A static decoupling method, device, electronic equipment and storage medium

Abstract

The application relates to a static decoupling method, a static decoupling device, electronic equipment and a storage medium. The static decoupling method comprises the following steps: acquiring voltage data output by a target sensing object to be statically decoupled, the target sensing object being subjected to an acting force in any preset axial direction, and the target sensing object being used for detecting the size of the acting force; inputting the voltage data into a preset static decoupling model, so that the static decoupling model outputs a predicted value corresponding to the acting force; wherein the static decoupling model is obtained by introducing an attention mechanism algorithm into a neural network algorithm and optimizing model parameters by using a genetic algorithm. The application embodiment can accelerate the convergence speed of the composite model RBF-AM-GA, improve the decoupling precision and decoupling efficiency, thereby improving the prediction effect of the neural network, improving the robustness and generalization ability of the composite model RBF-AM-GA, and further realizing accurate decoupling of a six-dimensional force sensor. The algorithm is simple and reliable, and the result is accurate and effective.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a static decoupling method, apparatus, electronic device, and storage medium. Background Technology

[0002] Six-dimensional force / torque sensors are now widely used in various applications to provide real-time information for feedback control of robots and automation systems, such as contour tracking, zero-force teaching, flexible automated assembly, remote robot operation, multi-arm robot collaboration, robotic surgery, and rehabilitation training.

[0003] Due to limitations in the mechanical manufacturing process, patch technology, strain gauge lateral effects, and detection methods of six-dimensional force sensors, almost every force / torque component acting on the sensor in each dimension affects the sensor's output signals. This inevitably leads to some mutual interference—inter-dimensional coupling—between the sensors. Different degrees of inter-dimensional signal coupling significantly impact the detection accuracy of six-dimensional force sensors and constrain their application and development in high-precision and cutting-edge fields. Summary of the Invention

[0004] To address the technical problem of low detection accuracy in six-dimensional force sensors, this application provides a static decoupling method, apparatus, electronic device, and storage medium.

[0005] Firstly, this application provides a static decoupling method, including:

[0006] The voltage data output by the target sensing object to be statically decoupled is obtained. The target sensing object is subjected to a force in any preset axial direction. The target sensing object is used to detect the magnitude of the force.

[0007] The voltage data is input into a preset static decoupling model so that the static decoupling model outputs a predicted value corresponding to the force.

[0008] The static decoupling model is obtained by introducing an attention mechanism algorithm into the neural network algorithm and using a genetic algorithm to optimize the model parameters.

[0009] Optionally, the static decoupling model includes: an input layer, a hidden layer, an attention mechanism module, and an output layer;

[0010] The input layer is connected to the hidden layer through a radial basis function layer, and the radial basis function layer introduces a genetic algorithm to optimize the model parameters;

[0011] The hidden layer is connected to the output layer through a feedforward network, and the attention mechanism module is used to apply attention weights to each neuron in the output layer.

[0012] Optionally, the input layer includes a first number of neurons, with different neurons corresponding to voltage data in different preset axis directions, and the parameters of the encoding string corresponding to each neuron in the input layer are determined according to a genetic algorithm.

[0013] Optionally, the hidden layer includes a second number of neurons, and the value of each neuron in the hidden layer is determined based on the neurons associated with it in the input layer and the radial basis function;

[0014] The output layer includes a third number of neurons, and the value of each neuron in the output layer is determined based on the values ​​of the neurons associated with it in the hidden layer, a linear function, and attention weights.

[0015] Optionally, the formula for calculating the length of the encoded string is:

[0016]

[0017] Where L is the length of the encoded string, R is the first quantity, S1 is the second quantity, and S3 is the third quantity. , For attention weights, is the input vector of the attention mechanism module.

[0018] Optionally, the individual fitness function of the genetic algorithm is:

[0019]

[0020]

[0021] Where f is the individual fitness function, SE is the sum of the mean squared errors of the target expected value and the actual predicted value of the output layer, n represents the number of neurons in the output layer, and y i Let x be the target expected value of the i-th neuron. i This is the actual predicted value for the i-th neuron.

[0022] Optionally, in the genetic algorithm, the expression for the probability of an individual being selected is:

[0023]

[0024] Among them, f i Let f be the individual fitness function of the i-th neuron. sum It is the sum of the individual fitness functions of multiple neurons.

[0025] Secondly, this application provides a static decoupling device, comprising:

[0026] The acquisition module is used to acquire the voltage data output by the target sensing object to be statically decoupled. The target sensing object is subjected to a force in any preset axial direction, and the target sensing object is used to detect the magnitude of the force.

[0027] The input module is used to input the voltage data into a preset static decoupling model so that the static decoupling model outputs a predicted value corresponding to the force.

[0028] The static decoupling model is obtained by introducing an attention mechanism algorithm into the neural network algorithm and using a genetic algorithm to optimize the model parameters.

[0029] Optionally, the static decoupling model includes: an input layer, a hidden layer, an attention mechanism module, and an output layer;

[0030] The input layer is connected to the hidden layer through a radial basis function layer, and the radial basis function layer introduces a genetic algorithm to optimize the model parameters;

[0031] The hidden layer is connected to the output layer through a feedforward network, and the attention mechanism module is used to apply attention weights to each neuron in the output layer.

[0032] Optionally, the input layer includes a first number of neurons, with different neurons corresponding to voltage data in different preset axis directions, and the parameters of the encoding string corresponding to each neuron in the input layer are determined according to a genetic algorithm.

[0033] Optionally, the hidden layer includes a second number of neurons, and the value of each neuron in the hidden layer is determined based on the neurons associated with it in the input layer and the radial basis function;

[0034] The output layer includes a third number of neurons, and the value of each neuron in the output layer is determined based on the values ​​of the neurons associated with it in the hidden layer, a linear function, and attention weights.

[0035] Optionally, the formula for calculating the length of the encoded string is:

[0036]

[0037] Where L is the length of the encoded string, R is the first quantity, S1 is the second quantity, and S3 is the third quantity. , For attention weights, is the input vector of the attention mechanism module.

[0038] Optionally, the individual fitness function of the genetic algorithm is:

[0039]

[0040]

[0041] Where f is the individual fitness function, SE is the sum of the mean squared errors of the target expected value and the actual predicted value of the output layer, n represents the number of neurons in the output layer, and y i Let x be the target expected value of the i-th neuron. i This is the actual predicted value for the i-th neuron.

[0042] Optionally, in the genetic algorithm, the expression for the probability of an individual being selected is:

[0043]

[0044] Among them, f i Let f be the individual fitness function of the i-th neuron. sum It is the sum of the individual fitness functions of multiple neurons.

[0045] Thirdly, this application provides an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus;

[0046] Memory, used to store computer programs;

[0047] When a processor executes a program stored in memory, it implements the static decoupling method described in any of the first aspects.

[0048] Fourthly, this application provides a computer-readable storage medium storing a program for a static decoupling method, wherein when the program for the static decoupling method is executed by a processor, it implements the steps of the static decoupling method described in any of the first aspects.

[0049] The technical solutions provided in this application have the following advantages compared with the prior art:

[0050] This application embodiment obtains voltage data output by a target sensing object to be statically decoupled. The target sensing object is subjected to a force in any preset axial direction, and the target sensing object is used to detect the magnitude of the force. The voltage data is input into a preset static decoupling model, which can make the static decoupling model output a predicted value corresponding to the force. The static decoupling model is obtained by introducing an attention mechanism algorithm into a neural network algorithm and using a genetic algorithm to optimize the model parameters.

[0051] This application's embodiments introduce an attention mechanism (AM) into the RBF neural network to assign different attention to voltage signals along different axes, thereby obtaining the degree of coupling between force / torque dimensions in each direction and solving them sequentially in descending order of influence. This accelerates the convergence speed of the composite model RBF-AM-GA, improves decoupling accuracy and efficiency, and enhances the prediction performance of the neural network. Simultaneously, it incorporates the stochastic global search and optimization capabilities of the Genetic Algorithm (GA), leveraging its ability to find the global optimum in data fitting to fine-tune the nonlinear model parameters. This improves the robustness and generalization ability of the composite model RBF-AM-GA, ultimately achieving precise decoupling of the six-dimensional force sensor. The algorithm is simple, reliable, and yields accurate and effective results. Attached Figure Description

[0052] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0053] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0054] Figure 1 A flowchart of a static decoupling method provided in an embodiment of this application;

[0055] Figure 2 This application provides a schematic diagram of the applied forces in each axial direction of a six-dimensional force sensor.

[0056] Figure 3 This is a schematic diagram of the applied force in each axial direction of another six-dimensional force sensor provided in this application embodiment;

[0057] Figure 4 A model structure diagram of a static decoupling model provided in an embodiment of this application;

[0058] Figure 5 This is an overall flowchart of a static decoupling method provided in an embodiment of this application;

[0059] Figure 6 A schematic diagram comparing the curves of a training network and a test network is provided as an embodiment of this application;

[0060] Figure 7 This is another schematic diagram comparing the curves of a training network and a test network provided in an embodiment of this application;

[0061] Figure 8 A structural diagram of a static decoupling device provided in an embodiment of this application;

[0062] Figure 9 This is a structural diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0063] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0064] Due to limitations in the mechanical manufacturing process, patch technology, strain gauge lateral effects, and detection methods of six-dimensional force sensors, almost every force / torque component acting on the sensor in each dimension affects the sensor's output signals. This inevitably leads to mutual interference—inter-dimensional coupling—between dimensions. Different degrees of inter-dimensional signal coupling significantly impact the detection accuracy of six-dimensional force sensors and constrain their application and development in high-precision and cutting-edge fields. Therefore, this application provides a static decoupling method, apparatus, electronic device, and storage medium.

[0065] like Figure 1 As shown, the static decoupling method provided in this application embodiment may include the following steps:

[0066] Step S101: Obtain the voltage data output by the target sensing object to be statically decoupled;

[0067] In this embodiment, the target sensing object is a six-dimensional force sensor or other multi-dimensional force sensor, fiber optic sensor, etc., that exhibits coupling. The target sensing object is subjected to a force along any preset axis direction. The target sensing object is used to detect the magnitude of the force. The preset axis direction can be as follows: Figure 2 and Figure 3 As shown, after a force is applied to the target sensing object, the target sensing object outputs voltage data.

[0068] Step S102: Input the voltage data into a preset static decoupling model so that the static decoupling model outputs a predicted value corresponding to the force.

[0069] The static decoupling model is pre-trained using multiple sets of applied force values ​​and corresponding voltage data in different axial directions, so that the static decoupling model can predict the force value based on the applied force direction and voltage data during the inference stage.

[0070] When training the static decoupling model, voltage data can be obtained through data loading experiments on the target sensing object, in which a set force value is applied to the target sensing object. Figure 2 and Figure 3 The lateral forces FX and FY, axial force FZ, bending moment forces MX and MY, and torque force MZ are shown. The corresponding force values ​​are recorded as the output of the static decoupling model. The voltage data output by the target sensing object is used as the input of the static decoupling model. The purpose is to make the force value predicted by the static decoupling model based on the input closer to the recorded force value. This is because when a force of 5 Nm is actually applied in any axial direction, the force value detected by the target sensing object in that axial direction may not be 5 Nm. This is caused by coupling. The coupling phenomenon cannot be completely eliminated. However, the static decoupling model provided in this application embodiment can make the predicted force value closer to the actual applied force value to a great extent, ensuring that the measured value of the target sensing object is within a reasonable error range.

[0071] In this embodiment, the static decoupling model is a composite RBF-AM-GA model. That is, the static decoupling model is obtained by introducing an attention mechanism into the neural network algorithm and using a genetic algorithm to fine-tune the model parameters. This application improves the prediction performance of the neural network by introducing an attention mechanism (AM) into the Radial Basis Function Neural Network (RBF neural network) and integrates a genetic algorithm (GA) to fine-tune the model parameters, thereby achieving precise decoupling of the six-dimensional force sensor.

[0072] Population initialization, individual encoding, setting neural network parameters, and determining the fitness function. The initial learning rate η is set to 0.1, the number of epochs is set to 30, the population size M is set to 80, and the number of iterations to terminate the GA algorithm is selected as 100.

[0073] The model structure after adding the attention mechanism is as follows Figure 4 As shown, Figure 4In this diagram, each circle represents a neuron, the basic structure of a neural network. Weights are one of the parameters of the neural network, and their values ​​are updated as the network is trained. Adding an attention mechanism improves the original calculation method by prioritizing the resolution of factors with high interdimensional coupling. For example, Mx may be affected by X, Y, Z, or even MY, MZ, but the degree of influence is different. With the attention mechanism, the more important interfering factors will be resolved first in each training iteration, resulting in a better final result than without the attention mechanism. Based on the current sample data, the accuracy is improved by at least 2%.

[0074] Figure 4 In this context, Cj is the vector of attention weights used to construct the content. Cj can be understood as the attention distribution, and j represents the decoding position. These are the special attention weights for the feedforward network from the hidden layer to the output layer.

[0075] In this embodiment of the application, the static decoupling model includes: an input layer, a hidden layer, an attention mechanism module, and an output layer;

[0076] The input layer is connected to the hidden layer through a radial basis function layer, and the radial basis function layer introduces a genetic algorithm to optimize the model parameters;

[0077] The hidden layer is connected to the output layer through a feedforward network, and the attention mechanism module is used to apply attention weights to each neuron in the output layer.

[0078] In one embodiment of this application, the input layer includes a first number of neurons, with different neurons corresponding to voltage data in different preset axis directions, and the parameters of the encoding string corresponding to each neuron in the input layer are determined according to a genetic algorithm.

[0079] In one embodiment of this application, the hidden layer includes a second number of neurons, and the value of each neuron in the hidden layer is determined based on the neurons associated with it in the input layer and the radial basis function;

[0080] The output layer includes a third number of neurons, and the value of each neuron in the output layer is determined based on the values ​​of the neurons associated with it in the hidden layer, a linear function, and attention weights.

[0081] Chromosome encoding / individual population encoding is one of the parameters of a genetic algorithm. When using a real-number encoding mechanism, there is no quantization error problem in encoding and decoding, which enables fine-tuning and can represent a wide range of data, facilitating the design of GA algorithms.

[0082] The encoded string of a chromosome (this is a biomimetic term, similar to that used in neural networks; see the principles of genetic algorithms for details) generally consists of four parts: hidden layer weights, hidden layer thresholds, output layer weights, and output layer thresholds. The parameters in the RBF network algorithm are: the number of input layer nodes is R; the number of hidden layer nodes is S1; the number of output layer nodes is S2 (the various layers here are basic structural terms for neural networks). Figure 4 (Both have been indicated).

[0083] In one embodiment of this application, the formula for calculating the length L of the encoded string is:

[0084]

[0085] Where L is the length of the encoded string, R is the first quantity, S1 is the second quantity, and S3 is the third quantity. , For attention weights, The input vector of the attention mechanism module, that is Figure 4 C j The row of circles in front.

[0086] The purpose of using the GA algorithm to optimize the initial weights and thresholds of the RBF neural network is to minimize the mean square error between the predicted value and the target expected value of the RBF algorithm. Therefore, in one embodiment of this application, the individual fitness function of the genetic algorithm is:

[0087]

[0088]

[0089] Where f is the individual fitness function, SE is the sum of the mean squared errors between the target expected value and the true predicted value of the output layer, and the mean squared error is the average of the sum of the squared distances of each data point from the true value. This value is used to calculate f (in this invention, the reciprocal of the sum of mean squared errors is used as the fitness function in the GA algorithm). f is the fitness function in the GA algorithm, and it is the most critical value used to evaluate the performance of the network model. n represents the number of neurons in the output layer, and y... i Let x be the target expected value of the i-th neuron. i This is the actual predicted value for the i-th neuron.

[0090] In one embodiment of this application, the selection probability in the genetic algorithm—the probability that an individual can be selected—is based on its fitness level. The expression for the probability of an individual being selected is:

[0091]

[0092] Among them, f i is the individual fitness function of the i-th neuron, and f sum is the sum of the individual fitness functions of multiple neurons.

[0093] Crossover probability: Excessive crossover results in unnecessary waste of the solution space. If the crossover probability is set too small, it may reduce the newborn ability of individuals, leading to iteration stagnation. In this invention, the crossover probability is set to Pc = 0.5;

[0094] Mutation probability: The selection of the mutation probability affects the number of mutations. A reasonable mutation probability enables newborn individuals to retain good genes and discard bad genes. An unreasonable mutation probability will reduce the evolution speed and destroy the population gene pattern. In this invention, the mutation probability is set to Pm = 0.05.

[0095] In another embodiment of this application, when dealing with the overfitting problem, it is necessary to add a parameter constraint term based on the principle of empirical risk minimization to achieve the effect of constraining the loss function. The calculation formula of the loss function J in this invention is as follows:

[0096]

[0097] Among them, represents the initial value of the initial loss function; <00002​​​​​​​​​​​​​​​​​​​​​​​​​​​​​After the network has been continuously reconstructed and trained, the result is output after the convergence termination condition is met.

[0105] After completing the simulation experiment based on the static decoupling optimization scheme of the six-dimensional force sensor, the six-dimensional force sensor was connected to a controller with the decoupling algorithm already programmed in. Data loading experiments were then conducted again to collect sample data and train the network, outputting the decoupling results after applying the optimized scheme. The results of the two training iterations were compared, and the results are as follows: Figure 6 and Figure 7 (train: used to represent the training status with the latest sample dataset; val: used to represent the training status with the initial sample data. This verifies the superiority of the improved algorithm's static decoupling.)

[0106] This application's embodiments introduce an attention mechanism (AM) into the RBF neural network to assign different attention to voltage signals along different axes, thereby obtaining the degree of coupling between force / torque dimensions in each direction and solving them sequentially in descending order of influence. This accelerates the convergence speed of the composite model RBF-AM-GA, improves decoupling accuracy and efficiency, and enhances the prediction performance of the neural network. Simultaneously, it incorporates the stochastic global search and optimization capabilities of the Genetic Algorithm (GA), leveraging its ability to find the global optimum in data fitting to fine-tune the nonlinear model parameters. This improves the robustness and generalization ability of the composite model RBF-AM-GA, ultimately achieving precise decoupling of the six-dimensional force sensor. The algorithm is simple, reliable, and yields accurate and effective results.

[0107] In another embodiment of this application, a static decoupling device is also provided, such as... Figure 8 As shown, it includes:

[0108] The acquisition module 11 is used to acquire the voltage data output by the target sensing object to be statically decoupled. The target sensing object is subjected to a force in any preset axial direction. The target sensing object is used to detect the magnitude of the force.

[0109] Input module 12 is used to input the voltage data into a preset static decoupling model so that the static decoupling model outputs a predicted value corresponding to the force.

[0110] The static decoupling model is obtained by introducing an attention mechanism algorithm into the neural network algorithm and using a genetic algorithm to optimize the model parameters.

[0111] Optionally, the static decoupling model includes: an input layer, a hidden layer, an attention mechanism module, and an output layer;

[0112] The input layer is connected to the hidden layer through a radial basis function layer, and the radial basis function layer introduces a genetic algorithm to optimize the model parameters;

[0113] The hidden layer is connected to the output layer through a feedforward network, and the attention mechanism module is used to apply attention weights to each neuron in the output layer.

[0114] Optionally, the input layer includes a first number of neurons, with different neurons corresponding to voltage data in different preset axis directions, and the parameters of the encoding string corresponding to each neuron in the input layer are determined according to a genetic algorithm.

[0115] Optionally, the hidden layer includes a second number of neurons, and the value of each neuron in the hidden layer is determined based on the neurons associated with it in the input layer and the radial basis function;

[0116] The output layer includes a third number of neurons, and the value of each neuron in the output layer is determined based on the values ​​of the neurons associated with it in the hidden layer, a linear function, and attention weights.

[0117] Optionally, the formula for calculating the length of the encoded string is:

[0118]

[0119] Where L is the length of the encoded string, R is the first quantity, S1 is the second quantity, and S3 is the third quantity. , For attention weights, is the input vector of the attention mechanism module.

[0120] Optionally, the individual fitness function of the genetic algorithm is:

[0121]

[0122]

[0123] Where f is the individual fitness function, SE is the sum of the mean squared errors of the target expected value and the actual predicted value of the output layer, n represents the number of neurons in the output layer, and y i Let x be the target expected value of the i-th neuron. i This is the actual predicted value for the i-th neuron.

[0124] Optionally, in the genetic algorithm, the expression for the probability of an individual being selected is:

[0125]

[0126] Among them, f iLet f be the individual fitness function of the i-th neuron. sum It is the sum of the individual fitness functions of multiple neurons.

[0127] In another embodiment of this application, an electronic device is also provided, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus.

[0128] Memory, used to store computer programs;

[0129] When the processor executes a program stored in memory, it implements the static decoupling method described in any of the foregoing method embodiments.

[0130] The electronic device provided in this invention, through its processor executing a program stored in memory, introduces an attention mechanism (AM) into the RBF neural network to assign different attention to voltage signals along different axes, thereby acquiring the degree of coupling between force / torque dimensions in each direction and solving them sequentially in descending order of influence. This accelerates the convergence speed of the composite model RBF-AM-GA, improves decoupling accuracy and efficiency, and enhances the prediction performance of the neural network. Simultaneously, it incorporates the stochastic global search and optimization capabilities of the Genetic Algorithm (GA), leveraging its ability to find the global optimum in data fitting to fine-tune the nonlinear model parameters. This improves the robustness and generalization ability of the composite model RBF-AM-GA, ultimately achieving precise decoupling of the six-dimensional force sensor. The algorithm is simple, reliable, and yields accurate and effective results.

[0131] The communication bus 1140 mentioned in the above-mentioned electronic device can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus 1140 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 9 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0132] The communication interface 1120 is used for communication between the above-mentioned electronic device and other devices.

[0133] The memory 1130 may include random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0134] The processor 1110 mentioned above can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0135] In another embodiment of this application, a computer-readable storage medium is provided, on which a program for a static decoupling method is stored. When the program for the static decoupling method is executed by a processor, it implements the steps of the static decoupling method described in any of the foregoing method embodiments.

[0136] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0137] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A static decoupling method, characterized in that, include: The voltage data output by the target sensing object to be statically decoupled is obtained. The target sensing object is subjected to a force in any preset axial direction. The target sensing object is used to detect the magnitude of the force. The voltage data is input into a preset static decoupling model so that the static decoupling model outputs a predicted value corresponding to the force. The static decoupling model includes: an input layer, a hidden layer, an attention mechanism module, and an output layer; The input layer is connected to the hidden layer through a radial basis function layer, and the radial basis function layer introduces a genetic algorithm to optimize the model parameters; The input layer includes a first number of neurons, with different neurons corresponding to voltage data in different preset axis directions. The parameters of the encoding string corresponding to each neuron in the input layer are determined according to a genetic algorithm. The hidden layer is connected to the output layer through a feedforward network, and the attention mechanism module is used to apply attention weights to each neuron in the output layer. The hidden layer includes a second number of neurons, and the value of each neuron in the hidden layer is determined based on the neurons associated with it in the input layer and the radial basis function. The output layer includes a third number of neurons, and the value of each neuron in the output layer is determined based on the values ​​of the neurons associated with it in the hidden layer, the linear function, and the attention weights. The static decoupling model is obtained by introducing an attention mechanism algorithm into the neural network algorithm and using a genetic algorithm to optimize the model parameters.

2. The method according to claim 1, characterized in that, The individual fitness function of the genetic algorithm is: Where f is the individual fitness function, SE is the sum of the mean squared errors of the target expected value and the actual predicted value of the output layer, n represents the number of neurons in the output layer, and y i Let x be the target expected value of the i-th neuron. i This is the actual predicted value for the i-th neuron.

3. The method according to claim 1, characterized in that, In the genetic algorithm, the expression for the probability of an individual being selected is: Among them, f i Let f be the individual fitness function of the i-th neuron. sum It is the sum of the individual fitness functions of multiple neurons.

4. A static decoupling device, characterized in that, include: The acquisition module is used to acquire the voltage data output by the target sensing object to be statically decoupled. The target sensing object is subjected to a force in any preset axial direction, and the target sensing object is used to detect the magnitude of the force. The input module is used to input the voltage data into a preset static decoupling model so that the static decoupling model outputs a predicted value corresponding to the force. The static decoupling model includes an input layer, a hidden layer, an attention mechanism module, and an output layer. The input layer is connected to the hidden layer via a radial basis function layer, which incorporates a genetic algorithm to fine-tune the model parameters. The input layer includes a first number of neurons, with different neurons corresponding to voltage data along different preset axes. The parameters of the encoding strings corresponding to each neuron in the input layer are determined by the genetic algorithm. The hidden layer is connected to the output layer via a feedforward network, and the attention mechanism module applies attention weights to each neuron in the output layer. The hidden layer includes a second number of neurons, and the value of each neuron in the hidden layer is determined based on its associated neurons in the input layer and the radial basis function. The output layer includes a third number of neurons, and the value of each neuron in the output layer is determined based on the values ​​of its associated neurons in the hidden layer, a linear function, and the attention weights. The static decoupling model is obtained by introducing an attention mechanism algorithm into the neural network algorithm and using a genetic algorithm to optimize the model parameters.

5. An electronic device, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; When a processor executes a program stored in memory, it implements the static decoupling method as described in any one of claims 1 to 3.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program for a static decoupling method, which, when executed by a processor, implements the steps of the static decoupling method according to any one of claims 1-3.

Images