Flexible robot arm system identification method and apparatus based on wiener model

Abstract

The application provides a flexible mechanical arm system identification method and device based on a Wiener model, and the method comprises the following steps: training a four-layer neural fuzzy network through sample data to obtain a nominal model of the flexible mechanical arm system, wherein the sample data comprises input variables torque and output variables acceleration of the flexible mechanical arm system; constructing a Wiener model of the flexible mechanical arm system, wherein the Wiener model comprises a dynamic linear module and a static nonlinear module; taking a combined signal formed by a separable signal and an actual torque as the input of the Wiener model, taking the output of the separable signal through the nominal model and the actual acceleration as the output of the Wiener model, and identifying the parameters of the dynamic linear module and the static nonlinear module in the Wiener model; and predicting the acceleration of the flexible mechanical arm system through the identified Wiener model. The application can reduce the complexity of Wiener model calculation, improve the parameter identification accuracy, effectively control the flexible mechanical arm system, and improve the working efficiency of the flexible mechanical arm system.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial control, in particular to a flexible robot system identification method based on Wiener model and a flexible robot system identification device based on Wiener model. BACKGROUND

[0002] The flexible robot has the advantages of light structure, low energy consumption, strong adaptability to working environment, flexible operation and high working efficiency, and therefore has great development space in high-tech fields such as medical treatment, aerospace and material science. The control technology of the flexible robot system is mostly based on a model, and the precision of the model identified directly affects the design of the subsequent controller, so the selection of the model and the precision of the parameter identification play a crucial role in the subsequent control.

[0003] In actual industrial production process, the flexible robot system often has inherent complex mechanism problems such as nonlinearity, uncertainty and large time delay. The traditional mathematical mechanism model is difficult to accurately describe the flexible robot system, so it is difficult to effectively control the flexible robot, which leads to the inability to improve the production efficiency.

[0004] In recent years, in the identification modeling research of the flexible robot system, many important achievements have been made, but there are still problems, the main reason is that: in the existing main identification method, there are often product items of parameters, which need to further use parameter decomposition technology to separate mixed parameters, so as to increase the complexity of calculation and reduce the precision of parameter identification. SUMMARY

[0005] The present application provides a flexible robot system identification method and device based on Wiener model, which can reduce the complexity of Wiener model calculation, improve the precision of parameter identification, so as to effectively control the flexible robot system and improve the working efficiency of the flexible robot system.

[0006] The technical scheme adopted by the present application is as follows:

[0007] The application discloses a flexible mechanical arm system identification method based on a Wiener model, and comprises the following steps: training a four-layer neural fuzzy network through sample data to obtain a nominal model of the flexible mechanical arm system, wherein the sample data comprises input variables torque and output variables acceleration of the flexible mechanical arm system; constructing a Wiener model of the flexible mechanical arm system, wherein the Wiener model comprises a dynamic linear module and a static nonlinear module; taking a combined signal formed by a separable signal and actual torque as input of the Wiener model, taking an output of the nominal model and actual acceleration of the separable signal as output of the Wiener model, and identifying parameters of the dynamic linear module and the static nonlinear module in the Wiener model; and predicting the acceleration of the flexible mechanical arm system through the identified Wiener model.

[0008] The nominal model is:

[0009]

[0010] wherein k represents a sampling time, u(k) represents torque of the flexible mechanical arm system, torque of the flexible mechanical arm system is taken as input of the nominal model, y(k) represents acceleration of the flexible mechanical arm system, acceleration of the flexible mechanical arm system is taken as output of the nominal model, w l (l=1,...,L) is a weight of the neural fuzzy network, L is a number of fuzzy rules, and is a Gaussian membership function, c l is a center of the Gaussian membership function, and sigma l is a width of the Gaussian membership function.

[0011] The Wiener model of the flexible mechanical arm system is constructed, and specifically comprises the following steps: a time-delay state space model is adopted to fit the dynamic linear module of the Wiener model; and a polynomial model is adopted to fit the static nonlinear module of the Wiener model.

[0012] The time-delay state space model is:

[0013] x(k+1)=Ax(k)+Bx(k-1)+hu(k)

[0014] v(k)=cx(k)

[0015] wherein torque of the flexible mechanical arm system is taken as input of the time-delay state space model, x(k) is a state variable, A∈R n×n , B∈R n×n , h∈R n×1 and c∈R 1×nare system parameter matrices, and v(k) is the output of the time-delay state space model,

[0016] Based on the characteristics of the unit backward shift operator: x(k-1) = z -1 x(k) and x(k+1) = zx(k), the time-delay state space model is converted into the following input-output relationship:

[0017]

[0018] where a(z) = z -2n det[z 2 I-Az-B] = 1 + a1z -1 +…+a 2n z -2n , b(z) = z -2n+1 cadj[z 2 I-Az-B]h = b1z -1 +…+b 2n-1 z -(2n-1) .

[0019] The polynomial model is:

[0020] y(k) = f(v(k)) = c1v(k) + c2v 2 (k) +…+c r v r (k)

[0021] where c1, c2, …, c r are coefficients of the polynomial model, and r is the model order.

[0022] The expression of the Wiener model is:

[0023]

[0024] y(k) = f(v(k)) + e(k)

[0025] where e(k) represents white noise, f(·) represents a static nonlinear module, and is described by a polynomial model, n a and n b represent the model order.

[0026] Identifying the parameters of the dynamic linear module and the static nonlinear module in the Wiener model, specifically comprising: taking the combined signal composed of separable signal u1(k) and actual torque u2(k) as the input of the Wiener model, taking the output y1(k) of the separable signal through the nominal model and the actual acceleration y2(k) as the output of the Wiener model; based on the input and output data {u1(k), y1(k)} of the separable signal, the correlation analysis method is used to identify the coefficients a i (i = 1, 2,..., n a ), b j (j = 1, 2,..., n b ) of the polynomial model; based on the torque input data and acceleration output data {u2(k), y2(k)} of the flexible manipulator system, the recursive least squares method is used to identify the coefficients c r .

[0027] The separable signal includes a binary signal, a Gaussian signal and a sinusoidal signal.

[0028] The acceleration of the flexible manipulator system is predicted through the identified Wiener model, specifically comprising: inputting the torque of the flexible manipulator system to be predicted into the identified Wiener model to output the predicted acceleration.

[0029] A flexible manipulator system identification device based on a Wiener model, comprising: a training module, the training module is used for training a four-layer neural fuzzy network through sample data to obtain a nominal model of the flexible manipulator system, wherein the sample data includes input variables torque and output variables acceleration of the flexible manipulator system; a construction module, the construction module is used for constructing a Wiener model of the flexible manipulator system, wherein the Wiener model includes a dynamic linear module and a static nonlinear module; an identification module, the identification module is used for taking a combined signal composed of a separable signal and an actual torque as the input of the Wiener model, taking the output of the separable signal through the nominal model and the actual acceleration as the output of the Wiener model, and identifying the parameters of the dynamic linear module and the static nonlinear module in the Wiener model; a prediction module, the prediction module is used for predicting the acceleration of the flexible manipulator system through the identified Wiener model.

[0030] The beneficial effects of the present application are:

[0031] This invention obtains a nominal model of a flexible robotic arm system by training a neural fuzzy network, constructs a Wiener model of the flexible robotic arm system, and identifies the parameters of the dynamic linear module and the static nonlinear module in the Wiener model, thereby predicting the acceleration of the flexible robotic arm system. This reduces the computational complexity of the Wiener model, improves the accuracy of parameter identification, and thus enables effective control of the flexible robotic arm system, thereby improving the working efficiency of the flexible robotic arm system. Attached Figure Description

[0032] Figure 1 This is a flowchart of the flexible robotic arm system identification method based on the Wiener model according to an embodiment of the present invention;

[0033] Figure 2 This is a schematic diagram of a flexible robotic arm system identification method based on the Wiener model according to an embodiment of the present invention;

[0034] Figure 3 This is a comparison graph of the predicted acceleration and the actual acceleration in a specific embodiment of the present invention;

[0035] Figure 4 This is a block diagram of a flexible robotic arm system identification device based on the Wiener model according to an embodiment of the present invention. Detailed Implementation

[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] Figure 1 This is a flowchart of a flexible robotic arm system identification method based on the Wiener model, according to an embodiment of the present invention.

[0038] like Figure 1 As shown, the flexible robotic arm system identification method based on the Wiener model in this embodiment of the invention includes the following steps:

[0039] S1. The four-layer neural fuzzy network is trained using sample data to obtain the nominal model of the flexible robotic arm system. The sample data includes the input variable torque and the output variable acceleration of the flexible robotic arm system.

[0040] In one embodiment of the present invention, the four-layer neural fuzzy network is a four-layer neural fuzzy network composed of a fuzzy system and a radial basis function neural network, specifically divided into an input layer, a Gaussian membership function layer, a fuzzy rule layer, and an output layer. The first layer is the input layer, where neurons directly transmit the input signal to the next layer. The second layer is the Gaussian membership function layer, which receives the signal from the input layer and calculates the membership function of the input signal, i.e., Where L represents the number of fuzzy rules, c l σ is the center of the membership function. l The width of the membership function is denoted by . The third layer is the fuzzy rule layer, where each neuron represents a fuzzy rule. The fourth layer is the output layer, where the output signal is represented by the output signal of a single neuron; that is, the output is . Among them, w l (l=1,…,L) represents the weights of the neural fuzzy model.

[0041] The nominal model obtained by training a four-layer neural fuzzy network is as follows:

[0042]

[0043] Where k represents the sampling time, u(k) represents the torque of the flexible robotic arm system (using the torque of the flexible robotic arm system as the input of the nominal model), y(k) represents the acceleration of the flexible robotic arm system (using the acceleration of the flexible robotic arm system as the output of the nominal model), and w l (l=1,…,L) represents the weights of the neural fuzzy network, and L represents the number of fuzzy rules. and Let c be the Gaussian membership function. l σ is the center of the Gaussian membership function. l is the width of the Gaussian membership function.

[0044] In one embodiment of the present invention, the center c of the Gaussian membership function can be calculated using a clustering algorithm. l and width σ l Specifically, firstly, the input data u(1) of the random signal can be used as the first cluster, and the cluster center c1 = u(1) of the first cluster can be set. Furthermore, the similarity criterion can be used to determine the cluster center. Calculate the similarity between the k-th input data u(k) and each cluster center to determine whether to add a new cluster, and then proceed according to... Adjust the cluster centers, repeating this step until all input data from the random signal are assigned to the appropriate clusters, then... Calculate the width σ of the membership function lFinally, the weights w of the neural fuzzy model can be updated by using an iterative method l (l = 1, …, L). Wherein, N represents the total number of input data of the random signal, e represents an exponential function, N L represents the number belonging to the Lth cluster, λ ∈ [0, 1] represents an adjustable parameter, ρ represents an overlap parameter, and usually 1 ≤ ρ ≤ 2.

[0045] S2, a Wiener model of the flexible robot arm system is constructed, wherein the Wiener model comprises a dynamic linear module and a static nonlinear module.

[0046] Specifically, a time-delay state space model can be used to fit the dynamic linear module of the Wiener model, and a polynomial model can be used to fit the static nonlinear module of the Wiener model.

[0047] In an embodiment of the present application, the time-delay state space model is:

[0048] x(k+1) = Ax(k) + Bx(k-1) + hu(k)

[0049] v(k) = cx(k)

[0050] Wherein, the torque of the flexible robot arm system is taken as the input of the time-delay state space model, x(k) is a state variable, A ∈ R n×n , B ∈ R n×n , h ∈ R n×1 , and c ∈ R 1×n are all system parameter matrices, and v(k) is the output of the time-delay state space model.

[0051] Based on the characteristics of the unit backward shift operator: x(k-1) = z -1 x(k) and x(k+1) = zx(k), the time-delay state space model is converted into the following input-output relationship:

[0052]

[0053] Wherein, L(·) represents a dynamic linear module described by the time-delay state space model, a(z) = z -2n det[z 2 I-Az-B] = 1 + a1z -1 + … + a2nz -2n , b(z) = z -2n+1 cadj[z 2 I-Az-B]h = b1z -1 + … + b 2n-1 z -(2n-1) .

[0054] In an embodiment of the present application, the polynomial model is:

[0055] y(k) = f(v(k)) = c1v(k) + c2v 2 (k) + … + c r v r (k)

[0056] wherein c1, c2, …, c r are coefficients of the polynomial model, and r is the model order.

[0057] In an embodiment of the present application, the Wiener model dynamic linear module fitted by the above time delay state space model and the static nonlinear module of the Wiener model fitted by the polynomial model can obtain the expression of the Wiener model as follows:

[0058]

[0059] y(k) = f(v(k)) + e(k)

[0060] wherein e(k) represents white noise, f(·) represents the static nonlinear module, and is described by the polynomial model, n a and n b represent the model order.

[0061] S3, the combination signal composed of the separable signal and the actual torque is taken as the input of the Wiener model, the output of the separable signal through the nominal model and the actual acceleration are taken as the output of the Wiener model, and the parameters of the dynamic linear module and the static nonlinear module in the Wiener model are identified.

[0062] Specifically, first, the combination signal composed of the separable signal u1(k) and the actual torque u2(k) is taken as the input of the Wiener model, and the output y1(k) of the separable signal through the nominal model and the actual acceleration y2(k) are taken as the output of the Wiener model.

[0063] Secondly, based on the input and output data {u1(k), y1(k)} of the separable signal, the unknown parameters of the dynamic linear module, i.e. the coefficients a i (i = 1, 2, …, n a ), b j (j = 1, 2, …, n b). Wherein, the separable signal includes binary signal, Gaussian signal and sinusoidal signal. In an embodiment of the application, the Gaussian signal can be used as the separable signal, and the Gaussian input signal u1(k) is obtained through the nominal model to get the output y1(k), and the autocorrelation function of the input u1(k) and the dynamic linear module are used to replace the cross-correlation function of the output y1(k) and the input u1(k), that is , so as to realize the separation of the dynamic linear module and the static nonlinear submodule identification. Wherein, b0=E(y1(k)v(k)) / E(v(k)v(k)) is a constant, τ is a time constant, and the cross-correlation function The autocorrelation function is Specifically:

[0064] From It can be obtained:

[0065]

[0066] Let Get:

[0067]

[0068] The correlation analysis method is used to identify the dynamic linear module parameters, and it is assumed that τ=1,2,...,P(P≥n a +n b ), and P=7 in the patent, so:

[0069]

[0070] Wherein,

[0071]

[0072] And It is calculated by And Specifically, based on N1(N1=1000) groups of Gaussian signals with mean value 0 and variance 0.2 2 , under the condition that the order of the system matrix is n=2, n a =4, n b =3, and the correlation analysis method is used to identify the parameters of the dynamic linear module as

[0073] Finally, based on the torque input data and acceleration output data {u2(k), y2(k)} of the flexible manipulator system, the recursive least squares method can be used to identify the coefficients c r of the polynomial model. Specifically, the flexible manipulator system based on the Wiener model is written in the regression form:

[0074]

[0075] in, c = [c1, c2, c3] T

[0076] The parameter vector c of the polynomial model is obtained using the recursive least squares method:

[0077]

[0078]

[0079]

[0080]

[0081]

[0082] This invention is based on N2 (N2 = 1000) datasets {u2(k), y2(k)}, and sets parameter vectors. initial value Obtain the coefficients of the static nonlinear module

[0083] S4 predicts the acceleration of the flexible robotic arm system using the identified Wiener model.

[0084] The identification and prediction steps in steps S3 and S4 are as follows: Figure 2 As shown, firstly, the combined signal consisting of the separable signal u1(k) and the actual torque u2(k) is used as the input to the Wiener model. The output y1(k) of the separable signal after passing through the nominal model and the actual acceleration y2(k) are used as the outputs of the Wiener model. Further, based on the input and output data {u1(k), y1(k)} of the separable signal, correlation analysis is used to identify the unknown parameters of the dynamic linear module. Even further, based on the torque input data and acceleration output data {u2(k), y2(k)} of the flexible robotic arm system, the coefficients c of the polynomial model can be identified using the recursive least squares method. r Finally, the torque of the flexible robotic arm system to be predicted is input into the identified Wiener model to output the predicted acceleration.

[0085] Figure 3 This is a comparison chart of the predicted acceleration and the actual acceleration in a specific embodiment of the present invention. Figure 3 It can be seen that the flexible robotic arm system identification method based on the Wiener model in the embodiments of the present invention has high accuracy in predicting the acceleration of the flexible robotic arm system.

[0086] The Wiener model-based flexible mechanical arm system identification method according to the embodiment of the present application can obtain the nominal model of the flexible mechanical arm system by training the neural fuzzy network, construct the Wiener model of the flexible mechanical arm system, and identify the parameters of the dynamic linear module and the static nonlinear module in the Wiener model, thereby predicting the acceleration of the flexible mechanical arm system. Thus, the complexity of the Wiener model calculation can be reduced, the parameter identification accuracy can be improved, the flexible mechanical arm system can be effectively controlled, and the working efficiency of the flexible mechanical arm system can be improved.

[0087] To implement the Wiener model-based flexible mechanical arm system identification method of the above embodiment, the present application further provides a Wiener model-based flexible mechanical arm system identification device.

[0088] As shown in Figure 4 the Wiener model-based flexible mechanical arm system identification device according to the embodiment of the present application comprises a training module 10, a construction module 20, an identification module 30, and a prediction module 40. The training module 10 is configured to train a four-layer neural fuzzy network by sample data to obtain the nominal model of the flexible mechanical arm system, wherein the sample data comprises the input variable torque and the output variable acceleration of the flexible mechanical arm system. The construction module 20 is configured to construct the Wiener model of the flexible mechanical arm system, wherein the Wiener model comprises a dynamic linear module and a static nonlinear module. The identification module 30 is configured to take the combination signal composed of the separable signal and the actual torque as the input of the Wiener model, take the output of the separable signal through the nominal model and the actual acceleration as the output of the Wiener model, and identify the parameters of the dynamic linear module and the static nonlinear module in the Wiener model. The prediction module 40 is configured to predict the acceleration of the flexible mechanical arm system by the identified Wiener model.

[0089] In one embodiment of the present application, the four-layer neural fuzzy network is a four-layer neural fuzzy network composed of a fuzzy system and a radial basis neural network, which can be specifically divided into an input layer, a Gaussian membership function layer, a fuzzy rule layer, and an output layer. The first layer is the input layer, and the neurons in the input layer directly transmit the input signal to the next layer. The second layer is the Gaussian membership function layer, which is configured to receive the signal input by the input layer and calculate the membership function of the input signal, i.e. wherein L represents the number of fuzzy rules, c l is the center of the membership function, and σ l is the width of the membership function. The third layer is the fuzzy rule layer, and each neuron of the layer represents a fuzzy rule. The fourth layer is the output layer, and the output layer outputs the signal represented by a single neuron, i.e. the output is wherein wl (l = 1,..., L) are weights of the neural fuzzy model.

[0090] The nominal model obtained by training the four-layer neural fuzzy network by the training module 10 is:

[0091]

[0092] where k represents the sampling time, u(k) represents the torque of the flexible robot system, the torque of the flexible robot system is taken as the input of the nominal model, y(k) represents the acceleration of the flexible robot system, the acceleration of the flexible robot system is taken as the output of the nominal model, w l (l = 1,..., L) are weights of the neural fuzzy network, L is the number of fuzzy rules, and is a Gaussian membership function, c l is the center of the Gaussian membership function, and σ l is the width of the Gaussian membership function.

[0093] In an embodiment of the present application, the center c l and the width σ l of the Gaussian membership function can be calculated by using a clustering algorithm. Specifically, first, the input data u(1) of the random signal can be taken as the first cluster, and the cluster center of the first cluster can be set as c1 = u(1). Further, the similarity of the kth input data u(k) with each cluster center can be calculated according to the similarity criterion , so as to determine whether a new cluster is added, and the cluster center is adjusted according to , and the step is repeated until all the input data in the random signal are assigned to the corresponding cluster. Then, the width σ of the membership function can be calculated according to l . Finally, the weights w l (l = 1,..., L) of the neural fuzzy model can be updated by using an iterative method. Wherein, N represents the total number of the input data of the random signal, e represents the exponential function, N L represents the number belonging to the Lth cluster, λ ∈ [0, 1] represents an adjustable parameter, and ρ represents an overlap parameter, usually 1 ≤ ρ ≤ 2.

[0094] In an embodiment of the present application, the constructing module 20 can adopt a time-delay state space model to fit the dynamic linear module of the Wiener model, and adopt a polynomial model to fit the static nonlinear module of the Wiener model.

[0095] In an embodiment of the present application, the time-delay state space model is:

[0096] x(k+1) = Ax(k) + Bx(k-1) + hu(k)

[0097] v(k) = cx(k)

[0098] wherein the torque of the flexible manipulator system is taken as the input of the time-delay state space model, x(k) is a state variable, A ∈ R n×n , B ∈ R n×n , h ∈ R n×1 and c ∈ R 1×n are system parameter matrices, and v(k) is the output of the time-delay state space model.

[0099] Based on the characteristics of the unit backward shift operator: x(k-1) = z -1 x(k) and x(k+1) = zx(k), the time-delay state space model is converted into the following input-output relationship:

[0100]

[0101] wherein L(·) represents a dynamic linear module described by the time-delay state space model, a(z) = z -2n det[z 2 I-Az-B] = 1 + a1z -1 +…+a2nz -2n , b(z) = c -2n+1 adj[z 2 I-Az-B]h = b1z -1 +…+b 2n -1z -(2n-1) .

[0102] In an embodiment of the present application, the polynomial model is:

[0103] y(k) = f(v(k)) = c1v(k) + c2v 2 (k) +…+c r v r (k)

[0104] wherein c1, c2, …, c r are coefficients of the polynomial model, and r is the order of the model.

[0105] In an embodiment of the present application, the expression of the Wiener model can be obtained by fitting the dynamic linear module of the Wiener model through the above time-delay state space model and fitting the static nonlinear module of the Wiener model through the polynomial model.

[0106]

[0107] y(k) = f(v(k)) + e(k)

[0108] Where e(k) represents white noise, and f(·) represents a static nonlinear module, described by a polynomial model. n a and n b Indicates the model order.

[0109] In one embodiment of the present invention, firstly, the identification module 30 takes the combined signal consisting of the separable signal u1(k) and the actual torque u2(k) as the input of the Wiener model, and takes the output y1(k) of the separable signal through the nominal model and the actual acceleration y2(k) as the output of the Wiener model.

[0110] Secondly, based on the input and output data {u1(k), y1(k)} of the separable signals, correlation analysis is used to identify the unknown parameters of the dynamic linear module, i.e., the coefficients a of the polynomial model. i (i = 1, 2, ..., n) a b j (j = 1, 2, ..., n) b The separable signals include binary signals, Gaussian signals, and sinusoidal signals. In one embodiment of the present invention, a Gaussian signal can be used as the separable signal. The Gaussian input signal u1(k) is processed by a nominal model to obtain the output y1(k). The cross-correlation function of the input u1(k) is replaced by the product of the autocorrelation function of the input u1(k), a constant, and the dynamic linear module. This achieves the separation of dynamic linear modules and static nonlinear sub-modules. Here, b0 = E(y1(k)v(k)) / E(v(k)v(k)) is a constant, τ is the time constant, and the cross-correlation function is... Autocorrelation function Specifically:

[0111] Depend on We can obtain:

[0112]

[0113] make get:

[0114]

[0115] The parameters of the dynamic linear module are identified using correlation analysis, where τ = 1, 2, ..., P (P ≥ n). a +n b In the patent, P = 7, then:

[0116]

[0117] in,

[0118]

[0119] and By using and The calculation is obtained. Specifically, the application is based on N1 (N1 = 1000) group mean value 0, variance 0.2 2 Gaussian signal, under the condition of system matrix order n = 2, n a = 4, n b = 3, using correlation analysis method to identify the parameters of dynamic linear module

[0120] Finally, based on the torque input data and acceleration output data {u2(k), y2(k)} of the flexible manipulator system, the recursive least squares method can be used to identify the coefficients c r of the polynomial model. Specifically: write the flexible manipulator system based on Wiener model into regression form:

[0121]

[0122] Among them, c = [c1, c2, c3] T .

[0123] The parameter vector c of the polynomial model is obtained by recursive least squares method:

[0124]

[0125]

[0126]

[0127]

[0128]

[0129] The application is based on N2 (N2 = 1000) group data set {u2(k), y2(k)}, and the initial value of the parameter vector is set to The coefficients of the static nonlinear module are obtained

[0130] In one embodiment of the application, as Figure 2As shown, first, the separable signal and the actual torque constitute a combined signal, which is input to the Wiener model through the identification module 30, and the output y1(k) of the nominal model and the actual acceleration y2(k) are output from the Wiener model, so as to identify the parameters of the dynamic linear module and the static nonlinear module in the Wiener model. Finally, the prediction module 40 predicts the acceleration of the flexible robot system through the identified Wiener model.

[0131] According to the flexible robot system identification device based on the Wiener model, the nominal model of the flexible robot system is obtained by training the neural fuzzy network through the training module, the Wiener model of the flexible robot system is constructed through the construction module, and the parameters of the dynamic linear module and the static nonlinear module in the Wiener model are identified through the identification module, so that the prediction module predicts the acceleration of the flexible robot system. Therefore, the complexity of the Wiener model calculation can be reduced, the parameter identification accuracy can be improved, the flexible robot system can be effectively controlled, and the working efficiency of the flexible robot system can be improved.

[0132] In the description of the present application, the terms "first", "second" are only for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first", "second" can explicitly or implicitly include one or more of the features. The meaning of "plurality" is two or more, unless otherwise specifically limited.

[0133] In the present application, unless otherwise specifically defined and limited, the terms "mounting", "connection", "connection", "fixing" and the like should be understood in a broad sense, for example, it can be fixed connection, or detachable connection, or integrated; it can be mechanical connection, or electrical connection; it can be directly connected, or indirectly connected through intermediate medium; it can be the internal communication of two elements or the interaction relationship between two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0134] In the present application, unless otherwise specifically defined and limited, the first feature "on" or "under" the second feature can be that the first and second features are in direct contact, or the first and second features are in indirect contact through an intermediate medium. Moreover, the first feature "above", "above" and "above" the second feature can be that the first feature is directly above or obliquely above the second feature, or only indicates that the horizontal height of the first feature is higher than that of the second feature. The first feature "below", "below" and "below" the second feature can be that the first feature is directly below or obliquely below the second feature, or only indicates that the horizontal height of the first feature is less than that of the second feature.

[0135] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0136] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of the invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which embodiments of the invention pertain.

[0137] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0138] It should be understood that each part of the present application can be realized by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be realized by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if realized by hardware, and as in another embodiment, any one or a combination of the following technologies known in the art can be used: discrete logic circuit with logic gate circuit for implementing logic functions on data signals, application specific integrated circuit with suitable combination logic gate circuit, programmable gate array (PGA), field programmable gate array (FPGA), etc.

[0139] Those skilled in the art of the present technology can understand that all or part of the steps carried out by the above-mentioned embodiment methods can be completed by a program instructing the relevant hardware, and the program can be stored in a computer readable storage medium, and when the program is executed, it includes one of the steps of the method embodiment or a combination thereof.

[0140] In addition, each functional unit in each embodiment of the present application can be integrated into one processing module, or each unit can exist physically alone, or two or more units can be integrated into one module. The integrated module can be realized in the form of hardware or in the form of a software functional module. The integrated module, if realized in the form of a software functional module and sold or used as an independent product, can also be stored in a computer readable storage medium.

[0141] Although the embodiments of the present application have been shown and described above, it should be understood that the above-mentioned embodiments are exemplary and cannot be understood as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above-mentioned embodiments within the scope of the present application.

Claims

1. A flexible robot arm system identification method based on Wiener model, characterized in that, The method comprises the following steps: Training a four-layer neural fuzzy network through sample data to obtain a nominal model of the flexible manipulator system, wherein the sample data comprises input variable torque and output variable acceleration of the flexible manipulator system; Constructing a Wiener model of the flexible manipulator system, wherein the Wiener model comprises a dynamic linear module and a static nonlinear module; Taking a combined signal composed of a separable signal and actual torque as input of the Wiener model, taking output of the separable signal through the nominal model and actual acceleration as output of the Wiener model, and identifying parameters of the dynamic linear module and the static nonlinear module in the Wiener model; Predicting acceleration of the flexible manipulator system through the identified Wiener model, The nominal model is: wherein, k denotes a sampling time, denotes a moment of the flexible robot arm system, the moment of the flexible robot arm system being an input of the nominal model, denotes an acceleration of the flexible robot arm system, the acceleration of the flexible robot arm system being an output of the nominal model, is a weight of the neuro-fuzzy network, L is a number of fuzzy rules, , and is a Gaussian membership function, is a center of the Gaussian membership function, is a width of the Gaussian membership function, The Wiener model of the flexible manipulator system is constructed, specifically comprising: fitting a dynamic linear module of the Wiener model by using a time-delay state space model; fitting a static nonlinear module of the Wiener model by using a polynomial model, The time-delay state space model is: wherein a torque of the flexible robot arm system is taken as an input of the time-delay state space model, is a state variable, , , and are system parameter matrices, is an output of the time-delay state space model, Based on the properties of the unit-delay operator: and the time-delay state-space model translates into the following input-output relationship: wherein , .

2. The Wiener model-based flexible manipulator system identification method of claim 1, wherein The polynomial model is: wherein, denotes a static nonlinear module, is a coefficient of the polynomial model, r is a model order.

3. The Wiener model based flexible manipulator system identification method of claim 2, wherein, The expression of the Wiener model is: wherein represents white noise, , , n a and n b denotes the model order.

4. The Wiener model based flexible manipulator system identification method of claim 3, wherein, Identifying parameters of the dynamic linear module and the static nonlinear module in the Wiener model specifically comprises: a separable signal and an actual torque as input to the Wiener model, the separable signal being passed through the output of the nominal model and an actual acceleration as output of the Wiener model; Input-output data based on the separable signal , identifying coefficients of the polynomial model using a correlation analysis method 、 ; based on torque input data and acceleration output data of the flexible robot arm system coefficients of the polynomial model are identified using a recursive least squares method .

5. The Wiener model based flexible manipulator system identification method of claim 4, wherein, The separable signal comprises a binary signal, a Gaussian signal and a sinusoidal signal.

6. The Wiener model based flexible manipulator system identification method of claim 5, wherein, Predicting acceleration of the flexible manipulator system through the identified Wiener model specifically comprises: Inputting torque of the flexible manipulator system to be predicted into the identified Wiener model to output predicted acceleration.

7. A Wiener model based flexible manipulator system identification apparatus implementing the method of any one of claims 1-6, characterized by, The method comprises: A training module is configured to train a four-layer neural fuzzy network through sample data to obtain a nominal model of the flexible manipulator system, wherein the sample data comprises input variable torque and output variable acceleration of the flexible manipulator system; A construction module is configured to construct a Wiener model of the flexible manipulator system, wherein the Wiener model comprises a dynamic linear module and a static nonlinear module; An identification module is configured to take a combined signal composed of a separable signal and actual torque as input of the Wiener model, take output of the separable signal through the nominal model and actual acceleration as output of the Wiener model, and identify parameters of the dynamic linear module and the static nonlinear module in the Wiener model; A prediction module is configured to predict acceleration of the flexible manipulator system through the identified Wiener model.

Images