CN115225139A - A planning method for satellite network SDN multi-control domains - Google Patents

Abstract

The invention relates to a planning method for multiple control domains of a satellite network SDN, comprising the following steps: S1: coding satellite controller nodes and switch nodes in a LEO satellite network, and randomly initializing a planning scheme of multiple control domains under a constraint relationship; S2: Calculate the individual fitness value; S3: Determine the binding energy and binding state of each level; S4: Introduce an adaptive weight and reverse learning mechanism in the update strategy of the photon emission and absorption parts to update the electronic state; S5: Steps S3 to S4 are looped for iterative optimization until the maximum number of iterations is reached, and the optimal binding energy and binding state are output; S6: Decoding to form an optimal planning strategy for multiple control domains of the satellite network SDN. The invention uses the atomic orbit search algorithm to optimize and improve, and introduces adaptive weight and reverse learning mechanism to update the electronic state to be suitable for the planning problem of satellite network multi-control domains, so as to improve network delay performance and realize network load balance.

Description

A planning method for satellite network SDN multi-control domains

technical field

The invention belongs to satellite network technology, in particular to a planning method of satellite network SDN multi-control domains based on an improved atomic orbit search algorithm. This method needs to build a satellite network delay and network load balancing model based on SDN-based satellite network architecture, optimize the effect of multi-control domain planning through an improved atomic orbit search algorithm, further reduce network delay and improve controller performance. Load balancing effect.

Background technique

Software Defined Network (SDN) decouples the control plane and data plane, can allocate network resources from a global perspective, formulate effective resource allocation strategies, and has the advantages of flexible centralized control. SDN abstracts the different, distinguishable layers of the network, making the network agile and flexible. SDN consists of application plane, data plane and control plane. The application plane is composed of various network applications; the data plane is composed of physical switches in the network; the control plane, as the core of SDN, manages the policies and traffic of the entire network. The three planes of the SDN mainly communicate through the API interface. The northbound interface establishes the communication between the application plane and the control plane, and the communication and interaction between the control plane and the data plane is mainly realized by the southbound interface. The emergence of SDN has changed the idea of traditional network distributed management. The controller of the control plane can manage the underlying hardware devices, orchestrate network services, reduce operation and maintenance costs, and improve operation and maintenance efficiency.

With the rapid development of satellite Internet technology and the rapid growth of satellite network users, the disadvantages of traditional satellite networks, such as inflexible network configuration, coexistence of multiple protocols, and heterogeneity of different networks, have gradually emerged. Drawing on the mature experience of SDN in terrestrial networks, it is considered to introduce SDN into satellite networks, so that satellites in the data plane only need to complete simple data forwarding and hardware configuration, while satellites in the control plane complete functions such as flow table distribution and resource allocation, which simplifies the Satellite network configuration reduces network costs.

Atomic orbital search (AOS) is a new type of intelligent optimization algorithm proposed in 2021. The algorithm is mainly based on quantum atom theory. The algorithm draws on electron density configuration and atomic absorption of energy. Iterative optimization based on the basic principle of launch has the advantages of strong optimization ability and fast convergence speed. An electron in the atomic orbital search algorithm represents an individual population. The energy state of the electron corresponds to the objective function value of the individual population. The photon rate represents the probability of the photon acting on the electron. If the random probability is greater than the photon rate, it means that the photon has a certain effect on the electron. Otherwise, it is considered that photons have no effect on electrons, which may be particles or magnetic fields. The effect of photons on electrons can be judged based on the binding energy of each layer. When the energy state of the electron is greater than the average of all electron energy states, it reflects the emission effect of photons, otherwise it is the absorption effect of photons.

Even if the atomic orbit search algorithm is adopted, the planning problem of the existing satellite network SDN multi-control domain is still reflected in the satellite network time extension and the load is not balanced enough.

Based on this, the present invention is proposed.

SUMMARY OF THE INVENTION

The purpose of the present invention is to solve the planning problem of satellite network multi-control domains under the influence of satellite network delay and load balancing factors, and propose a satellite network SDN multi-control domain planning method based on an improved atomic orbit search algorithm. The invention uses the atomic orbit search algorithm to optimize and improve, introduces adaptive weight and reverse learning mechanism to update the electronic state to be suitable for the planning problem of multiple control domains of satellite networks, and matches the relationship between the controller and the switch according to the planning method of multiple control domains. , to improve network latency performance and achieve network load balancing.

The invention provides a planning method for multiple control domains of a satellite network SDN. The satellite network architecture based on SDN is adopted, and the satellite network delay and load balancing model are solved based on the improved atomic orbit algorithm. According to the output of electrons with the lowest energy state, Decoding into a planning method for multiple control domains of a satellite network, which includes the following steps:

S1: Code the satellite controller nodes and switch nodes in the LEO satellite network, and randomly initialize the multi-control domain planning scheme under constraints;

S2: Calculate the individual fitness value;

S3: Determine the binding energy and binding state of each level;

S4: Introduce an adaptive weight and reverse learning mechanism in the update strategy of the photon emission and absorption parts to update the electronic state;

S5: Loop steps S3 to S4 for iterative optimization until the maximum number of iterations is reached, and output the best binding energy and binding state;

S6: Decode to form an optimal planning strategy for multiple control domains of the satellite network SDN.

Further, the control plane of the satellite network SDN adopts a distributed deployment mode, the main controller is deployed on the ground station, the regional controller is deployed on the GEO satellite, and some LEO satellites are selected to deploy the slave controllers, using the LEO satellite Iridium network topology. structure and treat all LEO satellites as switch nodes.

Further, the satellite network delay includes the inter-satellite propagation delay, the queuing delay in the control domain, and the task processing delay of the controller node.

Further, the load balancing situation is represented by the load balancing parameter BL, which is used to evaluate the difference degree of the LEO satellite controller node load, which is represented by formula (10):

In the formula, f _i represents the traffic request situation in the control domain of the controller c _i , and f _p represents the traffic request situation in the control domain of the controller c _p .

Further, in the step S3, the average value of the electronic coding sequences in each layer is recorded as the binding state, and the binding energy is the average value of each individual fitness value.

Further, in the S4, the adaptive weight is expressed by formula (12)

Among them, mgen is the maximum number of iterations, and gen is the current number of iterations.

Further, in the S4, the emission of photons is represented by formula (13)

The absorption of photons is expressed by formula (14)

与

分别为第k层第i个电子的当前个体和更新的个体，α_i、γ_i为(0，1)之间的随机数向量，LE为原子中最低能量状态的个体，BS为原子的结合态；LE^k为第k层中最低能量状态的个体，BS^k为第k层中的结合态，w为自适应权重。in,

and

are the current individual and the updated individual of the ith electron in the k-th layer, respectively, α _i and γ _i are random number vectors between (0, 1), LE is the individual with the lowest energy state in the atom, and BS is the combination of atoms LE ^k is the individual with the lowest energy state in the k-th layer, BS ^k is the binding state in the k-th layer, and w is the adaptive weight.

Further, the reverse learning mechanism in the S4 includes:

1) For the overall optimal individual, perform reverse learning on LE, as shown in Equation (15)

代表对整体最优个体LE(原子中最低能量状态的个体)进行反向学习后的个体，lb为最小控制域的编号，这里是1，m为控制域的数量，r为(0,1)的随机数，LE为原子中最低能量状态的个体；in,

Represents the individual after reverse learning of the overall optimal individual LE (the individual with the lowest energy state in the atom), lb is the number of the smallest control domain, here is 1, m is the number of control domains, and r is (0,1) The random number of , LE is the individual with the lowest energy state in the atom;

2) For the optimal individuals in each layer, reverse learning is performed on LE ^k , as shown in Equation (16)

为对第k层最优个体LE^k(第k层中最低能量状态的个体)进行反向学习后的个体，lb为最小控制域的编号，这里是1，m为控制域的数量，r为(0,1)的随机数，LE^k为第k层中最低能量状态的个体；in,

is the individual after reverse learning for the k-th layer optimal individual LE ^k (the individual with the lowest energy state in the k-th layer), lb is the number of the minimum control domain, here is 1, m is the number of control domains, and r is (0,1) random number, LE ^k is the individual with the lowest energy state in the kth layer;

与

的适应度值，保留适应度值较小的个体作为

(

代表第k层第i个电子更新后变成的第k层第(i+1)个电子的个体)；对于整体最优个体和各层最优个体进行反向学习，通过比较得到第(i+1)个电子的个体

但

可能会破坏了个体的约束条件，当出现这种情况时，要对更新后破坏了约束条件的个体进行约束修复(约束修复是指随机生成一组新的满足约束条件的多控制域规划方法)，使其适用于卫星网络多控制域的规划。3) By comparing

and

The fitness value of , and the individual with the smaller fitness value is reserved as the

(

Represents the (i+1)-th electron of the k-th layer after the update of the i-th electron of the k-th layer); reverse learning is performed for the overall optimal individual and the optimal individual of each layer, and the (i-th electron is obtained by comparison) +1) electron individual

but

The constraints of individuals may be destroyed. When this happens, constraint repair should be performed on the individuals whose constraints are broken after the update. , making it suitable for the planning of satellite networks with multiple control domains.

The advantages of the present invention: the satellite network SDN control domain planning method mainly integrates the advantages of LEO satellites with low delay and excellent signal quality, focuses on the planning method of the LEO satellite control domain, adopts the classic LEO satellite iridium network topology, and Consider all LEO satellites as switch nodes, build a model with the goal of optimizing satellite network delay and network load balancing, use the improved atomic orbit search algorithm to obtain the optimal control domain planning method, and match the controller according to the planning method Relationship with switches to improve network latency performance and achieve network load balancing.

Description of drawings

FIG. 1 is a flowchart of a method for planning multiple control domains of a satellite network SDN according to the present invention.

FIG. 2 is a schematic diagram of an SDN-based satellite network architecture adopted in the present invention.

FIG. 3 is a schematic diagram of the electronic coding series adopted in the present invention.

FIG. 4 is a schematic diagram of the variation of the adaptive weight w used in the present invention.

Figure 5 is a comparison chart of the running results of the three algorithms.

Figure 6 is an analysis graph of objective function values for different numbers of control domains.

Figure 7 is an analysis diagram of network load balancing parameters for different numbers of control domains.

Figure 8 is a graph of network delay analysis for different numbers of control domains.

Detailed ways

A method for planning multiple control domains of a satellite network SDN according to the present invention will be further described below with reference to the accompanying drawings 1-8. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than an exhaustive list of all the embodiments. It should be noted that the embodiments in this solution and the features of the embodiments may be combined with each other under the condition of no conflict.

As shown in FIG. 1, it is a schematic flowchart of a planning method for a satellite network SDN multi-control domain provided by the present invention. The present invention adopts an SDN-based satellite network architecture, and based on an improved atomic orbit algorithm, the satellite network delay and load balancing model are analyzed. Solve, according to the output of the electron with the lowest energy state, decode it into a planning method for the multi-control domain of the satellite network, which includes the following steps:

S1: Code the satellite controller nodes and switch nodes in the LEO satellite network, and randomly initialize the multi-control domain planning scheme under constraints;

S2: Calculate the individual fitness value;

S3: Determine the binding energy and binding state of each level;

S4: Improve the update strategy of the photon emission and absorption parts, introduce adaptive weights and reverse learning mechanisms, and update the electronic state;

S5: Loop steps S3 to S4 for iterative optimization until the maximum number of iterations is reached, and output the best binding energy and binding state;

S6: Decode to form an optimal planning strategy for multiple control domains of the satellite network SDN.

Specific Example 1

1. SDN-based satellite network architecture

As shown in FIG. 2 , it is a schematic diagram of the SDN-based satellite network architecture adopted in the present invention. The control plane of the SDN-based satellite network adopts a distributed deployment method. The main controller is deployed on the ground station, the regional controller is deployed on the GEO satellite, and some LEO satellites are selected to deploy the slave controller. The integrated LEO satellite has the advantages of low delay and excellent signal quality. An embodiment of the present invention takes the planning method of the LEO satellite control domain as an example, adopts the classic LEO satellite Iridium network topology, and regards all LEO satellites as switch node.

Second, SDN control domain planning method model analysis

In this embodiment, the planning method of the LEO satellite control domain is analyzed, and a problem model is constructed according to the satellite network delay and network load balancing, so as to provide a basis for the planning of the LEO satellite control domain.

(1) Constraint description

In order to form an effective satellite control domain planning scheme, it is necessary to combine the current LEO satellite network status and the SDN principle to comprehensively analyze the constraints existing in the control domain planning problem. The constraints are described as follows

① Control domain management constraints. Each control domain is managed by each controller and each control domain does not overlap with each other.

In the formula, the controller set C=[c ₁ , c ₂ ... c _i ... c _m ], the control domain set D=[D ₁ , D ₂ ... D _i ... D _m ], D _a represents the control domain of the controller c _a , D _b represents the control domain of the controller c _a , each control domain corresponds to a corresponding controller, and each control domain can have multiple switches.

②Control constraints of control domains and switches. X _ij represents the association between the control domain and the switch, and X _ij can be “0” or “1”. If the switch s _j is in the c _i control domain, it is represented by “1”, and if the switch s _j is not in the c _i control domain, it is represented by “1”. 0" means. Any switch can only belong to the management and control of a single control domain. S=[s ₁ , s ₂ ,...s _j ...s _n ] is a set of switch nodes.

Additionally, each controller can handle multiple switches.

③ Controller processing capacity constraints. The total load in any control domain cannot exceed the processing capacity of the current controller. μ _j is the data flow request rate of the switch s _j , and λ _i is the processing capability of the controller c _i .

(2) Model construction

1) The satellite network delay is mainly composed of the inter-satellite propagation delay, the queuing delay in the control domain and the task processing delay of the controller node.

① Inter-satellite propagation delay. Due to the large scale and complex _topology of the satellite network, the propagation delay _caused by the inter-satellite link distance cannot be ignored. The propagation delay Tcc between the LEO satellite controller nodes consists of:

In the formula, d _ij represents the shortest link distance between node i and node j, and similarly _dir represents the shortest link distance between node i and node r.

② Control the queuing delay in the domain. The queuing delay Q _i in the LEO satellite control domain can be obtained by the Little principle:

③ Task processing delay. The satellite controller node c _i is the time generated by the controller to process the data flow in its control domain, and the task processing delay W _i is expressed as

To sum up, the average total delay of the satellite network is expressed as T:

2) The load balancing situation of the current network can be represented by the load balancing parameter BL, which can evaluate the difference degree of the LEO satellite controller node load:

In the formula, f _i represents the traffic request situation in the control domain of the controller c _i , and f _p represents the traffic request situation in the control domain of the controller c _p .

The LEO satellite network multi-controller deployment problem model established in this embodiment to optimize satellite network delay and network load balancing is as follows:

min F=α*BL+β*T (11)

In the formula, α+β=1, 0≤α, β≤1.

3. Implementation steps of the satellite network SDN control domain planning method based on the improved atomic orbit search algorithm

The overall flow of the planning method of satellite network SDN multi-control domains based on the atomic orbit search algorithm is shown in Figure 1.

Based on the improved atomic orbit algorithm, the satellite network delay and load balancing model are solved, and the output electrons with the lowest energy state are decoded into a planning method of satellite network multi-control domains.

1) Code the satellite controller nodes and switch nodes in the LEO satellite network, and randomly initialize the multi-control domain planning scheme under the constraints.

According to the Iridium network topology, there are 66 LEO satellites in the Iridium constellation, so these 66 LEOs are regarded as satellite switch nodes, and the set of satellite switch nodes can be expressed as S=[s ₁ , s ₂ ... s _j ... s ₆₆ ], satellite controller node set C=[c ₁ , c ₂ ... c _i ... c _m ], satellite controller domain set D=[D ₁ , D ₂ ... D _i ...D _m ], the number of satellite controller nodes is m. In order to allocate 66 switches to m controller nodes, that is, to plan 66 switches to m control domains, it is achieved by setting an array with a length of 66 and an array element size of 1 to m, and based on this embodiment The constraint relationship of , after the constraints are restricted, an effective array is formed, that is, an electron coding sequence, and an electron represents a feasible solution. As shown in Figure 3, an electron coding sequence X=[m,3,3,2,4 ,1…5,4,m]. Then, according to the population number npop, npop electronic coding sequences are generated.

In Figure 3, "X[1]=m" indicates that the first switch is assigned to the control domain D _m , "X[2]=3" indicates that the second switch is assigned to the control domain _D3 , Perform control domain planning in sequence.

2) Calculation of individual fitness value

Since each electron has an energy state, which corresponds to the fitness value of an individual, the individual fitness value of the population is calculated.

3) Determine the binding energy and binding state of each level

The average value of the electronic coding sequences in each layer is recorded as the binding state, and the binding energy is the average value of each individual fitness value.

4) Update electronic status

The action of a photon, particle or magnetic field on an electron changes the individual electron and its energy state. Since the original algorithm is based on the photon rate to determine whether the photon has an effect on the electron, if the generated random number is greater than the photon rate, it is considered to be the effect of the photon effect. Photons can absorb and emit electrons, and judging the effects of these two is based on the binding energy of this level. When it is greater than the binding energy of this level, it is the emission effect, otherwise it is the absorption effect. In order to adapt to the multi-control domain planning method of the present invention, the update strategy of the photon emission and absorption parts is improved.

①Photon action

Equation (12) indicates that w is the adaptive weight, where mgen is the maximum number of iterations, gen is the current number of iterations, and w decreases from 1 to 0 with the increase of the number of iterations. The amplitude increases, and the change is shown in Figure 4.

与

分别为第k层第i个电子的当前个体和更新的个体，α_i、γ_i为(0，1)之间的随机数向量，w为自适应权重。式(13)表示为光子的发射作用，其中LE为原子中最低能量状态的个体，BS为原子的结合态。式(14)表示为光子的吸收作用，其中LE^k为第k层中最低能量状态的个体，BS^k为第k层中的结合态。考虑到原子轨道算法在后期的局部搜索中易出现早熟现象，且电子个体的更新主要在原始个体上围绕整体最优个体或是各层最优个体进行随机探索，因此考虑在最低能量状态个体上加自适应权重，以提升算法收敛速度和精度。本实施例将自适应权重主要应用在光子对电子作用阶段如式(13)和(14)所示，迭代前期w变化幅度较小，电子个体主要围绕最优个体或是各层最优个体进行变化，迭代后期w变化幅度加大，避免陷入局部极值。In formula (13) (14),

and

are the current individual and the updated individual of the i-th electron in the k-th layer, respectively, α _i , γ _i are random number vectors between (0, 1), and w is the adaptive weight. Equation (13) is expressed as the emission of photons, where LE is the individual in the lowest energy state in the atom, and BS is the binding state of the atom. Equation (14) is expressed as the absorption of photons, where LE ^k is the individual in the lowest energy state in the k-th layer, and BS ^k is the bound state in the k-th layer. Considering that the atomic orbital algorithm is prone to premature phenomenon in the later local search, and the update of the electronic individual is mainly based on the original individual to randomly explore the overall optimal individual or the optimal individual at each layer, so consider the lowest energy state individual. Add adaptive weights to improve the convergence speed and accuracy of the algorithm. In this embodiment, the adaptive weight is mainly applied to the action stage of photons on electrons, as shown in equations (13) and (14). In the early stage of iteration, the variation range of w is small, and the electron individuals mainly focus on the optimal individual or the optimal individual at each layer. changes, the variation of w in the later stage of the iteration increases, so as to avoid falling into local extreme values.

②The action of particles or magnetic fields

When the generated random number is smaller than the photon rate, it means that the action of photons on electrons is impossible. At this time, the movement of electrons between different layers around the nucleus is based on the action of particles or magnetic fields, and energy absorption will also occur at this time. or launch. Since the electrons change randomly under the action of particles or magnetic fields in the traditional atomic orbital search algorithm, the present invention changes the traditional update formula, introduces a reverse learning mechanism, and at the same time reverses the overall optimal individual and the optimal individual at each layer. Learn, further expand the search scope, and increase the global search capability.

For the overall optimal individual, reverse learning is performed on LE, as shown in the formula

In formula (15), lb is the number of the smallest control domain, here is 1, m is the number of control domains, and r is a random number of (0,1).

For the optimal individual at each layer, reverse learning is performed on LE ^k , as shown in the formula

In formula (16), lb is the number of the minimum control domain, which is 1 here, m is the number of control domains, and r is a random number of (0, 1).

与

的适应度值，保留适应度值较小的个体作为

Perform reverse learning for the overall optimal individual and the optimal individual at each layer, and then perform constraint repair to make it suitable for the planning of multiple control domains of satellite networks. By comparison

and

The fitness value of , and the individual with the smaller fitness value is reserved as the

5) Output the best binding energy and binding state

Loop steps 3) to 4) for iterative optimization until the maximum number of iterations mgen is reached, and the optimal binding energy and binding state are output.

6) Decoding

According to the final output of the best combined state and decoding, the optimal planning strategy of the multi-control domain of the satellite network is formed.

Specific embodiment 2

Simulation and Performance Evaluation

The present invention is a satellite network SDN control domain planning method based on the atomic orbit search algorithm, and by improving the traditional atomic orbit search algorithm, the reasonable planning of the satellite network SDN control domain can be solved. The improved Atomic Orbital Search Algorithm (IAOS) proposed in the present invention is simulated with Particle Swarm Algorithm (PSO) and Atomic Orbital Search Algorithm (AOS). When the number m of satellite control domains is 12, the running results of different algorithms are shown in Figure 5. Compared with PSO and AOS, IAOS has the best optimization results. Compared with AOS, IAOS is significantly improved in terms of global exploration ability and convergence speed.

By setting different numbers of control domains and randomly selecting the simulation results once. As shown in Figure 6-8, with the increase of the number of control domains, the objective function value and network delay all show a downward trend, and the load balancing parameters show an upward trend. This is because the inter-satellite controller nodes and switch nodes can be connected. The number of paths increases as the number of control domains increases, and the selectivity of the controller to process switch requests increases. The proximity principle is used to process switch requests, and the load balancing parameters between controllers become larger. From the perspective of overall performance, the IAOS search The advantages are not affected by the number of control domains, and a better planning method for satellite network SDN control domains can still be obtained.

Obviously, the above-mentioned embodiments of the present invention are only examples for clearly illustrating the present invention, rather than limiting the embodiments of the present invention. Changes or changes in other different forms cannot be exhausted here, and all obvious changes or changes derived from the technical solutions of the present invention are still within the protection scope of the present invention.

Claims (8)

1. a planning method for satellite network SDN multiple control domains, is characterized in that, comprises the steps: S1: Code the satellite controller nodes and switch nodes in the LEO satellite network, and randomly initialize the multi-control domain planning scheme under constraints; S2: Calculate the individual fitness value; S3: Determine the binding energy and binding state of each level; S4: Introduce an adaptive weight and reverse learning mechanism in the update strategy of the photon emission and absorption parts to update the electronic state; S5: Loop steps S3 to S4 for iterative optimization until the maximum number of iterations is reached, and output the best binding energy and binding state; S6: Decode to form an optimal planning strategy for multiple control domains of the satellite network SDN. 2. The planning method of satellite network SDN multi-control domain as claimed in claim 1, is characterized in that, the control plane of described satellite network SDN adopts distributed deployment mode, and main controller is deployed in ground station, in GEO satellite deployment area controller, and select some LEO satellites to deploy slave controllers, adopt the LEO satellite Iridium network topology, and regard all LEO satellites as switch nodes. 3. The planning method of satellite network SDN multiple control domains as claimed in claim 1, wherein the time delay of the satellite network comprises inter-satellite propagation delay, queuing delay in the control domain and task processing of the controller node time delay. 4. The planning method of satellite network SDN multi-control domains as claimed in claim 1, is characterized in that, the load balancing situation of described satellite network is represented by load balancing parameter BL, and load balancing parameter is used for evaluating LEO satellite controller node The degree of difference in load, which is represented by formula (10):

In the formula, f _i represents the traffic request situation in the control domain of the controller c _i , and f _p represents the traffic request situation in the control domain of the controller c _p . 5. The planning method of satellite network SDN multi-control domains as claimed in claim 1, is characterized in that, in described S3, the average value of the electronic coding sequence in each layer is recorded as the combined state, and the combined energy is adapted for each individual The average value of the degree value. 6. The planning method for satellite network SDN multi-control domains according to claim 1, wherein in the S4, the adaptive weight is expressed by formula (12)

Among them, mgen is the maximum number of iterations, and gen is the current number of iterations. 7. The planning method for satellite network SDN multi-control domains according to claim 1, wherein in said S4, the emission effect of photons is expressed by formula (13)

The absorption of photons is expressed by formula (14)

in,

and

are the current individual and the updated individual of the ith electron in the k-th layer, respectively, α _i and γ _i are random number vectors between (0, 1), LE is the individual with the lowest energy state in the atom, and BS is the combination of atoms state; LE ^k is the individual in the lowest energy state in the k-th layer, and BS ^k is the binding state in the k-th layer. 8. The planning method for satellite network SDN multiple control domains as claimed in claim 1, wherein the reverse learning mechanism in the S4 comprises: 1) For the overall optimal individual LE, perform reverse learning on LE, as shown in Equation (15)

in,

Represents the individual after reverse learning of the overall optimal individual LE, lb is the number of the smallest control domain, here is 1, m is the number of control domains, r is a random number of (0, 1), and LE is the lowest among atoms the energy state of the individual; 2) For the optimal individual LE ^k of each layer, perform reverse learning on LE ^k , as shown in Equation (16)

in,

For the individual after reverse learning of the k-th layer optimal individual LE ^k , lb is the number of the minimum control domain, here is 1, m is the number of control domains, r is a random number of (0, 1), LE ^k is the individual with the lowest energy state in the kth layer; 3) By comparing

and

The fitness value of , and the individual with the smaller fitness value is reserved as the

Perform reverse learning for the overall optimal individual and the optimal individual at each layer, and obtain the individual with the i+1th electron through comparison

but

The constraints of individuals may be destroyed. When this happens, the individuals whose constraints are destroyed after the update should be repaired to make them suitable for the planning of multiple control domains of satellite networks.

Images