A robot and its control method, device and medium

Through the series-parallel robotic arm and full-body motion control algorithm, the problems of unstable movement and insufficient load capacity of the robotic arm in the prior art are solved, and higher stability and load capacity are achieved.

CN116945124BActive Publication Date: 2024-11-22SHENZHEN PUDU TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210415958.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-04-20
Publication Date
2024-11-22
Estimated Expiration
2042-04-20

AI Technical Summary

Technical Problem

The existing leg foot and wheeled mobile robotic arms have unstable movement and small load capacity due to the chain connection method, which lacks effective solutions.

Method used

The series-parallel robot arm is used to collect data through sensors and perform equivalent series fitting. The movement of the base and robot arm is controlled based on the whole-body motion control algorithm to realize the desired overall configuration of the robot when executing task instructions.

Benefits of technology

It improves the robot's motion stability and load capacity, significantly improving the overall performance of the robot.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116945124B_ABST
    Figure CN116945124B_ABST
Patent Text Reader

Abstract

The present application discloses a robot and its control method, equipment and medium, belonging to the field of control engineering technology. The robot includes a base, a serial-parallel type mechanical arm, a sensor, a memory and a processor. The processor is used to implement the following steps when executing the computer-readable instructions stored in the memory: when receiving a task instruction, the data information of the base and the mechanical arm collected by the sensor is obtained, and the current state of the base and the equivalent serial configuration state of the mechanical arm are obtained; then, the current overall configuration of the robot is obtained, and the expected overall configuration of the robot corresponding to the task instruction is obtained according to the current overall configuration of the robot and the whole body motion control algorithm; the base and the mechanical arm are controlled to move so that the robot is in the expected overall configuration. Because the connection structure of the mechanical arm in the robot is a serial-parallel type, the robot can not only improve the stability of the robot during movement, but also significantly improve the load capacity of the robot.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of control engineering, and in particular to a robot, a control method, a device and a medium for the robot. Background Art

[0002] With the continuous development of social economy, the application of robots is becoming more and more extensive. Among them, the composite robot refers to a new type of robot that integrates the functions of hands and feet. Now people usually combine the robot arm with a mobile platform (legged robot or wheeled robot) to meet the relevant service needs of the catering industry. However, whether it is a legged or wheeled mobile robot arm, it adopts a chain connection method, which makes the legged and wheeled mobile robot arms have the defects of unstable movement and small load capacity. At present, there is no more effective solution to this technical problem. Summary of the invention

[0003] In view of this, the purpose of the present invention is to provide a robot and a control method, device and medium for the robot, so as to improve the robot's motion stability and the robot's load capacity. The specific scheme is as follows:

[0004] A robot, comprising a base, a mechanical arm, a sensor, a memory and a processor, wherein the mechanical arm is connected to the base, the mechanical arm is a serial-parallel type mechanical arm, the memory stores computer-readable instructions that can be run on the processor, and the processor is used to implement the following steps when executing the computer-readable instructions:

[0005] When a task instruction is received, data information of the base and the robotic arm collected by the sensor is acquired, and current states of the base and the robotic arm are obtained based on the data information;

[0006] Performing equivalent series fitting on the current state of the robotic arm to obtain an equivalent series configuration state of the robotic arm;

[0007] Based on the current state of the base and the equivalent serial configuration state of the robot arm, obtain the current overall configuration of the robot;

[0008] Based on the current overall configuration and the whole body motion control algorithm, obtaining the expected overall configuration of the robot when responding to the task instruction;

[0009] The base and the robotic arm are controlled to move so that the robot is in the desired overall configuration.

[0010] Preferably, the processor implements the step of obtaining the desired overall configuration of the robot when executing the task instruction based on the current overall configuration and the whole body motion control algorithm, comprising:

[0011] Obtaining a Jacobian matrix of the robot according to the current overall configuration;

[0012] The whole-body motion control algorithm formula is obtained according to the task instruction and the Jacobian matrix, and the expected overall configuration of the robot when executing the task instruction is obtained by using the whole-body motion control algorithm formula.

[0013] Preferably, the processor implements the step of acquiring the whole body motion control algorithm formula according to the task instruction and the Jacobian matrix, including:

[0014] When the task instruction is a multi-task instruction, performing task priority processing on multiple tasks in the multi-task instruction to obtain a main task in the multi-task instruction;

[0015] The Jacobian matrix is ​​used to project all tasks except the main task in the multi-task instruction into the null space of the main task to obtain the whole-body motion control algorithm formula.

[0016] Preferably, the processor implements the step of using the whole body motion control algorithm formula to obtain the desired overall configuration of the robot when executing the task instruction, including:

[0017] Based on the task instruction, obtaining the expected output position of the robot;

[0018] Inputting the desired output position into an active disturbance rejection control algorithm to obtain a target compensation amount;

[0019] Obtaining a target speed of the robot when executing a target task in the multi-task instruction using the target compensation amount, the Jacobian matrix and the expected output position;

[0020] The desired overall configuration of the robot when executing the multi-task instruction is obtained according to the whole-body motion control algorithm formula and the target speed.

[0021] Preferably, the robotic arm includes some parallel joints, and the processor implements the step of performing equivalent series fitting on the current state of the robotic arm, including:

[0022] Acquire a first joint angle of the parallel partial joint in a parallel configuration;

[0023] The first joint angle is processed using a preset forward kinematics analysis model to obtain a second joint angle of the parallel partial joint in a series configuration; wherein the preset forward kinematics analysis model is a deep learning model trained by preset training samples, and the preset training samples are a set constructed according to an inverse kinematics analysis process.

[0024] Preferably, the processor implements the step of creating a preset forward kinematics analysis model, comprising:

[0025] Obtaining the range of motion of the parallel joints in the series configuration;

[0026] Selecting a preset number of series joint angles in the motion range to obtain a series joint angle set;

[0027] According to the inverse kinematics analysis process, a parallel joint angle corresponding to each serial joint angle is obtained to obtain a parallel joint angle set;

[0028] Constructing the preset training samples according to the series joint angle set and the parallel joint angle set; wherein each training sample in the preset training samples includes a set of series joint angles and corresponding parallel joint angles;

[0029] The deep learning model in the initial state is trained using the preset training samples to obtain the preset forward kinematics analysis model.

[0030] Preferably, the processor implements the step of training the deep learning model in the initial state using the preset training samples, including:

[0031] Acquire a generative adversarial network model; wherein the generative adversarial network model includes a generator and a discriminator;

[0032] Inputting all parallel joint angles in the preset training sample into the generator to obtain a generated result;

[0033] The generated result and all the serial joint angles in the preset training sample are input into the discriminator for training.

[0034] Correspondingly, the present invention also discloses a control method for a robot, the robot comprising a base, a mechanical arm, and a sensor, the mechanical arm being connected to the base, the mechanical arm being a serial-parallel type mechanical arm, comprising:

[0035] When a task instruction is received, data information of the base and the robotic arm collected by the sensor is acquired, and current states of the base and the robotic arm are obtained based on the data information;

[0036] Performing equivalent series fitting on the current state of the robotic arm to obtain an equivalent series configuration state of the robotic arm;

[0037] Based on the current state of the base and the equivalent serial configuration state of the robot arm, obtain the current overall configuration of the robot;

[0038] Based on the current overall configuration and the whole body motion control algorithm, obtaining the expected overall configuration of the robot when responding to the task instruction;

[0039] The base and the robotic arm are controlled to move so that the robot is in the desired overall configuration.

[0040] Correspondingly, the present invention also discloses a robot control device, comprising:

[0041] Memory for storing computer programs;

[0042] A processor is used to implement the steps of a robot control method disclosed above when executing the computer program.

[0043] Correspondingly, the present invention also discloses a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of a robot control method disclosed above are implemented.

[0044] It can be seen that in the present invention, when the robot processor receives the task instruction, it first collects the data information of the base and the manipulator through the sensor, and obtains the current state of the base and the manipulator based on the data information of the base and the manipulator; then, the current state of the manipulator is equivalently fitted in series to obtain the equivalent series configuration state of the manipulator, and based on the current state of the base and the equivalent series configuration state of the manipulator, the current overall configuration of the robot is obtained; then, based on the current overall configuration of the robot and the whole body motion control algorithm, the expected overall configuration of the robot when executing the task instruction is obtained, and the base and manipulator of the robot are controlled to move so that the robot is in the expected overall configuration. Obviously, compared with the prior art, because the connection structure of the manipulator in the robot is a series-parallel type, this method can not only improve the stability of the robot during the movement, but also significantly improve the load capacity of the robot. Correspondingly, a control method, device and medium of a robot provided by the present invention also have the above-mentioned beneficial effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0046] Figure 1 A flowchart of operation steps performed by a processor in a robot provided by an embodiment of the present invention;

[0047] Figure 2 A schematic diagram of a robot provided by an embodiment of the present invention;

[0048] Figure 3 A schematic diagram for obtaining the current overall configuration of the robot;

[0049] Figure 4 A control architecture block diagram of a robot provided by an embodiment of the present invention when executing multi-task instructions;

[0050] Figure 5 This is a schematic diagram of using the active disturbance rejection control algorithm to obtain the target compensation amount;

[0051] Figure 6 Schematic diagram of the adversarial generation network model;

[0052] Figure 7 A structural diagram of a robot control device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] An embodiment of the present invention provides a robot, which includes a base, a mechanical arm, a sensor, a memory and a processor. The mechanical arm is connected to the base, and the mechanical arm is a serial-parallel type mechanical arm. The memory stores computer-readable instructions that can be run on the processor. When the processor is used to execute the computer-readable instructions, the following is realized: Figure 1 See the process steps shown. Figure 1 , Figure 1 A flowchart of operation steps performed by a processor in a robot provided by an embodiment of the present invention, the operation flow includes:

[0055] Step S11: when a task instruction is received, data information of the base and the robotic arm collected by the sensor is obtained, and the current status of the base and the robotic arm are obtained based on the data information;

[0056] Step S12: performing equivalent series fitting on the current state of the robot arm to obtain an equivalent series configuration state of the robot arm;

[0057] Step S13: based on the current state of the base and the equivalent series configuration state of the robot arm, obtain the current overall configuration of the robot;

[0058] Step S14: based on the current overall configuration and the whole body motion control algorithm, obtaining the expected overall configuration of the robot when it corresponds to the task instruction;

[0059] Step S15: Control the base and the robot arm to move so that the robot is in a desired overall configuration.

[0060] In this embodiment, a robot is provided, which can not only improve the stability of the robot during movement, but also significantly improve the load capacity of the robot. The robot includes a base, a mechanical arm, a sensor, a memory, and a processor. The base is a movable structure, and can also be set to a non-movable structure as needed. The mechanical arm of the robot is connected to the base, and the mechanical arm of the robot is a serial-parallel type mechanical arm. It can be imagined that because the connection structure of the mechanical arm in the robot is a serial-parallel type, compared with the serial type mechanical arm, the serial-parallel type mechanical arm can not only improve the stability of the robot during movement, but also significantly improve the load capacity of the robot. Please refer to Figure 2 , Figure 2 The schematic diagram of a robot provided by an embodiment of the present invention includes a base 100 and a mechanical arm 200, wherein the mechanical arm 200 includes a connected parallel joint 201 and a serial joint 202. The hardware system of the robot can be simply described as a differentially driven base equipped with a serial-parallel mechanical arm. In general, the mechanical arm of the robot has 6 degrees of freedom.

[0061] Because the robot's memory stores computer-readable instructions that can be run on the processor, the processor can execute corresponding instruction operations according to the computer-readable instructions stored in the memory. Specifically, when the robot's processor receives a task instruction sent by the user, the robot first collects data information of the base and the robotic arm through the sensors set inside it, and obtains the current status of the base and the robotic arm respectively based on the data information of the base and the robotic arm. It should be noted that the current status of the base includes at least the current position information and posture information of the base, and the current status of the robotic arm includes at least the current position information and posture information of the robotic arm.

[0062] In this embodiment, the current state of the base is obtained from the data information of the base and the robot arm: b =[x b ,y b ,φ] T , the current state of the robot is: q m =[θ1,θ2,θ3,θ4,θ5,θ6] T , where x b is the moving distance of the base on the x-axis, y b is the moving distance of the base on the y-axis, φ is the rotation angle of the base, and θ1, θ2, θ3, θ4, θ5, and θ6 are the joint angles of the robot in the six degrees of freedom.

[0063] It can be understood that since the configuration of the robot's manipulator arm is obtained in the form of parallel configuration + series configuration, and the control algorithms of existing robots are all based on the series configuration, in order to facilitate the calculation of subsequent processes, after obtaining the manipulator arm state of the manipulator arm, it is also necessary to perform equivalent fitting on the states of the parallel part of the joints of the manipulator arm, so as to obtain the equivalent manipulator arm state in the series state.

[0064] In this embodiment, assuming that the first three joints of the robot arm are in parallel configuration and the last three joints are in series configuration, the equivalent state of the robot arm in the series state can be expressed as: q me =[θ 1e ,θ 2e ,θ 3e ,θ 4e ,θ 5e ,θ 6e ] T , where θ 1e ,θ 2e ,θ 3e ,θ 4e ,θ 5e ,θ 6e are the equivalent series joint angles of the robot arm in six degrees of freedom, and θ 1e ,θ 2e ,θ 3e ,θ 4e ,θ 5e ,θ 6e In fact, since the joint angles of the robot in series are equivalent to the joint angles in parallel, in this embodiment, we can obtain: θ4 = θ 4e 、θ5=θ 5e ,θ6=θ 6eOf course, in practical applications, the robot can also have other forms of series and parallel connection, which will not be described in detail here.

[0065] Then, based on the current state of the base in the robot and the equivalent series configuration state of the robot arm, the current overall configuration of the robot is obtained q = [x b ,y b ,φ,θ 1e ,θ 2e ,θ 3e ,θ 4e ,θ 5e ,θ 6e ] T . See Figure 3 , Figure 3 This is a schematic diagram for obtaining the current overall configuration of the robot. In this embodiment, in order to enable the robot to achieve better performance tracking, when obtaining the current overall configuration of the robot, the current overall configuration of the robot and the whole body motion control algorithm are also used to obtain the expected overall configuration of the robot when executing the task instruction.

[0066] It can be understood that because the desired overall configuration of the robot can be divided into two parts, one part is the configuration speed of the robot's manipulator arm, and the other part is the configuration speed of the robot's base, therefore, by using the desired overall configuration of the robot to control the robot's base and manipulator arm, the robot can implement task instructions.

[0067] Specifically, since the configuration speed of the robot arm is usually the speed of the robot's equivalent joints, the speed of the serial joint part of the robot arm can be directly output, and the speed of the parallel joint part can be calculated and obtained through the inverse kinematics model, and the actual joint speed of each joint on the robot arm can be obtained. In this way, the actual joint speed of the robot is input into the joint controller to drive the movement of the joint motor. The configuration speed of the base is usually the Cartesian space speed, so the driving speed and steering speed of the robot can be obtained through speed transformation, and finally input into the base controller to drive the movement of the base.

[0068] It can be seen that in this embodiment, when the robot processor receives the task instruction, it first collects the data information of the base and the manipulator through the sensor, and obtains the current state of the base and the manipulator based on the data information of the base and the manipulator; then, the current state of the manipulator is equivalently fitted in series to obtain the equivalent series configuration state of the manipulator, and based on the current state of the base and the equivalent series configuration state of the manipulator, the current overall configuration of the robot is obtained; then, based on the current overall configuration of the robot and the whole body motion control algorithm, the expected overall configuration of the robot when executing the task instruction is obtained, and the base and manipulator of the robot are controlled to move so that the robot is in the expected overall configuration. Obviously, compared with the prior art, because the connection structure of the manipulator in the robot is a series-parallel type, this method can not only improve the stability of the robot during the movement, but also significantly improve the load capacity of the robot.

[0069] Based on the above embodiments, this embodiment further illustrates and optimizes the technical solution. As a preferred implementation, the processor implements the steps of obtaining the desired overall configuration of the robot when executing the task instruction based on the current overall configuration and the whole body motion control algorithm, including:

[0070] Obtain the Jacobian matrix of the robot according to the current overall configuration;

[0071] The whole-body motion control algorithm formula is obtained according to the task instruction and the Jacobian matrix, and the whole-body motion control algorithm formula is used to obtain the expected overall configuration of the robot when executing the task instruction.

[0072] It can be understood that since the position and posture expression of the robot end effector can be obtained according to the sensor data, it is assumed that the position and posture expression of the end effector is: x = [x0, y0, z0, r x ,r y ,r z ] T , where x0, y0 and z0 are the robot's moving distances in the forward, lateral and longitudinal directions respectively, r x is the rotation angle of the robot around the x-axis, r y is the rotation angle of the robot around the y-axis, r z is the rotation angle of the robot around the z-axis.

[0073] Because the expression of the whole body forward kinematics of the robot manipulator is: x=f(q), where q is the current overall configuration of the robot. According to the expression of the whole body forward kinematics of the robot x=f(q), its corresponding differential form can be obtained.

[0074] That is, in, is the differential of x, is the differential of q, and J0 is the Jacobian matrix of the manipulator in the robot. Because the base is differentially driven, it satisfies the nonholonomic constraints Where, J1=[cosφ,shiφ,0,0,0,0,0,0,0] T , is the Jacobian matrix of the robot base. Combining the above two formulas, we can get the whole body positive kinematics expression of the robot in the extended form: Among them, J e is the Jacobian matrix of the robot, J e =[J0,J1] T ,

[0075] When the Jacobian matrix of the robot is obtained, the whole-body motion control algorithm formula can be obtained according to the robot's task instructions and the Jacobian matrix, and the whole-body motion control algorithm formula can be used to obtain the expected overall configuration of the robot when executing the task instructions.

[0076] As a preferred implementation, the processor implements the step of obtaining the whole body motion control algorithm formula according to the task instruction and the Jacobian matrix, including:

[0077] When the task instruction is a multi-task instruction, multiple tasks in the multi-task instruction are processed according to their task priorities to obtain a main task in the multi-task instruction;

[0078] The Jacobian matrix is ​​used to project all tasks except the main task in the multi-task instructions into the null space of the main task to obtain the whole-body motion control algorithm formula.

[0079] It is understandable that, under normal circumstances, the robot will receive multi-task instructions to coordinate and control multiple targets, where the multi-tasks performed by the robot include the end operation task of the manipulator, the base movement task, the collision avoidance task, the joint limit task, and subsequent expandable tasks, etc. The whole body motion control algorithm (WBC) can coordinate and control multiple targets, so in practical applications, the whole body motion control algorithm formula can be obtained based on the whole body motion control algorithm when the robot executes multi-task instructions.

[0080] In the whole body motion control algorithm, each target can be described in the form of a Jacobian matrix. Assume that the i-th priority target in the multi-task instruction can be defined as in, for The expansion of J i is the Jacobian matrix (which can be viewed as the same as J e equal), is the whole body joint velocity of the i-th priority target. Then the joint velocity input of the low-priority task will be mapped to the null space of the high-priority task through layer-by-layer projection, and finally projected to the null space of the main task, so that the whole body motion control algorithm formula of the robot can be obtained.

[0081] In other words, when the robot receives a multi-task instruction, it performs task priority processing on multiple tasks in the multi-task instruction, and thus obtains the main task in the multi-task instruction; then, the Jacobian matrix is ​​used to project all tasks except the main task in the multi-task instruction into the null space of the main task, and the whole-body motion control algorithm formula can be obtained.

[0082] Among them, the expression of the whole body motion control algorithm formula is:

[0083]

[0084] Where N i is the null space projection matrix, i=1,2,3…, is the whole body joint velocity of the i-th priority target.

[0085] In the whole body motion control algorithm formula, Among them, J i is the Jacobian matrix, is the Jacobian matrix J i The pseudo-inverse of , E is the identity matrix. Where W is the weight matrix. And the whole body joint velocity of the i-th priority target can be determined as:

[0086] See also Figure 4 , Figure 4 A control architecture block diagram of a robot provided in an embodiment of the present invention when executing a multi-task instruction. When the robot receives a multi-task instruction, a processor arranged inside it processes the multi-task instruction using a task priority processor, and determines the main task in the multi-task instruction, and other tasks except the main task are projected layer by layer into the null space of the main task, thereby obtaining the configuration speed of the robot when executing the multi-task instruction. Since the configuration speed of the robot when executing the multi-task instruction can be decomposed into the robot arm speed and the base speed, the motion state of the robot's robot arm and base can be controlled by the robot's joint speed controller and the base speed controller.

[0087] As a preferred embodiment, the processor implements the step of using the whole body motion control algorithm formula to obtain the desired overall configuration of the robot when executing the task instruction, including:

[0088] Based on the task instructions, obtain the expected output position of the robot;

[0089] The desired output position is input into the ADRC algorithm to obtain the target compensation amount;

[0090] The target speed of the robot when executing the target task in the multi-task instruction is obtained by using the target compensation, Jacobian matrix and expected output position;

[0091] The expected overall configuration of the robot when executing multi-task instructions is obtained according to the whole-body motion control algorithm formula and the target speed.

[0092] In this embodiment, in order to enable the robot to achieve better performance tracking, the whole body joint speed of the i-th priority target in the above whole body motion control algorithm formula is converted into Rewrite as in, is the target speed of the robot when executing the target task in the multi-task instruction, u d is the target compensation amount.

[0093] See also Figure 5 , Figure 5 This is a schematic diagram of using the ADRC algorithm to obtain the target compensation. Figure 5 The input of the Active Disturbance Rejection Control (ADRC) algorithm is v = x ed , x ed For x e The output is the expected output position of the robot, which can be marked as y = x e , the target compensation u can be obtained through the control law d Since in practical applications, the position of the robot is easier to obtain than its speed, Figure 5 The robot's position can be used instead of its velocity.

[0094] When the target speed of the robot is obtained when executing the target task in the multi-task instruction When executing multi-task instructions, the expected overall configuration of the robot can be obtained based on the whole-body motion control algorithm formula and the expression of the target speed.

[0095] It should be noted that the ADRC algorithm is usually composed of three parts: Tracking Differentiator (TD), Extended State Observer (ESO) and Nonlinear State Error Feedback (NLSEF). Since the ADRC algorithm is well known to those skilled in the art, it will not be described in detail in this embodiment. For details, please refer to the relevant content in the prior art, and no detailed description will be given here.

[0096] Based on the above embodiment, this embodiment further illustrates and optimizes the technical solution. As a preferred implementation, the robot arm includes some parallel joints, and the processor implements the steps of performing equivalent series fitting on the current state of the robot arm, including:

[0097] Obtaining the first joint angle of the parallel partial joint in the parallel configuration;

[0098] The first joint angle is processed using a preset forward kinematics analysis model to obtain a second joint angle of the parallel joints in a series configuration; wherein the preset forward kinematics analysis model is a deep learning model trained by preset training samples, and the preset training samples are a set constructed according to the inverse kinematics analysis process.

[0099] It can be understood that, since there is a corresponding functional conversion relationship between the joint angles of the parallel part of the robot and the equivalent series joint angles, in this embodiment, we define the expression of forward kinematics as g1:θ1,θ2,θ3→θ 1e ,θ 2e ,θ 3e , the inverse kinematics expression is: g2:θ 1e ,θ 2e ,θ 3e →θ1,θ2,θ3.

[0100] In this embodiment, in order to perform equivalent series fitting on the states of the parallel joints of the robot arm, the first joint angles (θ1, θ2, θ3) of the parallel joints of the robot arm in the parallel configuration are obtained, and then the first joint angles (θ1, θ2, θ3) are processed using the preset forward kinematics analysis model g1 to obtain the second joint angles (θ1, θ2, θ3) of the parallel joints of the robot arm in the series configuration. 1e ,θ 2e ,θ 3e ).

[0101] The preset forward kinematics analysis model g1 is a deep learning model trained by preset training samples, and the preset training samples are a set constructed according to the inverse kinematics analysis process. It can be understood that because the inverse kinematics analysis model has an analytical solution, while the forward kinematics analysis model does not have an analytical solution, and there is a mapping relationship between the inverse kinematics analysis model and the forward kinematics analysis model, the preset forward kinematics analysis model can be solved by using a deep learning model.

[0102] As a preferred implementation, the processor implements the steps of creating a preset forward kinematics analysis model, including:

[0103] Obtain the range of motion of the parallel joints in the series configuration;

[0104] Selecting a preset number of series joint angles in the motion range to obtain a series joint angle set;

[0105] According to the inverse kinematics analysis process, the parallel joint angle corresponding to each serial joint angle is obtained to obtain a set of parallel joint angles;

[0106] Constructing preset training samples according to the series joint angle set and the parallel joint angle set; wherein each training sample in the preset training samples includes a set of series joint angles and corresponding parallel joint angles;

[0107] The deep learning model in the initial state is trained using the preset training samples to obtain a preset positive kinematics analysis model.

[0108] When creating a preset forward kinematics analysis model, the first step is to obtain the motion range of the parallel joints of the robot arm in the series configuration. Assume that the motion range of the parallel joints of the robot arm in the series configuration is: θ 1e ∈[minθ 1e ,maxθ 1e ],θ 2e ∈[minθ 2e ,maxθ 2e ]、θ 3e ∈[minθ 3e ,maxθ 3e ]; then, a preset number of serial joint angles are selected in the motion range to obtain a set of serial joint angles; then, according to the inverse kinematics analysis process g2:θ 1e ,θ 2e ,θ 3e→θ1,θ2,θ3 obtain the parallel joint angles θ1,θ2,θ3 corresponding to each serial joint angle to obtain a set of parallel joint angles; then, construct a preset training sample based on the serial joint angle set and the parallel joint angle set, and use the preset training samples to train the deep learning model of the initial state to obtain a preset forward kinematics analysis model.

[0109] As a preferred implementation, the processor implements the step of training the deep learning model in the initial state using the preset training samples, including:

[0110] Get the adversarial generative network model;

[0111] Among them, the adversarial generative network model includes a generator and a discriminator;

[0112] Input all the parallel joint angles in the preset training samples into the generator to obtain the generated results;

[0113] The generated results and all the serial joint angles in the preset training samples are input into the discriminator for training.

[0114] Specifically, in this embodiment, a generative adversarial network (GAN) is used to create a deep learning model. Of course, in practical applications, other deep learning algorithms can also be used to generate deep learning models, which will not be described in detail here.

[0115] See also Figure 6 , Figure 6 Schematic diagram of the adversarial generative network model. The adversarial generative network model includes: a generator G and a discriminator D. Specifically, the generator G is used to train and learn from a low-dimensional potential vector z(z~p z (z)) to the real data x; the discriminator D is used to train and learn to distinguish data from real data x (x~P data (x)) is still the generated data G(z); the adversarial generative network model adjusts the generator G and the discriminator D through the optimization process.

[0116] Among them, the objective function of the adversarial generation network model is:

[0117]

[0118] Based on the above theoretical basis, in this embodiment, all the parallel joint angles (θ1, θ2, θ3) in the preset training sample are input into the generator G to obtain the generated result (θ 1eg ,θ 2eg ,θ 3eg ), and the result (θ 1eg,θ 2eg ,θ 3eg ) and all the serial joint angles in the preset training sample are input into the discriminator for training, thereby obtaining the preset forward kinematics analysis model g1:θ1,θ2,θ3→θ 1e ,θ 2e ,θ 3e .

[0119] Correspondingly, an embodiment of the present invention further discloses a control method of a robot, wherein the robot includes a base, a mechanical arm, and a sensor, the mechanical arm is connected to the base, and the mechanical arm is a serial-parallel type mechanical arm, and the control method includes:

[0120] When receiving a task instruction, the data information of the base and the robotic arm collected by the sensor is obtained, and the current status of the base and the robotic arm are obtained based on the data information;

[0121] Perform equivalent series fitting on the current state of the robot arm to obtain the equivalent series configuration state of the robot arm;

[0122] Based on the current state of the base and the equivalent serial configuration state of the robotic arm, the current overall configuration of the robot is obtained;

[0123] Based on the current overall configuration and the whole body motion control algorithm, the expected overall configuration of the robot corresponding to the task instruction is obtained;

[0124] Control the movement of the base and the robotic arm to put the robot in the desired overall configuration.

[0125] A robot control method provided in an embodiment of the present invention corresponds to a robot disclosed in the aforementioned embodiment. For relevant details, reference may be made to the contents disclosed in the aforementioned embodiment, and no detailed description will be given here.

[0126] A robot control method provided by an embodiment of the present invention has the beneficial effects of the robot disclosed above.

[0127] See also Figure 7 , Figure 7 A structural diagram of a robot control device provided by an embodiment of the present invention, the device includes:

[0128] A memory 31, used for storing computer programs;

[0129] The processor 32 is used to implement the steps of a robot control method disclosed above when executing a computer program.

[0130] A robot control device provided by an embodiment of the present invention has the beneficial effects of the robot control method disclosed above.

[0131] Correspondingly, an embodiment of the present invention further discloses a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of a robot control method disclosed above are implemented.

[0132] A computer-readable storage medium provided by an embodiment of the present invention has the beneficial effects of the aforementioned robot control method disclosed above.

[0133] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0134] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0135] The robot, the control method, the device and the medium of the robot provided by the present invention are introduced in detail above. The principle and implementation mode of the present invention are explained in this article by using specific examples. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for the general technical personnel in this field, according to the idea of ​​the present invention, there will be changes in the specific implementation mode and the scope of application. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A robot, characterized in that: The robot comprises a base, a mechanical arm, a sensor, a memory and a processor. The mechanical arm is connected to the base. The mechanical arm is a serial-parallel type mechanical arm. The memory stores computer-readable instructions that can be run on the processor. The processor is used to implement the following steps when executing the computer-readable instructions: When a task instruction is received, the data information of the base and the robotic arm collected by the sensor is obtained, and the current states of the base and the robotic arm are obtained based on the data information; the current state of the base includes at least the current position information and posture information of the base, and the current state of the robotic arm includes at least the current position information and posture information of the robotic arm; Performing equivalent series fitting on the current state of the robotic arm to obtain an equivalent series configuration state of the robotic arm; Based on the current state of the base and the equivalent serial configuration state of the robot arm, obtain the current overall configuration of the robot; Obtaining a Jacobian matrix of the robot according to the current overall configuration; When the task instruction is a multi-task instruction, performing task priority processing on multiple tasks in the multi-task instruction to obtain a main task in the multi-task instruction; Projecting all tasks except the main task in the multi-task instruction into the null space of the main task using the Jacobian matrix to obtain a whole-body motion control algorithm formula; Based on the task instruction, obtaining the expected output position of the robot; Inputting the desired output position into an active disturbance rejection control algorithm to obtain a target compensation amount; Obtaining a target speed of the robot when executing a target task in the multi-task instruction using the target compensation amount, the Jacobian matrix and the expected output position; Obtaining a desired overall configuration of the robot when executing the multi-task instruction according to the whole-body motion control algorithm formula and the target speed; Controlling the base and the robotic arm to move so that the robot is in the desired overall configuration; Wherein, the expression of the whole body motion control algorithm formula is: ; In the formula, is the null space projection matrix, , For the The whole body joint velocities of the priority targets, , is the Jacobian matrix, is the Jacobian matrix The pseudo-reversal, is the identity matrix; Rewrite as , The multi-tasking instruction priority goals, is the target compensation amount, for The first derivative of for expectations, , is the expected output position of the robot; The robotic arm includes some parallel joints, and the processor implements the step of performing equivalent series fitting on the current state of the robotic arm, including: Acquire a first joint angle of the parallel partial joint in a parallel configuration; The first joint angle is processed by using a preset forward kinematics analysis model to obtain a second joint angle of the parallel joint in a series configuration; wherein the preset forward kinematics analysis model is a deep learning model trained by preset training samples, and the preset training samples are a set constructed according to an inverse kinematics analysis process; The processor implements the steps of creating a preset forward kinematics analysis model, including: Obtaining the range of motion of the parallel joints in the series configuration; Selecting a preset number of series joint angles in the motion range to obtain a series joint angle set; According to the inverse kinematics analysis process, a parallel joint angle corresponding to each serial joint angle is obtained to obtain a parallel joint angle set; Constructing the preset training samples according to the series joint angle set and the parallel joint angle set; wherein each training sample in the preset training samples includes a set of series joint angles and corresponding parallel joint angles; The deep learning model in the initial state is trained using the preset training samples to obtain the preset forward kinematics analysis model.

2. The robot according to claim 1, characterized in that: The processor implements the step of training the deep learning model in the initial state by using the preset training sample, including: Acquire a generative adversarial network model; wherein the generative adversarial network model includes a generator and a discriminator; Inputting all parallel joint angles in the preset training sample into the generator to obtain a generated result; The generated result and all the serial joint angles in the preset training sample are input into the discriminator for training.

3. A robot control method, characterized in that: The robot comprises a base, a mechanical arm, and a sensor. The mechanical arm is connected to the base. The mechanical arm is a serial-parallel type mechanical arm, and comprises: When a task instruction is received, the data information of the base and the robotic arm collected by the sensor is obtained, and the current states of the base and the robotic arm are obtained based on the data information; the current state of the base includes at least the current position information and posture information of the base, and the current state of the robotic arm includes at least the current position information and posture information of the robotic arm; Performing equivalent series fitting on the current state of the robotic arm to obtain an equivalent series configuration state of the robotic arm; Based on the current state of the base and the equivalent serial configuration state of the robot arm, obtain the current overall configuration of the robot; Obtaining a Jacobian matrix of the robot according to the current overall configuration; When the task instruction is a multi-task instruction, performing task priority processing on multiple tasks in the multi-task instruction to obtain a main task in the multi-task instruction; Projecting all tasks except the main task in the multi-task instruction into the null space of the main task using the Jacobian matrix to obtain a whole-body motion control algorithm formula; Based on the task instruction, obtaining the expected output position of the robot; Inputting the desired output position into an active disturbance rejection control algorithm to obtain a target compensation amount; Obtaining a target speed of the robot when executing a target task in the multi-task instruction using the target compensation amount, the Jacobian matrix and the expected output position; Obtaining a desired overall configuration of the robot when executing the multi-task instruction according to the whole-body motion control algorithm formula and the target speed; Controlling the base and the robotic arm to move so that the robot is in the desired overall configuration; Wherein, the expression of the whole body motion control algorithm formula is: ; In the formula, is the null space projection matrix, , For the The whole body joint velocities of the priority targets, , is the Jacobian matrix, is the Jacobian matrix The pseudo-reversal, is the identity matrix; Rewrite as , The multi-tasking instruction priority goals, is the target compensation amount, for The first derivative of for expectations, , is the expected output position of the robot; The robotic arm includes some parallel joints, and the step of performing equivalent series fitting on the current state of the robotic arm includes: Acquire a first joint angle of the parallel partial joint in a parallel configuration; The first joint angle is processed by using a preset forward kinematics analysis model to obtain a second joint angle of the parallel joint in a series configuration; wherein the preset forward kinematics analysis model is a deep learning model trained by preset training samples, and the preset training samples are a set constructed according to an inverse kinematics analysis process; The steps of creating the preset forward kinematics analysis model include: Obtaining the range of motion of the parallel joints in the series configuration; Selecting a preset number of series joint angles in the motion range to obtain a series joint angle set; According to the inverse kinematics analysis process, a parallel joint angle corresponding to each serial joint angle is obtained to obtain a parallel joint angle set; Constructing the preset training samples according to the series joint angle set and the parallel joint angle set; wherein each training sample in the preset training samples includes a set of series joint angles and corresponding parallel joint angles; The deep learning model in the initial state is trained using the preset training samples to obtain the preset forward kinematics analysis model.

4. A robot control device, characterized in that: include: Memory for storing computer programs; A processor is used to implement the steps of a robot control method as claimed in claim 3 when executing the computer program.

5. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by the processor, the steps of the robot control method according to claim 3 are implemented.