TWI385971B - Network congestion control parameter measurement system and its method and proportional integral differential control module design method - Google Patents

Description

Network congestion control parameter measurement system and method thereof and proportional integral differential control module design method

A measurement system and method thereof, in particular to a network congestion control parameter measurement system and a proportional integral thereof, which combines a genetic algorithm and quickly calculates an optimal parameter value of a proportional integral differential control module of a control network The design method of the differential control module.

According to the current network control theory, even if there are many different methods, it can be used to design various types of network control systems to solve various complex network control problems.

As far as the current network control system is based on the proportional integral derivative controller (PID Controller), the structure is simple, the cost is low, and the maintenance is relatively easy.

However, the prior art has an inevitable deficiency: First, the adjustment methods proposed by Ziegler-Nichols (Z-N) and Cohen-Coon (C-C) are often cited in the design of PID controllers. However, for more complex nonlinear systems and systems without specific delay characteristics, such as network congestion control systems, and some higher-order systems, these methods are not applicable.

Second, the network system is mostly a complex nonlinear system control system. If the optimal parameters of the PID controller are adjusted by try-error or according to the rule of thumb, the controlled system may be set due to poor parameters. , making the entire network system more unstable and difficult to control accurately. Moreover, the trial and error method itself has unpredictability of adjustment time, and it is likely to cost a lot. The cost of time, in turn, makes the measurement of network congestion control parameters slower.

In view of this, the problem to be solved by the present invention is to provide an error in the state of the controlled network and the ideal value of the designer, to introduce a genetic algorithm, to obtain the optimal parameters of the PID controller, and to make the PID controller effective. Controlling the controlled network allows the network system to quickly control the designer's ideal system response.

In order to solve the above system problem, the technical means provided by the present invention discloses a network congestion control parameter measurement system. The system comprises a controlled network module, a computing module, a proportional integral differential control module and a calculus module.

The controlled network module provides a momentary sequence length; the computing module calculates a systematic error based on an ideal sequence length and the instantaneous sequence length; the calculus module takes the systematic error as input and calculates a system based on a genetic algorithm. Proportional gain, one integral gain and one differential gain; the proportional integral differential control module takes the proportional gain, the integral gain and the differential gain as input conditions, and calculates the control input variable to correct the instantaneous sequence length of the controlled network module.

In order to solve the above method problem, the technical means provided by the present invention discloses a network congestion control parameter measurement method, which is applicable to a network congestion control parameter measurement system. The method first constructs a controlled network module, and calculates the instantaneous sequence length through the data packet transmission amount of the controlled network module; and uses a computing module according to the ideal sequence length and the instantaneous sequence length Calculate a systematic error; take the systematic error as input, calculate a proportional gain, an integral gain and a differential gain according to a genetic algorithm; input proportional gain, integral gain and differential gain into a proportional integral differential control module to generate a Control the input variable; input the control input variable to the controlled network module to correct the instantaneous sequence length; and then judge whether the network congestion control parameter measurement system reaches equilibrium convergence according to the change of the system error to determine whether to record the proportional gain and the integral gain. And differential gain.

The invention further discloses a design method of a proportional integral differential control module using a gene algorithm, which firstly uses a random sequence method to generate a plurality of chromosomes to form a group; calculating an adaptation value of each chromosome and determining a maximum A good chromosome, the adaptation value is used as the first adaptation value of the ethnic group, and then the robin-type selection method is used to generate a plurality of chromosomes of the next generation to update the ethnic group; the ethnic group is mated and mutated, and the second adaptive value of one of the ethnic groups is calculated. Using the second fitness value and the first fitness value to determine whether to retain the mating and mutation population; and determining whether the first fitness value converges to determine whether to convert the proportional gain, the integral gain and the differential gain or recalculate the adaptation of each chromosome value.

The invention has the effects that the prior art cannot achieve: First, by genetic algorithm, the system can form a system balance convergence in a short time for analyzing all control, setting and state parameters, and changes of the controlled network module.

Second, through the genetic algorithm, it is easy to get the best proportional integral differential control module parameters, and make the control performance index (IAE) reach the most Small value, so that the system responds to meet the needs of the designer.

Thirdly, the Proportional-Integral-Derivative Controller (PIDController) is used for control, which has the advantages of simple structure, low cost and easy maintenance, and maintains good performance for the control system.

Fourth, the proportional integral differential control module designed by the genetic algorithm can control the controlled network module, in addition to maintaining the system balance for a long time, even if the transmission network module has a very large difference in transmission volume. Compared with the previous technology, it can still maintain a certain quality of network congestion control.

In order to further understand the object, structural features and functions of the present invention, the following detailed description is given in conjunction with the related embodiments and drawings: Please refer to FIG. 1, which is a block diagram of a network congestion control parameter measurement system according to an embodiment of the present invention. The system includes a controlled network module 110, a computing module 120, a proportional integral derivative control module 140, and a calculus module 130.

The controlled network module 110 is a network model in which a plurality of network devices are connected to the Internet through a network controller such as a router. The controlled network module 110 can calculate the instantaneous sequence length according to the transmission amount of the data packet. The calculation module 120 can obtain an ideal sequence length of the external reception, and calculate a systematic error of the ideal sequence length and the instantaneous sequence length, which is the value of the ideal network state determined by the designer. The systematic error is used by the calculus module 130 to calculate a ratio through a genetic algorithm. Gain, an integral gain and a differential gain. The proportional integral derivative control module 140 calculates a control input variable by using the proportional gain, the integral gain, and the differential gain, so that the controlled network module 110 corrects the instantaneous sequence length according to the control input variable.

Please refer to FIG. 2 , which is a flowchart of a network congestion control parameter measurement method according to an embodiment of the present invention, which is applicable to the network congestion control parameter measurement system. This method includes: A controlled network module 110 is constructed, and an instantaneous sequence length is calculated through the data packet transmission amount of the controlled network module 110 (step S210). The controlled network module 110 is a virtual network model established according to the actual network system structure.

The construction method is mainly the application of Hollot et al., TCP/AQM (Transmission Control Protocol/Active Queue Management) network congestion model proposed in 2001, as shown in Figure 7, The following non-linear systems with no specific delay characteristics: Where w is the average window size (packets); q is the instantaneous sequence length (packets); T _p is the transfer delay time (second); R is the round-trip time (RTT) and can be equivalent to q / C + T _p ; where C is the network connection capacity (packets/sec or bps); N is the number of network users; p is the packet marking or loss rate used to control the packet transmission speed and the management sequence length. All of the above variables are positive numbers. (1.a) describes the characteristics of the additive-increase & multiplicative-decrease (AIMD) of TCP congestion control. The equation (1.b) is a dynamic behavior indicating the length of the instantaneous sequence transmitted by the controlled network module 110.

Because the control input of the system, the packet mark or loss rate is between 0 and 1, therefore, the formula (1) is rewritten as a non-linear delay system with input saturation.

Where the input saturation u ( t )= p (t- R ( t )) can be replaced by the following nonlinear term

Where w is the average window size, q is the instantaneous sequence length, T _p is the transit delay time, R ( t ) is the data packet round trip time and can be equivalent C is the network connection capacity, N is the number of network users of the controlled network module 110, sat( u ( t )) is the data packet label or loss rate, u _min =0 and u _max =1.

The proportional integral derivative control module 140 is of a continuous type, and its standard type is as follows:

Where e ( t )= q ( t )- q _d is the systematic error, q is the instantaneous sequence length, q _d is the ideal sequence length, u ( t ) is the control input variable, K _P is the proportional gain, and T _D is the differential time Constant, T _I is the integral time constant, and equation (4) can also be expressed as

among them , K _D = K _P T _D are the integral gain and the differential gain, respectively.

A system error is calculated by a calculation module 120 according to the ideal sequence length and the instantaneous sequence length (step S220). As mentioned earlier, the systematic error is calculated by e ( t ) = q ( t ) - q _d .

Taking the systematic error as an input, a proportional gain, an integral gain, and a differential gain are calculated according to a genetic algorithm (step S230).

The derivation of these parameters involves the design of the proportional integral derivative control module 140. The main problem of this design method is how to obtain the appropriate parameters of K _P , K _I and K _{D to} make the system have better control performance. Typical output specifications often include maximum overshoot, rise time, settling time, and steady-state error. The more common performance indicators are the error squared integral (ISE) and the absolute value of the error (IAE). The definitions are as follows:

As shown in Figure 3, it can be known that the smaller the IAE (ie, the shaded area of Figure 3), the smaller the control error, that is, the better the control result, so how to design an easy-to-use method to make the designer It is easy to determine K _P , K _I , K _D , and the system has the best (small) performance index (IAE), which is one of the main purposes of the present invention.

In general, the gene algorithm (GA) consists of three steps: 1. Generating the initial population 2. Group competition 3. Mating, mutation. Moreover, in order to avoid the system getting into the local optimum value during the search process, the selection of the initial ethnic group is also very important, and the Quasi-Random Sequence (QRS) is often used. Because the QRS distribution is more random and uniform, it is used to generate the initial population to avoid the system being trapped in local optimum values. The difference between QRS and general random variables can be understood from Figure 4 and Figure 5. Generally, random sampling under the limited individual conditions may result in local aggregation, and the QRS is more evenly distributed.

Therefore, if we substitute the network mathematics mode (2a and 2b) to be controlled into the control system block in Figure 1, and use the genetic algorithm to evolve, we can easily get K _P , K _I , K _{D .} Three parameters, and according to the spirit of the gene algorithm, the proportional integral derivative control module 140 can minimize the performance index IAE of the controlled system.

By incorporating the concept of gene algorithm, let the variable S = [ K _P , K _I , K _D ], then here, the optimization problem in the present invention can be defined as follows: a set of parameters is found To minimize the performance index IAE of the closed loop system, more precisely, this optimization problem can be mathematically described as follows:

minimize.

The method for performing the gene algorithm by the calculus module 130 is as shown in FIG. 6, and the steps are described in detail as follows: A plurality of chromosomes are generated by a random sequence-like method to form a group, and an adaptive function is defined. Each chromosome has a proportional parameter, an integral parameter and a differential parameter (step S710).

First use QRS to generate the starting group , the population size is N _p , wherein each chromosome S _i , i =1, 2, ..., N _p has a fixed length binary string ( M ₁ + M ₂ + M ₃ ) length, respectively The proportional parameter K _p , the integral parameter K _I , and the differential parameter K _D are encoded, so that the following chromosome can be obtained

Then define the adaptation function as shown below:

Where h ( S ) is the performance index IAE. When the IAE is smaller, the larger the f ( S ) value is, the higher the fitness value is. In order to facilitate the operation, the adaptation function of the above formula can be replaced by the normalized result of equation (11).

Where f _min is the smallest of all fitness values and f _max is the largest of all fitness values. In order to calculate the fitness value of each chromosome according to the adaptive function, and then determine an optimal chromosome based on all the fitness values, and use the fitness value of the best chromosome as the first fitness value of one of the ethnic groups, and then according to the ratio of all the fitness values, The roulette selection method generates a plurality of chromosomes of the next generation to update the ethnic group (step S720).

For the k-th generation P ^k, k calculated fitness value of each chromosome generation of S _i, k and find the best chromosome generations Make Think of it as the first fitness value. If there is more than one optimal chromosome, choose either one as the best chromosome. Finally, according to the proportional allocation, the new children are copied in the way of roulette wheel selection to update the group.

The population is mated and mutated, and one of the second adaptation values of the population is calculated, and whether the second fitness value is greater than the first fitness value is compared, and whether the mating and the mutated population are retained is determined (step S730).

Establish a mating pool and calculate the mating rate p _c and the mutation rate p _m from the following formula (12), and the chromosomes after mating and mutation Said.

among them The largest of all fitness values after formalization, In the second mother chromosome to be mated after normalization, the fitness value is larger; After normalization, it is the adaptive value of the chromosome at the time of mutation; The average of all fitness values after normalization; k ₁ , k ₂ , k ₃ , k ₄ are positive numbers less than 1 to limit the mating rate p _c and the mutation rate p _m between [0, 1]. p _{c _ last} and p _{m _ last} are the mating rate and mutation-rate of the previous generation; a is the weight; δ is a small constant, avoiding the situation that the above formula cannot be solved. Comparing the first fitness value of chromosome S _i with The second adaptation value. If , the gene combination of the original chromosome S _i is retained, otherwise the chromosome after mating and mutation process As a combination of genes for the new ethnic group.

It is judged whether the first fitness value converges (step S740), and if the f' ( S ) value has not converge, the process returns to step S720 to calculate the fitness value of each chromosome of the next generation.

If the value of f' ( S ) converges to a minimum value, a chromosome matching the first fitness value is obtained, and the proportional parameter, the integral parameter, and the differential parameter of the chromosome are converted into a proportional gain, an integral gain, and a differential gain (step S750). That is, s ^* is the best value, and its gene That is, the optimal parameters of the proportional integral derivative control module 140, which will minimize the system performance index IAE.

The proportional gain, integral gain and differential gain are input to a proportional integral derivative control module 140 to generate a control input variable (step S240). The control input variable is then input to the controlled network module 110 to correct the instantaneous sequence length (step S250).

The system error is recalculated by the calculation module 120 (step S260), and it is determined whether the system error is less than a limit value (step S270). This limit value is set by the user to determine the system error should be within the range of values when the system is in the best state.

If it is judged to be less than the limit value, determine the network congestion control parameter measurement system The system balance is achieved, and the proportional gain, the integral gain, and the differential gain are recorded (step S280). Proportional integral differential control parameters can be applied to real network systems based on these parameters.

If it is determined that the value is not less than the limit value, the process returns to step S230. Until the entire system is balanced and the best K _P , K _I , K _D parameters are obtained.

From the above discussion, the parameters of the controlled network module 110 are first substituted into the formula (2), and the control block diagram of the network congestion control parameter measurement system of the present invention is established by using SIMULINK (simulation program), as shown in FIG. 7 Show:

In Figure 7, the number of network users is set to N=100. The external link has a bandwidth capacity of 10 Mbps (the corresponding link capacity is 1250 packets/sec, and the packet size is 1000 bytes), and the window size is a parameter generated by the network operation. In addition, the transmission delay time T _p is 0.08 sec, and the propagation delay parameter is the time delay of data packet transmission between the two places; the ideal sequence length is q _d =150; the upper and lower limits of the control input variable saturation are respectively u _Min =0 and u _max =1. In the part of the gene algorithm, the population size N _p = 50, the mating rate and mutation rate of the first generation with In addition, other variables are given as follows. k ₁ = k ₃ =0.8, k ₂ = k ₄ =0.3, a=0.6 and δ=0.01 and use the tools provided by the network state simulation program such as Matlab 6.5, Simulink and NS-2 (Network Simulator ver.2) The function uses the gene algorithm described above to solve the optimization problem of the network system with the proportional integral derivative control module 140 as the main axis. According to the above gene algorithm, the QRS is first used to generate the starting ethnic group. , wherein each chromosome s _i has a length of 30 bits (1.e. M _j =10, j =1, 2, 3). After a series of competitive evolutions, the system's IAE convergence curve is shown in Figure 8. It can be seen from Figure 8 that the system's IAE converges after the evolution of 1000 generations, and its optimal value is f' ( S ^* ) = 0.6644. The optimal parameters of the proportional integral derivative control module 140 can be obtained as follows: . We substitute this optimal parameter into the network congestion control system, and its state response situation is shown in Figure 9.

The proportional integral derivative control module 140 in FIG. 9 is ' ' It can be observed that the response of the system is very good in terms of response rise time, maximum overshoot, and steady-state error and settling time.

From FIG. 10A to FIG. 10D, the proportional integral differential control module 140 of the gene algorithm type, regardless of the network system state (including the normal state, the number of network users, the low transmission delay state, and the high transmission delay) State), a balanced system steady state can be achieved. There will be no unstable network transmission due to multiple data packets, or special circumstances. As shown in FIG. 10A to FIG. 10D, the general proportional integral control module (the PI curve in the figure) is very unstable in response to the congestion control system, and the proportional integral differential control module 140 (the GA-PID curve in the figure) of the present invention is It can be stably maintained at a value of the ideal sequence length q _d = 150, thus verifying the feasibility and progress of the present invention.

While the present invention has been described above in terms of the preferred embodiments thereof, it is not intended to limit the invention, and the equivalent of the modification and retouching of the present invention is still within the spirit and scope of the present invention. This hair Within the scope of patent protection.

110‧‧‧Controlled network module

120‧‧‧Computation Module

130‧‧‧ calculus module

140‧‧‧Proportional Integral Differential Control Module

1 is a diagram of a network congestion control parameter measurement system according to an embodiment of the present invention; 2 is a flow chart of measuring network congestion control parameters according to an embodiment of the present invention; 3 is a schematic diagram of performance indicators of an embodiment of the present invention; 4 is a schematic diagram of a general random variable of an embodiment of the present invention; FIG. 5 is a schematic diagram of a random variable such as an embodiment of the present invention; FIG. 6 is a flow chart of a gene algorithm according to an embodiment of the present invention; 7 is a control block diagram of a network congestion control parameter measurement system according to an embodiment of the present invention; 8 is a schematic diagram of an IAE convergence curve according to an embodiment of the present invention; 9 is a schematic diagram of a system response curve according to an embodiment of the present invention; 10A is a diagram showing a normal response of a congestion control system according to an embodiment of the present invention; FIG. 10B is a network user of a congestion control system according to an embodiment of the present invention; FIG. Response map under change; 10C is a response diagram of the congestion control system of the embodiment of the present invention in a low transmission delay state; Figure 10D is a response diagram of the congestion control system of the embodiment of the present invention in a high transmission delay state.

110‧‧‧Controlled network module

120‧‧‧Computation Module

130‧‧‧ calculus module

140‧‧‧Proportional Integral Differential Control Module

Claims (18)

A network congestion control parameter measurement system, comprising: a controlled network module calculates the instantaneous sequence length through the transmission amount of the data packet, and corrects the length of the instantaneous sequence according to a control input variable; A computing module obtains an ideal sequence length of external reception, and calculates a systematic error according to the ideal sequence length and the length of the instantaneous sequence; A proportional integral derivative control module calculates the control input variable by using a proportional gain, an integral gain and a differential gain as input conditions; A calculus module calculates the proportional gain, the integral gain and the differential gain according to a genetic algorithm by using the systematic error as an input. The network congestion control parameter measurement system described in claim 1, wherein the configuration of the controlled network module can be expressed as Where w is the average window size, q is the length of the instant sequence, T _p is the transit delay time, R ( t ) is the data packet round trip time and can be equivalent , C is the network connection capacity, N is the number of network users of the controlled network module, sat( u ( t )) is the data packet label or loss rate, u _min =0 and u _max =1. The network congestion control parameter measurement system described in claim 1, wherein the proportional integral differential control module is constructed Where e ( t )= q ( t )- q _d is the systematic error, q ( t ) is the length of the instantaneous sequence, q _{d is} the length of the ideal sequence, u ( t ) is the control input variable, and K _{P is} the Proportional gain, T _D is the differential time constant, and T _I is the integral time constant. The network congestion control parameter measurement system described in claim 1, wherein the proportional integral differential control module is constructed Where e ( t )= q ( t )- q _d is the systematic error, q ( t ) is the length of the instantaneous sequence, q _{d is} the length of the ideal sequence, u ( t ) is the control input variable, and K _{P is} the Proportional gain, K _I is the integral gain, and K _D is the differential gain. The network congestion control parameter measurement system described in item 1 of the patent application scope The method in which the calculus module executes the gene algorithm includes the following steps: Using a random sequence method to generate a plurality of chromosomes to form a group, defining an adaptation function, each chromosome having a proportional parameter, an integral parameter and a differential parameter; Calculating the fitness value of each chromosome according to the adaptation function, and determining an optimal chromosome according to the adaptation values, and using the fitness value of the optimal chromosome as the first fitness value of the ethnic group, and then according to the fitness values The ratio of robin-type selection to generate a plurality of chromosomes of the next generation to update the ethnic group; Performing mating and mutation on the population, calculating a second fitness value of the one of the ethnic groups, and comparing whether the second fitness value is greater than the first fitness value to determine whether to retain the mating and the mutant population; Determining whether the first fitness value converges, and if it is determined that the convergence is not converged, returning the step of calculating the fitness value of each chromosome, and if it is determined to be convergent, obtaining the chromosome matching the first fitness value, and proportion of the chromosome The parameters, integral parameters, and differential parameters are converted to the proportional gain, the integral gain, and the differential gain. For example, the network congestion control parameter measurement system described in claim 5, wherein the representative expression of the group is , the population size is N _p , each chromosome represents S _i and has a coding length of ( M ₁ + M ₂ + M ₃ ), i =1, 2, ..., N _p , so each chromosome can be expressed as The network congestion control parameter measurement system according to claim 5, wherein the adaptation function is expressed as Where h ( S ) is a performance indicator, K _{P is} the proportional parameter, K _{I is} the integral parameter, and K _{D is} the differential parameter. The network congestion control parameter measurement system described in claim 7, wherein the performance indicator can be expressed as . The network congestion control parameter measurement system described in claim 5, wherein the mating rate of the population is mating And the mutation rate of the mutation in this group , The largest of all fitness values after formalization, In the second mother chromosome to be mated after normalization, the fitness value is larger; After normalization, it is the adaptive value of the chromosome at the time of mutation; The average of all fitness values after normalization, k ₁ , k ₂ , k ₃ , k ₄ is a positive number less than 1 to limit the mating rate p _c and the mutation rate p _m between [0, 1], p _{c_last} and p _{M_last} is the mating rate and mutation rate of the previous generation, and a is a small constant for the weight and δ . A network congestion control parameter measurement method is applicable to a network congestion control parameter measurement system, which comprises: constructing a controlled network module, and calculating a data packet transmission amount through the controlled network module Instantly calculating the length of the sequence; using a computing module to calculate a systematic error according to the length of the ideal sequence and the length of the instantaneous sequence; using the systematic error as an input, calculating according to a genetic algorithm a proportional gain, an integral gain and a differential gain; the proportional gain, the integral gain and the differential gain are input to a proportional integral differential control module to generate a control input variable; and the control input variable is input to the controlled network The module is configured to correct the length of the instantaneous sequence; and recalculate the system error to determine whether the system error is less than a limit value, and if it is determined to be less than the limit value, determine that the network congestion control parameter measurement system reaches a system balance, and The proportional gain, the integral gain, and the differential gain are recorded, and if it is determined that the value is not less than the limit value, the systematic error is returned as an input step. The method for measuring a network congestion control parameter according to claim 10, wherein the configuration of the controlled network module can be expressed as Where w is the average window size, q is the length of the instant sequence, T _p is the transit delay time, R ( t ) is the data packet round trip time and can be equivalent , C is the network connection capacity, N is the number of network users of the controlled network module, sat( u ( t )) is the data packet label or loss rate, u _min =0 and u _max =1. For example, the network congestion control parameter measurement method described in claim 10, wherein the proportional integral differential control module is constructed Where e ( t )= q ( t )- q _d is the systematic error, q ( t ) is the length of the instantaneous sequence, q _{d is} the length of the ideal sequence, u ( t ) is the control input variable, and K _{P is} the Proportional gain, T _D is the differential time constant, and T _I is the integral time constant. For example, the network congestion control parameter measurement method described in claim 10, wherein the proportional integral differential control module is constructed Where e ( t )= q ( t )- q _d is the systematic error, q ( t ) is the length of the instantaneous sequence, q _{d is} the length of the ideal sequence, u ( t ) is the control input variable, and K _{P is} the Proportional gain, K _I is the integral gain, and K _D is the differential gain. The method for measuring a network congestion control parameter according to claim 10, wherein the method for performing the gene algorithm comprises the following steps: generating a plurality of chromosomes by using a random sequence method to form a group, defining An adaptive function, each chromosome has a proportional parameter, an integral parameter and a differential parameter; calculating an adaptive value of each chromosome according to the adaptive function, and then determining an optimal chromosome based on the adapted values, and using the optimal one The fitness value of the chromosome is taken as one of the first adaptation values of the ethnic group, and then according to the adaptation values a ratio of robin-type selection to generate a plurality of chromosomes of the next generation to update the ethnic group; mating and mutating the population, calculating a second adaptation value of the one of the ethnic groups, and comparing whether the second fitness value is greater than the first An adaptation value to determine whether to retain the mating and the mutation of the ethnic group; and determining whether the first fitness value converges, and if it is determined that the convergence is not converged, returning the step of calculating the fitness value of each chromosome, if it is determined to be convergent, The chromosome matching the first fitness value is obtained, and the proportional parameter, the integral parameter and the differential parameter of the chromosome are converted into the proportional gain, the integral gain and the differential gain. The method for measuring a network congestion control parameter according to claim 14, wherein the representative formula of the group is P ⁰ = { S _i , S ₂ , ..., S _Np }, and the population size is N _p , each The chromosome represents S _i and has a coding length of ( M ₁ + M ₂ + M ₃ ), i = 1 , 2 ,..., N _p , so each chromosome can be expressed as . The method for measuring a network congestion control parameter according to claim 5, wherein the adaptation function is expressed as Where h ( S ) is a performance indicator, K _{P is} the proportional parameter, K _{I is} the integral parameter, and K _{D is} the differential parameter. The method for measuring a network congestion control parameter according to claim 16 of the patent application scope, wherein the performance indicator can be expressed as . The method for measuring the network congestion control parameter described in claim 14 of the patent application, wherein the mating rate of the population is mating And the mutation rate of the mutation in this group , The largest of all fitness values after formalization, In the second mother chromosome to be mated after normalization, the fitness value is larger; After normalization, it is the adaptive value of the chromosome at the time of mutation; For the normalization of all the fitness values after normalization, k1, k2, k3, and k4 are positive numbers less than 1 to limit the mating rate pc and the mutation rate pm between [0, 1], p _{c_last} and p _{m_last} respectively. The mating rate and mutation rate, a is the weight and δ is a small constant.