Control method and device of flexible robotic arm and flexible robotic arm
The angle and speed of the flexible robot arm are corrected through the neural network compensation model, which solves the control accuracy problem caused by nonlinear deformation of the flexible element and improves the control accuracy of the flexible robot arm.
Patent Information
- Application Number
- CN202410459064.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-04-16
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2044-04-16
AI Technical Summary
When performing tasks, the nonlinear deformation of the flexible components leads to complex deformation, which affects control accuracy, especially in large deformations and vibrations, making it difficult to directly model the impact of joint flexibility on the control algorithm.
Using the neural network compensation model, the angle and speed correction is performed by obtaining the actual angle and speed information of the flexible robot arm, combining the expected position and speed information, and compensating with the RBF radial basis function and BP backpropagation subnet to eliminate dynamics Coupling effect.
The control accuracy of the flexible robot arm is improved, the impact of joint flexibility on control accuracy is reduced, and the accuracy of movement is achieved.
Smart Images

Figure CN118254177B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of flexible robotic arm control, and in particular to a control method and device for a flexible robotic arm and a flexible robotic arm. Background Art
[0002] The arm or joint of the flexible robot arm is composed of elastic and flexible elements, which has good flexibility. However, when performing target operation tasks, the nonlinear deformation of the flexible elements makes the motion deformation of the flexible robot arm more complicated, especially large deformation and vibration during movement, so the joint flexibility will affect the control accuracy of the robot arm. Summary of the invention
[0003] In order to solve the above problems, the present application provides a control method and device for a flexible robotic arm and a flexible robotic arm.
[0004] According to a first aspect of the present application, a control method for a flexible robotic arm is provided, comprising:
[0005] Acquire the actual angle value and actual angular velocity value of each flexible joint of the flexible robotic arm, as well as the actual position information and actual velocity value of the end of the flexible robotic arm;
[0006] Determining the desired position information and desired velocity value of the end of the flexible mechanical arm;
[0007] The actual position information of the terminal, the actual speed value of the terminal, the expected position information of the terminal and the expected speed value of the terminal are input into a preset neural network compensation model to obtain the expected angle correction value and the expected angular velocity correction value of each flexible joint; wherein the neural network compensation model has learned to obtain the ability to compensate and correct the expected angle value and the expected angular velocity value of each flexible joint based on the actual position information of the terminal, the actual speed value of the terminal, the expected position information of the terminal and the expected speed value of the terminal of the flexible manipulator;
[0008] Determine an angle error value and an angular velocity error value of each of the flexible joints according to the actual angle value, the actual angular velocity value, the expected angle correction value and the expected angular velocity correction value of each of the flexible joints;
[0009] The flexible robotic arm is controlled based on the angle error value and the angular velocity error value.
[0010] In some embodiments of the present application, the neural network compensation model includes an RBF radial basis function subnetwork and a BP back propagation subnetwork.
[0011] As a possible implementation manner, the actual terminal position information, the actual terminal velocity value, the expected terminal position information and the expected terminal velocity value are input into a preset neural network compensation model to obtain the expected angle correction value and the expected angular velocity correction value of each flexible joint, including:
[0012] Inputting the actual position information of the terminal, the actual speed value of the terminal, the expected position information of the terminal and the expected speed value of the terminal into the RBF subnetwork to obtain the flexibility feature information of the flexible manipulator;
[0013] The flexible feature information is input into the BP sub-network to obtain the expected angle correction value and the expected angular velocity correction value of each flexible joint.
[0014] In some embodiments of the present application, the neural network compensation model is pre-trained by the following steps:
[0015] Based on a preset flexible robotic arm simulation model, a first training set is obtained; wherein the first training set includes a terminal actual position information sample, a terminal actual velocity value sample, a terminal expected position information sample, a terminal expected velocity value sample, and an expected angle label value and an expected angular velocity label value of each flexible joint corresponding to the flexible robotic arm simulation model;
[0016] Pre-training the initial neural network compensation model based on the first training set to obtain a pre-trained neural network compensation model;
[0017] Applying the pre-trained neural network compensation model to the control of the flexible robotic arm to obtain control data generated during the control process;
[0018] Determine a second training set according to the control data; wherein the second training set includes a sample of actual position information of the end of the flexible manipulator, a sample of actual velocity value of the end, a sample of expected position information of the end, a sample of expected velocity value of the end, a label of expected angle correction value of each flexible joint, and a label of expected angular velocity correction value;
[0019] The pre-trained neural network compensation model is trained based on the second training set to obtain the neural network compensation model.
[0020] According to a second aspect of the present application, a control device for a flexible robotic arm is provided, comprising:
[0021] A first acquisition module is used to acquire the actual angle value and the actual angular velocity value of each flexible joint of the flexible robotic arm, as well as the actual position information and the actual velocity value of the end of the flexible robotic arm;
[0022] A first determination module is used to determine the desired position information and desired velocity value of the end of the flexible manipulator;
[0023] A second acquisition module is used to input the actual position information of the terminal, the actual speed value of the terminal, the expected position information of the terminal and the expected speed value of the terminal into a preset neural network compensation model to obtain the expected angle correction value and the expected angular velocity correction value of each flexible joint; wherein the neural network compensation model has learned to obtain the ability to compensate and correct the expected angle value and the expected angular velocity value of each flexible joint based on the actual position information of the terminal, the actual speed value of the terminal, the expected position information of the terminal and the expected speed value of the terminal of the flexible manipulator;
[0024] A second determination module is used to determine the angle error value and angular velocity error value of each of the flexible joints according to the actual angle value, the actual angular velocity value, the expected angle correction value and the expected angular velocity correction value of each of the flexible joints;
[0025] A control module is used to control the flexible robotic arm based on the angle error value and the angular velocity error value.
[0026] In some embodiments of the present application, the neural network compensation model includes an RBF radial basis function subnetwork and a BP back propagation subnetwork.
[0027] As a possible implementation manner, the second acquisition module is specifically used to:
[0028] Inputting the actual position information of the terminal, the actual speed value of the terminal, the expected position information of the terminal and the expected speed value of the terminal into the RBF subnetwork to obtain the flexibility feature information of the flexible manipulator;
[0029] The flexible feature information is input into the BP sub-network to obtain the expected angle correction value and the expected angular velocity correction value of each flexible joint.
[0030] In some embodiments of the present application, the device further includes a training module; the training module is used to:
[0031] Based on a preset flexible robotic arm simulation model, a first training set is obtained; wherein the first training set includes a terminal actual position information sample, a terminal actual velocity value sample, a terminal expected position information sample, a terminal expected velocity value sample, and an expected angle label value and an expected angular velocity label value of each flexible joint corresponding to the flexible robotic arm simulation model;
[0032] Pre-training the initial neural network compensation model based on the first training set to obtain a pre-trained neural network compensation model;
[0033] Applying the pre-trained neural network compensation model to the control of the flexible robotic arm to obtain control data generated during the control process;
[0034] Determine a second training set according to the control data; wherein the second training set includes a sample of actual position information of the end of the flexible manipulator, a sample of actual velocity value of the end, a sample of expected position information of the end, a sample of expected velocity value of the end, a label of expected angle correction value of each flexible joint, and a label of expected angular velocity correction value;
[0035] The pre-trained neural network compensation model is trained based on the second training set to obtain the neural network compensation model.
[0036] According to a third aspect of the present application, an electronic device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method described in the first aspect is implemented.
[0037] According to a fourth aspect of the present application, a computer-readable storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, the method described in the first aspect is implemented.
[0038] According to the technical solution of the present application, by obtaining the actual angle value and the actual angular velocity value of each flexible joint of the flexible manipulator, as well as the actual position information and the actual velocity value of the end of the flexible manipulator, and inputting the actual position information, the actual velocity value, the expected position information and the expected velocity value of the end into a preset neural network compensation model, the expected angle correction value and the expected angular velocity correction value of each flexible joint are obtained, and the angle error value and the angular velocity error value of each flexible joint are determined according to the actual angle value, the actual angular velocity value, the expected angle correction value and the expected angular velocity correction value of each flexible joint, and the flexible manipulator is controlled based on the angle error value and the angular velocity error value. This solution compensates and corrects the expected angle value and the expected angular velocity value of each flexible joint of the flexible manipulator through a neural network compensation model to eliminate the influence of joint flexibility on the dynamic coupling of the manipulator, thereby improving the control accuracy of the manipulator.
[0039] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS
[0040] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:
[0041] Figure 1A flow chart of a control method for a flexible robotic arm provided in an embodiment of the present application;
[0042] Figure 2 A flowchart of a training process of a neural network compensation model in an embodiment of the present application;
[0043] Figure 3 A structural block diagram of a control device for a flexible robotic arm provided in an embodiment of the present application;
[0044] Figure 4 A structural block diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0045] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.
[0046] It should be noted that the arm or joint of the flexible manipulator is composed of elastic and flexible elements, and has good flexibility. When the manipulator performs the target operation task, it will inevitably generate instantaneous contact force on the contact surface with the target. When the contact force is transmitted to the flexible element, the flexible element will be elastically deformed, so that the stiffness or damping coefficient of the mechanism will change, thereby buffering the contact force and making the operation process smoother. When performing the target operation task, the nonlinear deformation of the flexible element makes the motion deformation of the flexible manipulator more complicated, especially large deformation and vibration during movement, so the joint flexibility will affect the control accuracy of the manipulator. Because the strong nonlinearity of joint flexibility is difficult to directly model with the control algorithm, it is difficult to quantify the impact of joint flexibility on the control of the manipulator, and it is currently difficult to improve the control accuracy of the flexible manipulator.
[0047] In order to solve the above problems, the present application provides a control method and device for a flexible robotic arm and a flexible robotic arm.
[0048] Figure 1 This is a flow chart of a control method for a flexible robotic arm provided in an embodiment of the present application. It should be noted that the control method for a flexible robotic arm in an embodiment of the present application can be applied to a control device for a flexible robotic arm in an embodiment of the present application, and the device can be configured in a flexible robotic arm. Figure 1 As shown, the method may include the following steps:
[0049] Step 101, obtaining the actual angle value and actual angular velocity value of each flexible joint of the flexible robotic arm, as well as the actual position information and actual velocity value of the end of the flexible robotic arm.
[0050] In some embodiments of the present application, the actual angle value, actual speed value, actual position information of the end and actual speed value of each flexible joint of the flexible robotic arm can be obtained based on the sensor configured by the flexible robotic arm. Among them, the actual position information of the end of the flexible robotic arm can be the actual coordinate information of the end of the flexible robotic arm.
[0051] As an example, the actual angle value, actual speed value, actual position information of the end and actual speed value of the end of each flexible joint of the flexible robotic arm obtained in step 101 can all be real-time values, for example, the actual angle value, actual speed value, actual position information of the end and actual speed value of the end of each flexible joint of the flexible robotic arm at the current moment.
[0052] Step 102, determining the desired position information and desired velocity value of the end of the flexible robotic arm.
[0053] The expected position information of the end of the flexible robotic arm refers to the coordinate information of the end of the flexible robotic arm at the current moment, which is determined based on the expected trajectory information of the end of the flexible robotic arm.
[0054] Step 103, input the actual position information of the terminal, the actual velocity value of the terminal, the expected position information of the terminal and the expected velocity value of the terminal into a preset neural network compensation model to obtain the expected angle correction value and the expected angular velocity correction value of each flexible joint; wherein the neural network compensation model has learned the ability to compensate and correct the expected angle value and the expected angular velocity value of each flexible joint based on the actual position information of the terminal, the actual velocity value of the terminal, the expected position information of the terminal and the expected velocity value of the terminal of the flexible robotic arm.
[0055] Since the strong nonlinearity of joint flexibility is difficult to directly model with the control algorithm, and neural networks have powerful fitting and parallel processing capabilities, the inventors of this application use a neural network algorithm that does not require a precise mathematical model to compensate for the control accuracy, and use the powerful approximation ability of the neural network to compensate for the dynamic model, thereby reducing the impact of the dynamic coupling of the flexible joint and improving the position accuracy.
[0056] That is to say, the neural network compensation model is used to compensate and correct the expected angle values and angular velocity values of each flexible joint of the robot arm according to the actual position information of the end, the actual velocity value of the end, the expected position information of the end and the expected velocity value of the end, so as to achieve the purpose of feedforward compensation correction.
[0057] In some embodiments of the present application, the neural network compensation model may include an RBF radial basis function subnetwork and a BP back propagation subnetwork. The RBF radial basis function subnetwork is used to extract the flexible feature information of the flexible manipulator based on the actual position information of the end, the actual velocity value of the end, the expected position information of the end, and the expected velocity value of the end of each flexible joint. The BP back propagation subnetwork is used to further perform feature extraction based on the flexible feature information, and to compensate and correct the expected angle value and expected angular velocity value of each flexible joint based on the extracted feature information, so as to obtain the expected angle correction value and expected angular velocity correction value of each flexible joint.
[0058] As a possible implementation method, the actual position information of the terminal, the actual velocity value of the terminal, the expected position information of the terminal and the expected velocity value of the terminal are input into a preset neural network compensation model to obtain the expected angle correction value and the expected angular velocity correction value of each flexible joint. The implementation process includes: inputting the actual position information of the terminal, the actual velocity value of the terminal, the expected position information of the terminal and the expected velocity value of the terminal into the RBF subnetwork to obtain the flexible feature information of the flexible robotic arm; inputting the flexible feature information into the BP subnetwork to obtain the expected angle correction value and the expected angular velocity correction value of each flexible joint.
[0059] In some embodiments of the present application, a training sample can be determined based on the historical control data of the flexible robotic arm; wherein the training sample includes an expected position information sample of the end of the flexible robotic arm, an expected velocity value sample of the end, an actual position information sample of the end, an actual velocity value sample of the end, an expected angle value label of each flexible joint, and an expected angular velocity value label of each flexible joint; the expected angle value label and the expected angular velocity value label of each flexible joint can be the expected angle value and the expected angular velocity value with the highest control accuracy of the flexible robotic arm determined by continuously adjusting the expected angle value and the expected angular velocity value of each flexible joint through a large number of control experiments; the training sample is input into the initial neural network compensation model to obtain the expected angle prediction value of each flexible joint and the expected angular velocity prediction value of each flexible joint; the loss value is calculated based on the expected angle prediction value and the expected angular velocity prediction value of each flexible joint, as well as the expected angle value label and the expected angular velocity value label of each flexible joint in the training sample, and the model parameters of the neural network compensation model are continuously updated based on the loss value to obtain the trained neural network compensation model.
[0060] Step 104 , determining the angle error value and angular velocity error value of each flexible joint according to the actual angle value, actual angular velocity value, expected angle correction value and expected angular velocity correction value of each flexible joint.
[0061] In some embodiments of the present application, for each flexible joint, the difference between the actual angle value of the flexible joint and the expected angle correction value can be determined as the angle error value of the flexible joint, and the difference between the actual angular velocity value of the flexible joint and the expected angular velocity correction value can be determined as the angular velocity error value of the flexible joint.
[0062] Step 105: Control the flexible robotic arm based on the angle error value and the angular velocity error value.
[0063] That is to say, when controlling the flexible robotic arm, the embodiment of the present application does not determine the angle error value and the angular velocity error value based on fixed expected angle value and expected angular velocity value, but determines the angle error value and the angular velocity error value based on the corrected expected angle correction value and the expected angular velocity correction value, which is equivalent to a feedforward compensation correction of the control process to improve the control accuracy of the flexible robotic arm.
[0064] In some embodiments of the present application, the process of controlling the flexible robotic arm based on the angle error value and the angular velocity error value can be consistent with the control process in the related technology, for example, it can be a PID control method, or it can be controlled by a sliding membrane controller based on the flexible robotic arm, etc.
[0065] According to the control method of the flexible manipulator of the embodiment of the present application, by obtaining the actual angle value and the actual angular velocity value of each flexible joint of the flexible manipulator, as well as the actual position information and the actual velocity value of the end of the flexible manipulator, and inputting the actual position information, the actual velocity value, the expected position information and the expected velocity value of the end into a preset neural network compensation model, the expected angle correction value and the expected angular velocity correction value of each flexible joint are obtained, and the angle error value and the angular velocity error value of each flexible joint are determined according to the actual angle value, the actual angular velocity value, the expected angle correction value and the expected angular velocity correction value of each flexible joint, and the flexible manipulator is controlled based on the angle error value and the angular velocity error value. This scheme compensates and corrects the expected angle value and the expected angular velocity value of each flexible joint of the flexible manipulator through the neural network compensation model to eliminate the influence of the joint flexibility on the dynamic coupling of the manipulator, thereby improving the control accuracy of the manipulator.
[0066] Next, the training process of the neural network compensation model will be introduced in detail.
[0067] Figure 2 This is a flowchart of a training method for a neural network compensation model in an embodiment of the present application.
[0068] like Figure 2 As shown, the training method of the neural network compensation model may include the following steps:
[0069] Step 201, based on a preset flexible robotic arm simulation model, obtain a first training set; wherein the first training set includes actual terminal position information samples, actual terminal velocity value samples, expected terminal position information samples, expected terminal velocity value samples, expected angle label values and expected angular velocity label values of each flexible joint corresponding to the flexible robotic arm simulation model.
[0070] Among them, the flexible robotic arm simulation model can be a flexible robotic arm simulation model in the relevant technology, or it can be a simulation model built by a technician in this field based on the technical solution of the present application. The number of flexible robotic arm simulation models can be multiple. It should be noted that the flexible robotic arm simulation model can be based on the inverse dynamics model, and according to the actual terminal position information samples, actual terminal velocity value samples, expected terminal position information samples, and expected terminal velocity value samples corresponding to the flexible robotic arm simulation model, the expected angle value label value and the expected angular velocity label value of each flexible joint are determined. According to the actual terminal position information samples, actual terminal velocity value samples, expected terminal position information samples, and expected terminal velocity value samples corresponding to the flexible robotic arm simulation model, it can be the real data in the flexible robotic arm control process corresponding to the flexible robotic arm simulation model.
[0071] Step 202: pre-train the initial neural network compensation model based on the first training set to obtain a pre-trained neural network compensation model.
[0072] Step 203: Apply the pre-trained neural network compensation model to the control of the flexible robotic arm to obtain control data generated during the control process.
[0073] That is to say, the pre-trained neural network compensation model is applied to the actual control of the flexible manipulator, and the control data generated within a period of time is obtained. The control data includes the actual position information of the end of the flexible manipulator at each moment, the actual velocity value of the end, the expected position information of the end, the expected velocity value of the end, the expected angle correction value of each flexible joint, the expected angular velocity correction value, the actual angle value and the actual angular velocity value, etc.
[0074] Step 204, determine a second training set based on the control data; wherein the second training set includes samples of actual position information of the end of the flexible robotic arm, samples of actual velocity value of the end, samples of expected position information of the end, samples of expected velocity value of the end, expected angle correction value labels of each flexible joint, and expected angular velocity correction value labels.
[0075] It can be understood that the pre-trained neural network compensation model has learned the ability to supplement and correct the expected angle values and expected angular velocity values of each joint of the flexible robotic arm, but it is not applicable to the flexible robotic arm of the embodiment of the present application. Therefore, it is necessary to continue to train the pre-trained neural network compensation model based on the training samples determined by the flexible robotic arm of the embodiment of the present application to obtain a neural network compensation model that meets the joint flexibility of the flexible robotic arm of the embodiment of the present application.
[0076] In some embodiments of the present application, the control accuracy corresponding to each group of control data can be determined, and the first control data whose control accuracy meets the preset requirements, the actual position information of the terminal at each moment in the first control data is determined as the actual position information sample of the terminal in the second training set, the actual speed value of the terminal at each moment in the first control data is determined as the actual speed value sample of the terminal in the second training set, the expected position information of the terminal at each moment in the first control data is determined as the expected position information sample of the terminal in the second training set, the expected speed value of the terminal at each moment in the first control data is determined as the expected speed value sample of the terminal in the second training set, the expected angle correction value of each flexible joint output by the pre-trained neural network compensation model is determined as the expected angle label value of each flexible joint in the second training set, and the expected angular velocity correction value of each flexible joint output by the pre-trained neural network compensation model is determined as the expected angular velocity label value of each flexible joint in the second training set.
[0077] Step 205: Train the pre-trained neural network compensation model based on the second training set to obtain a neural network compensation model.
[0078] Next, the process of training the pre-trained application network compensation model based on the second training set will be introduced by taking the neural network compensation model as an RBP-BP combination neural network as an example, where the RBP-BP combination neural network includes an RBP sub-network and a BP sub-network. The process of training the pre-trained neural network compensation model based on the second training set to obtain the neural network compensation model includes:
[0079] (1) Parameter initialization
[0080] Set the maximum number of iterations of the neural network, the target loss value, and the weights of the RBP-BP combination neural network.
[0081] (2) Calculation process of RBF subnetwork
[0082] The training samples in the second training set are input into the RBF subnetwork in the pre-trained neural network compensation model, and the kernel function of the hidden layer node of the RBF subnetwork is designed as the Gaussian function shown in formula (1). The Gaussian function is the transfer function expression of the hidden layer node of the RBF subnetwork as follows:
[0083]
[0084] In the formula, j is the number of nodes in the hidden layer, u j (x) is the output of the jth hidden layer node, x is the input sample data, c j is the center of the Gaussian kernel function at the jth node in the neural network, is the scale function of the Gaussian kernel function at the jth node. The Gaussian kernel function is radially symmetric. The closer the input is to the node center, the larger the output is. The output of the RBF subnetwork is a linear combination of its hidden layer nodes, that is, the following formula (2)
[0085]
[0086] Where N1 is the number of nodes in the hidden layer of the RBF subnetwork, w 1,j is the weight vector of hidden layer j in the RBF sub-network, and y is the output of the RBF sub-network.
[0087] (3) Calculation process of BP sub-network
[0088] The transfer function of the hidden layer nodes of the BP sub-network in the RBF-BP combination network is set to Sigmoid type, and the expression is
[0089]
[0090] The output is:
[0091]
[0092] Where Y(m) is the output of the mth node in the output layer, i.e., the expected angle correction value and expected angular velocity correction value of the mth flexible joint; w 2,m (k) is the weight vector from the kth hidden layer node to the mth node of the output layer of the BR sub-network in the RBF-BP combination network, and N2 is the number of nodes in the BR sub-network.
[0093] (4) Calculate the deviation vector
[0094] According to the network output and expected value, the RBF-BP network deviation vector can be calculated as follows (5):
[0095] ERR m =Y(m)(1-Y(m))(T m -Y(m)) (5)
[0096] Among them, ERR m is the deviation vector of the mth flexible joint; T mis the expected angle correction value label and expected angular velocity correction value label of the mth flexible joint; Y(m) is the output of the mth node of the output layer, that is, the expected angle correction value and expected angular velocity correction value of the mth flexible joint;
[0097] (5) Perform back propagation based on the deviation vector, update the weights in the RBF subnetwork and the BP subnetwork according to formula (6), and determine whether the training is completed based on the number of iterations and the deviation vector. If the training is not completed, return to step (2) to continue training.
[0098] Updated w i,m =w i,m +α·ERR m ·Y(m) (6)
[0099] Among them, w i,m is the weight of the mth node in the ith subnetwork, when i=1, the ith subnetwork is the RBF subnetwork, when i=2, the ith subnetwork is the BP subnetwork; α is the update step size.
[0100] In order to implement the above embodiments, the present application provides a control device for a flexible robotic arm.
[0101] Figure 3 A structural block diagram of a control device for a flexible robotic arm provided in an embodiment of the present application.
[0102] like Figure 3 As shown, the device comprises:
[0103] The first acquisition module 301 is used to acquire the actual angle value and the actual angular velocity value of each flexible joint of the flexible robotic arm, as well as the actual position information and the actual velocity value of the end of the flexible robotic arm;
[0104] The first determination module 302 is used to determine the desired position information and desired velocity value of the end of the flexible manipulator;
[0105] The second acquisition module 303 is used to input the actual position information of the terminal, the actual speed value of the terminal, the expected position information of the terminal and the expected speed value of the terminal into a preset neural network compensation model to obtain the expected angle correction value and the expected angular velocity correction value of each flexible joint; wherein the neural network compensation model has learned to obtain the ability to compensate and correct the expected angle value and the expected angular velocity value of each flexible joint based on the actual position information of the terminal, the actual speed value of the terminal, the expected position information of the terminal and the expected speed value of the terminal of the flexible manipulator;
[0106] A second determination module 304 is used to determine an angle error value and an angular velocity error value of each flexible joint according to the actual angle value, the actual angular velocity value, the expected angle correction value and the expected angular velocity correction value of each flexible joint;
[0107] The control module 305 is used to control the flexible robotic arm based on the angle error value and the angular velocity error value.
[0108] In some embodiments of the present application, the neural network compensation model includes an RBF radial basis function subnetwork and a BP back-propagation subnetwork.
[0109] As a possible implementation manner, the second acquisition module 303 is specifically used for:
[0110] The actual position information of the end, the actual speed value of the end, the expected position information of the end and the expected speed value of the end are input into the RBF sub-network to obtain the flexible feature information of the flexible manipulator;
[0111] The flexible feature information is input into the BP sub-network to obtain the expected angle correction value and expected angular velocity correction value of each flexible joint.
[0112] In some embodiments of the present application, the apparatus further includes a training module 306; the training module 306 is used to:
[0113] Based on a preset flexible robotic arm simulation model, a first training set is obtained; wherein the first training set includes a terminal actual position information sample, a terminal actual velocity value sample, a terminal expected position information sample, a terminal expected velocity value sample, an expected angle label value and an expected angular velocity label value of each flexible joint corresponding to the flexible robotic arm simulation model;
[0114] Pre-training the initial neural network compensation model based on the first training set to obtain a pre-trained neural network compensation model;
[0115] Apply the pre-trained neural network compensation model to the control of the flexible robotic arm and obtain the control data generated by the control process;
[0116] Determine a second training set according to the control data; wherein the second training set includes a sample of actual position information of the end of the flexible manipulator, a sample of actual velocity value of the end, a sample of expected position information of the end, a sample of expected velocity value of the end, a label of expected angle correction value of each flexible joint, and a label of expected angular velocity correction value;
[0117] The pre-trained neural network compensation model is trained based on the second training set to obtain a neural network compensation model.
[0118] According to the control device of the flexible manipulator of the embodiment of the present application, by obtaining the actual angle value and the actual angular velocity value of each flexible joint of the flexible manipulator, as well as the actual position information and the actual velocity value of the end of the flexible manipulator, and inputting the actual position information, the actual velocity value, the expected position information and the expected velocity value of the end into a preset neural network compensation model, the expected angle correction value and the expected angular velocity correction value of each flexible joint are obtained, and the angle error value and the angular velocity error value of each flexible joint are determined according to the actual angle value, the actual angular velocity value, the expected angle correction value and the expected angular velocity correction value of each flexible joint, and the flexible manipulator is controlled based on the angle error value and the angular velocity error value. This scheme compensates and corrects the expected angle value and the expected angular velocity value of each flexible joint of the flexible manipulator through the neural network compensation model to eliminate the influence of the joint flexibility on the dynamic coupling of the manipulator, thereby improving the control accuracy of the manipulator.
[0119] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.
[0120] Figure 4 A block diagram of a flexible robotic arm provided in an embodiment of the present application. The flexible robotic arm can implement the flexible robotic arm control method of the above embodiment. The components, their connections and relationships, and their functions shown herein are merely examples and are not intended to limit the implementation of the present application described and / or required herein.
[0121] like Figure 4 As shown, the flexible robotic arm includes: one or more processors 401, a memory 402, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. The various components are connected to each other using different buses and can be installed on a common mainboard or installed in other ways as needed. The processor can process instructions executed in the flexible robotic arm, including instructions stored in or on the memory to display graphical information of the GUI on an external input / output device (such as a display device coupled to the interface). In other embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple flexible robotic arms can be connected, and each device provides part of the necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 4 A processor 401 is taken as an example.
[0122] The memory 402 is the non-transitory computer-readable storage medium provided in the present application. The memory stores instructions executable by at least one processor so that the at least one processor executes the control method of the flexible robotic arm provided in the present application. The non-transitory computer-readable storage medium of the present application stores computer instructions, which are used to cause a computer to execute the control method of the flexible robotic arm provided in the present application. The computer program product of the present application includes a computer program, which, when executed by the processor 401, implements the control method of the flexible robotic arm proposed in the present application.
[0123] The memory 402 is a non-transient computer-readable storage medium that can be used to store non-transient software programs, non-transient computer executable programs and modules, such as program instructions / modules corresponding to the control method in the embodiment of the present application. The processor 401 executes various functional applications and data processing of the server by running the non-transient software programs, instructions and modules stored in the memory 402, that is, the control method of the flexible robotic arm in the above method embodiment is realized.
[0124] The memory 402 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the flexible robotic arm for speech recognition processing, etc. In addition, the memory 402 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory 402 may optionally include a memory remotely arranged relative to the processor 401, and these remote memories may be connected to the flexible robotic arm of the data processing method via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.
[0125] The flexible robotic arm may further include: an input device 403 and an output device 404. The processor 401, the memory 402, the input device 403 and the output device 404 may be connected via a bus or other means. Figure 4 The example of connecting through bus is taken in the following.
[0126] The input device 403 can receive input digital or character information, as well as user settings of the flexible mechanical arm and key signal input related to function control, such as a touch screen, a keypad, a mouse, a track pad, a touch pad, an indicator rod, one or more mouse buttons, a trackball, a joystick and other input devices. The output device 404 may include a display device, an auxiliary lighting device (e.g., an LED) and a tactile feedback device (e.g., a vibration motor), etc. The display device may include, but is not limited to, a liquid crystal display (LCD), a light emitting diode (LED) display and a plasma display. In some embodiments, the display device may be a touch screen.
[0127] In order to implement the above embodiments, the present disclosure also proposes a storage medium.
[0128] When the instructions in the storage medium are executed by the processor of the flexible robotic arm, the flexible robotic arm can execute the control method of the flexible robotic arm as described above.
[0129] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations 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 one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.
[0130] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.
[0131] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.
[0132] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute the instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.
[0133] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.
[0134] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.
[0135] In addition, each functional unit in each embodiment of the present application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into one module. The above-mentioned integrated module can be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. The above-mentioned storage medium can be a read-only memory, a disk or an optical disk, etc.
[0136] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in the field can change, modify, replace and modify the above embodiments within the scope of the present application.
Claims
1. A control method for a flexible robotic arm, characterized in that: include: Acquire the actual angle value and actual angular velocity value of each flexible joint of the flexible robotic arm, as well as the actual position information and actual velocity value of the end of the flexible robotic arm; Determining the desired position information and desired velocity value of the end of the flexible mechanical arm; The actual position information of the terminal, the actual speed value of the terminal, the expected position information of the terminal and the expected speed value of the terminal are input into a preset neural network compensation model to obtain the expected angle correction value and the expected angular velocity correction value of each flexible joint; wherein the neural network compensation model has learned to obtain the ability to compensate and correct the expected angle value and the expected angular velocity value of each flexible joint based on the actual position information of the terminal, the actual speed value of the terminal, the expected position information of the terminal and the expected speed value of the terminal of the flexible manipulator; Determine an angle error value and an angular velocity error value of each of the flexible joints according to the actual angle value, the actual angular velocity value, the expected angle correction value and the expected angular velocity correction value of each of the flexible joints; Controlling the flexible robotic arm based on the angle error value and the angular velocity error value; The neural network compensation model is pre-trained through the following steps: Based on a preset flexible robotic arm simulation model, a first training set is obtained; wherein the first training set includes a terminal actual position information sample, a terminal actual velocity value sample, a terminal expected position information sample, a terminal expected velocity value sample, and an expected angle label value and an expected angular velocity label value of each flexible joint corresponding to the flexible robotic arm simulation model; Pre-training the initial neural network compensation model based on the first training set to obtain a pre-trained neural network compensation model; Applying the pre-trained neural network compensation model to the control of the flexible robotic arm to obtain control data generated during the control process; Determine a second training set according to the control data; wherein the second training set includes a sample of actual position information of the end of the flexible manipulator, a sample of actual velocity value of the end, a sample of expected position information of the end, a sample of expected velocity value of the end, a label of expected angle correction value of each flexible joint, and a label of expected angular velocity correction value; The pre-trained neural network compensation model is trained based on the second training set to obtain the neural network compensation model.
2. The method according to claim 1, characterized in that The neural network compensation model includes a RBF subnetwork and a BP subnetwork.
3. The method according to claim 2, characterized in that The step of inputting the actual terminal position information, the actual terminal speed value, the expected terminal position information and the expected terminal speed value into a preset neural network compensation model to obtain the expected angle correction value and the expected angular speed correction value of each flexible joint includes: Inputting the actual position information of the terminal, the actual speed value of the terminal, the expected position information of the terminal and the expected speed value of the terminal into the RBF subnetwork to obtain the flexibility feature information of the flexible manipulator; The flexible feature information is input into the BP sub-network to obtain the expected angle correction value and the expected angular velocity correction value of each flexible joint.
4. A control device for a flexible robotic arm, characterized in that: include: A first acquisition module is used to acquire the actual angle value and the actual angular velocity value of each flexible joint of the flexible robotic arm, as well as the actual position information and the actual velocity value of the end of the flexible robotic arm; A first determination module is used to determine the desired position information and desired velocity value of the end of the flexible manipulator; A second acquisition module is used to input the actual position information of the terminal, the actual speed value of the terminal, the expected position information of the terminal and the expected speed value of the terminal into a preset neural network compensation model to obtain the expected angle correction value and the expected angular velocity correction value of each flexible joint; wherein the neural network compensation model has learned to obtain the ability to compensate and correct the expected angle value and the expected angular velocity value of each flexible joint based on the actual position information of the terminal, the actual speed value of the terminal, the expected position information of the terminal and the expected speed value of the terminal of the flexible manipulator; A second determination module is used to determine the angle error value and angular velocity error value of each of the flexible joints according to the actual angle value, the actual angular velocity value, the expected angle correction value and the expected angular velocity correction value of each of the flexible joints; A control module, configured to control the flexible robotic arm based on the angle error value and the angular velocity error value; The device further comprises a training module; the training module is used for: Based on a preset flexible robotic arm simulation model, a first training set is obtained; wherein the first training set includes a terminal actual position information sample, a terminal actual velocity value sample, a terminal expected position information sample, a terminal expected velocity value sample, and an expected angle label value and an expected angular velocity label value of each flexible joint corresponding to the flexible robotic arm simulation model; Pre-training the initial neural network compensation model based on the first training set to obtain a pre-trained neural network compensation model; Applying the pre-trained neural network compensation model to the control of the flexible robotic arm to obtain control data generated during the control process; Determine a second training set according to the control data; wherein the second training set includes a sample of actual position information of the end of the flexible manipulator, a sample of actual velocity value of the end, a sample of expected position information of the end, a sample of expected velocity value of the end, a label of expected angle correction value of each flexible joint, and a label of expected angular velocity correction value; The pre-trained neural network compensation model is trained based on the second training set to obtain the neural network compensation model.
5. The device according to claim 4, characterized in that The neural network compensation model includes a RBF subnetwork and a BP subnetwork.
6. The device according to claim 5, characterized in that The second acquisition module is specifically used for: Inputting the actual position information of the terminal, the actual speed value of the terminal, the expected position information of the terminal and the expected speed value of the terminal into the RBF subnetwork to obtain the flexibility feature information of the flexible manipulator; The flexible feature information is input into the BP sub-network to obtain the expected angle correction value and the expected angular velocity correction value of each flexible joint.
7. A flexible robotic arm, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the method according to any one of claims 1 to 3 when executing the program.
8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 3 is implemented.
