A multi-controller deployment method for SD-MANET based on improved Pied Kingfisher algorithm

Abstract

The present invention discloses a method for deploying multiple controllers in an SD-MANET (Mobile Self-Organizing Network) based on an improved Pied Kingfisher algorithm, belonging to the technical field of mobile self-organizing networks. The method comprises: obtaining network topology information of the SD-MANET; constructing an undirected graph representing the network topology; defining the controller deployment problem as a multi-objective optimization problem; solving the multi-objective optimization problem using the improved Pied Kingfisher optimization algorithm to obtain an optimal controller deployment solution; and deploying controllers and their management domains in the SD-MANET environment based on the optimal controller deployment solution. The present invention simultaneously optimizes three key metrics: link failure rate, latency, and load balancing, achieving significant results in reducing link failure rate and latency, prioritizing link stability and latency performance, and thereby improving the overall reliability and efficiency of the SD-MANET network.

Description

Technical Field

[0001] The present invention relates to a SD-MANET controller deployment method, in particular to a SD-MANET multi-controller deployment method based on an improved Pied Kingfisher algorithm, and belongs to the technical field of mobile self-organizing networks. Background Art

[0002] With the rapid development of information technology, especially the widespread application of emerging technologies such as big data, cloud computing, and artificial intelligence, the nature of modern warfare is undergoing profound changes. Against this backdrop, manned and unmanned collaborative combat systems have emerged as a crucial innovation in modern warfare. Collaborative perception is a crucial component of this system. It requires different combat units to continuously share battlefield information and status in real time, ensuring that every decision is based on global, real-time data. However, complex battlefield environments, such as strong adversarial forces and volatile climates, can easily damage combat units and disrupt communication links, making it impossible to provide real-time and continuous battlefield information. Traditional mobile ad hoc networks (MANETs) also struggle to address these challenges. Therefore, software-defined networking (SDN) has been introduced into the MANET architecture, resulting in software-defined mobile ad hoc networks (SD-MANETs). By separating the control and data planes, SDN provides centralized control and intelligent resource scheduling, enabling the network to dynamically adjust resource allocation from a global perspective and adapt to changing battlefield conditions. However, in SD-MANET networks, the deployment location and number of controllers is an important research issue, because combat units in battlefield environments are vulnerable to damage, communication links are also vulnerable to damage, and some nodes may receive a large amount of data, resulting in excessive load and failure, which in turn makes it impossible to transmit data in real time and continuously, making it impossible for individual nodes to make correct judgments.

[0003] Existing SDN controller deployment research generally suffers from a one-sided optimization objective. Most approaches focus on optimizing a single or two metrics: latency, load, or reliability. They lack a joint consideration and optimization mechanism for the inherent conflicts and synergies between these three factors, making it difficult to arrive at a deployment solution with optimal overall performance. Furthermore, these studies primarily focus on local area networks (LANs), aviation, and satellite networks. However, specialized research on controller deployment in software-defined mobile ad hoc networks (SD-MANETs) is even scarcer, making existing approaches difficult to directly apply and effectively address the unique challenges of tactical scenarios. Currently, most approaches to controller deployment in SD-MANETs focus solely on latency, without comprehensively considering other performance metrics. Summary of the Invention

[0004] Purpose of the invention: In response to the above problems, the purpose of the present invention is to provide an SD-MANET multi-controller deployment method based on the improved Pied Kingfisher algorithm, with the goal of reducing the control link failure rate and network latency and ensuring controller load balancing. By improving the Pied Kingfisher optimization algorithm, a multi-controller deployment algorithm is designed that can automatically determine the number and location of controllers.

[0005] Technical solution: The present invention provides a SD-MANET multi-controller deployment method based on an improved Pied Kingfisher algorithm, comprising the following steps:

[0006] Step 1: Obtain the network topology information of SD-MANET, including node set, node coordinates and link set;

[0007] Step 2: Construct an undirected graph representing the network topology, calculate the physical distance of the direct link using the node coordinates, and apply the Dijkstra algorithm to calculate the shortest path distance between any two nodes in the network;

[0008] Step 3: Define the controller deployment problem as a multi-objective optimization problem to determine the optimal controller deployment location set and the mapping relationship between each node and the controller;

[0009] Step 4: Use the improved Pied Kingfisher optimization algorithm to solve the multi-objective optimization problem and obtain the optimal controller deployment plan. Based on the optimal controller deployment plan, the controller and its management domain are deployed in the SD-MANET environment.

[0010] Furthermore, step 3 includes:

[0011] Calculate the propagation delay from the data plane node to the controller using the formula:

[0012] ,

[0013] Where V represents the data node A collection of, C represents the controller A collection of represents the i-th node, represents the j-th controller, Representation node With controller The shortest path distance between

[0014] Calculate the transmission delay from the data plane node to the controller using the following formula:

[0015] ,

[0016] Where, Representation node To the controller transmission delay;

[0017] Calculate the propagation delay between controllers using the formula:

[0018] ,

[0019] Where, Representation Controller With controller The shortest distance between

[0020] Calculate the transmission delay between controllers using the following formula:

[0021] ,

[0022] Where, Representation Controller To the controller transmission delay;

[0023] The total communication delay of the controller domain is:

[0024] .

[0025] Furthermore, step 3 also includes:

[0026] According to the failure probability of nodes in the control domain of each controller and link failure probability , computing controller The reliability of the control link within the domain is expressed as:

[0027] ,

[0028] Where, Indicates the node managed by the jth controller gather, represents the set of links e in the domain of controller j;

[0029] Calculate the control plane link failure rate of the SD-MANET network, expressed as:

[0030] ,

[0031] Where m represents the number of controllers in the SD-MANET network.

[0032] Furthermore, step 3 also includes:

[0033] Calculate the number of controllers in the SD-MANET network Load , the formula is:

[0034] ,

[0035] Where, Indicates the number of stream requests for the node, Representation node and Is there a link between them? Indicates that a link exists. Indicates that no link exists;

[0036] Compute Controller The utilization rate is:

[0037] ,

[0038] Where, Representation Controller Maximum load capacity;

[0039] The average utilization rate of all controllers in the SD-MANET network is calculated as follows:

[0040] ,

[0041] Calculate the load balancing degree of the controller using the following formula:

[0042] .

[0043] Furthermore, step 3 also includes:

[0044] A multi-objective optimization function F for controller deployment is constructed based on the total communication delay, control plane link failure rate, and load balancing rate within the controller domain, with minimizing this function as the optimization goal. The formula for the multi-objective optimization function is:

[0045] ,

[0046] Where, Represent the total communication delay , control plane link failure rate , controller load balancing The weight coefficient of

[0047] The constraints for constructing the multi-objective optimization function are:

[0048] ,

[0049] ,

[0050] ,

[0051] ,

[0052] ,

[0053] Where V represents the data plane node A collection of Indicates the controllers;

[0054] The first constraint states that each node can only be managed by one controller, the second constraint states that a controller must manage at least one node, the third constraint states that the number of controllers in the SD-MANET network is m; the fourth constraint states that the controller cannot exceed its load capacity; the fifth constraint states that the weighted sum of the weight coefficients of the multi-objective optimization function is 1.

[0055] Furthermore, step 4 includes:

[0056] Step 41: Set the maximum number of iterations, use the specific deployment location of each controller as an individual in the population, and introduce the Chebyshev chaos map to initialize the population. The formula is:

[0057] ,

[0058] Where, represents the position of the i-th individual in the j-th dimension, Respectively represent The upper and lower boundaries of the search range of the dimensional search space, represents the value within (-1,1) generated by the Chebyshev chaotic map, N represents the total number of individuals, Indicates the number of dimensions;

[0059] Step 42: In the exploration phase, the Levy flight strategy is introduced to update the population position. The update formula is:

[0060] ,

[0061] Where, is the position of the individual at the current iteration number s, is another individual randomly selected from the population position, used to guide Movement and exploration, is the state parameter, is a random pointing parameter, for Random values from a normal distribution; is the final step length of Levy flight, and the formula is:

[0062] ,

[0063] Where, and are all random numbers generated from a normal distribution. The parameters for Levy's flight;

[0064] Step 43: In the development phase, a Gaussian mutation strategy is introduced to update the population position. The update formula is:

[0065] ,

[0066] Where, In probability The new position obtained under the influence, is the standard deviation, and the formula is , is a coefficient used to control the size of the standard deviation. Respectively represent the upper and lower boundaries of the search range; rand represents a random number in [0,1]. is the mutation probability, indicating that it has a certain probability Perturb individual variables; represents the new position of the individual after the introduction of Gaussian mutation;

[0067] Step 44, in the symbiotic stage, the population position is updated at the current number of iterations. The update formula is:

[0068] ,

[0069] Where, Indicates the position of 2 individuals randomly selected from the population at the current iteration number s, The hunting efficiency of the kingfisher is Indicates hunting ability, Represents a random number in [0,1];

[0070] Step 45: merge the updated new population, calculate the fitness value of each individual in the new population, sort the fitness values from small to large, and take the individual corresponding to the smallest fitness value as the optimal feasible solution under the current number of iterations; determine whether it is greater than the number of iterations. If so, output the optimal feasible solution, otherwise return to the exploration stage and solve the iteration again.

[0071] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages:

[0072] (1) This paper addresses the controller deployment problem in SD-MANET networks and proposes a multi-objective optimization algorithm that aims to simultaneously optimize three key indicators: link failure rate, latency, and load balancing.

[0073] (2) The present invention improves the population initialization method of the Pied Kingfisher algorithm and combines Levy flight and Gaussian strategies to effectively improve the convergence and search efficiency of the algorithm, and can automatically determine the optimal number and location of controller deployment;

[0074] (3) The present invention has achieved remarkable results in reducing link failure rate and latency, giving priority to ensuring link stability and latency performance, thereby improving the overall reliability and efficiency of the SD-MANET network. In particular, its advantages are more obvious in scenarios with extremely high requirements for latency and reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] Figure 1 It is a flowchart of a multi-controller deployment method for SD-MANET based on the improved Pied Kingfisher algorithm;

[0076] Figure 2 is the original topology;

[0077] Figure 3 This is a flowchart of the improved Pied Kingfisher optimization algorithm;

[0078] Figure 4 It is the topology diagram corresponding to the controller deployment solution after the EPKO algorithm solves it;

[0079] Figure 5 This is a comparison of the fitness curves of EPKO, PKO, PSO, and SCSSA when the number and location of controller deployment are optimal;

[0080] Figure 6 This is a comparison chart of the link failure rates of EPKO, PKO, PSO, and SCSSA when the number and locations of controllers deployed are optimal;

[0081] Figure 7 This is a comparison of the total delay of EPKO, PKO, PSO and SCSSA when the number and location of controllers deployed are optimal;

[0082] Figure 8 This is a comparison chart of the load balancing degree of EPKO, PKO, PSO, and SCSSA when the number and location of controllers deployed are optimal. DETAILED DESCRIPTION

[0083] In order to make the purpose, technical solutions and advantages of this application more clear, this application is further described in detail below with reference to the accompanying drawings and embodiments.

[0084] This paper describes a multi-controller deployment method for SD-MANET based on an improved Pied Kingfisher algorithm. With the goal of reducing control link failure rates and network latency while ensuring balanced controller load, this method, through an improved Pied Kingfisher optimization algorithm, proposes a multi-controller deployment algorithm that automatically determines the number and location of controllers. This deployment method consists of three main phases. In the first phase, the SD-MANET network topology is acquired, storing information about nodes and links in the network. In the second phase, based on basic network information, an optimization objective is constructed to minimize the control plane link failure rate, latency, and load. In the third phase, the improved Pied Kingfisher optimization algorithm is used to solve the problem, ensuring load balancing while minimizing the control plane link failure rate and latency, and determining the optimal number and location of controllers to deploy.

[0085] Specifically, the SD-MANET multi-controller deployment method based on the improved Pied Kingfisher algorithm described in this embodiment is as follows: Figure 1 As shown, the deployment method includes the following steps:

[0086] Step 1: Obtain the network topology information of SD-MANET, including node set, node coordinates and link set.

[0087] A software-defined mobile ad hoc network (SD-MANET) is a self-organizing network based on mobile nodes. The SD-MANET consists of multiple nodes and communication links connecting them. Real-time or preset network topology information for the SD-MANET is obtained, including a node set V consisting of nodes, the geographic or logical coordinates of each node, and a link set E consisting of all communication links.

[0088] Step 2: Construct an undirected graph representing the network topology, calculate the physical distance of the direct link using the node coordinates, and apply the Dijkstra algorithm to calculate the distance between any two nodes in the network. The shortest path distance between .

[0089] In one example, if Figure 2 The figure shows the original topology of 20 nodes within a 2000m×2000m area, that is, the undirected graph G=(V,E), where serial numbers 1 to 20 represent nodes.

[0090] Step 3: Define the controller deployment problem as a multi-objective optimization problem to determine the optimal controller deployment location set and the mapping relationship between each node and the controller.

[0091] The mapping relationship between each node and the controller refers to which controller should manage the remaining nodes after the controller position is determined, that is, which controller the remaining nodes are assigned to.

[0092] In this example, a comprehensive objective function is constructed that includes the following core indicators: , with minimizing the function value as the optimization goal:

[0093] Total communication delay :Comprehensively consider the propagation delay and transmission delay between the data plane node and its management controller, as well as the synchronization propagation delay and transmission delay between controllers. Among them, the propagation delay is based on the shortest path distance The transmission delay is calculated based on the data packet size and the link data transmission rate.

[0094] Control plane link failure rate : Based on the estimated probability of node failure and link failure probability , calculate the reliability of each controller domain and the failure rate of all control domain links , to quantify the overall link failure risk of the network;

[0095] Controller load balancing : Based on the flow request rate of each node and the maximum processing capacity of the controller, the load and utilization of each controller are calculated, and the standard deviation or variance of the controller utilization is used to measure the degree of load balancing. The goal is to minimize the difference.

[0096] The objective function is the weighted sum of the above three indicators, namely At the same time, constraints need to be set: a) Each node must be managed by and can only be managed by one controller; b) Each deployed controller manages at least one node; c) The total number of controllers deployed in the network is fixed to m; d) The load of each controller must not exceed its capacity; e) The objective function weight coefficient Non-negative and sum to 1.

[0097] Specifically, step 3 includes:

[0098] Calculate the propagation delay from the data plane node to the controller using the formula:

[0099] ,

[0100] Where V represents the data node A collection of, C represents the controller A collection of represents the i-th node, represents the j-th controller, Representation node With controller The shortest path distance between

[0101] Calculate the transmission delay from the data plane node to the controller using the following formula:

[0102] ,

[0103] Where, Representation node To the controller The transmission delay is:

[0104] ,

[0105] Where, Represents a data plane node The packet size, Represents a data plane node With controller The transmission rate between is:

[0106] ,

[0107] Where B represents bandwidth, Represents a data plane node The transmission power, represents the variance of Gaussian white noise;

[0108] Calculate the propagation delay between controllers using the formula:

[0109] ,

[0110] Where, Representation Controller With controller The shortest distance between

[0111] Calculate the transmission delay between controllers using the following formula:

[0112] ,

[0113] Where, Representation Controller To the controller The transmission delay is:

[0114] ,

[0115] Where, Representation Controller The packet size, Representation Controller With controller The transmission rate between them is:

[0116] ,

[0117] Where B represents bandwidth, Representation Controller The transmission power, represents the variance of Gaussian white noise;

[0118] The total communication delay of the controller domain is:

[0119] .

[0120] Specifically, step 3 also includes:

[0121] According to the failure probability of nodes in the control domain of each controller and link failure probability , computing controller The reliability of the control link within the domain is expressed as:

[0122] ,

[0123] Where, Indicates the node managed by the jth controller gather, represents the set of links e in the domain of controller j;

[0124] Calculate the control plane link failure rate of the SD-MANET network, expressed as:

[0125] ,

[0126] Where m represents the number of controllers in the SD-MANET network.

[0127] Specifically, step 3 also includes:

[0128] Calculate the number of controllers in the SD-MANET network Load , the formula is:

[0129] ,

[0130] Where, Indicates the number of stream requests for the node, Representation node and Is there a link between them? Indicates that a link exists. Indicates that no link exists;

[0131] Compute Controller The utilization rate is:

[0132] ,

[0133] Where, Representation Controller Maximum load capacity;

[0134] The average utilization rate of all controllers in the SD-MANET network is calculated as follows:

[0135] ,

[0136] Calculate the load balancing degree of the controller using the following formula:

[0137] .

[0138] Specifically, step 3 also includes:

[0139] A multi-objective optimization function F for controller deployment is constructed based on the total communication delay, control plane link failure rate, and load balancing rate within the controller domain, with minimizing this function as the optimization goal. The formula for the multi-objective optimization function is:

[0140] ,

[0141] Where, Represent the total communication delay , control plane link failure rate , controller load balancing The weight coefficient of

[0142] The constraints for constructing the multi-objective optimization function are:

[0143] ,

[0144] ,

[0145] ,

[0146] ,

[0147] ,

[0148] Where V represents the data plane node A collection of Indicates the controllers;

[0149] The first constraint states that each node can only be managed by one controller, the second constraint states that a controller must manage at least one node, the third constraint states that the number of controllers in the SD-MANET network is m; the fourth constraint states that the controller cannot exceed its load capacity; the fifth constraint states that the weighted sum of the weight coefficients of the multi-objective optimization function is 1.

[0150] Step 4: Use the improved Pied Kingfisher optimization algorithm to solve the multi-objective optimization problem and obtain the optimal controller deployment plan. Based on the optimal controller deployment plan, the controller and its management domain are deployed in the SD-MANET environment.

[0151] This example improves the Pied Kingfisher optimization algorithm. The key improvements include:

[0152] (1) Chaos initialization enhancement: The Chebyshev chaos map is used to replace the random initialization strategy of the standard algorithm to generate the initial population, aiming to improve the diversity and ergodicity of the initial solution and avoid premature convergence.

[0153] (2) Levy flight enhanced exploration: During the exploration phase of the algorithm, the Levy flight mechanism is introduced to adjust the individual position update step size. The occasional large jump characteristics of Levy flight are utilized to enhance the global search capability of the algorithm and help escape the local optimal trap.

[0154] (3) Gaussian mutation enhanced development: During the development phase of the algorithm, a Gaussian mutation strategy is introduced to the positions obtained by the original update rule. A small perturbation that conforms to the Gaussian distribution is applied to the individual positions with a certain probability, aiming to improve the local search accuracy of the algorithm within the neighborhood of the optimal solution. The algorithm iteratively executes the exploration, development, and symbiosis phases, continuously updating the positions of individuals in the population (representing the controller deployment plan) until the termination condition is met, such as reaching the maximum number of iterations or the accuracy requirement of the solution, and outputs the best fitness, that is, the objective function value. The smallest individual is regarded as the optimal solution.

[0155] Combine Figure 3 As shown, step 4 includes:

[0156] Step 41: Set the maximum number of iterations, use the specific deployment location of each controller as an individual in the population, and introduce the Chebyshev chaos map to initialize the population. The formula is:

[0157] ,

[0158] Where, represents the position of the i-th individual in the j-th dimension, Respectively represent The upper and lower boundaries of the search range of the dimensional search space, represents the value within (-1, 1) generated by the Chebyshev chaotic mapping, N represents the total number of individuals, represents the number of dimensions;

[0159] The population initialization formula of the Chebyshev chaotic mapping is as follows:

[0160] ,

[0161] When the dimension is 1, it provides the initial seed value for the Chebyshev chaotic mapping, and the value is between 0 and 1. When the dimension is greater than 1, the values of the subsequent dimensions are generated through the Chebyshev mapping using the values of the previous dimension. In this example , ensuring that no matter how close the initial value selections are, the iterated sequences are uncorrelated, that is, they are chaotic and ergodic within this range;

[0162] Step 42, in the exploration phase when the random floating-point number Rand < 0.8, introduce the Levy flight strategy to update the population position. The large-step jumps of Levy flight can help the algorithm explore the search space more effectively, jump out of the local optimal trap, and discover potential optimal solutions in a wider area; the update formula is:

[0163] ,

[0164] In the formula, is the position of the individual at the current iteration s, is the position of another individual randomly selected from the population [[ID=三十一]]to guide the movement and exploration of , is the state parameter, is the random direction parameter, is a random value within the normal distribution; is the final step size of the Levy flight, and the formula is:

[0165] ,

[0166] In the formula, and are both random numbers generated from the normal distribution, is the parameter of the Levy flight. In the example, it can be set ;

[0167] The exploration phase strategies of the pied kingfisher are divided into the perching strategy and the hovering strategy, and the main difference lies in the state parameter value . When 0.5 < Rand < 0.8, enter the perching strategy. When Rand At 0.5, enter the hovering strategy.

[0168] Habitat strategy The value formula is as follows:

[0169] ,

[0170] Where, represents the maximum number of iterations, is a constant, set to 8, for Random number;

[0171] Hover strategy The value formula is as follows:

[0172] ,

[0173] Where, For the Hedi The fitness of the kingfisher, the fitness function is the multi-objective optimization function F;

[0174] Step 43: In the development phase, when the random floating point number Rand ≥ 0.8, a Gaussian mutation strategy is introduced to update the population position. The update formula is:

[0175] ,

[0176] Where, In probability The new position obtained under the influence, is the standard deviation, and the formula is , is a coefficient used to control the size of the standard deviation. Respectively represent the upper and lower boundaries of the search range; rand represents a random number in [0,1]. is the mutation probability, indicating that it has a certain probability Perturb individual variables; represents the new position of the individual after the introduction of Gaussian mutation;

[0177] Step 44, in the symbiotic stage, the population position is updated at the current number of iterations. The update formula is:

[0178] ,

[0179] Where, Indicates the position of 2 individuals randomly selected from the population at the current iteration number s, The hunting efficiency of the kingfisher is Indicates hunting ability, Represents a random number in [0,1];

[0180] ,

[0181] Where, are the maximum and minimum values of the kingfisher's predation efficiency;

[0182] Step 45: merge the updated new population, calculate the fitness value of each individual in the new population, sort the fitness values from small to large, and take the individual corresponding to the smallest fitness value as the optimal feasible solution under the current number of iterations; determine whether it is greater than the number of iterations. If so, output the optimal feasible solution, otherwise return to the exploration stage and solve the iteration again.

[0183] In order to further illustrate the effectiveness and excellence of the SD-MANET multi-controller deployment method based on the improved Pied Kingfisher algorithm described in the present invention, the following examples are used to illustrate.

[0184] Table 1 shows the selected parameters and their corresponding values for calculating transmission delay in adhoc mode under the IEEE 802.11ag protocol.

[0185] Table 1 Parameter selection

[0186]

[0187] Figure 4 This is the topology corresponding to the controller deployment solution solved by the EPKO algorithm. The controllers are located at nodes 3, 12, and 18. The controller at node 3 manages nodes 1 and 17; the controller at node 12 manages nodes 4, 6, 9, 13, and 19; the remaining nodes are managed by the controller at node 18.

[0188] The method of the present invention is referred to as EPKO. Figure 5 This is a comparison of the fitness curves of the EPKO algorithm with the Pied Kingfisher algorithm (PKO), particle swarm optimization algorithm (PSO), and sparrow optimization algorithm with sine-cosine strategy and Cauchy mutation (SCSSA) when the number and location of controller deployment are optimal. It can be seen that the performance of the EPKO algorithm is better than that of the PKO algorithm, PSO algorithm, and SCSSA algorithm, and the controller deployment location solved is better.

[0189] Figure 6 This figure compares the control plane link failure rates of the EPKO algorithm, the PKO algorithm, the PSO algorithm, and the SCSSA algorithm when solving the optimal controller deployment solution. The lower the failure rate, the better. It can be seen that the EPKO algorithm has a better link failure rate when solving the controller deployment solution.

[0190] Figure 7This is a latency comparison chart of the EPKO algorithm, PKO algorithm, PSO algorithm, and SCSSA algorithm when solving the optimal controller deployment solution. The lower the latency, the better. It can be seen that the EPKO algorithm has a better latency effect when solving the controller deployment solution.

[0191] Figure 8 The figure below compares the load balancing performance of the EPKO algorithm with the PKO algorithm, PSO algorithm, and SCSSA algorithm when solving the optimal controller deployment solution. Although the EPKO algorithm has a certain gap compared with other algorithms in terms of load balancing, this is the result of a trade-off in multi-objective optimization, which prioritizes link stability and latency performance, thereby improving the overall reliability and efficiency of the SD-MANET network.

Claims

1. A SD-MANET multi-controller deployment method based on the improved Pied Kingfisher algorithm, characterized in that: The following steps are involved: Step 1: Obtain the network topology information of SD-MANET, including node set, node coordinates and link set; Step 2: Construct an undirected graph representing the network topology, calculate the physical distance of the direct link using the node coordinates, and apply the Dijkstra algorithm to calculate the shortest path distance between any two nodes in the network; Step 3: Define the controller deployment problem as a multi-objective optimization problem to determine the optimal controller deployment location set and the mapping relationship between each node and the controller; Step 4: Use the improved Pied Kingfisher optimization algorithm to solve the multi-objective optimization problem and obtain the optimal controller deployment plan. Deploy the controller and its management domain in the SD-MANET environment based on the optimal controller deployment plan. Step 3 includes: Calculate the propagation delay from the data plane node to the controller using the formula: ， Where V represents the data node A collection of C represents the controller A collection of represents the i-th node, represents the j-th controller, Representation node With controller The shortest path distance between Calculate the transmission delay from the data plane node to the controller using the following formula: ， Where, Representation node To the controller transmission delay; Calculate the propagation delay between controllers using the formula: ， Where, Representation Controller With controller The shortest distance between Calculate the transmission delay between controllers using the following formula: ， Where, Representation Controller To the controller transmission delay; The total communication delay of the controller domain is: ； Step 3 also includes: According to the failure probability of nodes in the control domain of each controller and link failure probability , computing controller The reliability of the control link within the domain is expressed as: ， Where, Indicates the node managed by the jth controller gather, represents the set of links e in the domain of controller j; Calculate the control plane link failure rate of the SD-MANET network, expressed as: ， Where m represents the number of controllers in the SD-MANET network.

2. The SD-MANET multi-controller deployment method based on the improved Pied Kingfisher algorithm according to claim 1 is characterized in that: Step 3 also includes: Calculate the number of controllers in the SD-MANET network Load , the formula is: ， Where, Indicates the number of stream requests for the node, Representation node and Is there a link between them? Indicates that a link exists. Indicates that no link exists; Compute Controller The utilization rate is: ， Where, Representation Controller Maximum load capacity; The average utilization rate of all controllers in the SD-MANET network is calculated as follows: ， Calculate the load balancing degree of the controller using the following formula: 。 3. The SD-MANET multi-controller deployment method based on the improved Pied Kingfisher algorithm according to claim 2 is characterized in that: Step 3 also includes: A multi-objective optimization function F for controller deployment is constructed based on the total communication delay, control plane link failure rate, and load balancing rate within the controller domain, with minimizing this function as the optimization goal. The formula for the multi-objective optimization function is: ， Where, Represent the total communication delay , control plane link failure rate , controller load balancing The weight coefficient of The constraints for constructing the multi-objective optimization function are: ， ， ， ， ， Where V represents the data plane node A collection of Indicates the controllers; The first constraint states that each node can only be managed by one controller, the second constraint states that a controller must manage at least one node, the third constraint states that the number of controllers in the SD-MANET network is m; the fourth constraint states that the controller cannot exceed its load capacity; the fifth constraint states that the weighted sum of the weight coefficients of the multi-objective optimization function is 1.

4. The SD-MANET multi-controller deployment method based on the improved Pied Kingfisher algorithm according to claim 3 is characterized in that: Step 4 includes: Step 41: Set the maximum number of iterations, use the specific deployment location of each controller as an individual in the population, and introduce the Chebyshev chaos map to initialize the population. The formula is: ， Where, represents the position of the i-th individual in the j-th dimension, Respectively represent The upper and lower boundaries of the search range of the dimensional search space, represents the value within (-1,1) generated by the Chebyshev chaotic map, N represents the total number of individuals, Indicates the number of dimensions; Step 42: In the exploration phase, the Levy flight strategy is introduced to update the population position. The update formula is: ， Where, is the position of the individual at the current iteration number s, is another individual randomly selected from the population position, used to guide Movement and exploration, is the state parameter, is a random pointing parameter, for Random values from a normal distribution; is the final step length of Levy flight, and the formula is: ， Where, and are all random numbers generated from a normal distribution. The parameters for Levy's flight; Step 43: In the development phase, a Gaussian mutation strategy is introduced to update the population position. The update formula is: ， Where, In probability The new position obtained under the influence, is the standard deviation, and the formula is , is a coefficient used to control the size of the standard deviation. Respectively represent the upper and lower boundaries of the search range; rand represents a random number in [0,1]. is the mutation probability, indicating that it has a certain probability Perturb individual variables; represents the new position of the individual after the introduction of Gaussian mutation; Step 44, in the symbiotic stage, the population position is updated at the current number of iterations. The update formula is: ， Where, Indicates the position of 2 individuals randomly selected from the population at the current iteration number s, The hunting efficiency of the kingfisher is Indicates hunting ability, Represents a random number in [0,1]; Step 45: merge the updated new population, calculate the fitness value of each individual in the new population, sort the fitness values from small to large, and take the individual corresponding to the smallest fitness value as the optimal feasible solution under the current number of iterations; determine whether it is greater than the number of iterations. If so, output the optimal feasible solution, otherwise return to the exploration stage and solve the iteration again.

Images