Method for optimizing six-dimensional force sensor decoupling based on improved grey wolf algorithm

Abstract

The invention relates to a six-dimensional force sensor decoupling method based on improved grey wolf algorithm optimization, which relates to the technical field of six-dimensional force sensors, and comprises the following steps: calibrating a six-dimensional force sensor to obtain data under force input in different directions; then, a BP neural network model is established, output voltage signals of six channels serve as input vectors of a BP neural network, and force acting on the sensor serves as an output vector; and finally, a grey wolf optimization algorithm is improved by using three strategies of dynamic distribution parameter adjustment, dynamic weight position updating strategy, diversity threshold setting and diversity-based population re-initialization, and accurate decoupling of calibration data is realized. According to the method, in the improved grey wolf algorithm, the diversity of the population is maintained or improved, so that effective exploration of the algorithm in a global search space is promoted, local optimum is prevented, the probability of finding a global optimal solution is increased, the measurement precision and adaptability of the six-dimensional force sensor can be effectively improved, and inter-dimensional coupling is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of six-dimensional force sensors, and in particular to a six-dimensional force sensor decoupling method based on an improved grey wolf algorithm. Background Art

[0002] With the advancement of science and technology, six-dimensional force sensors have urgent needs and broad application prospects in high-tech fields such as industrial engineering. In the field of aerospace, the "Tiangong-1" and "Shenzhou-9" are helped by six-dimensional force sensors to achieve their stable space docking. They can also be used to simulate the aerodynamic measurement of aircraft propellers, and can accurately measure the changing six-dimensional forces of the aircraft under the action of aerodynamic loads; in the field of robotics, six-dimensional force sensors are often installed in the joint movements of the robot's fingers, wrists, etc. to achieve intelligent control of the robot. Therefore, the development and application of multi-dimensional force sensors have important theoretical and practical significance. Because the six-dimensional force sensor is affected by its own structure and processing errors, the force applied in one dimension of the sensor will also have output in other directions, resulting in serious inter-dimensional coupling in the six-dimensional force sensor, and inter-dimensional coupling is an important factor affecting its accuracy. In order to reduce the influence of coupling errors, the decoupling algorithm of the six-dimensional force sensor is studied by combining error analysis, theoretical derivation and experimental verification.

[0003] When using intelligent algorithms to decouple six-dimensional force sensors, commonly used algorithms include BP neural network algorithm (BPNN), extreme learning machine (ELM), support vector regression (SVR) algorithm, etc. When applying the above algorithms to decouple six-dimensional force sensors, problems such as oscillation, slow convergence, premature convergence, and falling into local extremes are prone to occur during the training process, resulting in decoupling failure, which brings difficulties to research in this field. A more intelligent and universal method is urgently needed to solve the problem of sensor inter-dimensional coupling. Summary of the invention

[0004] In order to overcome the shortcomings of the above-mentioned prior art that oscillation, slow convergence, premature convergence and falling into local extrema easily occur during training, which lead to decoupling failure, the main purpose of the present invention is to provide a decoupling method for optimizing a six-dimensional force sensor based on an improved grey wolf algorithm, and to improve the grey wolf optimization algorithm (Improve Grey Wolf Optimization) by using three strategies, namely, adjusting dynamic distribution parameters, nonlinear convergence factors, setting diversity thresholds and reinitializing populations based on diversity, so as to solve the problems existing in the prior art that traditional algorithms are prone to falling into local optimality and slow convergence.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A decoupling method for a six-dimensional force sensor based on an improved grey wolf algorithm is characterized in that: the output voltage signal U of the six channels of the six-dimensional force sensor is U = [U Fx , U Fy , U Fz 、M Fx 、M Fy 、M Fz ], as the input vector of the neural network, acting on the sensor F = [F x 、F y 、F z 、M x 、M y 、M z ], as the output vector, to establish a six-dimensional force sensor decoupling model; the Grey Wolf Optimization algorithm (Improve GreyWolfOptimization) was improved by using three strategies including adjusting dynamic distribution parameters, nonlinear convergence factor, setting diversity threshold and reinitializing population based on diversity. The improved Grey Wolf Optimization algorithm was used to optimize the BP neural network to decouple the six-dimensional force sensor model.

[0007] In the example of the present invention, the six-dimensional force sensor data is first calibrated, and the data is collected and saved using a data acquisition card. Next, a BP neural network decoupling model is established, as follows:

[0008] The six-dimensional force sensor is statically loaded. The hardware system includes USB3200 data acquisition card, weights, six-dimensional force test platform, etc. First, install the sensor on the calibration platform and fix it. After completing all the hardware and software equipment required for calibration, and loading the calibration table, confirming that the system connection is normal and can be used, the experimental preparation is completed and calibration begins. In the calibration loading, follow the principle of small to large, and start loading from the minimum loading point, and then increase the calibration load step by step until the maximum range. After loading is completed, the value is calibrated and saved. When the sensor is calibrated in multiple directions, the calibration loading space becomes a three-dimensional space, and three aspects of force and torque are loaded simultaneously, which is convenient for recording the nonlinear mechanical characteristics under the simultaneous action of multi-dimensional loads to prevent the occurrence of singular values. It is necessary to calibrate three times in each direction and take the average value to ensure the accuracy of the data. Complete the loading of force and torque in all directions. Record and save all data to complete the calibration experiment.

[0009] In an example of the present invention, a neural network decoupling model is first constructed, and then the parameters are adjusted through orthogonal experiments, and the neural network decoupling ability evaluation model is used to determine whether the optimized weights and thresholds can achieve more accurate decoupling, and the final decoupling model is established. The gray wolf optimization algorithm GWO is improved by adjusting dynamic distribution parameters, nonlinear convergence factors, setting diversity thresholds, and reinitializing populations based on diversity, as follows:

[0010] (1) Dynamically adjust distribution parameters

[0011] In order to improve the performance of the gray wolf optimization algorithm in high-dimensional and complex search spaces, the method of dynamically adjusting the distribution parameters is adopted, as shown in formula (3), to generate the initial position of the search agent (wolf pack) and ensure that the initialized wolf pack position does not exceed the predetermined boundary, as shown in formula (6)

[0012]

[0013] Among them, L represents the number of layers of the wolf pack, S refers to the number of search agents, and L e represents the number of agents in each layer, C represents the standard deviation, and l represents the total number of wolf pack layers. Initialize the agent position. For even dimensions, the position P is initialized using the normal distribution model, as shown in (4):

[0014] P(idx,d)=Ib(d)-(ub(d)-Ib(d))·n(0.5,C) (4)

[0015] For odd dimensions, the position P is initialized using a uniform distribution model, as shown in formula (5):

[0016] P(idx,d)=Ib(d)+(ub(d)-Ib(d))·h (5)

[0017] Ensure that the location generated by each distribution is within the search space, as shown in formula (6):

[0018] P(idx,d)=max(Ib(d),min·(P(idx,d),ub(d))) (6)

[0019] Where ub(d) and Ib(d) represent the upper and lower boundaries respectively, and h represents a random number.

[0020] (2) Nonlinear convergence factor

[0021] The update mechanism of the control parameter a has a decisive influence on the overall performance of the algorithm. The traditional linear convergence cannot give full play to the optimization advantage, as shown in formula (7). Therefore, the convergence factor of the gray wolf algorithm is nonlinearly adjusted, such as Figure 1 , as shown in formula (8):

[0022]

[0023] α=a max -(α min -cos(π·t / T max ))·k0-0.5 (8)

[0024] Among them, bmax =2, b min =0, t is the current iteration number, T is the maximum iteration number, k0 is the adjustment coefficient, and its value range is (0, 1).

[0025] (3) Setting diversity threshold and reinitializing population based on diversity

[0026] Setting a diversity threshold and reinitializing the population based on diversity can improve the diversity of the gray wolf population in the gray wolf optimization algorithm and promote the efficient search of the algorithm in the global search space. The specific steps are as follows:

[0027] First, the standard deviation of the wolf pack's position in each dimension is calculated, and then the standard deviation of the position in each dimension is taken to evaluate the diversity of the population. If the diversity value is lower than the pre-set threshold (indicating that the individuals in the wolf pack may be too concentrated and the search range is limited), a part of the population is selected for initialization to increase the diversity of the population, thereby improving the search efficiency and avoiding the algorithm from converging to the local optimal solution too early.

[0028] The improved gray wolf algorithm is a new meta-heuristic search algorithm. The basic idea is to find the optimal solution through strict social hierarchy, encirclement, tracking and attacking prey in the population. The specific steps are as follows:

[0029] 1. First, determine the neural network model and transform the output voltage signal U of the six channels of the sensor into Fx , U Fy , U Fz 、M Fx 、M Fy 、M Fz ], as the input vector of the BP neural network, acting on the sensor F = [F x 、F y 、F z 、M x 、M y 、M z ] as the output vector, that is, the input layer parameter of the neural network is m = 6, the output layer parameter is n = 6, and the hidden layer parameter of the neural network is given by the empirical formula (a is an integer between 1 and 10) and trial and error method, the number of nodes in the neural network input layer and the neural network output layer is set to 6, the data is loaded and normalized, and the number of training samples, test samples, target error, etc. are set.

[0030] 2. Dynamically adjust the distribution parameters to initialize the wolf pack size, individual position, number of iterations, and the values ​​of a, A, and C.

[0031] 3. Calculate the fitness value between the gray wolf and the prey, and set the first three wolves with the shortest individual distance to the prey as α wolf, β wolf, and δ wolf.

[0032] The mathematical model of the algorithm is:

[0033] D=|C·X P (t)-X(t)| (9)

[0034] X(t+1)=X P (t)-A·D (10)

[0035] Where: D is the distance vector between the individual gray wolf and the prey; t is the current iteration number; X p (t) is the position vector of the prey; X(t) is the position vector of the individual gray wolf; the convergence factor A = 2a·r1-a, C = 2·r2, r1, r2 are random vectors in [0,1], and a is the control parameter.

[0036] When encircling the prey, wolf α, wolf β, and wolf δ are the wolves closest to and most aware of the prey. In the abstract search space, the position vector of ω is updated by calculating the position vectors of α, β, and δ. Its mathematical expression is:

[0037]

[0038] Where: X represents the current position of the gray wolf; C1, C2, C3 are random disturbance vectors; X α , X β , X δ are the positions of α, β, and δ respectively; D α , D β , D δ The following formula defines the step length and direction of the wolf ω in the wolf pack moving toward the wolf α, wolf β and wolf δ, as well as the final position of the new generation of gray wolves.

[0039]

[0040] In the Grey Wolf Algorithm, the update mechanism of the control parameter a has a decisive influence on the overall performance of the algorithm. However, the linear adjustment method fails to fully exert the potential of balancing between finding the global optimal solution and accurately searching for the local optimal solution. The formula is as follows:

[0041]

[0042] Therefore, the invention proposes a nonlinear update strategy, through which the adjustment of a will no longer follow a single linear path, but adopt a more complex nonlinear pattern, aiming to more accurately regulate the exploration and utilization behavior of the wolf pack, thereby achieving a better balance between global and local search. The specific mathematical expression of this nonlinear strategy is as follows:

[0043] α=a max -(α min -cos(π·t / T max ))·k0-0.5 (14)

[0044] Among them, b max =2, b min =0, t is the current iteration number, T is the maximum iteration number, k0 is the adjustment coefficient, and its value range is (0, 1).

[0045] Using this formula, the early deceleration rate of the control parameter a is reduced, enhancing the global search capability; the later deceleration rate is accelerated, improving the convergence speed of the algorithm's local optimization problem. This adjustment mechanism is conducive to balancing exploration and development at different stages of the algorithm and improving optimization efficiency and effectiveness.

[0046] 4. Use the update formula α=a max -(α min -cos(π·t / T max ))·k0-0.5, convergence factor formula: A=2a·r1-a, and random vector formula: C=2·r2 Update the values ​​of a, A, and C

[0047] 5. Improve the position update mechanism and dynamic weights and set diversity thresholds to improve global search capabilities and avoid falling into local optimality;

[0048] Output the optimal solution X α And the optimal fitness value J α ;

[0049] 6. Calculate the distance between the new generation of wolves and their prey, update the individual positions and new fitness values ​​of α wolf, β wolf, and δ wolf according to the fitness value and formula X(t+1)=(X1+X2+X3) / 3, and determine whether the maximum number of iterations has been reached. If so, output the optimization result: the optimal solution X α , and end the program. If it is not reached, return to recalculate.

[0050] 7. Reversely map the global optimal solution obtained in the entire improved gray wolf algorithm to the neural network, that is, use the gray wolf algorithm to solve the initial parameter values ​​of the number of hidden layer neurons, neuron connection weights and thresholds, establish a BP neural network, and use normalized training sample data for learning and training. Use the normalized test sample data set to test the trained network, output the denormalized test result value, and round the output result. The algorithm ends.

[0051] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention establishes a decoupling model with the output and input of the six-dimensional force sensor. The Grey Wolf Optimization Algorithm GWO is improved by using three strategies, namely, adjusting the dynamic distribution parameters, nonlinear convergence factor, setting the diversity threshold, and reinitializing the population based on diversity. The results show that the improved algorithm controls the type I and type II errors within 0.21% and 0.068% respectively (specifically, Figure 5 and Figure 6 ), showing higher decoupling accuracy and faster convergence speed than traditional BP neural network, least squares method and other intelligent algorithms. The versatility and stability of the improved GWO-BP algorithm are excellent under different data sets, providing an effective way to optimize complex sensor systems. Verify the correctness and superiority of the model proposed in this paper. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The drawings described herein are used to provide further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute improper limitations on the present application.

[0053] Figure 1 Improve the gray wolf optimization algorithm for the present invention to optimize the BP neural network solution process

[0054] Figure 2 Before and after comparison to improve the convergence factor

[0055] Figure 3 Iteration curves of five optimization algorithms

[0056] Figure 4 Convergence curves of five optimization algorithms

[0057] Figure 5 Comparison curve of type I error of different optimization algorithms

[0058] Figure 6 Comparison curve of type II error of different optimization algorithms DETAILED DESCRIPTION

[0059] The present invention will be further described below in conjunction with the accompanying drawings and implementation modes.

[0060] Example:

[0061] See also Figure 1-Figure 6The present invention provides a method for optimizing the decoupling of six-dimensional force sensors based on an improved gray wolf algorithm, which solves the problem that when the existing algorithm is used to decouple the six-dimensional force sensor, oscillation, slow convergence, premature convergence, and local extrema are prone to occur during the training process, resulting in decoupling failure. It also brings difficulties to the research in this field. A more intelligent and universal method is urgently needed to solve the problem of inter-dimensional coupling of sensors. The method can be summarized as the following steps:

[0062] S1. Data calibration of six-dimensional force sensor; S2. Construction of BP neural network model; S3. Improved grey wolf algorithm; S4. Example analysis.

[0063] Among them, S1 includes two steps:

[0064] S1.1. First, statically load the six-dimensional force sensor. The hardware system includes USB3200 data acquisition card, weights, six-dimensional force test platform, etc. Then install the sensor on the calibration platform and fix it. After completing all the hardware and software equipment required for calibration and loading the calibration table, finally, confirm that the system connection is normal and can be used, the experimental preparation is completed and calibration begins.

[0065] S1.2. In the calibration loading, follow the principle of starting from small to large, and start loading from the minimum loading point, then gradually increase the calibration load until the maximum range. After loading is completed, calibrate and save the value. When calibrating the sensor in multiple directions, the calibration loading space becomes a three-dimensional space, and three aspects of force and torque are loaded simultaneously to facilitate recording the nonlinear mechanical characteristics under the simultaneous action of multi-dimensional loads and prevent the occurrence of singular values. It is necessary to calibrate three times in each direction and take the average value to ensure the accuracy of the data. Complete the loading of forces and torques in all directions. Record and save all data to complete the calibration experiment.

[0066] S2 includes a step:

[0067] The input force F of the six-dimensional force sensor is [F x 、F y 、F z 、M x 、M y 、M z ] and output voltage U=[U Fx , U Fy , U Fz 、M Fx 、M Fy 、M Fz ], a six-dimensional force sensor decoupling model is established, that is, the input layer parameter m of the neural network is 6, the output layer parameter n is 6, and the hidden layer parameter of the neural network is given by the empirical formula (a is an integer between 1 and 10) and trial and error.

[0068] S3 consists of seven steps:

[0069] S3.1. According to the fitness value, individuals are divided into four levels: α, β, δ, and ω. Among them, α is the optimal solution, β and δ are the second and third optimal solutions respectively, and the candidate solution is ω;

[0070] S3.2, initialize the relevant parameters of the improved gray wolf algorithm, set the population size, adjustment coefficient, maximum number of iterations, etc.;

[0071] S3.3, calculate the fitness value between the gray wolf individual and the prey, and set the first three wolves with the shortest individual distance to the prey as α wolf, β wolf, and δ wolf;

[0072] S3.4, update the individual positions of wolf α, wolf β, and wolf δ, and calculate the fitness value of the current gray wolf individual according to the formula: X(t+1)=(X1+X2+X3) / 3;

[0073] S3.5, generate a according to the improved nonlinear factor convergence control strategy, and calculate the coordination parameters A and C according to the coordination parameter adaptive optimization strategy;

[0074] S3.6, improve the position update mechanism and dynamic weight and set the diversity threshold to enhance the global search capability and avoid falling into the local optimum;

[0075] Output the optimal solution X α And the optimal fitness value J α ;

[0076] S3.7, the optimal solution X α Save the optimal weights and biases as the BP neural network, use the optimized neural network to decouple the calibration data, and determine the optimal solution X α If it is feasible, the algorithm ends; otherwise, the calculation is repeated.

[0077] In step S2, the established decoupled data model is divided into a training set and a test set, 80% of which is used as a training set and 20% is used as a test set. The data set is divided randomly using the randperm function in Matlab to verify the stability of the algorithm and enhance the credibility of the algorithm through multiple calculations.

[0078] The step S3.2 is to initialize the dynamically adjusted distribution parameters of the improved grey wolf algorithm, and the specific steps are as follows:

[0079] In order to improve the performance of the gray wolf optimization algorithm in high-dimensional and complex search spaces, the method of dynamically adjusting the distribution parameters is adopted, as shown in formula (3), to generate the initial position of the search agent (wolf pack) and ensure that the initialized wolf pack position does not exceed the predetermined boundary, as shown in formula (6)

[0080]

[0081]

[0082] Among them, L represents the number of layers of the wolf pack, S refers to the number of search agents, and L e represents the number of agents in each layer, C represents the standard deviation, and l represents the total number of wolf pack layers. Initialize the agent position. For even dimensions, the position P is initialized using the normal distribution model, as shown in (4):

[0083] P(idx,d)=Ib(d)-(ub(d)-Ib(d))·n(0.5,C) (4)

[0084] For odd dimensions, the position P is initialized using a uniform distribution model, as shown in formula (5):

[0085] P(idx,d)=Ib(d)+(ub(d)-Ib(d))·h (5)

[0086] Ensure that the location generated by each distribution is within the search space, as shown in formula (6):

[0087] P(idx,d)=max(Ib(d),min·(P(idx,d),ub(d))) (6)

[0088] Where ub(d) and Ib(d) represent the upper and lower boundaries respectively, and h represents a random number.

[0089] In step S3.5, the improved gray wolf algorithm parameters are initialized. The control parameter a needs to be optimized. The traditional linear convergence cannot give full play to the optimization advantage. The formula is as follows (7). Therefore, the convergence factor of the gray wolf algorithm is nonlinearly adjusted, such as Figure 2 , as shown in formula (8):

[0090]

[0091] α=a max -(α min -cos(π·t / T max ))·k0-0.5 (8)

[0092] Among them, b max =2, b min=0, t is the current iteration number, T is the maximum iteration number, k0 is the adjustment coefficient, and its value range is (0, 1).

[0093] Setting the diversity threshold and reinitializing the population based on diversity in step S3.6 can improve the diversity of the gray wolf population in the gray wolf optimization algorithm and promote the efficient search of the algorithm in the global search space. The specific steps are as follows:

[0094] First, the standard deviation of the wolf pack's position in each dimension is calculated, and then the standard deviation of the position in each dimension is taken to evaluate the diversity of the population. If the diversity value is lower than the pre-set threshold (indicating that the individuals in the wolf pack may be too concentrated and the search range is limited), finally a part of the population is selected for initialization to increase the diversity of the population, thereby improving the search efficiency and avoiding the algorithm from converging to the local optimal solution too early.

[0095] S4. Example Analysis

[0096] Taking the calibration data of the six-dimensional force sensor as an example, there are 160 sets of calibration data, of which 80% are used as training sets and 20% are used as test sets. This study takes into account the situation of random loading of forces in various directions. The data set is divided randomly using the randperm function in Matlab to verify the stability of the algorithm, and the average value is taken through multiple calculations to enhance the credibility of the algorithm. The setting parameters of the IGWO algorithm are as follows: the number of gray wolf populations N = 50, the maximum number of iterations T max =200, the maximum value of the convergence factor a max =2, minimum value a min =0, set the diversity threshold to 0.2.

[0097] Programming calculations were performed in Matlab software, and the IGWO algorithm was used to optimize the BP neural network to decouple the six-dimensional force sensor model. The fitness curve of the IGWO algorithm is shown in Figure 3 For comparative analysis, Figure 3 The convergence curves of BP neural network algorithm, genetic algorithm (GA), particle swarm optimization algorithm (PSO), GWO algorithm and IGWO algorithm for solving the model are given in Figure 3 It can be seen that the IGWO-BP algorithm converges to the optimal fitness value after about 29 generations, and the number of iterations is the least.

[0098] Figure 4A comparison chart of the training error results of the five algorithms is given. The IGWO-BP algorithm shows fewer iterations and faster running time than other algorithms when achieving the same mean square error, effectively overcoming the shortcomings of other algorithms in terms of convergence speed and local optimality. The correctness of adjusting dynamic distribution parameters, nonlinear convergence factors, setting diversity thresholds, and reinitializing population strategies based on diversity is verified.

[0099] Table 1 gives the errors of the six algorithms. Figure 5 and Figure 6 It can be concluded that the linear decoupling ability of the least squares method is relatively poor, and the maximum type I error in each direction after decoupling is 11.11%. The maximum type I errors in each direction of BP, GA-BP, and improved GWO-BP are 3.268%, 1.3%, and 0.341%, respectively, and the type I error is greatly reduced. The improved GWO-BP algorithm controls the type I error within 0.267%, and the type I errors in different directions are smaller than those of other algorithms; in terms of type II error, the maximum error of the improved GWO-BP algorithm is 0.049%. Compared with other algorithms, the decoupling accuracy is greatly improved, especially in M x 、M y 、M z In terms of direction, the Type II error is controlled within 0.05% and the coupling error is well controlled.

[0100] Table 1 - Errors of various algorithms

[0101]

[0102] In summary, the decoupling of the six-dimensional force sensor by optimizing the BP neural network according to the improved gray wolf algorithm of the present invention can effectively reduce the inter-dimensional coupling of the six-dimensional force sensor, thereby improving the measurement accuracy of the sensor. By using three strategies such as adjusting the dynamic distribution parameters, nonlinear convergence factors, setting the diversity threshold, and reinitializing the population based on diversity, the convergence of the algorithm is effectively improved, the search space is expanded, and the probability of the algorithm jumping out of the local optimal solution is increased. The experimental results not only prove the advantages of this algorithm in the decoupling control of the six-dimensional force sensor, but also show that it is suitable for the decoupling control of other models.

[0103] It should be noted that, in the present invention, relational terms such as the I error and the II error are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0104] The above embodiments are merely examples of the present invention and do not limit the protection scope of the present invention. All designs that are the same or similar to the present invention fall within the protection scope of the present invention.

Claims

1. The decoupling method of the six-dimensional force sensor is optimized based on the improved grey wolf algorithm, characterized in that: The following steps are involved: Calibrate the six-dimensional force sensor data, collect and save the data using a data acquisition card, and build a BP neural network decoupling model; The gray wolf algorithm is improved by adjusting dynamic distribution parameters, nonlinear convergence factors, setting diversity thresholds, and reinitializing the population based on diversity. The improved grey wolf algorithm is used to obtain the optimal weight matrix and threshold vector, and the BP neural network decoupling model is optimized to obtain the optimized BP neural network decoupling model; The six-dimensional force sensor data is input into the optimized BP neural network decoupling model to obtain the decoupled force components.

2. The method for optimizing the decoupling of a six-dimensional force sensor based on an improved grey wolf algorithm as claimed in claim 1, characterized in that: The construction of the BP neural network decoupling model comprises: The input force F based on the six-dimensional force sensor is F x 、F y 、F z 、M x 、M y 、M z ] and output voltage U=[U Fx , U Fy , U Fz 、M Fx 、M Fy 、M Fz ], determine the input layer m = 6, the output layer n = 6 and the number of neurons in the hidden layer of the neural network by And trial and test method, where a is an integer between 1 and 10, to build the structure of the BP neural network model.

3. The method for optimizing the decoupling of a six-dimensional force sensor based on an improved grey wolf algorithm as claimed in claim 1, characterized in that: The optimal weight matrix and threshold vector include the number of hidden layer neurons, neuron connection weights and initial parameter values ​​of the threshold solved by the Grey Wolf Algorithm.

4. The method for optimizing the decoupling of a six-dimensional force sensor based on an improved grey wolf algorithm as claimed in claim 1, characterized in that: The improved gray wolf algorithm comprises the following steps: According to the fitness value, individuals are divided into four levels: α, β, δ, ω, where α is the optimal solution, β and δ are the second and third optimal solutions respectively, and the candidate solution is ω; Initialize the relevant parameters of the improved gray wolf algorithm, set the population size, adjustment coefficient, and maximum number of iterations; Calculate the fitness value between the gray wolf individual and the prey, and set the first three wolves with the shortest individual distance to the prey as α wolf, β wolf, and δ wolf; Update the individual positions of wolf α, wolf β, and wolf δ, and calculate the fitness value of the current gray wolf individual according to the formula: X(t+1)=(X1+X2+X3) / 3; Generate a according to the improved nonlinear factor convergence control strategy, and calculate the coordination parameters A and C according to the coordination parameter adaptive optimization strategy; Improvements and setting of diversity thresholds based on location update mechanisms and dynamic weights; Output the optimal solution X α And the optimal fitness value J α .

5. The method for optimizing the decoupling of a six-dimensional force sensor based on an improved grey wolf algorithm as claimed in claim 1, characterized in that: The improvement of the grey wolf algorithm by adjusting the dynamic distribution parameters includes: The method of dynamically adjusting the distribution parameters is used to generate the initial position of the search wolf pack to ensure that the initialized wolf pack position does not exceed the predetermined boundary. The formula is as follows: Among them, L represents the number of layers of the wolf pack, S refers to the number of search wolves, and L e represents the number of agents in each layer, C represents the standard deviation, and l represents the total number of wolf pack layers; initialize the agent position. For even dimensions, the position P is initialized using the normal distribution model, as shown in formula (4): P(idx,d)=Ib(d)-(ub(d)-Ib(d))·n(0.5,C) (4) For odd dimensions, the position P is initialized using a uniform distribution model, as shown in formula (5): P(idx,d)=Ib(d)+(ub(d)-Ib(d))·h (5) Ensure that the location generated by each distribution is within the search space, as shown in formula (6): P(idx,d)=max(Ib(d),min·(P(idx,d),ub(d))) (6) Among them, ub(d) and Ib(d) represent the upper and lower boundaries respectively, and h represents a random number; The nonlinear convergence factor improves the Grey Wolf Algorithm, comprising the following steps: a=a max -(a min -cos(π·t / T max ))·k0-0.5 (8) Among them, b max =2, b min =0, t is the current iteration number, T is the maximum iteration number, k0 is the adjustment coefficient, and its value range is (0, 1); The method of setting a diversity threshold and reinitializing the population based on diversity to improve the gray wolf algorithm includes the following steps: Obtain the standard deviation of the wolf pack's position in each dimension; The diversity of the population is evaluated according to the standard deviation of the position in each dimension, where the diversity value is lower than a pre-set threshold and a part of the population is selected for initialization.

Images