CN105335784B - A kind of method of the optimal soft protection of dsp system of selection based on genetic algorithm - Google Patents

Abstract

The invention belongs to the technical field of DSP system soft protection, and specifically relates to a method for selecting an optimal DSP system soft protection based on a genetic algorithm, comprising step 1, establishing a target optimization model; step 2, adopting a genetic algorithm to solve the target model; 3. Combination output of protection methods Establish the target model of DSP system error rate and resource parameters, the method of selecting the optimal DSP system soft protection based on genetic algorithm, and constrain the resource overhead; use genetic pairs to divide program blocks The benefits of using different protection methods are optimally solved, and the globally optimal combination of protection methods is obtained. The genetic algorithm starts to find the optimal solution from multiple points in space, and is not easy to fall into a local optimum; and for different combinations of protection methods, the genetic algorithm is used to The algorithm can perform parallel operation on the adaptive value, which can save the optimal solution search time.

Description

A Method of Selecting Optimal DSP System Soft Protection Based on Genetic Algorithm

technical field

The invention belongs to the technical field of DSP system soft protection, and in particular relates to a method for selecting an optimal DSP system soft protection based on a genetic algorithm.

Background technique

DSP chip, also known as digital signal processor, is a kind of microprocessor that is especially suitable for digital signal processing operations. Its main application is to realize various digital signal processing algorithms in real time and quickly; A technical discipline that uses digital methods to analyze, transform, filter, detect, modulate, demodulate, and perform fast algorithms on signals. But many people think that digital signal processing is mainly about digital filtering technology, discrete transform fast algorithm and spectral analysis method. With the development of digital circuit and system technology and computer technology, digital signal processing technology has also been developed accordingly, and its application fields are very extensive.

Digital Signal Processor (DSP), with its high computing speed, has been more and more widely used in modern technology products, especially in cutting-edge technology, aerospace and other fields, playing an increasingly important role. For aerospace technology, the application of DSP devices has to face the complex space environment, such as the radiation and impact of high-energy particles. Devices manufactured by special processes will weaken the relevant performance of the device to some extent, and the cost is high. With the continuous improvement of device operating speed and the improvement of device integration, the frequency of its interference by outer space particles is increasing. Therefore, in order to ensure the stable operation of the processor and the stable operation of the system, a series of tests will be carried out for DSP devices. Soft protection works.

During the operation of DSP systems in space, single event effects are prone to occur. In actual applications, there are many protection methods for DSP. It is often difficult for a single protection method to meet the requirements of the system for resources and reliability. Therefore, in actual design In general, a variety of protection methods are used to integrate each other to protect the system. The benefits and cost models of various protection methods are different, which makes it difficult to weigh in the process of selecting a protection method, and it is difficult for designers to intuitively know whether the currently selected scheme has room for optimization.

So how can these methods be combined so that the system can achieve the probability of allowing errors without taking up too many resources? Is the combined defense approach we use an optimal design?

In view of this situation, the present invention provides a kind of optimal design method of DSP soft protection scheme based on genetic algorithm, by modeling to system error rate and resource overhead (number of instructions), solve optimal solution by genetic algorithm, obtain The optimal protection combination method maximizes the performance of the system.

Contents of the invention

The purpose of the present invention is to overcome the problem in the prior art that the different benefit and cost models of various DSP system protection methods make it difficult to balance in the process of selecting a protection method.

For this reason, the invention provides a kind of method based on genetic algorithm selection optimal DSP system soft protection, comprises the steps:

Step 1: Establish a target optimization model;

Step 2, using a genetic algorithm to solve the target model;

Step 3: output the protection method combination.

The above step 1, establishing the target optimization model includes the following steps:

(1) _Divide the engineering code into _m blocks with sequential execution sequence, _denoted as X ₁ , X ₂ , X ₃ . is P _s , the error probability of program block X _s after being protected by the _{k sth} protection method is The total number of signals of program block X _s is recorded as N _s , the error propagation probability of the jth signal of program block X _s is recorded as P _s ^j , the number of code instructions of program block X _s is recorded as C _s , and the data size is recorded as D _s , the execution time is recorded as T _s , after adopting the k _sth protection method, the number of code instructions is The data size is recorded as The execution time is recorded as

Then the calculation formula of the system error rate P can be obtained as follows:

After obtaining the system protection, the calculation formula of the total code size C is as follows:

After obtaining the system protection, the calculation formula of the total data size D is as follows:

The calculation formula of the total execution time T after obtaining the system protection is as follows:

(2) Establish the objective function optimization model:

Among them, C _Z represents the maximum allowable code capacity, D _Z represents the maximum allowable data space size, T _Z represents the maximum allowable time delay, a, b, and c are resource weight coefficients, where a+b+c=1 , a≥0, b≥0, c≥0, the designer can allocate the values of a, b, and c according to the actual situation in the protection, so as to weigh the important relationship of the three resource reference quantities.

Get the optimization function equation system:

The second step above is to use the genetic algorithm to solve the target model, including the following steps:

(1) Determine the coding scheme;

For the total number of modules divided by the DSP code is m, there are n types of protection methods used in total. For any module, one of the n protection methods can be selected, or no protection can be selected. The mode that can be selected There are (n+1) kinds, represented by binary data, then the encoding length l should satisfy the following relationship:

2 ^l ≥ n+1

which is Therefore, for each solution, the length of the binary code is L=m×l, and the population size is N, generally N=10;

(2) Determine the fitness value function;

According to the coding law in (1), the corresponding relationship between each coding and the protection method of each module is determined, and the benefit and resource overhead of the corresponding protection method are queried. By setting the weight value of the resource reference quantity, it can be calculated The value of the objective function under the coding sample; the optimized function equations obtained in step 1 are used as the fitness value function; generally take Then the fitness value function is as follows:

(3) population initialization;

Under the premise of ensuring that there is no invalid code, generate N binary random numbers with a length of L, so as to encode each individual in the population, and complete the initialization of the population, and the length of the coded individual is L=m×l;

(4) Calculation of sample fitness value;

Bring the values of N individuals into the fitness function in (2), and calculate the fitness value function value of each group of codes respectively; among them, the sample fitness obtained by calculating the i (i=1, 2, 3...N) value The value is recorded as f _i ; if the initial value of the population does not satisfy the limiting conditions, the individual will be deleted from the population. In order to maintain the stability of the population size, a new sample will be generated in step (8);

(5) Termination judgment;

Construct a sample fitness value difference array with a length of N: compare the size of the sample fitness value obtained in (4), take out the minimum value f _min of the sample fitness value, and make a difference between all sample fitness values and f _min , and set The difference is stored in the sample fitness value difference array; if w consecutive 0s appear in the array, it is determined that this value is the optimal solution, and the encoded value is output, otherwise, continue to execute;

Among them, the w value is determined according to the population size N value, which can be set as

(6) Calculation of relative fitness value;

Calculate the size of the sample fitness value through (4), and use the following formula to get the relative fitness value of each initial coding value,

(7) Population reproduction;

The relative fitness value represents the proportion of the fitness value function value under various combination protection methods in the sample population; among them, the higher the proportion, the lower the fitness value function value of the corresponding sample value, and the better the benefit of the protection combination method ;Use the method of roulette to determine the new population based on the relative fitness value;

In the new population, the number B _i of the i-th individual is obtained by the following formula:

Among them, the function g(x) means that x is rounded to an integer; in this way, the number of each individual in the new population is calculated, B _i =0, the individual will be eliminated in the new population; if the number of individuals in the new population exceeds The number of the original population, the individual with the largest sample fitness value will be eliminated; if the number of individuals in the new population is less than the number of the original population, the individual with the smallest sample fitness value will be added to the new population;

(8) Genetic operator design;

Mutation is to randomly select a certain binary bit from the new population to invert, that is, the mutation operation, to obtain a new population member; take the mutation probability p _m , that is, to select N×p _m binary bits from the temporary population after hybridization to mutate, Generally take the mutation probability p _m =0.02;

Hybridization is to randomly select a certain number of individuals from the new population, randomly perform bit exchange, and obtain new population members; take the hybridization probability p _c , that is, select N×p _c individuals from the temporary population for hybridization, and obtain a new population after hybridization, Generally, the hybridization probability p _c =0.75p _m is taken;

When performing genetic operator operations, during the process of hybridization and mutation, invalid codes cannot appear; if so, re-execute the genetic operator calculations;

(9) Use the coding value of the new population as the initial value, and return to (2) for execution.

The above (2) determines the fitness value function, while satisfying a+b+c=1, a≥0, b≥0, c≥0

Under the premise, the weights a, b, and c are set according to the resource reference.

The above-mentioned step three, combined output of the protection method, is to decode the code obtained by the genetic algorithm, the code is grouped by l digits, numbered from left to right as 1 to m, and the binary number of each group is converted into a decimal system respectively The generated m decimal numbers are the protection schemes corresponding to each program block, and the combination of the protection schemes of all program blocks is the optimal protection scheme for the entire program.

Beneficial effects of the present invention: the method for selecting the optimal DSP system soft protection based on genetic algorithm provided by the present invention includes step 1, establishing a target optimization model; step 2, adopting genetic algorithm to solve the target model; step 3, The combined output of protection methods establishes the target model of DSP system error rate and resource parameters. The method of selecting the optimal soft protection of DSP system based on genetic algorithm also constrains the resource overhead; using genetic pairs to divide different protections between program blocks The use of the method benefits from the optimal solution to obtain the globally optimal combination of protection methods. The genetic algorithm starts to find the optimal solution from multiple points in space, and is not easy to fall into a local optimum; and for different combinations of protection methods, the genetic algorithm can Parallel calculation of the fitness value can save the time to find the optimal solution.

The present invention will be described in further detail below in conjunction with the accompanying drawings.

Description of drawings

Fig. 1 is the genetic algorithm flow chart of the method for selecting the optimal DSP system soft protection based on the genetic algorithm.

Detailed ways

In order to further illustrate the technical means and effects adopted by the present invention to achieve the intended purpose, the specific implementation, structural features and effects of the present invention will be described in detail below in conjunction with the accompanying drawings and examples.

Example 1:

In order to overcome the different benefit and cost models of various DSP system protection methods in the prior art, it is difficult to make a trade-off in the process of selecting a protection method.

For this reason, the invention provides a kind of method based on genetic algorithm selection optimal DSP system soft protection, comprises the steps:

Step 1: Establish a target optimization model;

Step 2, using a genetic algorithm to solve the target model;

Step 3: output the protection method combination.

The above step 1, establishing the target optimization model includes the following steps:

(1) _Divide the engineering code into _m blocks with sequential execution sequence, _denoted as X ₁ , X ₂ , X ₃ . is P _s , the error probability of program block X _s after being protected by the _{k sth} protection method is The total number of signals of program block X _s is recorded as N _s , the error propagation probability of the jth signal of program block X _s is recorded as P _s ^j , the number of code instructions of program block X _s is recorded as C _s , and the data size is recorded as D _s , and the execution time is recorded as T _s , after adopting the _ki -th protection method, the number of code instructions is The data size is recorded as The execution time is recorded as

Then the calculation formula of the system error rate P can be obtained as follows:

After obtaining the system protection, the calculation formula of the total code size C is as follows:

After obtaining the system protection, the calculation formula of the total data size D is as follows:

The calculation formula of the total execution time T after obtaining the system protection is as follows:

(2) Establish the objective function optimization model:

Among them, C _Z represents the maximum allowable code capacity, D _Z represents the maximum allowable data space size, T _Z represents the maximum allowable time delay, a, b, and c are resource weight coefficients, where a+b+c=1 , and a≥0, b≥0, c≥0, the designer can allocate the values of a, b, and c according to the actual situation in the protection, so as to weigh the important relationship of the three resource reference quantities.

Get the optimization function equation system:

As shown in Figure 1, the second step above is to use the genetic algorithm to solve the target model, including the following steps:

(1) Determine the coding scheme;

For the total number of modules divided by the DSP code is m, there are n types of protection methods used in total. For any module, one of the n protection methods can be selected, or no protection can be selected. The mode that can be selected There are (n+1) kinds, represented by binary data, then the encoding length l should satisfy the following relationship:

2 ^l ≥ n+1

which is Therefore, for each solution, the length of the binary code is L=m×l; the population size is N, generally N=10;

(2) Determine the fitness value function;

According to the coding law in (1), the corresponding relationship between each coding and the protection method of each module is determined, and the benefit and resource overhead of the corresponding protection method are queried. By setting the weight value of the resource reference quantity, it can be calculated The value of the objective function under the coding sample; the optimized function equations obtained in step 1 are used as the fitness value function; generally take Then the fitness value function is as follows:

(3) population initialization;

Under the premise of ensuring that there is no invalid code, generate N binary random numbers with a length of L, so as to encode each individual in the population, and complete the initialization of the population, and the length of the coded individual is L=m×l;

(4) Calculation of sample fitness value;

Bring the values of N individuals into the fitness function in (2), and calculate the fitness value function value of each group of codes respectively; among them, the sample fitness obtained by calculating the i (i=1, 2, 3...N) value The value is recorded as f _i ; if the initial value of the population does not satisfy the limiting conditions, the individual will be deleted from the population. In order to maintain the stability of the population size, a new sample will be generated in step (8);

(5) Termination judgment;

Construct a sample fitness value difference array with a length of N: compare the size of the sample fitness value obtained in (4), take out the minimum value f _min of the sample fitness value, and make a difference between all sample fitness values and f _min , and set The difference is stored in the sample fitness value difference array; if w consecutive 0s appear in the array, it is determined that this value is the optimal solution, and the encoded value is output, otherwise, continue to execute;

Among them, the w value is determined according to the population size N value, which can be set as

(6) Calculation of relative fitness value;

Calculate the size of the sample fitness value through (4), and use the following formula to get the relative fitness value of each initial coding value,

(7) Population reproduction;

The relative fitness value represents the proportion of the fitness value function value under various combination protection methods in the sample population; among them, the higher the proportion, the lower the fitness value function value of the corresponding sample value, and the better the benefit of the protection combination method ;Use the method of roulette to determine the new population based on the relative fitness value;

In the new population, the number B _i of the i-th individual is obtained by the following formula:

Among them, the function g(x) means that x is rounded to an integer; in this way, the number of each individual in the new population is calculated, B _i =0, the individual will be eliminated in the new population; if the number of individuals in the new population exceeds The number of the original population, the individual with the largest sample fitness value will be eliminated; if the number of individuals in the new population is less than the number of the original population, the individual with the smallest sample fitness value will be added to the new population;

(8) Genetic operator design;

Mutation is to randomly select a certain binary bit from the new population to invert, that is, the mutation operation, to obtain a new population member; take the mutation probability p _m , that is, to select N×p _m binary bits from the temporary population after hybridization to mutate, Generally take the mutation probability p _m =0.02;

Hybridization is to randomly select a certain number of individuals from the new population, randomly perform bit exchange, and obtain new population members; take the hybridization probability p _c , that is, select N×p _c individuals from the temporary population for hybridization, and obtain a new population after hybridization, Generally, the hybridization probability p _c =0.75p _m is taken;

When performing genetic operator operations, during the process of hybridization and mutation, invalid codes cannot appear; if so, re-execute the genetic operator calculations;

(9) Use the coding value of the new population as the initial value, and return to (2) for execution.

The above (2) determines the fitness value function, under the premise of satisfying a+b+c=1, while satisfying

On the premise of a+b+c=1, a≥0, b≥0, and c≥0, the weights a, b, and c are set according to the resource reference.

The above-mentioned step three, combined output of the protection method, is to decode the code obtained by the genetic algorithm, the code is grouped by l digits, numbered from left to right as 1 to m, and the binary number of each group is converted into a decimal system respectively The generated m decimal numbers are the protection schemes corresponding to each program block, and the combination of the protection schemes of all program blocks is the optimal protection scheme for the entire program.

Example 2:

Use the above method to divide the project code into 5 program blocks with sequential execution order. Each module can use 4 protection methods to calculate the optimal protection combination. The initial data is shown in Table 1 below.

Table 1 module related data and protection data

(1) Coding scheme

For the above data, the number of modules is 5, ie m=5, and each module has 4 protection methods, ie n=4, so, for each module, there are five options. Binary codes are used for these five options. Since two-digit binary codes can only represent four combinations at most, here we use three-digit binary codes. There are a total of eight combinations of three-bit binary codes, and we only use five of them. The specific combinations are shown in Table 2 below.

Table 2 Valid codes and corresponding methods

protection no protection method 1 Method 2 Method 3 Method 4 coding 000 001 010 011 100

For the combination output of the protection method of each module of the system, the code length L=M*1=5*3=15, that is, each effective combination output code is a 15-bit binary code.

(2) Adaptive value function

Here we use the objective function established above as the fitness value function, and regard each module as an independent module, that is, the error rate is accumulated. Take a=b=c=1/3, take constraint conditions C _Z =4000, D _Z =500, T _Z =5000.

(3) Population initialization

Here, N=10 is selected for the sample space, and 10 groups of 15-digit binary numbers are randomly selected, and every combination of three digits from left to right is required to be within the value range described in Table 2, otherwise, it will be regenerated after being eliminated. The randomly generated combinations are shown in Table 3 below.

Table 3 Initial population table

(4) Calculation of sample fitness value

For the initial population in Table 3, each group of codes is separated from left to right in groups of three, the first three-digit group represents module 1, and its protection method is determined according to Table 2, and then, by looking up Table 1 , substitute the error rate, protected code size, data size and execution time into the objective function optimization model, and calculate the fitness value of each encoding. Let's take the first group 001100010011100 as an example and split it. The split results are shown in Table 4.

Table 4 encoding split method

module 1 module 2 Module 3 Module 4 Module 5 001 100 010 011 100 method 1 Method 4 Method 2 Method 3 Method 4

By looking up Table 1, the combination of the first coded protection method is determined and substituted into the objective function optimization model to obtain f ₁ . The same method calculates each group of codes in the sample, and obtains the fitness value of each sample.

(5) Termination Judgment

Construct a sample fitness value difference array with length N. Compare the size of the sample fitness values obtained in step (4), take out the minimum value f _min of the sample fitness values, and make a difference between all sample fitness values and f _min , and store the difference in the sample fitness value difference array; If there are 8 consecutive 0s in the array, it is determined that this value is the optimal solution, and the encoded value is output; otherwise, continue to execute.

(6) Calculation of relative fitness value

Calculate the size of the sample fitness value by step (4), use the following formula to get the relative fitness value of each initial coding value,

(7) Population reproduction

The relative fitness value represents the ratio of the fitness value function values under various combined protection methods in the sample population. The higher the ratio, the lower the fitness value function value of the corresponding sample value, and the better the benefit of the protection combination method . Taking the relative fitness value as a reference, the roulette method is used to determine the new population. In the new population, the number B _i of the i-th individual is obtained by the following formula:

The function g(x) means that x is rounded to an integer; in this way, the number of each individual in the new population is calculated, B _i =0, and the individual will be eliminated in the new population. If the number of individuals in the new population exceeds the number of the original population, the individual with the largest sample fitness value is eliminated; if the number of individuals in the new population is less than the number of the original population, the individual with the smallest sample fitness value is added to the new population.

(8) Genetic operator design

The hybridization probability p _c =0.7 is taken, that is, 7 individuals in the population are hybridized to obtain a new population after hybridization. The mutation probability p _m =0.02 is taken, that is, one binary bit is flipped from the interim population after hybridization. When performing genetic operator operations, pay attention to the fact that invalid codes cannot appear during the process of hybridization and mutation; if so, the genetic operator calculation needs to be re-executed.

(9) Use the coding value of the new population as the initial value, and return to (2) for execution.

Finally, the optimal protection method is output.

The above content is a further detailed description of the present invention in conjunction with specific preferred embodiments, and it cannot be assumed that the specific implementation of the present invention is limited to these descriptions. For those of ordinary skill in the technical field of the present invention, without departing from the concept of the present invention, some simple deduction or replacement can be made, which should be regarded as belonging to the protection scope of the present invention.

Claims (5)

1. A kind of 1. method of the optimal soft protection of dsp system of selection based on genetic algorithm, it is characterised in that comprise the following steps:

   Step 1: establish objective optimization model；

   (1) engineering code is divided into some code blocks, each code block protected with different means of defences, and calculate and be It is total after total data size D, protection after total code size C, protection after the error rate P of system, protection to perform time T；

   (2) objective optimization equation group, i.e. objective optimization model are established using parameter P, C, D, T for having obtained：

   Wherein, k represents combines to the means of defence of different code block, model of f (k) representatives after means of defence combination k protection Desired value, C_ZRepresent maximum allowable code capacity, D_ZRepresent maximum allowable data space size, T_ZRepresent maximum allowable Time delay, a, b, c are the weight coefficient of resource, and meet a+b+c=1 simultaneously, a >=0, b >=0, c >=0, and designer can be with A, b, c value are allocated according to actual conditions in protection；

   Step 2: object module is solved using genetic algorithm；

   (1) binary coding scheme of solution is determined according to total block count m of division and total means of defence kind number n；

   (2) resource weight coefficient a, b, c value is determined, adaptive value letter is determined with reference to the objective optimization model established in step 1 Number；

   (3) initialization of population：All samples as in population determine initial code value；

   (4) according to encoding samples value and the adaptive value f for adapting to value function and calculating each sample proposed in (2)_i；

   (5) terminate and judge：According to the adaptive value f of each sample_iAnd the minimum value f in the adaptive value of all samples_minJudge i-th Whether encoding samples value is optimum code value, if it is exports the encoding samples value and stops algorithm, otherwise continues；

   (6) according to the adaptive value f of all samples in population_iCalculate the relative adaptation value of each sample

   (7) Population breeding：According to relative adaptation valuePopulation scale N calculates the quantity B of i-th of sample_i, by B_i=0 sample Deleted from population；

   (8) genetic operator, including crossover operator and mutation operator are performed to the sample in population, is substituted with caused new samples former The sample come produces population of new generation, returns to (2) and performs；

   Step 3: means of defence combination output；

   (1) the binary system optimum code obtained by step 2 is divided into m groups altogether using every l positions as one group, with m program block one by one It is corresponding, wherein,N is the kind number of means of defence；

   (2) caused m groups binary coding in (1) is converted into corresponding decimal number, the decimal number is each program block Corresponding means of defence, the means of defence are combined as optimal protectiving scheme；

   (3) optimal protectiving scheme, i.e., the combination of means of defence corresponding to caused decimal number in (2) are exported.
2. 2. the method for the optimal soft protection of dsp system of selection based on genetic algorithm as claimed in claim 1, it is characterised in that： It is described to comprise the following steps Step 1: establishing objective optimization model：

   (1) engineering code is divided into m blocks has the program block of priority execution sequence, is designated as X₁、X₂、X₃…X_m, program block it is each Kind means of defence is designated as k_s, X_sThe probability of error is designated as P_s, program block X_sUsing kth_sThe probability to be malfunctioned after kind means of defence protection ForProgram block X_sSignal sum be designated as N_s, program block X_sThe error propagation probability of j-th of signal is designated as P_s ^j, program block X_s Code command bar number scale be C_s, size of data is designated as D_s, perform the time be designated as T_s, using kth_sAfter kind means of defence, code Instruction strip number isSize of data is designated asThe execution time is designated as

   The calculating formula that system fault rate P can then be obtained is as follows：

   The calculating formula for obtaining total code size C after systematic protection is as follows：

   The calculating formula for obtaining total data size D after systematic protection is as follows：

   Obtain after systematic protection it is total perform time T calculating formula it is as follows：

   (2) objective function optimization model is established：

   Wherein, C_ZRepresent maximum allowable code capacity, D_ZRepresent maximum allowable data space size, T_ZRepresent maximum allowable Time delay, a, b, c are the weight coefficient of resource, and wherein a+b+c=1, and a >=0, b >=0, c >=0, designer can be anti- A, b, c value are allocated according to actual conditions in shield, so as to weigh the important relationship of three resource reference amounts, optimized Functional equation group：
3. 3. the method for the optimal soft protection of dsp system of selection based on genetic algorithm as claimed in claim 2, it is characterised in that： It is described Step 2: solved using genetic algorithm to object module, comprise the following steps：

   (1) encoding scheme is determined；

   Total number of modules for DSP code division is m, and the means of defence used altogether shares n kinds, for any one module Speech, can select one kind in this n kind means of defence, can also select not protect, you can means of defence for selection shares n + a kind, represented with binary data, then code length l, l is positive integer, and should meet relationship below：

   2^l≥n+1

   I.e.Therefore, for each solution, binary-coded length is L=m × l, population scale For N, N=10 is taken；

   (2) determine to adapt to value function；

   According to the encoding law in (1), it is determined that the corresponding relation of each coding and means of defence of each module, inquiry The benefit and resource overhead of corresponding means of defence, by being set to the weights of resource reference amount, it can calculate under the coded samples Target function value；Using the majorized function equation group obtained by step 1 as adaptation value function；Typically takeThen It is as follows to adapt to value function：

   (3) initialization of population；

   On the premise of no invalid code is ensured, the binary system random number that N number of length is L is produced, with this to each in population Individual is encoded, and completes initialization of population, encodes length L=m × l of individual；

   (4) sample adaptive value calculates；

   The fitness function that individual value is brought into (2), the adaptive value functional value of each group of coding is calculated respectively；Wherein, The sample adaptive value that i (i=1,2,3 ... N) individual value is calculated is designated as f_i；If it is unsatisfactory for qualifications in population initial value Person, will the individual deleted from population, in order to maintain the stabilization of population scale, new sample will produce in step (8)；

   (5) stop technology；

   The sample adaptive value difference array that one length of structure is N：Compare the size of the sample adaptive value obtained in (4), take out The minimum value f of sample adaptive value_min, and by all sample adaptive values and f_minIt is poor to make, and difference is stored in into sample adaptive value difference In array；If continuously occurring w 0 in array, judge that sample is optimal solution corresponding to the difference, exports the coding of the sample Value, otherwise, is continued executing with；

   Wherein, w values determine according to population scale N values, may be set to

   (6) relative adaptation value calculates；

   The size of sample adaptive value is calculated by (4), the relative adaptation value of each initial code value is drawn with below equation,

   (7) Population breeding；

   Relative adaptation value characterizes the ratio for adapting to value function value in sample population under various combination means of defences；Wherein, than Example is higher, illustrates that the adaptive value functional value of its corresponding sample value is lower, protects the benefit of combined method better；With relatively suitable It should be worth to refer to, new population is determined using the method for roulette；

   In new population, the quantity B of i-th of individual_iObtained with the formula of being calculated as below：

   Wherein function g (x) represents to carry out round number to x；So, number of each individual in new population is calculated Amount, B_i=0, the individual will be eliminated in new population；As quantity individual in new population exceed original seed group quantity, then by sample The maximum individual rejecting of this adaptive value；As the individual amount in new population be less than original seed group quantity, then by sample adaptive value most Small individual adds new population；

   (8) genetic operator designs；

   Variation is to randomly choose certain binary digit from new population to be negated, i.e. mutation operation, obtain new population into Member；Take mutation probability p_m, i.e., N × p is selected from the interim population after hybridization_mIndividual binary digit enters row variation, typically takes variation general Rate p_m=0.02；

   Hybridization randomly selects a number of individual from new population, enters line position exchange at random, obtains new population member；Take Probability of crossover p_c, i.e., N × p is selected in interim population_cIndividual is hybridized, the new population after being hybridized, and typically takes hybridization general Rate p_c=0.75p_m；

   When carrying out genetic operator computing, invalid code can not occur during hybridization and variation；If there is then holding again Row genetic operator calculates；

   (9) using the encoded radio of new population as initial value, (2) is returned and are performed.
4. 4. the method for the optimal soft protection of dsp system of selection based on genetic algorithm as claimed in claim 2, it is characterised in that： Described (2) determine to adapt to value function, meet a+b+c=1, a >=0, b >=0 at the same time, on the premise of c >=0, according to resource reference Weights a, b, c are set.
5. 5. the method for the optimal soft protection of dsp system of selection based on genetic algorithm as claimed in claim 2, it is characterised in that： It is described Step 3: means of defence combination output, is to be decoded the coding obtained by genetic algorithm, by the coding per l positions One group, number consecutively is 1 arrive m from left to right, and every group of binary number is converted into decimal number respectively, and caused m individual ten enters Number processed is protectiving scheme corresponding to each program block, the optimal protection for being combined as whole program of the protectiving scheme of all program blocks Scheme.