Network slice function deployment method of satellite-ground fusion network

Abstract

The invention provides a network slice function deployment method for a satellite-ground convergence network. The method comprises the following steps: acquiring network state information and network slice deployment requirements of the satellite-ground convergence network; based on the network state information and the network slice deployment requirement, executing a network slice function deployment algorithm to obtain a deployment result of a network slice instance; the network slice function deployment algorithm is a hybrid algorithm realized based on a multi-population particle swarm and a genetic algorithm, the network slice function deployment algorithm is composed of a plurality of sub-populations, each sub-population represents a variable of an optimization problem, and each sub-population adopts different update strategies and optimization parameters according to the variable type of the optimization problem; and sending a deployment result of the network slice instance to each network node to complete deployment. According to the network slice function deployment method of the satellite-ground fusion network, the proper satellite or ground node can be selected according to the service requirement to flexibly deploy the network slice function, so that the network slice performance is effectively improved.

Description

Technical Field

[0001] The present invention belongs to the field of wireless communication technology, and specifically relates to a method for deploying network slicing functions in a satellite-ground fusion network. Background Art

[0002] With the advancement of communications technology, the convergence of satellite and terrestrial networks has become a trend in future communications. Network slicing technology can partition multiple logical networks within a satellite-ground converged network, utilizing different network slices to provide customized network services and effectively meet diverse business needs. Based on software-defined networking and network function virtualization technologies, network operators can deploy network functions (such as access and mobility management functions (AMF), session management functions (SMF), and user plane functions (UPF)) on satellites. Compared to traditional terrestrial deployment of network functions, onboard deployment reduces signaling transmission latency, thereby reducing network slicing latency. However, the highly dynamic nature of satellite-ground converged networks, coupled with factors such as network topology changes caused by high-speed satellite movement, the uneven global distribution of users, and the differentiated needs of network slicing, makes traditional terrestrial network function deployment algorithms impractical for onboard deployment.

[0003] Based on the above background, it is necessary to develop a new network slicing function deployment method to achieve flexible deployment of network slicing functions based on satellite or ground nodes, thereby effectively improving network slicing performance. Summary of the Invention

[0004] The purpose of the present invention is to provide a network slicing function deployment method for a satellite-ground fusion network, which can flexibly deploy network slicing functions by selecting appropriate satellites or ground nodes according to business needs, thereby effectively improving network slicing performance.

[0005] To achieve the above-mentioned purpose, the present invention provides a method for deploying network slicing functions in a satellite-ground fusion network, comprising:

[0006] S1: Obtain network status information and network slice deployment requirements of the satellite-ground fusion network;

[0007] S2: Based on the network status information and network slice deployment requirements in step S1, execute the network slice function deployment algorithm to obtain the deployment result of the network slice instance;

[0008] The network slicing function deployment algorithm is a hybrid algorithm based on multi-swarm particle swarm and genetic algorithms. The network slicing function deployment algorithm consists of multiple sub-populations, each of which represents a variable in the optimization problem. Each sub-population adopts different update strategies and optimization parameters based on the variable type of the optimization problem.

[0009] S3: Send the deployment result of the network slice instance obtained in step S2 to each network node to complete the deployment.

[0010] The network status information of the satellite-ground fusion network includes: the network topology of the satellite-ground fusion network, computing resource information of satellites or ground nodes, link delay, and user distribution; the network topology refers to the connection relationship between satellites and satellites and between satellites and ground nodes; the user distribution refers to the geographical location, density and service demand characteristics of terminal devices in the network;

[0011] Network slicing deployment requirements include: the service area of ​​the network slice instance, the number of deployed network slice instances, the type of network functions deployed, the number of deployed network functions, and the isolation requirements and performance requirements of each network slice instance.

[0012] The network status information of the network slice instance of the satellite-ground fusion network during the deployment cycle is calculated based on the historical network status and satellite operating parameters; satellite operating parameters include: satellite orbit type and altitude, satellite position; methods for calculating the network status information of the satellite-ground fusion network during the deployment cycle include: calculation based on satellite ephemeris information, prediction based on AI methods, or simulation and prediction based on digital twin technology.

[0013] The deployment results of the network slice instance specifically include the number of network function instances contained in each network slice instance, the network function instances allocated to each network slice instance, and the deployment location and allocated resources of each network function instance.

[0014] The step S2 specifically includes:

[0015] S21: Determine the variables and variable types of the optimization problem according to the optimization problem, and set the number of subpopulations to be equal to the number of variables;

[0016] S22: For each variable, determine the particle position and velocity update method in the PSO phase and the crossover and mutation method in the GA phase;

[0017] S23: setting optimization parameters and initializing subpopulations to generate initial subpopulations;

[0018] S24: The position of each particle in the subpopulation is taken as the solution to the optimization problem and substituted into the fitness function to calculate the fitness value of the solution;

[0019] S25: Update the individual and group optimal positions of each particle;

[0020] S26: Each subpopulation updates the velocity and position of each particle;

[0021] S27: Determine whether the maximum number of iterations has been reached. If so, output the optimal solution as the deployment result of the network slice instance. Otherwise, first execute steps S28 to S211 to implement the optimization in the GA phase, and then re-execute steps S24 to S27 to implement the optimization in the PSO phase.

[0022] S28: Each subpopulation generates offspring using the crossover and mutation method determined in step S22;

[0023] S29: Take the candidate solutions corresponding to all offspring genes of each subpopulation as the solution to the optimization problem and calculate its fitness value, and select excellent gene individuals from the offspring and parent generations to obtain the elite population;

[0024] S210: Determine whether the maximum number of iterations in the GA phase has been reached. If the maximum number of iterations has been reached, output the optimal solution of the GA phase optimization. If the maximum number of iterations has not been reached, execute steps S28 to S210 again.

[0025] S211: Determine whether the optimal solution optimized in the GA stage is better than the global optimal position of the PSO population. If so, update the global optimal position of the PSO population.

[0026] The variable types include integer variables and binary variables. The particle speed update method of integer variables and binary variables in the PSO stage adopts the same speed update method. The speed update formula of each particle is expressed as:

[0027] ,

[0028] in, Represents particles exist The speed of the iteration, Represents particles exist The position at the iteration, is the inertia weight, and are the individual and global learning factors, and for A random number uniformly distributed in the interval, and Represent particles The individual optimal position of and the global optimal position of all particles in the particle swarm;

[0029] The position update method of the particle of integer variable in the PSO stage satisfies the following formula:

[0030]

[0031] The position update method of the binary variable particles in the PSO stage satisfies the following formula:

[0032] ,

[0033] in, Represents particles exist The position at the iteration, Represents particles exist The speed of the iteration, is Random numbers generated in the interval, yes Substitution The value of the function;

[0034] The crossover method for integer variables and binary variables in the GA stage adopts the multi-point crossover method, the mutation method for integer variables in the GA stage adopts the Gaussian mutation method, and the mutation method for binary variables in the GA stage adopts the basic bit mutation method.

[0035] The optimization parameters include the number of algorithm iterations, the number of GA stage iterations, the number of PSO population particles, the inertia weight, the individual and global learning factors, the crossover probability and mutation probability in the GA stage.

[0036] When the optimization problem contains constraints, the fitness function is:

[0037]

[0038] in, represents an optimization function corresponding to the objective of the optimization problem; represents the constraint violation function, defined as the solution the number of constraint violations; is the constraint penalty, which is a constant; is the constraint penalty factor; Set to 200; Set to 10; represents the set of all feasible solutions that do not violate the constraints; represents the set of solutions that only violate the soft constraints; Represents a set that violates a hard constraint.

[0039] In step S3, the deployment result of the network slice instance obtained in step S2 is sent from the deployment location of the network slice function deployment algorithm to each network node through the control plane to update the network slice function deployment.

[0040] The deployment locations of the network slicing function deployment algorithm include: deployment in a ground control center or deployment on a satellite.

[0041] The execution method of the network slice function deployment algorithm includes: periodically running the network slice function deployment algorithm and regularly updating the network slice function deployment; and / or executing the network slice function deployment algorithm and updating the network slice function deployment when there is a new deployment request or user demand changes significantly.

[0042] The present invention proposes a hybrid algorithm based on a multi-population PSO and GA for deploying network slicing functions in a satellite-ground fusion network. This algorithm utilizes a multi-population optimization strategy to simultaneously optimize multiple variables in an optimization problem, improving optimization performance and speed. Each subpopulation can be flexibly customized based on the type of optimization variable, setting optimization parameters and methods suitable for solving each variable, thereby improving optimization performance. By optimizing the PSO's global optimal position using a genetic algorithm, the PSO can escape from local optimal solutions, enabling flexible deployment of network slicing functions by selecting appropriate satellites or ground nodes based on service needs, effectively improving network slicing performance.

[0043] In addition, the present invention provides a method for deploying network slicing functions in a satellite-ground fusion network, targeting the optimization deployment requirements of slicing functions in the satellite-ground fusion network, and is suitable for a fitness function representation method for solving constrained optimization problems. This method can assist the optimization algorithm in optimization, guiding the algorithm to prioritize satisfying system constraints before satisfying performance constraints, and can also quickly find a feasible solution through flexible definition of constraint violation functions. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is an example diagram of an application scenario of a network slicing function deployment method for a satellite-ground fusion network proposed in the present invention.

[0045] Figure 2 It is a flowchart of the network slicing function deployment method of the satellite-ground fusion network of the present invention.

[0046] Figure 3 This is the architecture diagram of the multi-population particle swarm genetic hybrid algorithm. DETAILED DESCRIPTION

[0047] The following is a detailed description of an embodiment of the present invention. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process. However, the protection scope of the present invention is not limited to the following embodiment.

[0048] The application scenarios of the network slicing function deployment method of the satellite-ground fusion network of the present invention are as follows: Figure 1As shown (taking three network slice instances as an example), this scenario primarily consists of a satellite network, a terrestrial core network, and users. Each network slice can deploy a different number of network function instances on the satellite or terrestrial core network based on its specific needs, thus forming a network slice instance. Users can access a network slice via satellite and sequentially access different network function instances within the slice instance according to the corresponding signaling process. A network function is a logical module that implements a specific function within the network, such as the access and mobility management function (AMF), session management function (SMF), and user plane function (UPF). A network function instance is a specific running instance of a network function and must be a running program instance. A network function instance is the actual deployment of a network function on physical or virtual resources. A single network function can create multiple network function instances. For example, a single AMF network function may be instantiated into multiple AMF network function instances. It should be noted that for multiple network function instances created by the same network function (e.g., multiple AMF network function instances created by an AMF network function), although the network function remains the same, each network function instance serves different users.

[0049] In this invention, the deployment of network slices follows the following basic principles:

[0050] (1) Each network function instance can only be deployed on one network node;

[0051] (2) The resources allocated to all network function instances deployed on a network node cannot exceed the resource capacity of the node.

[0052] (3) The resources allocated to each network function instance cannot be less than the minimum resources required for the deployment of the network function;

[0053] (4) Each network slice instance is assigned at least one network function instance of the network function it requires;

[0054] (5) When different network slice instances are isolated from each other, each network function instance can only be assigned to one network slice instance;

[0055] (6) The deployed network slice instance should meet the performance requirements of the network slice.

[0056] Network slice deployment requirements determine whether network slice instances are isolated from each other. When a network slice has isolation requirements, the network function instance of that network slice is exclusive to that network slice. When a network slice does not have isolation requirements, the network function instance of that network slice can be shared with other network slices.

[0057] Therefore, all network function instances deployed on the same network node can belong to different network slices, and each network function instance can be exclusively occupied by one network slice instance or shared by multiple network slice instances according to the isolation requirements.

[0058] like Figure 1 As shown, AMF1 and AMF2 are two network function instances of the network function AMF, which are assigned to network slice instance 1 and network slice instance 2 respectively. If a single network function instance is assigned to multiple network slice instances, taking the network function instance AMF1 as an example, multiple network slice instances share the network function instance AMF1.

[0059] like Figure 2 As shown, a network slicing function deployment method for a satellite-ground fusion network of the present invention specifically includes:

[0060] Step S1: Obtain network status information and network slice deployment requirements of the satellite-ground fusion network;

[0061] The network status information of the satellite-ground converged network includes, but is not limited to, the network topology, computing resource information of satellites or ground nodes, link latency, user distribution, etc. The network topology refers to the connection between satellites and between satellites and ground nodes; user distribution refers to the geographical location, density, and service demand characteristics of terminal devices in the network.

[0062] Among them, the network status information of the network slice instance of the satellite-ground fusion network during the deployment cycle is calculated through historical network status and satellite operating parameters.

[0063] Satellite operating parameters include but are not limited to: satellite orbit type and altitude, satellite position, etc. Satellite position refers to the spatial coordinates of the satellite at a certain moment, which can be calculated from orbital parameters (such as six-element numbers) or real-time ephemeris data (such as two-line orbital element numbers).

[0064] Methods for calculating network status information for satellite-ground converged networks during their deployment cycle include, but are not limited to, calculations based on satellite ephemeris information, predictions based on AI methods, and simulation and predictions based on digital twin technology. Satellite ephemeris information can be used to calculate network topology and link latency within the network status information; predictions based on AI methods can assist in predicting changes in network topology and link latency within the network status information, as well as predicting user distribution; and simulation and predictions based on digital twin technology can simulate network topology, link latency, and user distribution.

[0065] Network slicing deployment requirements include, but are not limited to, the service area of ​​the network slice instance, the number of deployed network slice instances, the types of network functions deployed, the number of each network function deployed, and the isolation and performance requirements for each network slice instance. The service area of ​​a network slice instance refers to the geographic area served by the network slice instance or the geographic area where the users of the network slice instance are distributed. The types of network functions deployed refer to core network functions (AMF, SMF, UPF, etc.) or other new network functions (such as AI agents). The isolation requirement of a network slice instance refers to whether the network slice instance exclusively occupies a network function instance or can share the same network function instance with other network slice instances. Performance requirements include the latency requirements, bandwidth requirements, and number of users served by the network slice instance. The difference between the network slice deployment requirements and the optimization problem described below (i.e., the slice function deployment problem for a satellite-ground converged network) is that the network slice deployment requirements are known, while the optimization problem requires solving.

[0066] Step S2: Based on the network status information and network slice deployment requirements of step S1, execute the network slice function deployment algorithm to obtain the deployment result of the network slice instance.

[0067] Among them, the deployment results of the network slice instance specifically include the number of network function instances contained in each network slice instance, the network function instances allocated to each network slice instance, the deployment location of each network function instance and the allocated resources, etc.

[0068] Among them, the network slicing function deployment algorithm is a hybrid algorithm based on multi-swarm particle swarm optimization (PSO) and genetic algorithm (GA). Its algorithm framework is as follows: Figure 3 The network slicing function deployment algorithm consists of multiple subpopulations, each of which represents a variable in the optimization problem. Each subpopulation adopts different update strategies and optimization parameters based on the variable type of the optimization problem, thereby achieving efficient optimization of multiple variables in the optimization problem.

[0069] Step S2 specifically includes:

[0070] Step S21: Determine the variables and variable types of the optimization problem according to the optimization problem, and set the number of subpopulations to be equal to the number of variables.

[0071] Therefore, the network slicing function deployment algorithm consists of multiple subpopulations, each representing a variable in an optimization problem. For each subpopulation, the positions of different particles represent the solution to the same variable in the optimization problem. All subpopulations form a population.

[0072] The variables of the optimization problem include but are not limited to the deployment location of the network function instance, whether the network function instance is assigned to a slice instance, the computing resources allocated to the network function instance, etc.

[0073] The variables and variable types of the optimization problem are determined according to the optimization problem (i.e., the slice function deployment problem of the satellite-ground integrated network); for example, if the means of the optimization problem include optimizing the deployment location of the network slice instance, then the variable of the optimization problem is the deployment location of the network slice instance, and the variable type is an integer variable, which is one of the variables; if the means of the optimization problem include optimizing the allocation of network function instances, then the variable of the optimization problem is whether the network function instance is allocated to a certain slice instance, and the variable type of the optimization problem is a 0, 1 binary variable, which is one of the variables; if the means of the optimization problem include optimizing the computing resources allocated to each network function instance, then the variable of the optimization problem is the computing resources allocated to the network function instance, and the variable type of the optimization problem is an integer variable, which is one of the variables.

[0074] In this embodiment, the number of variable types is represented as n, and thus the number of subpopulations is n.

[0075] Step S22: For each variable, determine the particle position and velocity update method of the variable in the PSO stage and the crossover and mutation method in the GA stage.

[0076] The position and velocity update methods of particles in the PSO phase, as well as the crossover and mutation methods in the GA phase, can be flexibly defined based on the optimization variable type and optimization requirements. The position of particles in the PSO phase corresponds to the solution of the variable, while the velocity of particles in the PSO phase represents the direction and magnitude of the solution update.

[0077] Taking the optimization problem of optimizing the deployment location of network slice instances (integer variables) and optimizing the allocation of network function instances (binary variables, i.e., whether to allocate to a certain network slice instance) as an example, a feasible variable type in the particle position and velocity update method in the PSO phase and the crossover and mutation method in the GA phase are as follows:

[0078] The speed update method of the particles of integer variables and binary variables in the PSO stage adopts the same speed update method. The speed update formula of each particle is expressed as:

[0079] ,

[0080] in, Represents particles exist The speed of the iteration, Represents particles exist The position at the iteration, is the inertia weight, and are the individual and global learning factors, and for A random number uniformly distributed in the interval, and Represent particles The individual optimal position of and the global optimal position of all particles in the particle swarm.

[0081] The position update method of the particle of integer variable in the PSO stage satisfies the following formula:

[0082] ,

[0083] The position update method of the binary variable particles in the PSO stage satisfies the following formula:

[0084] ,

[0085] in, Represents particles exist The position at the iteration, Represents particles exist The speed of the iteration, is Random numbers generated in the interval, yes Substitution The value of the function.

[0086] The crossover method for integer variables and binary variables in the GA stage adopts the multi-point crossover method, the mutation method for integer variables in the GA stage adopts the Gaussian mutation method, and the mutation method for binary variables in the GA stage adopts the basic bit mutation method.

[0087] Step S23: Set optimization parameters and initialize subpopulations to generate initial subpopulations.

[0088] The optimization parameters include the number of algorithm iterations, the number of GA stage iterations, the number of PSO population particles, the inertia weight, the individual and global learning factors, the crossover probability and mutation probability in the GA stage, etc.

[0089] The subpopulation is randomly initialized under the condition of satisfying the set optimization parameters to generate the initial subpopulation.

[0090] Step S24: The position of each particle in the sub-population is used as a solution to the optimization problem and substituted into the fitness function to calculate the fitness value of the solution.

[0091] As described above, the network slicing function deployment algorithm consists of multiple subpopulations, each representing a variable in the optimization problem. For each subpopulation, the positions of different particles represent different candidate solutions to the same variable in the optimization problem. Therefore, step S24 calculates the fitness value of the solution by substituting the particle positions into the fitness function as the solution to the optimization problem.

[0092] The optimization problem of a satellite-ground converged network usually includes constraints corresponding to the network slice deployment requirements. In one embodiment, when the optimization problem includes constraints, the fitness function is:

[0093]

[0094] in, represents an optimization function corresponding to the objective of the optimization problem; represents the constraint violation function, defined as the solution the number of constraint violations; is the constraint penalty, which is a constant; is the constraint penalty factor; represents the set of all feasible solutions that do not violate the constraints; Represents the set of solutions that only violate soft constraints (constraints that do not violate system constraints and only affect performance can be defined as soft constraints, such as the latency constraint of network slicing); Represents a set of violations of hard constraints (constraints that cannot be run if the constraint system is violated can be defined as hard constraints).

[0095] In one embodiment, the goal of the optimization problem is to optimize the average latency of the network slice. The corresponding fitness function and the constraints included in the optimization problem are:

[0096] ,

[0097] ,

[0098] in, Represents a collection of network slices; Indicates the number of network slices; Indicates the sequence number of the network slice; Indicates the latency of the network slice; In fact, it is the optimization function corresponding to the goal of the optimization problem; represents the constraint violation function, defined as the solution the number of constraint violations; Indicates network functionality; Represents a collection of network functions; Indicates the serial number of the network function instance; Indicates network functionality Examples ; Indicates network functionality A collection of instances of ; represents the Kronecker function; Indicates the node number; Represents a collection of network nodes; Indicates network functionality Examples Deployment location; Indicates network functionality Examples allocated computing resources; Represents a network node computing resources; Indicates network functionality The minimum resources required for deployment; Represents network slicing Delay requirements; Indicates network functionality Examples Whether to allocate to network slice .

[0099] The constraints are described as follows: The first constraint C1 indicates that the sum of the computing resources allocated to the network function instances deployed on each network node cannot exceed the computing resources of the network node; the second constraint C2 indicates that the computing resources allocated to the network function instance cannot be less than the minimum resources required for the deployment of the network function; the third constraint C3 indicates that the network slice latency should meet the latency requirements; the fourth constraint C4 indicates that each network slice instance is allocated at least one network function instance of the required network function; the fifth constraint C5 represents the network slice isolation constraint. This constraint exists when the network slice has isolation requirements, otherwise it does not exist. This constraint indicates that when the network slice is isolated, a network function instance can only be allocated to one network slice instance. For the above constraints, any system that violates constraints C1, C2, C4 and C5 cannot operate normally, which is expressed as a hard constraint; the system that violates constraint C3 can operate normally, but the requirements of the network slice instance cannot be met. This constraint is defined as a soft constraint.

[0100] Step S25: Update the individual and group optimal positions of each particle.

[0101] Step S26: Each subpopulation updates the speed and position of each particle.

[0102] Step S27: Determine whether the maximum number of iterations has been reached. If so, output the optimal solution as the deployment result of the network slice instance. Otherwise, first execute steps S28 to S211 to achieve optimization in the GA stage, and then re-execute steps S24 to S27 to achieve optimization in the PSO stage.

[0103] Among them, the optimal solution represents all the optimal variables of the optimization problem, and therefore includes the deployment results of the optimized network slice instance, that is, the number of network function instances contained in each network slice instance, the network function instances allocated to each network slice instance, the deployment location of each network function instance and the allocated resources, etc.

[0104] Step S28: Each subpopulation generates offspring using the crossover and mutation method determined in step S22;

[0105] Step S29: Take the candidate solutions corresponding to all offspring genes of each subpopulation as the solution to the optimization problem and calculate its fitness value, select excellent gene individuals from the offspring and parent generations to obtain the elite population for the next iteration.

[0106] Obtaining the elite population includes, but is not limited to: selecting according to the principle of optimal fitness or selecting based on a tournament algorithm, and the number of individuals in the elite population is equal to the number of particles in the PSO population.

[0107] Step S210: Determine whether the maximum number of iterations in the GA stage has been reached. If the maximum number of iterations has been reached, output the optimal solution of the optimization in the GA stage. If the maximum number of iterations has not been reached, execute steps S28 to S210 again.

[0108] Step S211: Determine whether the optimal solution optimized in the GA stage is better than the global optimal position of the PSO population. If so, update the global optimal position of the PSO population.

[0109] Step S3: Send the deployment result of the network slice instance obtained in step S2 to each network node to complete the deployment.

[0110] The deployment result of the network slice instance obtained in step S2 is sent from the deployment location (ground control center or satellite) of the network slice function deployment algorithm to each network node through the control plane to update the network slice function deployment.

[0111] The deployment location of the network slicing function deployment algorithm includes but is not limited to: deployment in a ground control center or deployment on a satellite.

[0112] The execution method of the network slice function deployment algorithm includes but is not limited to: periodically running the network slice function deployment algorithm and regularly updating the network slice function deployment; and / or executing the network slice function deployment algorithm and updating the network slice function deployment when there is a new deployment request or user demand changes significantly.

[0113] The present invention proposes a hybrid algorithm based on a multi-population PSO and GA for deploying network slicing functions in a satellite-ground fusion network. This algorithm utilizes a multi-population optimization strategy to simultaneously optimize multiple variables in an optimization problem, improving optimization performance and speed. Each subpopulation can be flexibly customized based on the type of optimization variable, setting optimization parameters and methods suitable for solving each variable, thereby improving optimization performance. By optimizing the PSO's global optimal position using a genetic algorithm, the PSO can escape from local optimal solutions, enabling flexible deployment of network slicing functions by selecting appropriate satellites or ground nodes based on service needs, effectively improving network slicing performance.

[0114] In addition, the present invention provides a method for deploying network slicing functions in a satellite-ground fusion network, targeting the optimization deployment requirements of slicing functions in the satellite-ground fusion network, and is suitable for a fitness function representation method for solving constrained optimization problems. This method can assist the optimization algorithm in optimization, guiding the algorithm to prioritize satisfying system constraints before satisfying performance constraints, and can also quickly find a feasible solution through flexible definition of constraint violation functions.

[0115] Experimental results:

[0116] The following is a detailed description of an embodiment of the present invention. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process. However, the protection scope of the present invention is not limited to the following embodiment.

[0117] This example demonstrates the implementation of two network slice instances deployed across the OneWeb constellation and three ground nodes. The OneWeb constellation consists of 720 satellites distributed across 18 orbital planes at an altitude of 1,200 kilometers, with 40 satellites per plane. The three ground nodes are located in Beijing, Urumqi, and Hainan.

[0118] A method for deploying network slicing functions in a satellite-ground fusion network according to the present invention specifically includes:

[0119] Step S1: Obtain the network status information and network slice deployment requirements of the satellite-ground fusion network.

[0120] The network topology and resource information for each network node within a two-hour deployment cycle for a network slice instance of a satellite-ground converged network are predicted and calculated based on historical network status and satellite operating parameters. Network slice deployment requirements include deploying two network slices, each containing network function instances of three network functions: AMF, SMF, and UPF. Twenty network function instances can be deployed for each type of network function. Network slices are isolated from each other, with latency requirements of 60 milliseconds and 100 milliseconds, respectively.

[0121] Step S2: Execute the network slice deployment algorithm to obtain the deployment result of the network slice.

[0122] Step S2 specifically includes:

[0123] Step 21: Determine the variables and variable types of the optimization problem based on the optimization problem, and set the number of subpopulations equal to the number of variables.

[0124] Among them, the optimization problem variables include three types of variables: the deployment location of the network function instance, the allocation of the network function instance, and the computing resource configuration of the network function instance. Three sub-populations are used for optimization.

[0125] Step S22: For each variable, determine the particle position and velocity update method of the variable in the PSO stage and the crossover and mutation method in the GA stage.

[0126] The deployment location of the network function instance and the computing resources allocated to the network function instance are both integer variables, and the same update strategy is adopted in the PSO and GA stages. However, whether the network function instance is allocated to a slice instance is a binary variable (if allocated, its value is 1, otherwise it is 0, so it is a binary variable), and a different update strategy is adopted.

[0127] The speed update method of the particles of integer variables and binary variables in the PSO stage adopts the same speed update method. The speed update formula of each particle is expressed as:

[0128] ,

[0129] in, Represents particles exist The speed of the iteration, Represents particles exist The position at the iteration, is the inertia weight, and are the individual and global learning factors, and for A random number uniformly distributed in the interval, and Represent particles The individual optimal position of and the global optimal position of all particles in the particle swarm.

[0130] The position update method of the particle of integer variable in the PSO stage satisfies the following formula:

[0131] ,

[0132] The position update method of the binary variable particles in the PSO stage satisfies the following formula:

[0133] ,

[0134] in, Represents particles exist The position at the iteration, Represents particles exist The speed of the iteration, is Random numbers generated in the interval, yes Substitution The value of the function.

[0135] The crossover method for integer variables and binary variables in the GA stage adopts the multi-point crossover method, the mutation method for integer variables in the GA stage adopts the Gaussian mutation method, and the mutation method for binary variables in the GA stage adopts the basic bit mutation method.

[0136] Step S23: Set optimization parameters and initialize subpopulations to generate initial subpopulations.

[0137] The optimization parameters include the number of algorithm iterations, the number of GA phase iterations, the number of PSO swarm particles, the inertia weight, the individual and global learning factors, the crossover probability and mutation probability in the GA phase, etc. The number of algorithm iterations is set to 100, the number of GA phase iterations is set to 30, the number of PSO swarm particles is set to 50, the inertia weight is set to 0.5, the individual and global learning factors are set to 1, and the crossover probability and mutation probability in the GA phase are set to 0.9 and 0.1, respectively.

[0138] Step S24: The position of each particle in the sub-population is used as a solution to the optimization problem and substituted into the fitness function to calculate the fitness value of the solution.

[0139] When the optimization problem contains constraints, the fitness function is:

[0140]

[0141] in, represents an optimization function corresponding to the objective of the optimization problem; represents the constraint violation function, defined as the solution the number of constraint violations; is the constraint penalty, which is a constant; is the constraint penalty factor; Set to 200; Set to 10; represents the set of all feasible solutions that do not violate the constraints; represents the set of solutions that only violate the soft constraints. Here, it is the solution that only violates the delay constraint of the network slice. Represents the set of violations of hard constraints, which here are solutions that violate all constraints except the network slice delay constraint.

[0142] Step S25: Update the individual and group optimal positions of each particle.

[0143] Step S26: Each subpopulation updates the speed and position of each particle.

[0144] Step S27: Determine whether the maximum number of iterations has been reached. If so, output the optimal solution as the deployment result of the network slice instance. Otherwise, first execute steps S28 to S211 to achieve optimization in the GA stage, and then re-execute steps S24 to S27 to achieve optimization in the PSO stage.

[0145] Step S28: Each subpopulation generates offspring using the crossover and mutation method determined in step S22;

[0146] Step S29: The candidate solutions corresponding to all offspring genes in each subpopulation are used as the solution to the optimization problem and their fitness values ​​are calculated. The elite population is selected from both the offspring and parent generations to form the elite population for the next iteration. The number of individuals in the elite population is 50. The elite population is selected based on the principle of optimal fitness.

[0147] Step S210: Determine whether the maximum number of iterations in the GA stage has been reached. If the maximum number of iterations has been reached, output the optimal solution of the optimization in the GA stage. If the maximum number of iterations has not been reached, execute steps S28 to S210.

[0148] Step S211: Determine whether the optimal solution optimized in the GA stage is better than the global optimal position of the PSO population. If so, update the global optimal position of the PSO population.

[0149] Step S3: Send the deployment result of the network slice instance obtained in step S2 to each network node to complete the deployment.

[0150] The network slice function deployment algorithm is deployed in the ground control center, which periodically runs the network slice function deployment algorithm and regularly updates the network slice function deployment. The deployment results of the network slice instance are sent from the ground control center to each network node via the control plane to complete the network slice deployment.

[0151] The above description is merely a preferred embodiment of the present invention and is not intended to limit its scope. Various variations are possible. For example, the laser sensor used in the present invention can be replaced with an ultrasonic sensor or other structure, and the colorimetry device model can be replaced. In other words, any simple, equivalent variations and modifications based on the claims and description of the present invention fall within the scope of protection of the present invention. Anything not fully described herein constitutes conventional technology.

Claims

1. A method for deploying network slicing functions in a satellite-ground fusion network, characterized in that: include: Step S1: Obtain network status information and network slice deployment requirements of the satellite-ground fusion network; Step S2: Based on the network status information and network slice deployment requirements of step S1, execute the network slice function deployment algorithm to obtain the deployment result of the network slice instance; The network slicing function deployment algorithm is a hybrid algorithm based on multi-swarm particle swarm and genetic algorithms. The network slicing function deployment algorithm consists of multiple sub-populations, each of which represents a variable in the optimization problem. Each sub-population adopts different update strategies and optimization parameters based on the variable type of the optimization problem. Step S3: Send the deployment result of the network slice instance obtained in step S2 to each network node to complete the deployment.

2. The method for deploying network slicing functions in a satellite-ground fusion network according to claim 1, characterized in that: The network status information of the satellite-ground fusion network includes: the network topology of the satellite-ground fusion network, computing resource information of satellites or ground nodes, link delay, and user distribution; the network topology refers to the connection relationship between satellites and satellites and between satellites and ground nodes; the user distribution refers to the geographical location, density and service demand characteristics of terminal devices in the network; Network slicing deployment requirements include: the service area of ​​the network slice instance, the number of deployed network slice instances, the type of network functions deployed, the number of deployed network functions, and the isolation requirements and performance requirements of each network slice instance.

3. The method for deploying network slicing functions in a satellite-ground fusion network according to claim 1, wherein: Calculate the network status information of the network slice instance of the satellite-ground fusion network during the deployment cycle based on the historical network status and satellite operating parameters; Satellite operating parameters include: satellite orbit type and altitude, satellite position; Methods for calculating network status information of a satellite-ground fusion network during its deployment cycle include: calculation based on satellite ephemeris information, prediction based on AI methods, or simulation and prediction based on digital twin technology.

4. The method for deploying network slicing functions in a satellite-ground fusion network according to claim 1, wherein: The deployment results of the network slice instance specifically include the number of network function instances contained in each network slice instance, the network function instances allocated to each network slice instance, and the deployment location and allocated resources of each network function instance.

5. The method for deploying network slicing functions in a satellite-ground fusion network according to claim 1, wherein: The step S2 specifically includes: Step S21: determining the variables and variable types of the optimization problem according to the optimization problem, and setting the number of subpopulations to be equal to the number of variables; Step S22: for each variable, determine the particle position and velocity update method of the variable in the PSO stage and the crossover and mutation method in the GA stage; Step S23: setting optimization parameters and initializing subpopulations to generate initial subpopulations; Step S24: taking the position of each particle in the sub-population as the solution to the optimization problem, substituting it into the fitness function to calculate the fitness value of the solution; Step S25: Update the individual and group optimal positions of each particle; Step S26: Each subpopulation updates the speed and position of each particle; Step S27: Determine whether the maximum number of iterations has been reached. If so, output the optimal solution as the deployment result of the network slice instance. Otherwise, first execute steps S28 to S211 to implement the optimization in the GA phase, and then re-execute steps S24 to S27 to implement the optimization in the PSO phase. Step S28: Each subpopulation generates offspring using the crossover and mutation method determined in step S22; Step S29: taking the candidate solutions corresponding to all offspring genes of each subpopulation as the solution to the optimization problem and calculating their fitness values, and selecting excellent gene individuals from the offspring and parent generations to obtain an elite population; Step S210: Determine whether the maximum number of iterations in the GA stage has been reached, and output the optimal solution of the optimization in the GA stage if the maximum number of iterations has been reached. If the maximum number of iterations has not been reached, execute steps S28 to S210 again. Step S211: Determine whether the optimal solution optimized in the GA stage is better than the global optimal position of the PSO population. If so, update the global optimal position of the PSO population.

6. The method for deploying network slicing functions in a satellite-ground fusion network according to claim 5, characterized in that: The variable types of the variables include integer variables and binary variables; The speed update method of the particles of integer variables and binary variables in the PSO stage adopts the same speed update method. The speed update formula of each particle is expressed as: ， in, Represents particles exist The speed of the iteration, Represents particles exist The position at the iteration, is the inertia weight, and are the individual and global learning factors, and for A random number uniformly distributed in the interval, and Represent particles The individual optimal position of and the global optimal position of all particles in the particle swarm; The position update method of the particle of integer variable in the PSO stage satisfies the following formula: ， The position update method of the binary variable particles in the PSO stage satisfies the following formula: ， in, Represents particles exist The position at the iteration, Represents particles exist The speed of the iteration, is Random numbers generated in the interval, yes Substitution The value of the function; The crossover method for integer variables and binary variables in the GA stage adopts the multi-point crossover method, the mutation method for integer variables in the GA stage adopts the Gaussian mutation method, and the mutation method for binary variables in the GA stage adopts the basic bit mutation method; The optimization parameters include the number of algorithm iterations, the number of GA stage iterations, the number of PSO population particles, the inertia weight, the individual and global learning factors, the crossover probability and mutation probability in the GA stage.

7. The method for deploying network slicing functions in a satellite-ground fusion network according to claim 5, characterized in that: When the optimization problem contains constraints, the fitness function is: ， in, represents an optimization function corresponding to the objective of the optimization problem; represents the constraint violation function, defined as the solution the number of constraint violations; is the constraint penalty, which is a constant; is the constraint penalty factor; Set to 200; Set to 10; represents the set of all feasible solutions that do not violate the constraints; represents the set of solutions that only violate the soft constraints; Represents a set that violates a hard constraint.

8. The method for deploying network slicing functions in a satellite-ground fusion network according to claim 1, wherein: In step S3, the deployment result of the network slice instance obtained in step S2 is sent from the deployment location of the network slice function deployment algorithm to each network node through the control plane to update the network slice function deployment.

9. The method for deploying network slicing functions in a satellite-ground fusion network according to claim 8, characterized in that: The deployment locations of the network slicing function deployment algorithm include: deployment in a ground control center or deployment on a satellite.

10. The method for deploying network slicing functions in a satellite-ground fusion network according to claim 1, wherein: The execution method of the network slice function deployment algorithm includes: periodically running the network slice function deployment algorithm and regularly updating the network slice function deployment; and / or executing the network slice function deployment algorithm and updating the network slice function deployment when there is a new deployment request or user demand changes significantly.

Images