission technology, take and reduce in the flow allocation method that the time delay that reorders between each connection is target, the method that following three kinds of better performances are arranged: based on media interviews, control MAC(Media Access Control) flow allocation method that layer is measured, flow allocation method and business separation mixed method based on feedback.Below introduce respectively it:

(1) flow allocation method of measuring based on the MAC layer: the method is that packet is passed through to the allocation criterion of the propagation delay time of this connection as connection resource, propagation delay time by this connection of periodic measurement, dynamically adjust shunt ratio, make it with to be connected propagation delay time inversely proportional, thereby the balance that keeps load, and then the time delay that reorders of reduction receiving terminal.But the method has only been considered two kinds of situations that connect parallel transmission, when linking number surpasses two, inapplicable.

(2) flow allocation method based on feedback: the method proposes a kind of dynamic traffic distribution ratio computational methods based on the receiving terminal feedback information, at first, transmitting terminal periodically sends the probe bag, then, metrical information according to receiving terminal, the data packet transmission time delay of link layer is periodically returned to transmitting terminal, and calculates for shunt ratio.And the shunting ratio calculated should make the data packet transmission time delay of link layer be consistent, the packet that so just can reduce the receiving terminal buffer area time delay that reorders.Assignment of traffic ratio in the method can periodically be adjusted along with the change of wireless connections capacity, and the method, after the ratio polymerization, can effectively reduce the time delay that reorders, still, and the polymerization time that it need to be longer.

(3) mixed method of business separation: this mixed method is comprised of following two kinds of methods: according to the optimal subset hop algorithm of the algorithm of ideal model computed segmentation ratio and the available connection of selection based on Fuzzy Multiple Attribute Decision Making Theory, by the ration of division theoretical value to calculating and the error between the actual transmissions value and the contrast between the error threshold value, when error during lower than error threshold, select the ideal model algorithm, and during higher than error threshold, only select the incompatible parallel transmission of optimal subset of available connection when error.This algorithm can significantly reduce the time delay that reorders, and still, because there is deviation in ideal model ration of division algorithm and actual conditions wherein, makes this algorithm performance in actual applications be suppressed.

Summary of the invention

In view of this, the purpose of this invention is to provide the flow allocation method in the multi-link parallel transmission of a kind of heterogeneous wireless network, the method is in the multi-link parallel transmission of heterogeneous wireless network, utilize the transmission performance parameter of each connection, calculate the optimization assignment of traffic ratio between each connection, and business data flow is configured to a plurality of sub data flows of parallel transmission according to this optimization assignment of traffic ratio, to reduce the receiving terminal time delay that reorders.Optimization flow rate ratio computational methods of the present invention are the poor MMDD(Minimize the of the maximum delay Maximum Delay Difference that minimize between connection) be optimization aim, by the propagation delay time that makes each connection, approach as much as possible, to reduce the time delay that reorders in the multi-link parallel transmission of heterogeneous wireless network.Because this optimization problem is the NP-hard problem, the present invention proposes a kind of method for solving based on the genetic Annealing algorithm, the optimization assignment of traffic ratio of gained can effectively reduce the time delay that reorders, and compared with the existing methods, aspect average Packet Error Ratio and throughput performance, is also having more advantage.

In order to achieve the above object, the invention provides the flow allocation method in the multi-link parallel transmission of a kind of heterogeneous wireless network, it is characterized in that: according to the transmission performance parameter of each connection in heterogeneous wireless network, the propagation delay time that the data estimator bag connects by difference, introduce again Global Genetic Simulated Annealing Algorithm GASA(Genetic Algorithm with Simulated Annealing) calculation optimization assignment of traffic ratio, consider the maximum transmission rate of each Internet resources simultaneously, mean transit delay, a plurality of different performance parameters of the error rate and maximum retransmission, a kind of feasible flow allocation method is provided, to realize minimizing the assignment of traffic of maximum transmitted delay inequality between each connection, described method comprises following operating procedure:

(1) the transmission performance parameter of each connection in the test heterogeneous wireless network;

(2) according to transmission performance parameter and the business data flow overall transmission rate of each connection, determine the parameters initial value of optimizing assignment of traffic ratio computational methods;

(3) carrying out based on minimizing between connection maximum delay poor is the optimization assignment of traffic ratio computational methods of optimization aim;

(4) transmitting terminal is according to optimizing assignment of traffic ratio distribution service data flow.

In heterogeneous wireless network of the present invention, the essence of the optimization method of completing the square of the flow rate ratio of multi-link parallel transmission is a nature of nonlinear integral programming problem, therefore, this is a NP difficult problem (being the NP-hard problem) (non-deterministic polynomial hard).Approach optimum solution in order effectively to obtain above-mentioned problem, the present invention has adopted genetic Annealing algorithm GASA to solve.This is because genetic algorithm (GA) has good ability of searching optimum and the speed of solution, but its ability aspect Local Search is poor, and easily " precocity " occur.And simulated annealing (SA) can be found best solution, but it easily is absorbed in the trap of local optimum, simultaneously its ability of searching optimum a little less than.So two kinds of algorithms of GA and SA have very strong complementary, both combinations-GASA algorithm is the effective ways that solve at present the NP-hard problem.

Innovative characteristics of the present invention and key technology are: the GASA algorithm is introduced in multi-link parallel transmission optimization flow allocation method first, wherein, Genetic Algorithms is set to the main frame of parallel search, and simulated annealing SA processes for the sudden change of genetic algorithm.Namely: at first use GA search overhead allocative decision space, calculate under each allocative decision each maximum delay between connecting poor, according to the fitness function of poor this allocative decision of calculating of maximum delay; Again filial generation fitness function value and parent fitness function value are compared.Then, the strategy according to SA obtains the present flow rate allocation result.While meeting the algorithm end condition, the near-optimization result that resulting assignment of traffic scheme is exactly assignment of traffic.In a word, the present invention adopts the method for solving based on genetic simulated annealing (GASA), utilizes the performance of climbing the mountain of Implicit Parallelism and the annealing algorithm of genetic algorithm, can effectively reduce computation complexity, greatly accelerate to solve speed and convergence rate.In addition, by rational convergence threshold value further equalization algorithm performance and elapsed time again are set.

The present invention is based on and minimize the poor optimization aim of maximum delay between each connection, the flow allocation method equal with time delay between each is connected compared, be more suitable for the concrete application scenarios of heterogeneous wireless network, with the method that reduces parallel transmission quantity and reduce the time delay that reorders, more take full advantage of available network connection resource, throughput is larger.Therefore, the present invention has good popularizing application prospect.

The accompanying drawing explanation

Fig. 1 is that the structure of heterogeneous wireless network forms schematic diagram.

Fig. 2 is the flow allocation method operating procedure flow chart that the present invention is based on multi-link parallel transmission in the heterogeneous wireless network of genetic Annealing algorithm.

Embodiment

For making the purpose, technical solutions and advantages of the present invention clearer, below in conjunction with accompanying drawing, the present invention is described in further detail.

Flow allocation method in the multi-link parallel transmission of heterogeneous wireless network of the present invention, it is the transmission performance parameter according to each connection in heterogeneous wireless network, the propagation delay time that the data estimator bag connects by difference, introduce again Global Genetic Simulated Annealing Algorithm GASA calculation optimization assignment of traffic ratio, consider a plurality of different performance parameters of maximum transmission rate, mean transit delay, the error rate and the maximum retransmission of each Internet resources simultaneously, provide a kind of feasible flow allocation method, to realize minimizing the assignment of traffic of maximum transmitted delay inequality between each connection.

Referring to Fig. 2, introduce the concrete operation step of the inventive method:

Step 1, the transmission performance parameter of each connection in the test heterogeneous wireless network.The content of operation of this step is: measure the sum M of all available connections under current network conditions, the overall transmission rate λ of business data flow; Measure again the transmission performance parameter of each available connection i: maximum transmission rate r _i, mean transit delay d _i, error rate q _iWith maximum retransmission e _iIn formula, positive integer i is that heterogeneous wireless network can be used the sequence number connected, and its maximum is M.

Step 2, according to transmission performance parameter and the business data flow overall transmission rate of each connection, determine the parameters initial value of optimizing assignment of traffic ratio computational methods.

This step comprises following content of operation:

(21) according to transmission performance parameter and the business data flow overall transmission rate of each connection, determine assignment of traffic ratio result vector

Initializing constraint be:

Wherein, ω _iBe the assignment of traffic ratio that the i bar connects, and have: 0≤ω _i≤ 1 He

The overall transmission rate that λ is business data flow, for guaranteeing system stability, distribute to the substream of data speed ω that the i bar connects _iλ should be no more than the maximum transmission rate r of this available connection _i, i.e. ω _iλ≤r _iBecause the smallest allocation unit in practical application is 1 packet, therefore the overall transmission rate λ of business data flow and distribute to the substream of data speed ω that the i bar connects _iλ can only get positive integer N, i.e. ω _iλ ∈ N, λ ∈ N.

(22) according to assignment of traffic ratio result vector

Constraints determine search volume U, then population scale PS, maximum iteration time GN, crossing-over rate P are set _cWith aberration rate P _m, simulated annealing initial temperature T ₀With coefficient of temperature drop r.

Because being initial phase, therefore initial temperature that current iteration time counter gn=1 and Current Temperatures are simulated annealing is set, be T=T simultaneously ₀.Wherein, positive integer PS is the individuality sum in population, and the maximum times that positive integer GN is iterative computation, as one of end condition calculated; Crossing-over rate P _cWith aberration rate P _mIn P all mean probability P robability, its subscript c and m are respectively the initial of the English crossover of intersection and variation mutation, P _cAnd P _mSpan be respectively [0.4,0.99] and [0.0001,0.1]; Simulated annealing initial temperature T ₀For set point, the span of coefficient of temperature drop r is (0,1).

(23) first determine the population initial value in Global Genetic Simulated Annealing Algorithm: select at random PS assignment of traffic ratio result vector value from the U of search volume, form initial population P={P ₁, P ₂..., P _Ps.Wherein, P is the initial of English population Population, and d in this population is individual

Be d the corresponding assignment of traffic ratio of population at individual result vector; ω _DiBe in d the corresponding assignment of traffic ratio of population at individual result vector, be assigned to the ratio that substream of data speed that the i bar connects accounts for the overall transmission rate of business data flow, the assignment of traffic ratio that the i bar connects, and 1≤d≤PS, 1≤i≤M.

Step 3, carry out based on minimize each connect between maximum delay poor be the optimization assignment of traffic ratio computational methods of optimization aim.This step comprises following content of operation:

(31) from current population P randomly combination of two go out Q to as the male parent list, to the every a pair of male parent group { P in this male parent list _s, P _t, respectively according to crossing-over rate P _cAfter carrying out interlace operation, then according to aberration rate P _mCarry out mutation operation, obtain filial generation { C _s, C _t.

In formula, current population quantity P={P ₁, P ₂..., P _Ps, subscript s and t are respectively two different male parent numberings choosing at random, and its maximum is PS, and s<t.Filial generation { C _s, C _tBe { P _s, P _tAfter the crossover and mutation operation, the new individuality of generation, i.e. next generation's individuality of current individuality, parent; The initial that C is English filial generation Children, with current individuality, be that parent p is corresponding.Male parent is the abbreviation of parent sample, and the parent sample is all individualities in parent, because population scale is PS, therefore the male parent list

(32) calculate respectively every couple of filial generation { C _s, C _tWith respect to its parent { P _s, P _tTwo fitness function increment sizes: Δ f=Fit (C _s)-Fit (P _s) and Δ f=Fit (C _t)-Fit (P _t); In formula, f is that fitness function Fit (x) writes a Chinese character in simplified form.

This fitness function Fit (P _d) calculating be emphasis, its computational methods can be divided into following three content of operation:

(32A) when assignment of traffic ratio result vector be P _dThe time,, calculate according to the following equation the propagation delay time T that the i bar connects _i(ω _Di):

$T_i\left({\mathrm{\&omega;}}_{\mathrm{di}}\right)=\underset{n=0}{\overset{q_i}{\mathrm{\&Sigma;}}}\left[{\left(f(e_i,{\mathrm{\&omega;}}_{\mathrm{di}}\mathrm{\&lambda;})\right)}^n(1-f(e_i,{\mathrm{\&omega;}}_{\mathrm{di}}\mathrm{\&lambda;}))(n+1)d_i\right]+{\left(f(e_i,{\mathrm{\&omega;}}_{\mathrm{ji}}\mathrm{\&lambda;})\right)}^{q_i+1}(q_i+1)d_i;$ In formula, the overall transmission rate that λ is business data flow (unit is packet/s); ω _DiFor assignment of traffic ratio result vector is P _dThe time, the assignment of traffic ratio that the i bar connects; ω _Diλ is assigned to the substream of data speed that the i bar connects; The substream of data speed be assigned in the connection of i bar is ω _DiPacket Error Ratio during λ

d _iBe the mean transit delay that the i bar connects, q _iAnd e _iBe respectively the error rate and maximum retransmission that the i bar connects; Positive integer i is the sequence number that heterogeneous wireless network connects, and its maximum is M.

(32B) when assignment of traffic ratio result vector be P _dThe time, calculate the time delay difference of relatively transmitting between any two connections in the connection of M bar $D_{\mathrm{diff}}\left(\stackrel{\&RightArrow;}{{\mathrm{\&omega;}}_d}\right):D_{\mathrm{diff}}\left(\stackrel{\&RightArrow;}{{\mathrm{\&omega;}}_d}\right)=\&ForAll;\left|{\mathrm{\&Delta;T}}_{\mathrm{ij}}({\mathrm{\&omega;}}_{\mathrm{di}},{\mathrm{\&omega;}}_{\mathrm{dj}})\right|=\&ForAll;|T_i\left({\mathrm{\&omega;}}_{\mathrm{di}}\right)-T_j\left({\mathrm{\&omega;}}_{\mathrm{dj}}\right)|.$

In formula, positive integer i is two different sequence numbers that are connected with j, and its maximum is M, and i<j; The initial that D is English propagation delay time Delay, Diff is writing a Chinese character in simplified form of English difference Difference, means relative difference herein, therefore

Δ means the difference of two numbers, Δ T _Ij(ω _Di, ω _Dj) for working as assignment of traffic ratio result vector, be P _dThe time, two connect i and j, distinguish corresponding assignment of traffic ratio ω _DiWith corresponding assignment of traffic ratio ω _DjBetween the propagation delay time difference, the propagation delay time T that namely the i bar connects _i(ω _Di) the propagation delay time T that is connected with the j bar _j(ω _Dj) difference, Δ T _Ij(ω _Di, ω _Dj) value be Arbitrary Digit, but its absolute value | Δ T _Ij(ω _Di, ω _Dj) |>=0.Function

Mean to get the arbitrary value in x.

(32C) calculate P _dFitness function Fit (Pd): it is P that fitness function is got assignment of traffic ratio result vector _dThe time any two connections between the relative peaked inverse of propagation delay time difference, $\mathrm{Fit}\left(P_d\right)={\left(\mathrm{Max}\left(D_{\mathrm{diff}}\left(\stackrel{\&RightArrow;}{{\mathrm{\&omega;}}_d}\right)\right)\right)}^{-1}.$

Wherein, function F it (x) means the fitness function of variable x in Global Genetic Simulated Annealing Algorithm, the initial that Fit is English fitness Fitness, and Max is the initial of English maximum Maximum, Max (x) means the maximum of getting x;

Mean that assignment of traffic ratio result vector is P _jThe time any two connections between relative propagation delay time difference, (x) ^-1The inverse that means x, i.e. 1/x.

It should be noted that: top with P _dFor example illustrates the fitness function Fit (P in this step _d) computational methods, and Fit (C _d) computational methods identical therewith, its difference only needs to change a variable P _dFor variable C _dGet final product.

(33) judge whether each Δ f is greater than 0, i.e. formula Δ f>0 whether set up; If, meaning that newborn filial generation fitness function value is greater than parent, i.e. filial generation more conforms than parent, now accepts at once this offspring individual C _sOr C _tFor current kind of group members, and the corresponding individuality of replacement parent; If not, Δ f≤0, mean that the fitness function value of filial generation is less than or equal to parent, now, do not abandon at once this filial generation, becomes current kind of group members but prepare to accept it with the acceptance probability of setting: namely first calculate C _sOr C _tAcceptance probability

After generation is positioned at the equally distributed pseudo random number random (0,1) on interval [0,1], then judge inequality P(Δ f)>whether random (0,1) set up; If accept this filial generation C _sOr C _tFor the member of current population, replace corresponding parent; Otherwise, abandon this new offspring individual, still retain the individual P of original parent _sOr P _tWherein, acceptance probability P (Δ f) depends on that fitness increment size, e are math constant, and T is Current Temperatures.

In this step (33), temperature T is higher when initial, therefore

Numerical value close to 1, have larger probability to accept the poor filial generation of fitness value.But, along with the reduction of Current Temperatures T,

Numerical value constantly diminish, finally almost no longer accept the poor filial generation of fitness function value.Adopt this acceptance probability as the tactful benefit of receiving filial generation to be: the trap of avoiding search procedure to be absorbed in prematurely local optimum, can obtain overall optimization assignment of traffic ratio vector with larger probability, this operating procedure has embodied the thought of simulated annealing.

(34) judge whether to meet one of following two stopping criterion for iteration:

(a) whether iterations reaches set point number, and the current iteration number of times has reached maximum iteration time: gn > GN; Perhaps

(b) whether the average fitness function difference of parent and filial generation is less than convergence threshold Th, and its computing formula is: filial generation is with respect to the average fitness increment Fit (f (C)) of parent-Fit (f (P))<Th;

As long as meet wherein any one stopping criterion for iteration, just finish the computational process of this assignment of traffic ratio, export the corresponding assignment of traffic ratio of the individuality result vector of fitness maximum in current population P, then using this numerical value as optimizing assignment of traffic ratio vector; If not, carry out subsequent step (35);

Wherein, the C={C in end condition (b) ₁, C ₂... C _PsMean whole individualities of filial generation, P={P ₁, P ₂..., P _PsMean whole individualities of parent (being current population); Filial generation C in Fit (f (C)) does not have subscript, means the average fitness function that it is filial generation, and its computational methods are:

Similarly, the parent P in Fit (f (P)) does not have subscript yet, means the average fitness function that it is parent, and its computational methods are $\mathrm{Fit}\left(f\left(P\right)\right)=\underset{i=1}{\overset{\mathrm{ps}}{\mathrm{\&Sigma;}}}\frac{\mathrm{fit}\left(f\left(P_{\mathrm{ps}}\right)\right)}{\mathrm{PS}}.$

In this step, it is 4～6% that the present invention arranges convergence threshold Th, when the average fitness increment less than 4～6% of population, and the termination of iterations calculating operation, export the corresponding assignment of traffic ratio of the individuality result vector of fitness function value maximum in current population, as optimizing the assignment of traffic ratio.Meanwhile, for preventing that computing time is long, greatest iteration algebraically GN is set for auxiliary end condition, be that current iteration number of times gn is while surpassing GN, also stop the iterative computation operation, the corresponding assignment of traffic ratio of the individuality result vector of fitness function value maximum in the current population of same output, as optimizing the assignment of traffic ratio.

(35) once just upgrade Current Temperatures one time because of the every evolution of population, also upgrade the current iteration number of times simultaneously; Therefore Current Temperatures T reduces gradually with coefficient of temperature drop r, i.e. T=T * r; The current iteration number of times increases progressively gradually: i.e. gn=gn+1; Now, return to execution step (31); Wherein, adopt index cooling strategy, the interval of coefficient of temperature drop r is (0,1).

Step 4, transmitting terminal is according to optimizing assignment of traffic ratio distribution service data flow.This step comprises following content of operation: transmitting terminal, in sending buffer area, is divided into a plurality of packet subflows according to the optimization assignment of traffic ratio calculated by business data packet stream, and by different connection parallel transmissions to receiving terminal; When packet arrives receiving terminal, because there being the data packet disorder phenomenon, therefore these packets are introduced in the buffer area that reorders, wait for and reordering, then receiving end receives.

The present invention has carried out the test of Multi simulation running embodiment, and the result of test is successfully, has realized goal of the invention.

The foregoing is only preferred embodiment of the present invention, in order to limit the present invention, within the spirit and principles in the present invention not all, any modification of making, be equal to replacement, improvement etc., within all should being included in the scope of protection of the invention.

Claims (9)

1. the flow allocation method in the multi-link parallel transmission of heterogeneous wireless network, it is characterized in that: according to the transmission performance parameter of each connection in heterogeneous wireless network, the propagation delay time that the data estimator bag connects by difference, introduce again Global Genetic Simulated Annealing Algorithm GASA calculation optimization assignment of traffic ratio, consider the maximum transmission rate of each Internet resources simultaneously, mean transit delay, a plurality of different performance parameters of the error rate and maximum retransmission, a kind of feasible flow allocation method is provided, to realize minimizing the assignment of traffic of maximum transmitted delay inequality between each connection, described method comprises following operating procedure:

(1) the transmission performance parameter of each connection in the test heterogeneous wireless network;

(2) according to transmission performance parameter and the business data flow overall transmission rate of each connection, determine the parameters initial value of optimizing assignment of traffic ratio computational methods;

(3) carrying out based on minimizing between connection maximum delay poor is the optimization assignment of traffic ratio computational methods of optimization aim;

(4) transmitting terminal is according to optimizing assignment of traffic ratio distribution service data flow.

2. method according to claim 1, it is characterized in that: the content of operation of described step (1) is: measure the sum M of all available connections under current network conditions, the overall transmission rate λ of business data flow; Measure again the transmission performance parameter of each available connection: maximum transmission rate r _i, mean transit delay d _i, error rate q _iWith maximum retransmission e _iIn formula, positive integer i is that heterogeneous wireless network can be used the sequence number connected, and its maximum is M.

3. method according to claim 1, it is characterized in that: described step (2) comprises following content of operation:

(21) according to transmission performance parameter and the business data flow overall transmission rate of each connection, determine assignment of traffic ratio result vector

Initializing constraint be:

Wherein, ω _iBe the assignment of traffic ratio that the i bar connects, and have: 0≤ω _i≤ 1 He

The overall transmission rate that λ is business data flow, for guaranteeing system stability, distribute to the substream of data speed ω that the i bar connects _iλ should be no more than the maximum transmission rate r of this connection _i, i.e. ω _iλ≤r _iBecause the smallest allocation unit in practical application is 1 packet, therefore the overall transmission rate λ of business data flow and distribute to the substream of data speed ω that the i bar connects _iλ can only get positive integer N, i.e. ω _iλ ∈ N, λ ∈ N;

(22) according to assignment of traffic ratio result vector

Constraints determine search volume U, then population scale PS, maximum iteration time GN, crossing-over rate P are set _cWith aberration rate P _m, simulated annealing initial temperature T ₀With coefficient of temperature drop r; Because being initial phase, therefore initial temperature that current iteration time counter gn=1 and Current Temperatures are simulated annealing is set, be T=T simultaneously ₀Wherein, positive integer PS is the individuality sum in population, and the maximum times that positive integer GN is iterative computation, as one of end condition calculated; Crossing-over rate P _cWith aberration rate P _mIn P all mean probability P robability, its subscript c and m are respectively the initial of the English crossover of intersection and variation mutation, P _cAnd P _mSpan be respectively [0.4,0.99] and [0.0001,0.1]; Simulated annealing initial temperature T ₀For set point, the span of coefficient of temperature drop r is (0,1);

(23) first determine the population initial value in Global Genetic Simulated Annealing Algorithm: select at random PS assignment of traffic ratio result vector value from the U of search volume, form initial population P={P ₁, P ₂..., P _Ps; Wherein, P is the initial of English population Population, and d in this population is individual

Be d the corresponding assignment of traffic ratio of population at individual result vector; ω _DiBe in d the corresponding assignment of traffic ratio of population at individual result vector, be assigned to the ratio that substream of data speed that the i bar connects accounts for the overall transmission rate of business data flow, the assignment of traffic ratio that the i bar connects, and 1≤d≤PS, 1≤i≤M.

4. method according to claim 1, it is characterized in that: described step (3) comprises following content of operation:

(31) from current population P randomly combination of two go out Q to as the male parent list, then to the every a pair of male parent group { P in this male parent list _s, P _t, respectively according to crossing-over rate P _cAfter carrying out interlace operation, then according to aberration rate P _mCarry out mutation operation, obtain filial generation { C _s, C _t;

In formula, current population quantity P={P ₁, P ₂..., P _Ps, subscript s and t are respectively two different male parent numberings choosing at random, and its maximum is PS, and s<t; Filial generation { C _s, C _tBe { P _s, P _tAfter the crossover and mutation operation, the new individuality of generation, i.e. next generation's individuality of current individuality; The initial that C is English filial generation Children, with parent, be that current individual p is corresponding; Described male parent is the abbreviation of parent sample, and the parent sample is all individualities in parent, because population scale is PS, therefore the male parent list

(32) calculate respectively every couple of filial generation { C _s, C _tWith respect to its parent { P _s, P _tTwo fitness function increment sizes: Δ f=Fit (C _s)-Fit (P _s) and Δ f=Fit (C _t)-Fit (P _t); In formula, f is that fitness function Fit (x) writes a Chinese character in simplified form;

(33) judge whether each Δ f is greater than 0, i.e. formula Δ f>0 whether set up; If, meaning that newborn filial generation fitness function value is greater than parent, i.e. filial generation more conforms than parent, now accepts at once this offspring individual C _sOr C _tFor current kind of group members, and the corresponding individuality of replacement parent; If not, Δ f≤0, mean that the fitness function value of filial generation is less than or equal to parent, now, do not abandon at once this filial generation, becomes current kind of group members but prepare to accept it with the acceptance probability of setting: namely first calculate C _sOr C _tAcceptance probability

After generation is positioned at the equally distributed pseudo random number random (0,1) on interval [0,1], then judge inequality P(Δ f)>whether random (0,1) set up; If accept this filial generation C _sOr C _tFor the member of current population, replace corresponding parent; Otherwise, abandon this new offspring individual, still retain the individual P of original parent _sOr P _tWherein, acceptance probability P (Δ f) depends on that fitness increment size, e are math constant, and T is Current Temperatures;

(34) judge whether to meet one of following two stopping criterion for iteration:

(a) whether iterations reaches set point number, and the current iteration number of times has reached maximum iteration time: gn > GN; Perhaps

(b) whether the average fitness function difference of parent and filial generation is less than convergence threshold Th, and its computing formula is: filial generation is with respect to the average fitness increment Fit (f (C)) of parent-Fit (f (P))<Th;

As long as meet wherein any one stopping criterion for iteration, just finish the computational process of this assignment of traffic ratio, export the corresponding assignment of traffic ratio of the individuality result vector of fitness maximum in current population P, then using this numerical value as optimizing assignment of traffic ratio vector; If not, carry out subsequent step (35);

Wherein, the C={C in end condition (b) ₁, C ₂... C _PsMean whole individualities of filial generation, P={P ₁, P ₂..., P _PsMean parent, be whole individualities of current population; Filial generation C in Fit (f (C)) does not have subscript, means the average fitness function that it is filial generation, and its computational methods are:

Similarly, the parent P in Fit (f (P)) does not have subscript yet, means the average fitness function that it is parent, and its computational methods are $\mathrm{Fit}\left(f\left(P\right)\right)=\underset{i=1}{\overset{\mathrm{ps}}{\mathrm{\&Sigma;}}}\frac{\mathrm{fit}\left(f\left(P_{\mathrm{ps}}\right)\right)}{\mathrm{PS}}.$

(35) once just upgrade Current Temperatures one time because of the every evolution of population, also upgrade the current iteration number of times simultaneously; Therefore Current Temperatures T reduces gradually with coefficient of temperature drop r, i.e. T=T * r; The current iteration number of times increases progressively gradually: i.e. gn=gn+1; Now, return to execution step (31); Wherein, the interval of coefficient of temperature drop r is (0,1).

5. method according to claim 4, is characterized in that: the fitness function Fit (P in described step (32) _d) calculating comprise following content of operation:

(32A) when assignment of traffic ratio result vector be P _dThe time,, calculate according to the following equation the propagation delay time T that the i bar connects _i(ω _Di):

$T_i\left({\mathrm{\&omega;}}_{\mathrm{di}}\right)=\underset{n=0}{\overset{q_i}{\mathrm{\&Sigma;}}}\left[{\left(f(e_i,{\mathrm{\&omega;}}_{\mathrm{di}}\mathrm{\&lambda;})\right)}^n(1-f(e_i,{\mathrm{\&omega;}}_{\mathrm{di}}\mathrm{\&lambda;}))(n+1)d_i\right]+{\left(f(e_i,{\mathrm{\&omega;}}_{\mathrm{ji}}\mathrm{\&lambda;})\right)}^{q_i+1}(q_i+1)d_i;$ In formula, the overall transmission rate that λ is business data flow; ω _DiFor assignment of traffic ratio result vector is P _dThe time, the assignment of traffic ratio that the i bar connects; ω _Diλ is assigned to the substream of data speed that the i bar connects; The substream of data speed be assigned in the connection of i bar is ω _DiPacket Error Ratio during λ

d _iBe the mean transit delay that the i bar connects, q _iAnd e _iBe respectively the error rate and maximum retransmission that the i bar connects; Positive integer i is that heterogeneous wireless network can be used the sequence number connected, and its maximum is M;

(32B) when assignment of traffic ratio result vector be P _dThe time, calculate the time delay difference of relatively transmitting between any two connections in the connection of M bar $D_{\mathrm{diff}}\left(\stackrel{\&RightArrow;}{{\mathrm{\&omega;}}_d}\right):D_{\mathrm{diff}}\left(\stackrel{\&RightArrow;}{{\mathrm{\&omega;}}_d}\right)=\&ForAll;\left|{\mathrm{\&Delta;T}}_{\mathrm{ij}}({\mathrm{\&omega;}}_{\mathrm{di}},{\mathrm{\&omega;}}_{\mathrm{dj}})\right|=\&ForAll;|T_i\left({\mathrm{\&omega;}}_{\mathrm{di}}\right)-T_j\left({\mathrm{\&omega;}}_{\mathrm{dj}}\right)|;$

In formula, positive integer i is two different sequence numbers that are connected with j, and its maximum is M, and i<j; The initial that D is English propagation delay time Delay, Diff is writing a Chinese character in simplified form of English difference Difference, means relative difference herein, therefore

Δ T _Ij(ω _Di, ω _Dj) for working as assignment of traffic ratio result vector, be P _dThe time, two connect i and j, distinguish corresponding assignment of traffic ratio ω _DiWith corresponding assignment of traffic ratio ω _DjBetween the propagation delay time difference, the propagation delay time T that namely the i bar connects _i(ω _Di) the propagation delay time T that is connected with the j bar _j(ω _Dj) difference, Δ T _Ij(ω _Di, ω _Dj) value be Arbitrary Digit, but its absolute value | Δ T _Ij(ω _Di, ω _Dj) |>=0; Function Mean to get the arbitrary value in x;

(32C) calculate P _dFitness function Fit (P _d): it is P that fitness function is got assignment of traffic ratio result vector _dThe time any two connections between the relative peaked inverse of propagation delay time difference, $\mathrm{Fit}\left(P_d\right)={\left(\mathrm{Max}\left(D_{\mathrm{diff}}\left(\stackrel{\&RightArrow;}{{\mathrm{\&omega;}}_d}\right)\right)\right)}^{-1};$

Wherein, function F it (x) means the fitness function of variable x in Global Genetic Simulated Annealing Algorithm, the initial that Fit is English fitness Fitness, and Max is the initial of English maximum Maximum, Max (x) means the maximum of getting x; Mean that assignment of traffic ratio result vector is P _jThe time any two connections between relative propagation delay time difference, (x) ^-1The inverse that means x, i.e. 1/x.

6. method according to claim 5, is characterized in that: the fitness function Fit (C in described step (32) _d) calculating operation content and Fit (P _d) the calculating operation content identical, its difference is just by corresponding variable P _dBe replaced by variable C _d.

7. method according to claim 4 is characterized in that: in described step (33), the temperature T when initial is higher, therefore Numerical value close to 1, have larger probability to accept the poor filial generation of fitness value; But, along with the reduction of Current Temperatures T, Numerical value constantly diminish, finally almost no longer accept the poor filial generation of fitness function value; Adopt this acceptance probability as the tactful benefit of receiving filial generation to be: the trap of avoiding search procedure to be absorbed in prematurely local optimum, can obtain overall optimization assignment of traffic ratio vector with larger probability, this operating procedure has embodied the thought of simulated annealing.

8. method according to claim 4, it is characterized in that: in described step (34), it is 4～6% that convergence threshold Th is set, when the average fitness increment less than 4～6% of population, the termination of iterations calculating operation, export the corresponding assignment of traffic ratio of the individuality result vector of fitness function value maximum in current population, as optimizing the assignment of traffic ratio; Simultaneously, for preventing that computing time is long, greatest iteration algebraically GN is set for auxiliary end condition, be that current iteration number of times gn is while surpassing GN, also stop the iterative computation operation, the corresponding assignment of traffic ratio of the individuality result vector of fitness function value maximum in the current population of same output, as optimizing the assignment of traffic ratio.

9. method according to claim 1, it is characterized in that: described step (4) comprises following content of operation: transmitting terminal is in sending buffer area, according to the optimization assignment of traffic ratio calculated, business data packet stream is divided into to a plurality of packet subflows, and by different connection parallel transmissions to receiving terminal; When packet arrives receiving terminal, because there being the data packet disorder phenomenon, therefore these packets are introduced in the buffer area that reorders, wait for and reordering, then receiving end receives.

Images