A non-line-of-sight communication signal error compensation method, system, device and medium

Abstract

The present invention discloses a non-line-of-sight communication signal error compensation method, system, device and medium, which relates to the field of machine learning technology. It includes: obtaining the ranging value of the original non-line-of-sight communication signal of the target object sending end; inputting the ranging value of the original non-line-of-sight communication signal into the IWO-SSA-KELM model to obtain the ranging value compensation result of the final non-line-of-sight communication signal of the target object receiving end. The present invention avoids falling into the problem of local optimal solution, and uses the IWO-SSA-KELM model corresponding to the regularization parameter C and the kernel width function σ after quadratic optimization to improve the non-line-of-sight communication signal error compensation effect.

Description

Technical Field

[0001] The present invention relates to the field of machine learning technology, and in particular to a non-line-of-sight communication signal error compensation method, system, device and medium. Background Art

[0002] Non-line-of-sight communication refers to indirect point-to-point communication between a receiver and a transmitter, where the line of sight is obstructed, the two parties cannot see each other, and more than 50% of the Fresnel zone is blocked. In this communication environment, the signal may be affected by various factors, such as attenuation, scattering, and reflection, which may lead to signal loss and distortion, and thus errors. Error compensation is a technology that aims to artificially create a new original error to offset the original original error that is currently a problem. In non-line-of-sight communication, error compensation technology can be used to improve the transmission quality and reliability of the signal.

[0003] In the prior art, non-line-of-sight communication signal error compensation is performed through signal processing technology. The specific process is: collect non-line-of-sight communication signals through sensors or other devices, analyze the pre-processed signals, and extract their key features; detect errors in the signals by comparing the original signals with the extracted signals, and evaluate their size and nature. The error detection results are input into the BP neural network to compensate the received signals.

[0004] The defects of the above-mentioned prior art are: the training cycle of the BP neural network is relatively long, and it is easy to fall into the problem of local optimal solution, resulting in poor error compensation effect of non-line-of-sight communication signals. Summary of the invention

[0005] Based on this, it is necessary to provide a non-line-of-sight communication signal error compensation method, system, device and medium to address the above technical problems.

[0006] An embodiment of the present invention provides a non-line-of-sight communication signal error compensation method, comprising:

[0007] Obtaining the ranging value of the original non-line-of-sight communication signal of the target object sending end;

[0008] The ranging value of the original non-line-of-sight communication signal is input into the IWO-SSA-KELM model to obtain the ranging value compensation result of the final non-line-of-sight communication signal at the receiving end of the target object;

[0009] The construction of the IWO-SSA-KELM model specifically includes:

[0010] Constructing an initial KELM model; wherein the initial KELM model includes a regularization parameter C and a kernel width function σ, and the initial KELM model is used to perform error compensation on the original non-line-of-sight communication signal ranging value;

[0011] The initial KELM model is iterated multiple times using the squirrel search optimization algorithm SSA with global optimization performance, and the error compensation information obtained by the initial KELM model each time is re-input into the initial KELM model as the ranging value of the original non-line-of-sight communication signal, to obtain multiple optimized regularization parameters C and kernel width functions σ, and the initial KELM model corresponding to the optimized regularization parameters C and kernel width function σ is used as an advanced KELM model; the advanced KELM model is a model with global optimization performance when performing error compensation on the ranging value of the original non-line-of-sight communication signal;

[0012] The advanced KELM model is iterated multiple times using the weeds algorithm IWO with the ability to jump out of the local optimum, and the quadratically optimized regularization parameter C and the kernel width function σ obtained by the advanced KELM model each time are re-input into the advanced KELM model to obtain multiple quadratically optimized regularization parameters C and kernel width functions σ; the KELM model corresponding to the quadratically optimized regularization parameter C and the kernel width function σ is used as the IWO-SSA-KELM model; the IWO-SSA-KELM model is a model with global optimization performance and the ability to jump out of the local optimum when performing error compensation on the ranging value of the original non-line-of-sight communication signal.

[0013] In addition, the initial KELM model includes a regularization parameter C and a kernel width function σ, which specifically includes:

[0014]

[0015] Among them, H is the hidden layer output matrix, H T Represents the Moore-Penrose pseudoinverse matrix of H, where T is the specified target vector.

[0016] In addition, the squirrel search optimization algorithm SSA with global optimization performance is used to iterate the initial KELM model multiple times, which specifically includes:

[0017] Initialize the parameters of the squirrel search algorithm SSA, including: initialize the position and speed of the squirrels, each squirrel has a current position X and speed V, and randomly generate the initial position;

[0018] Construct the objective function to evaluate the regularization parameter C and kernel width function σ of KELM;

[0019] Arrange the fitness values ​​of all squirrels in ascending order according to the objective function;

[0020] The squirrels were classified based on their location: from an acorn tree to a hickory tree, m1; from a common tree to an acorn tree, m2; and from a common tree to a hickory tree, m3.

[0021] Using the formula Determine the seasonal constant S c , where S c is a seasonal constant, t is the number of iterations, z is the category of the squirrel's location, d is the distance the squirrel slides, is the actual position of the squirrel in the k dimension in the tth iteration, FS ht，k is the reference position in the k dimension;

[0022] If the seasonal constant S c When the preset value is reached, the formula Iterate the squirrel position, where FS nt,i is the new position of the nth squirrel after updating at time t, Levy is the Levy flight formula, FS i，U -FS i，L is the difference between the upper and lower bounds of the search space, FS i，L is the current location of the squirrel;

[0023] When the maximum number of iterations is reached, the process stops. The squirrel currently on the hickory tree is the optimal solution for the optimized regularization parameter C and kernel width function σ. If the maximum number of iterations is not reached, the process proceeds to arranging the fitness values ​​of all squirrels in ascending order according to the objective function.

[0024] In addition, the IWO algorithm with the ability to escape from the local optimum is used to iterate the advanced KELM model multiple times, which specifically includes:

[0025] Initialize the invasive weed population position, population size and iteration times in IWO;

[0026] Arrange the fitness values ​​of invasive weeds in ascending order according to the objective function;

[0027] According to the formula To promote the growth and reproduction of invasive weeds, where f is the fitness of the current solution, f min is the minimum fitness of all solutions in the current population, f max is the maximum fitness of all solutions in the current population, seed max The maximum number of seeds that can be randomly generated for a single solution in each iteration, seed min The minimum number of seeds that can be randomly generated for a single solution during each iteration;

[0028] If the growth and reproduction of invasive weeds reaches the preset value, according to the formula Iterate the spatial spread of invasive weeds, where σ i is the standard deviation, iter max is the maximum number of iterations, σ init is the initial value of the standard deviation, σ finalis the final value of the standard deviation, n is the nonlinear adjustment coefficient;

[0029] After reaching the maximum number of iterations, the optimal solution of the regularization parameter C and the kernel width function σ after quadratic optimization is obtained.

[0030] In addition, a non-line-of-sight communication signal error compensation system includes:

[0031] An acquisition module, used to acquire the ranging value of the original non-line-of-sight communication signal of the target object sending end;

[0032] Model building module, used to build the IWO-SSA-KELM model;

[0033] The error compensation module is used to input the ranging value of the original non-line-of-sight communication signal into the IWO-SSA-KELM model to obtain the ranging value compensation result of the final non-line-of-sight communication signal at the receiving end of the target object.

[0034] In addition, the model building module includes:

[0035] An initial module, used to construct an initial KELM model; wherein the initial KELM model includes a regularization parameter C and a kernel width function σ, and the initial KELM model is used to perform error compensation on the original non-line-of-sight communication signal ranging value;

[0036] The initial iteration module is used to use the squirrel search optimization algorithm SSA with global optimization performance to iterate the initial KELM model multiple times, and the error compensation information obtained by the initial KELM model each time is re-input into the initial KELM model as the ranging value of the original non-line-of-sight communication signal, to obtain multiple optimized regularization parameters C and kernel width functions σ, and the initial KELM model corresponding to the optimized regularization parameters C and kernel width function σ is used as an advanced KELM model; the advanced KELM model is a model with global optimization performance when performing error compensation on the ranging value of the original non-line-of-sight communication signal;

[0037] The secondary iteration module is used to use the weeds algorithm IWO with the ability to jump out of the local optimum to iterate the advanced KELM model for multiple times, re-input the secondary optimized regularization parameter C and the kernel width function σ obtained by the advanced KELM model each time into the advanced KELM model, and obtain multiple secondary optimized regularization parameters C and kernel width functions σ; the KELM model corresponding to the secondary optimized regularization parameter C and the kernel width function σ is used as the IWO-SSA-KELM model; the IWO-SSA-KELM model is a model with global optimization performance and the ability to jump out of the local optimum when performing error compensation on the ranging value of the original non-line-of-sight communication signal.

[0038] In addition, a computer device includes a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the steps of the above-mentioned non-line-of-sight communication signal error compensation method are implemented.

[0039] In addition, a storage medium stores a computer program, which implements the steps of the above-mentioned non-line-of-sight communication signal error compensation method when executed by a processor.

[0040] The above-mentioned non-line-of-sight communication signal error compensation method, system, device and medium provided by the embodiments of the present invention have the following beneficial effects compared with the prior art:

[0041] In the prior art, in the process of performing error compensation for non-line-of-sight communication signals through a BP neural network, the training cycle of the BP neural network is relatively long, and it is easy to fall into the problem of a local optimal solution, resulting in poor error compensation effect for non-line-of-sight communication signals.

[0042] Therefore, the present invention iteratively optimizes the KELM model through the ability of the weed optimization algorithm IWO to jump out of the local optimum and the global optimization performance of the squirrel search algorithm SSA, solves the problem of the long training cycle of the BP neural network, and avoids the problem of falling into the local optimal solution. In the initial KELM model training process, the non-line-of-sight signal ranging value in the real-time communication process is introduced to determine the regularization parameter C and the kernel width function σ that are closer to the actual communication situation and more accurate. The IWO-SSA-KELM model corresponding to the regularization parameter C and the kernel width function σ after secondary optimization is used to improve the error compensation effect of the non-line-of-sight communication signal. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A schematic flow chart of a non-line-of-sight communication signal error compensation method provided in an embodiment;

[0044] Figure 2 A comparison diagram of NLOS ranging error compensation accuracy of different algorithms of a non-line-of-sight communication signal error compensation method provided in an embodiment;

[0045] Figure 3 The figure is a comparison chart of NLOS ranging error compensation stability of different algorithms of a non-line-of-sight communication signal error compensation method provided in an embodiment. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0047] In one embodiment, a non-line-of-sight communication signal error compensation method is provided, such as Figure 1 As shown, the method includes:

[0048] (1) Generation of training and validation sets

[0049] First, obtain the original non-line-of-sight communication signal ranging value generated by the target object's transmitter, and clear the environmental variables to ensure the cleanliness of data processing. Then, split the ranging value data set into a training set and a validation set, generally about 70%-80% for the training set and the rest for the validation set.

[0050] (2) Constructing the initial KELM model

[0051] 1. The initial KELM model includes a regularization parameter C and a kernel width function σ. The initial KELM model is used to perform error compensation based on the original communication signal ranging value.

[0052] 2. Input the input data X of the training set into the initial KELM model and evaluate the regularization parameter C and kernel function width parameter σ of KELM, which specifically includes:

[0053] The H in ELM is the output of the hidden layer calculated based on the hidden layer weight matrix W, and its formula is:

[0054] H=sin(X*W+b)

[0055] Among them, X is the input data of the training set, W is the hidden layer weight matrix, b is the hidden layer bias vector, H is the hidden layer output matrix, and sin is the activation function;

[0056] As an improved algorithm, KELM has fewer adjustment parameters and very stable result output compared to ELM.

[0057] KELM introduces the concept of kernel matrix, which can replace the random matrix HH in ELM with the kernel matrix Ω T ,Right now:

[0058] Ω KELM =HH T

[0059] The activation function of the extreme learning machine is expressed as a kernel function:

[0060] Ω i,j =h(x i )·h(x j )=k(x i ,x j )

[0061] In the extreme learning machine model, h(x) represents the output function, k(xi ,x j ) specifies the kernel function, and here we choose Gaussian Radial Basis Function (GRBF). The mathematical expression is:

[0062]

[0063] Where exp is the Gaussian radial basis function.

[0064] Subsequently, in order to ensure that HH T The eigenvalues ​​of a matrix are non-zero values, in HH T The parameter I / C is added to the diagonal position of the matrix to clarify the weight vector β * The value of . Replacing the random matrix process of ELM with the kernel matrix not only enhances the generalization ability of the model, but also effectively improves the stability of the model. Therefore, the output weight β of the extreme learning machine * It can be calculated by the following formula:

[0065] β * =H T (I / C+HH T ) -1 T

[0066] Among them, C is the regularization parameter, which aims to achieve a balance between the structural risk and empirical risk of the model, and I represents a diagonal matrix. Based on the above analysis, the specific formula of the output function of the kernel extreme learning machine can be derived as follows:

[0067]

[0068] Among them, H is the hidden layer output matrix, H T Represents the Moore-Penrose pseudoinverse matrix of H, where T is the specified target vector.

[0069] (3) Construction of IWO-SSA-ELM model

[0070] 1. Initialize the regularization parameter C and kernel width function σ of the KELM model.

[0071] 2. By combining the ability of IWO to jump out of the local optimum, the global optimization performance of SSA, the advantages of learning rate acceleration algorithm convergence and the good local characteristics of Gaussian kernel function, the regularization parameter C and kernel width function σ in the initial KELM model are iteratively optimized. The KELM model corresponding to the optimized regularization parameter C and kernel width function σ can improve the error compensation effect of non-line-of-sight communication signals. The specific steps are:

[0072] 2.1 Initialize the parameters of the squirrel search algorithm SSA, including: initialize the position and speed of the squirrels. Each squirrel has a current position X and speed V, and randomly generates an initial position;

[0073] 2.2 Construct the objective function and evaluate the regularization parameter C and kernel width function σ of KELM.

[0074] 2.3 Arrange the fitness values ​​of all squirrels in ascending order according to the objective function.

[0075] 2.4 Classify each squirrel based on its location: m1 for those from an acorn tree to a hickory tree, m2 for those from an ordinary tree to an acorn tree, and m3 for those from an ordinary tree to a hickory tree.

[0076] 2.5 Using the formula Determine the seasonal constant S c , where S c is a seasonal constant, t is the number of iterations, z is the category of the squirrel's location, d is the distance the squirrel slides, is the actual position of the squirrel in the k dimension in the tth iteration, FS ht，k is the reference position in the k dimension.

[0077] If the seasonal constant S c When the preset value is reached, the formula Iterate the squirrel position, where FS nt,i is the new position of the nth squirrel after updating at time t, Levy is the Levy flight formula, FS i，U -FS i，L is the difference between the upper and lower bounds of the search space, FS i，L The current location of the squirrel.

[0078] 2.6 When the maximum number of iterations is reached, the process stops. The squirrel currently on the hickory tree is the optimal solution for the optimized regularization parameter C and kernel width function σ. If the maximum number of iterations is not reached, the process proceeds to arranging the fitness values ​​of all squirrels in ascending order according to the objective function.

[0079] 2.7 Initialize the position, population size and iteration times of the invasive weed population in IWO;

[0080] 2.8 Arrange the fitness values ​​of invasive weeds in ascending order according to the objective function;

[0081] 2.9 According to the formula To promote the growth and reproduction of invasive weeds, where f is the fitness of the current solution, f min is the minimum fitness of all solutions in the current population, f max is the maximum fitness of all solutions in the current population, seed maxThe maximum number of seeds that can be randomly generated for a single solution in each iteration, seed min The minimum number of seeds that can be randomly generated for a single solution during each iteration.

[0082] If the growth and reproduction of invasive weeds reaches the preset value, according to the formula Iterate the spatial spread of invasive weeds, where σ i is the standard deviation, iter max is the maximum number of iterations, σ init is the initial value of the standard deviation, σ final is the final value of the standard deviation, and n is the nonlinear adjustment coefficient.

[0083] 2.10 After reaching the maximum number of iterations, the optimal solution of the regularization parameter C and the kernel width function σ after secondary optimization is obtained.

[0084] 3. The KELM model corresponding to the regularization parameter C and the kernel width function σ after quadratic optimization is used as the IWO-SSA-KELM model, and the ranging value of the original non-line-of-sight communication signal is input into the IWO-SSA-KELM model to obtain the ranging value compensation result of the final non-line-of-sight communication signal at the receiving end of the target object.

[0085] (4) IWO-SSA-KELM model simulation test

[0086] 4. Compare the uncompensated, BP neural network, KELM and IWO-SSA-KELM error compensation algorithms, and evaluate the performance of NLOS error compensation of the IWO-SSA-KELM model.

[0087] 4.1 First, the UWB DW1000 positioning device is used for experiments. The compensated ranging values ​​are used to evaluate the performance of uncompensated, BP neural network, KELM and IWO-SSA-KELM algorithms for NLOS error compensation.

[0088] 4.2 The NLOS ranging error compensation accuracy and NLOS ranging error compensation stability of the uncompensated, BP neural network, KELM and IWO-SSA-KELM error compensation algorithms are compared.

[0089] 4.3 Comparison results

[0090] like Figure 2 As shown in the figure, the average ranging error without compensation in the NLOS scenario is 0.91m, and the average ranging errors after compensation by BP, KELM and IWO-SSA-KELM are 0.16m, 0.13m and 0.09m respectively. It can be seen that the NLOS ranging error compensation performance of the IWO-SSA-KELM algorithm is better than that of the KELM and BP compensation algorithms.

[0091] like Figure 3 As shown in the figure, the average variance without compensation in the NLOS scenario is 2.46, and the average variances after BP, KELM, and IWO-SSA-KELM compensation are 0.60, 0.25, and 0.20, respectively. It can be seen that the NLOS error compensation algorithm based on IWO-SSA-KELM proposed in this chapter has the best stability.

[0092] in addition, Figure 3 The obvious peak change in indicates the corner point in the experiment, reflecting the sharp change of the target state. It can be seen that in extreme cases, even if the target state changes suddenly, the NLOS error compensation algorithm based on IWO-SSA-KELM still shows a good compensation effect, which verifies the effectiveness and robustness of this method.

[0093] (5) Model evaluation

[0094] For model performance evaluation in non-line-of-sight error compensation tasks, the average ranging error and variance can be used as evaluation indicators to understand the performance of the model.

[0095] The average ranging error is defined as the square of the difference between the measured value and the actual value. The smaller the average ranging error, the higher the error compensation accuracy of the algorithm.

[0096] The variance is defined as the square of the difference between the actual value and each distance measurement value divided by the number of distance measurement values. The smaller the variance change, the more stable the error compensation of the algorithm.

[0097] (6) Non-line-of-sight error compensation

[0098] The original non-line-of-sight communication signal ranging value is input into the IWO-SSA-KELM model to obtain the final non-line-of-sight communication signal ranging value compensation result at the target object receiving end.

[0099] In one embodiment, a non-line-of-sight communication signal error compensation system is provided, the system comprising:

[0100] The acquisition module is used to obtain the ranging value of the original non-line-of-sight communication signal of the target object sending end.

[0101] Model building module, used to build the IWO-SSA-KELM model. The model building module includes:

[0102] The initial module is used to construct an initial KELM model; wherein the initial KELM model includes a regularization parameter C and a kernel width function σ, and the initial KELM model is used to perform error compensation on the original non-line-of-sight communication signal ranging value.

[0103] The initial iteration module is used to use the squirrel search optimization algorithm SSA with global optimization performance to iterate the initial KELM model for multiple times, and the error compensation information obtained by the initial KELM model each time is re-input into the initial KELM model as the ranging value of the original non-line-of-sight communication signal, to obtain multiple optimized regularization parameters C and kernel width functions σ, and the initial KELM model corresponding to the optimized regularization parameters C and kernel width function σ is used as an advanced KELM model; the advanced KELM model is a model with global optimization performance when performing error compensation on the ranging value of the original non-line-of-sight communication signal.

[0104] The secondary iteration module is used to use the weeds algorithm IWO with the ability to jump out of the local optimum to iterate the advanced KELM model for multiple times, re-input the secondary optimized regularization parameter C and the kernel width function σ obtained by the advanced KELM model each time into the advanced KELM model, and obtain multiple secondary optimized regularization parameters C and kernel width functions σ; the KELM model corresponding to the secondary optimized regularization parameter C and the kernel width function σ is used as the IWO-SSA-KELM model; the IWO-SSA-KELM model is a model with global optimization performance and the ability to jump out of the local optimum when performing error compensation on the ranging value of the original non-line-of-sight communication signal.

[0105] The error compensation module is used to input the ranging value of the original non-line-of-sight communication signal into the IWO-SSA-KELM model to obtain the ranging value compensation result of the final non-line-of-sight communication signal at the receiving end of the target object.

[0106] A computer device is provided in one embodiment, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the steps of the non-line-of-sight communication signal error compensation method are implemented.

[0107] In one embodiment, a storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the non-line-of-sight communication signal error compensation method are implemented.

[0108] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A non-line-of-sight communication signal error compensation method, characterized in that: include: Obtaining the ranging value of the original non-line-of-sight communication signal of the target object sending end; The ranging value of the original non-line-of-sight communication signal is input into the IWO-SSA-KELM model to obtain the ranging value compensation result of the final non-line-of-sight communication signal at the receiving end of the target object; The construction of the IWO-SSA-KELM model specifically includes: Constructing an initial KELM model; wherein the initial KELM model includes a regularization parameter C and a kernel width function σ, and the initial KELM model is used to perform error compensation on the original non-line-of-sight communication signal ranging value; The initial KELM model is iterated multiple times using the squirrel search optimization algorithm SSA with global optimization performance, and the error compensation information obtained by the initial KELM model each time is re-input into the initial KELM model as the ranging value of the original non-line-of-sight communication signal, to obtain multiple optimized regularization parameters C and kernel width functions σ, and the initial KELM model corresponding to the optimized regularization parameters C and kernel width function σ is used as an advanced KELM model; the advanced KELM model is a model with global optimization performance when performing error compensation on the ranging value of the original non-line-of-sight communication signal; The advanced KELM model is iterated multiple times using the weeds algorithm IWO with the ability to jump out of the local optimum, and the quadratically optimized regularization parameter C and the kernel width function σ obtained by the advanced KELM model each time are re-input into the advanced KELM model to obtain multiple quadratically optimized regularization parameters C and kernel width functions σ; the KELM model corresponding to the quadratically optimized regularization parameter C and the kernel width function σ is used as the IWO-SSA-KELM model; the IWO-SSA-KELM model is a model with global optimization performance and the ability to jump out of the local optimum when performing error compensation on the ranging value of the original non-line-of-sight communication signal.

2. A non-line-of-sight communication signal error compensation method as claimed in claim 1, characterized in that: The initial KELM model includes a regularization parameter C and a kernel width function σ, which specifically includes: Ω i,j =h(x i )·h(x j )=k(x i ,x j ) Among them, H is the hidden layer output matrix, H T is the Moore-Penrose pseudo-inverse matrix of H, T is the specified target vector, C is the regularization parameter, I is the diagonal matrix, σ is the kernel function width parameter, h(x) is the output function, k(x) is the kernel function width parameter, i ,x j ) is the kernel function, and exp is the Gaussian radial basis function.

3. A non-line-of-sight communication signal error compensation method as claimed in claim 1, characterized in that: The initial KELM model is iterated multiple times using the squirrel search optimization algorithm SSA with global optimization performance, which specifically includes: Initialize the parameters of the squirrel search algorithm SSA, including: initialize the position and speed of the squirrels, each squirrel has a current position X and speed V, and randomly generate the initial position; Construct the objective function to evaluate the regularization parameter C and kernel width function σ of KELM; Arrange the fitness values ​​of all squirrels in ascending order according to the objective function; The squirrels were classified based on their location: from an acorn tree to a hickory tree, m1; from a common tree to an acorn tree, m2; and from a common tree to a hickory tree, m3. Using the formula Determine the seasonal constant S c , where S c is a seasonal constant, t is the number of iterations, z is the category of the squirrel's location, d is the distance the squirrel slides, is the actual position of the squirrel in the k dimension in the tth iteration, FS ht，k is the reference position in the k dimension; If the seasonal constant S c When the preset value is reached, the formula Iterate the squirrel position, where FS nt,i is the new position of the nth squirrel after updating at time t, Levy is the Levy flight formula, FS i，U -FS i，L is the difference between the upper and lower bounds of the search space, FS i，L is the current location of the squirrel; When the maximum number of iterations is reached, the process stops. The squirrel currently on the hickory tree is the optimal solution for the optimized regularization parameter C and kernel width function σ. If the maximum number of iterations is not reached, the process proceeds to arranging the fitness values ​​of all squirrels in ascending order according to the objective function.

4. A non-line-of-sight communication signal error compensation method as claimed in claim 1, characterized in that: The IWO algorithm with the ability to jump out of the local optimum is used to iterate the advanced KELM model multiple times, which specifically includes: Initialize the invasive weed population position, population size and iteration times in IWO; Arrange the fitness values ​​of invasive weeds in ascending order according to the objective function; According to the formula To promote the growth and reproduction of invasive weeds, where f is the fitness of the current solution, f min is the minimum fitness of all solutions in the current population, f max is the maximum fitness of all solutions in the current population, seed max The maximum number of seeds that can be randomly generated for a single solution in each iteration, seed min The minimum number of seeds that can be randomly generated for a single solution during each iteration; If the growth and reproduction of invasive weeds reaches the preset value, according to the formula Iterate the spatial spread of invasive weeds, where σ i is the standard deviation, iter max is the maximum number of iterations, σ init is the initial value of the standard deviation, σ final is the final value of the standard deviation, n is the nonlinear adjustment coefficient; After reaching the maximum number of iterations, the optimal solution of the regularization parameter C and the kernel width function σ after quadratic optimization is obtained.

5. A non-line-of-sight communication signal error compensation system, characterized in that: include: An acquisition module, used to acquire the ranging value of the original non-line-of-sight communication signal of the target object sending end; Model building module, used to build the IWO-SSA-KELM model; The model building module includes: An initial module, used to construct an initial KELM model; wherein the initial KELM model includes a regularization parameter C and a kernel width function σ, and the initial KELM model is used to perform error compensation on the original non-line-of-sight communication signal ranging value; The initial iteration module is used to use the squirrel search optimization algorithm SSA with global optimization performance to iterate the initial KELM model multiple times, and the error compensation information obtained by the initial KELM model each time is re-input into the initial KELM model as the ranging value of the original non-line-of-sight communication signal, to obtain multiple optimized regularization parameters C and kernel width functions σ, and the initial KELM model corresponding to the optimized regularization parameters C and kernel width function σ is used as an advanced KELM model; the advanced KELM model is a model with global optimization performance when performing error compensation on the ranging value of the original non-line-of-sight communication signal; A secondary iteration module is used to perform multiple iterations on the advanced KELM model using the weed algorithm IWO with the ability to jump out of the local optimum, re-input the secondary optimized regularization parameter C and the kernel width function σ obtained by the advanced KELM model each time into the advanced KELM model, and obtain multiple secondary optimized regularization parameters C and kernel width functions σ; the KELM model corresponding to the secondary optimized regularization parameter C and the kernel width function σ is used as the IWO-SSA-KELM model; the IWO-SSA-KELM model is a model with global optimization performance and the ability to jump out of the local optimum when performing error compensation on the original non-line-of-sight communication signal ranging value The error compensation module is used to input the ranging value of the original non-line-of-sight communication signal into the IWO-SSA-KELM model to obtain the ranging value compensation result of the final non-line-of-sight communication signal at the receiving end of the target object.

6. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the non-line-of-sight communication signal error compensation method described in any one of claims 1-4 are implemented.

7. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the non-line-of-sight communication signal error compensation method described in any one of claims 1 to 4 are implemented.

Images