WO2023046074A1 - Autonomous learning method for mechanical arm control method - Google Patents
Autonomous learning method for mechanical arm control method Download PDFInfo
- Publication number
- WO2023046074A1 WO2023046074A1 PCT/CN2022/120894 CN2022120894W WO2023046074A1 WO 2023046074 A1 WO2023046074 A1 WO 2023046074A1 CN 2022120894 W CN2022120894 W CN 2022120894W WO 2023046074 A1 WO2023046074 A1 WO 2023046074A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- linear motor
- mechanical arm
- actuator
- learning
- force feedback
- Prior art date
Links
- 238000000034 method Methods 0.000 title claims abstract description 63
- 230000033001 locomotion Effects 0.000 claims abstract description 109
- 239000011159 matrix material Substances 0.000 claims description 43
- 238000006073 displacement reaction Methods 0.000 claims description 33
- 238000004364 calculation method Methods 0.000 claims description 12
- 230000008569 process Effects 0.000 claims description 5
- 238000000354 decomposition reaction Methods 0.000 claims description 3
- 230000001960 triggered effect Effects 0.000 claims description 3
- 239000003550 marker Substances 0.000 claims 1
- 238000010187 selection method Methods 0.000 claims 1
- 125000004122 cyclic group Chemical group 0.000 abstract 1
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000001125 extrusion Methods 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/16—Programme controls
- B25J9/1602—Programme controls characterised by the control system, structure, architecture
Definitions
- the present invention relates to the field of manipulator control, and more specifically, to an autonomous learning method of a manipulator control method.
- the automatic movement of the robotic arm by letting the robotic arm execute the specified program has been applied in many fields.
- One of the application areas is that the robotic arm can be used to perform non-medical and medical procedures through remote control. .
- teleoperated surgical manipulators can be used to perform minimally invasive medical procedures. It is desirable in medical technology to reduce the amount of tissue damaged during medical procedures, thereby reducing recovery time, discomfort and unwanted side effects for patients.
- a robotic arm to move around a remote center of motion also known as a "remote center” in this field, the control accuracy of the robotic arm will decrease due to the influence of external tissues, resulting in a change in the actual posture of the robotic arm. Deviating from the posture or movement that the command really wants to output will cause errors in the movement of the robotic arm.
- the Chinese patent application with the publication number "CN111315309A” and the publication date of June 19, 2020 discloses a system and method for controlling a robot manipulator or related tools. It is believed that the external organization affects the control accuracy of the robotic arm. The resulting vibrations thus provide precision in the control of the robotic arm by reducing the vibrations to which the robotic arm is subjected.
- the present invention provides an autonomous learning method of the control method of the manipulator, and improves the control precision of the manipulator through learning feedback.
- the technical solution adopted by the present invention is to provide a self-learning method of a mechanical arm control method, which includes a mechanical arm and actuators installed on the mechanical arm, and the mechanical arm includes several driving actuators.
- the actuator also includes a force feedback sensor, and the self-learning method of the mechanical arm control method includes the following steps:
- Step 1 Initialize the robotic arm and force feedback sensor
- Step 2 Set the moving path of the actuator
- Step 3 Operate the actuator to move to the remote motion center point along the movement path, and make the actuator complete a rotation cycle within a given rotation range to obtain the remote motion center point data;
- Step 4 By establishing a learning network and given the number of learning steps n, the actuator starts to rotate n times on the moving path. After each step, the remote motion center point and the learning network are updated according to the force feedback sensor data.
- the movement displacement of the actuator is output through the learning network, and then the learning network is updated according to the force feedback of the force feedback sensor during the motion of the actuator. Readjust the movement displacement of the actuator, so that the actuator can continuously adjust its displacement according to the external environment during the movement, so that the actual movement of the actuator can be as consistent as possible with the movement output by the control command.
- the method for initializing the force feedback sensor is to read the force feedback signal and define it as the zero point when there is no set external force; during the online learning process, the force feedback signal is converted to a difference, and the difference with the zero point is used as output.
- the mechanical arm includes a clamping part for clamping the actuator, a first linear motor and a second linear motor for driving the clamping part to rotate, and the clamping part is provided with a device for driving the actuator
- a third linear motor that moves linearly along the clamping part; in the first step, the parameters of the mechanical arm are measured when the mechanical arm is initialized, including the third linear motor to the first linear motor or the second linear motor
- the motor is position calibrated and zeroed.
- the given rotation range is the positive and negative maximum rotation angles [ ⁇ , - ⁇ ] reached within the strokes of the first linear motor and the second linear motor; wherein, the end of the actuator
- the state of the vertical horizontal plane is defined as the zero point of the angle
- one rotation period is defined as the movement period in which the actuator starts to rotate to the angle ⁇ with the zero point of the angle as the initial position, reverses through the zero point of the angle to reach the - ⁇ angle, and then rotates back to the zero point of the angle in reverse.
- the remote motion center point is defined as a 3*k matrix, k is the total number of center point groups, and each column represents a group of center points (CL1 i , CL2 i , CL3 i ), where CL1, CL2, and CL3 are respectively
- CL1, CL2, and CL3 are respectively
- the marking positions of the first linear motor, the second linear motor and the third linear motor; the method of selecting the remote motion center point is to record the first linear motor, the second linear motor and the third linear motor during the movement of the actuator in step 3
- the actual range of travel specifically: [minL1, maxL1], [minL2, maxL2], [minL3, maxL3], then the k center points can be expressed as:
- the specific steps of updating the learning network in said step 4 are as follows:
- L1 and L2 are the current positions of the first linear motor and the second linear motor respectively;
- L3 is the current position of the third linear motor 3
- ⁇ L1, ⁇ L2, and ⁇ L3 are the estimated displacements of the first linear motor, the second linear motor, and the third linear motor respectively;
- x, y are the coordinates of the RCM (Remote Center of Motion, remote center of motion) point (x, y);
- ⁇ L1', ⁇ L2', ⁇ L3' are the target displacements of the first linear motor, the second linear motor and the third linear motor respectively;
- S is the matrix composed of the remote motion center points;
- b is a 3*1 matrix, indicating that each A bias value modified by the learning network in each displacement direction;
- ⁇ ' is the force feedback value of the actuator
- ⁇ x , ⁇ y are the decomposition forces of ⁇ ' in the x direction and y direction respectively;
- W' is the updated weight matrix
- W is the original weight matrix
- a is the current rotation direction mark (a ⁇ 1,-1 ⁇ )
- J 1 is the first column of the Jacobian matrix
- J 2 is the Jacobian matrix The second column;
- Steps S1-S7 are an update cycle of the learning network, and S1-S7 is re-executed in the next update cycle, and the parameters updated in the previous update cycle are used as parameters of the learning network.
- the first linear motor, the second linear motor and the third linear motor are calculated according to the Jacobian matrix of the end position of the actuator calculated in step 4 before each step of driving The target displacement amount, and complete the movement of the actuator according to the given target displacement amount.
- the data of the force feedback sensor is set with a threshold T.
- the force feedback value of the actuator exceeds the threshold T after this movement, the update of the learning network weight W is triggered; otherwise, no weight update is performed.
- the purpose of continuously reducing the force and reducing the degree of extrusion to the surroundings can be achieved through targeted updating of the learning network.
- the force feedback value of the actuator does not exceed the threshold T within one continuous rotation cycle, it is considered that the training process of the learning network has been completed. That is, after learning, it has been able to ensure that the needle tip performs remote motion center movement within the range of force less than T, thereby dynamically improving the accuracy of remote motion center movement.
- An autonomous learning system of a mechanical arm control method used to realize the above-mentioned autonomous learning method of a mechanical arm control method, comprising a mechanical arm and a controller module, the controller module is used to electrically connect with a mechanical arm to obtain a force feedback sensor data and control the movement of the robotic arm; the controller module includes a processor for performing calculations.
- the beneficial effect is: the real-time learning of the operating object in the actual environment, and the motion calibration of the remote motion center based on the force feedback of the actuator, this method makes the actuator highly consistent with the actual operating environment.
- Adaptation can take into account the deviation caused by the deformation of the environment, and dynamically guide the adjustment of the remote motion center motion, so that the accuracy of the robotic arm's motion execution can be greatly improved, and the standard motion between the actual motion of the robotic arm and the output of the control command can be reduced. error between.
- Fig. 1 is the structural representation of mechanical arm of the present invention
- Fig. 2 is a flow chart of an autonomous learning method of a mechanical arm control method in the present invention.
- FIG. 1-2 it is an embodiment of an autonomous learning method of a mechanical arm control method, which includes a mechanical arm 1 and an actuator 2 installed on the mechanical arm 1, and the mechanical arm 1 includes several driving actuators.
- the mechanical arm 1 includes a clamping part 3 for clamping the actuator 2, and a first linear motor for driving the clamping part 3 to rotate
- the motor 4 and the second linear motor 5, the first linear motor 4 and the second linear motor 5 are respectively hinged to the clamping part 3, and the clamping part 3 is provided with a device for driving the actuator 2 along the clamping part.
- Part 3 is a third linear motor 6 that performs linear motion; in the first step, the parameters of the mechanical arm 1 are measured when the mechanical arm 1 is initialized, including the third linear motor to the first linear motor 4 or the second linear motor.
- the linear motor 5 and the third linear motor perform position calibration and return to zero.
- the autonomous learning method of the control method of the mechanical arm 1 comprises the following steps:
- Step 1 Initialize the mechanical arm 1 and the force feedback sensor 201;
- Step 2 setting the movement path of the actuator 2
- Step 3 Operate the actuator 2 to move to the remote motion center point along the movement path, and make the actuator 2 complete a rotation cycle movement within a given rotation range to obtain the remote motion center point data;
- the given rotation range is the first The positive and negative maximum rotation angle [ ⁇ , - ⁇ ] reached within the stroke of a linear motor 4 and the second linear motor 5; wherein, the vertical horizontal plane state at the end of the actuator 2 is defined as the zero point of the angle, and one rotation period is defined as the actuator 2 Start to rotate to the ⁇ angle with the zero point of the angle as the initial position, reverse the rotation through the zero point of the angle to reach the - ⁇ angle, and then reverse the motion cycle of returning to the zero point of the angle.
- the remote motion center point is defined as a 3*k matrix, k is the total number of center point groups, and each column represents a group of center points CL1 i , CL2 i , CL3 i , where CL1, CL2, and CL3 are the first linear motor 4 , the marked positions of the second linear motor 5 and the third linear motor; the method for selecting the remote motion center point is to record the first linear motor 4, the second linear motor 5 and the third linear motor during the movement of the actuator 2 in step three
- the actual travel range of specifically: [minL1, maxL1], [minL2, maxL2], [minL3, maxL3], then the k remote motion center points can be expressed as:
- Step 4 By establishing a learning network and given the number of learning steps n, the actuator 2 starts to rotate n times on the moving path. After each step, the remote motion center point and the learning network are updated according to the data of the force feedback sensor 201. Specific steps are as follows:
- L1 and L2 are the current positions of the first linear motor 4 and the second linear motor 5 respectively;
- L3 is the current position of the third linear motor 3
- ⁇ L1, ⁇ L2, ⁇ L3 are the estimated displacements of the first linear motor 4, the second linear motor 5 and the third linear motor respectively;
- x, y are the coordinates x, y of the RCM point;
- ⁇ L1', ⁇ L2', ⁇ L3' are the target displacements of the first linear motor 4, the second linear motor 5 and the third linear motor respectively;
- S is a matrix composed of remote motion center points;
- b is a 3*1 matrix, Indicates the bias value modified by the learning network for each displacement direction;
- ⁇ ' is the force feedback value of actuator 2
- ⁇ x , ⁇ y are the decomposition forces of ⁇ ' in the x direction and y direction respectively;
- W' is the updated weight matrix
- W is the original weight matrix
- a is the current rotation direction mark a ⁇ ⁇ 1, -1 ⁇
- J 1 is the first column of the Jacobian matrix
- J 2 is the second column of the Jacobian matrix List
- Steps S1-S7 are an update cycle of the learning network, and S1-S7 is re-executed in the next update cycle, and the parameters updated in the previous update cycle are used as parameters of the learning network. That is, the weight matrix W of S4 in the nth cycle is the updated weight matrix W' obtained by S7 in the n-1th cycle.
- the data of the force feedback sensor 201 is set with a threshold T.
- the force feedback value of the actuator 2 exceeds the threshold T after this exercise, the update of the learning network weight W is triggered; otherwise, no weight update is performed. In this way, the purpose of continuously reducing the force and reducing the degree of extrusion to the surroundings can be achieved through targeted updating of the learning network. If the force feedback value of the actuator 2 does not exceed the threshold T within one continuous rotation cycle, it is considered that the training process of the learning network has been completed. That is, after learning, it has been able to ensure that the needle tip performs remote motion center movement within the range of force less than T, thereby dynamically improving the accuracy of remote motion center movement.
- the working principle of this embodiment when the mechanical arm 1 drives the actuator 2 to execute the remote motion center point movement along the set path, the movement displacement of the actuator 2 is output through the learning network, and then according to the force of the actuator 2 during the movement Feedback the force feedback of the sensor 201 to update the learning network to readjust the movement displacement of the actuator 2, so that the actuator 2 continuously adjusts its own displacement according to the external environment during the movement, so that the actual movement of the actuator 2 Keep as consistent as possible with the motion output by the control command.
- An autonomous learning system of a mechanical arm control method used to realize the autonomous learning method of the mechanical arm control method of embodiment 1, comprising a mechanical arm and a controller module, the controller module is used to electrically connect with the mechanical arm to obtain force Feedback sensor data and control the movement of the robotic arm; the controller module includes a processor for performing calculations.
Landscapes
- Engineering & Computer Science (AREA)
- Automation & Control Theory (AREA)
- Robotics (AREA)
- Mechanical Engineering (AREA)
- Manipulator (AREA)
Abstract
An autonomous learning method for a mechanical arm control method, comprising the following steps: step 1: initializing a mechanical arm (1) and a force feedback sensor (201); step 2: setting a motion path of an execution element (2); step 3: operating the execution element (2) to move to a remote motion center point along the motion path, and enabling the execution element (2) to complete a rotation cyclic motion within a given rotation range to obtain remote motion center point data; and step 4: giving a learning step number n by establishing a learning network, the execution element (2) starting n rotation motions on the motion path, and after each step of motion, updating the remote motion center point and the learning network according to data of the force feedback sensor (201). According to the autonomous learning method, the motion of the execution element (2) is adapted to an actual running environment, the deviation caused by the deformation of the environment to the motion is taken into consideration, and the motion adjustment of a remote motion center is dynamically guided, so that the motion execution precision of the mechanical arm (1) is improved, and the error between an actual motion and a standard motion of the mechanical arm (1) is reduced.
Description
本发明涉及机械臂控制领域，更具体地，涉及一种机械臂控制方法的自主学习方法。The present invention relates to the field of manipulator control, and more specifically, to an autonomous learning method of a manipulator control method.
通过让机械臂执行指定的程序来让机械臂实现自动运动的方式在许多的领域都得到了应用，其中的一个应用领域是，通过远程控制的机械臂能够被用于执行非医疗程序和医疗程序。作为一个具体使用方法，远程操作的手术操纵器能够被用于执行微创医疗程序。在医疗技术中希望减少医疗程序期间受损的组织数量，从而减少患者的恢复时间、不适性和有害的副作用。但是在该领域中使用机械臂执行绕着远程运动中心(也称为“远程中心”)移动的时候，由于会受到外部组织的影响而导致机械臂的控制精度下降，导致机械臂的实际姿势会偏离指令真正想要输出的姿势或运动，令机械臂的运动发生误差。The automatic movement of the robotic arm by letting the robotic arm execute the specified program has been applied in many fields. One of the application areas is that the robotic arm can be used to perform non-medical and medical procedures through remote control. . As one specific use, teleoperated surgical manipulators can be used to perform minimally invasive medical procedures. It is desirable in medical technology to reduce the amount of tissue damaged during medical procedures, thereby reducing recovery time, discomfort and unwanted side effects for patients. However, when using a robotic arm to move around a remote center of motion (also known as a "remote center") in this field, the control accuracy of the robotic arm will decrease due to the influence of external tissues, resulting in a change in the actual posture of the robotic arm. Deviating from the posture or movement that the command really wants to output will cause errors in the movement of the robotic arm.
如公开号为“CN111315309A”，公开日为2020年6月19日的中国专利申请公开了用于控制机器人操纵器或相关工具的***和方法，认为外部组织影响机械臂控制精度的因素为外部组织所产生的振动，因此通过减少机械臂受到的振动从而提供对机械臂的控制精度。For example, the Chinese patent application with the publication number "CN111315309A" and the publication date of June 19, 2020 discloses a system and method for controlling a robot manipulator or related tools. It is believed that the external organization affects the control accuracy of the robotic arm. The resulting vibrations thus provide precision in the control of the robotic arm by reducing the vibrations to which the robotic arm is subjected.
但是振动无法完全被消除，因此机械臂的精度在只能提高至一定的程度后就无法再提高，因此通过减少振动的方式所能提高的精度极其有限。However, vibration cannot be completely eliminated, so the accuracy of the robotic arm cannot be improved after it can only be improved to a certain extent, so the accuracy that can be improved by reducing vibration is extremely limited.
发明内容Contents of the invention
本发明为克服上述现有技术中机械臂远程运动中心控制精度低问题，提供一种机械臂控制方法的自主学习方法，通过学习反馈来提高机械臂的控制精度。In order to overcome the problem of low control precision of the remote motion center of the manipulator in the prior art, the present invention provides an autonomous learning method of the control method of the manipulator, and improves the control precision of the manipulator through learning feedback.
为解决上述技术问题，本发明采用的技术方案是：提供一种机械臂控制方法的自主学习方法，包括机械臂和安装于所述机械臂上的执行元件，所述机械臂包括若干个驱动执行元件运动的驱动电机；所述执行元件还包括力反馈传感器，机械臂控制方法的自主学习方法包括如下步骤：In order to solve the above-mentioned technical problems, the technical solution adopted by the present invention is to provide a self-learning method of a mechanical arm control method, which includes a mechanical arm and actuators installed on the mechanical arm, and the mechanical arm includes several driving actuators. The drive motor of component motion; The actuator also includes a force feedback sensor, and the self-learning method of the mechanical arm control method includes the following steps:
步骤一：对机械臂和力反馈传感器进行初始化；Step 1: Initialize the robotic arm and force feedback sensor;
步骤二：设定执行元件的移动路径；Step 2: Set the moving path of the actuator;
步骤三：操作执行元件沿着移动路径运动至远程运动中心点，并令执行元件在给定旋转范围内完成一个旋转周期运动，获得远程运动中心点数据；Step 3: Operate the actuator to move to the remote motion center point along the movement path, and make the actuator complete a rotation cycle within a given rotation range to obtain the remote motion center point data;
步骤四：通过建立学习网络，给定学习步数n，执行元件在移动路径上开始进行n次旋转运动，每步运动后根据力反馈传感器数据对远程运动中心点和学习网络进行更新。Step 4: By establishing a learning network and given the number of learning steps n, the actuator starts to rotate n times on the moving path. After each step, the remote motion center point and the learning network are updated according to the force feedback sensor data.
在机械臂带动执行元件沿着设定路径执行远程运动中心点运动的时候，通过学习网络输出执行元件的运动位移量，然后根据执行元件运动过程中力反馈传感器的受力反馈来更新学习网络来重新调整执行元件的运动位移量，使得执行元件在运动的过程中不断的根据外部环境调整自身的位移，从而令执行元件实际发生的运动与控制指令输出的运动尽量的保持一致。When the mechanical arm drives the actuator to move the remote motion center point along the set path, the movement displacement of the actuator is output through the learning network, and then the learning network is updated according to the force feedback of the force feedback sensor during the motion of the actuator. Readjust the movement displacement of the actuator, so that the actuator can continuously adjust its displacement according to the external environment during the movement, so that the actual movement of the actuator can be as consistent as possible with the movement output by the control command.
优选的，力反馈传感器进行初始化的方法为，在无设定外力作用时，读取力反馈信号并定义为零点；在线学习过程中对力反馈信号进行差值转换，将与零点的差值作为输出。Preferably, the method for initializing the force feedback sensor is to read the force feedback signal and define it as the zero point when there is no set external force; during the online learning process, the force feedback signal is converted to a difference, and the difference with the zero point is used as output.
优选的，所述机械臂包括用于夹持执行元件的夹持部，用于驱动夹持部转动的第一线性马达和第二线性马达，所述夹持部上设置有驱动所述执行元件沿所述夹持部做直线运动的第三线性电机；在所述步骤一中，对机械臂初始化的时候对机械臂的参数进行测量，包括第三线性马达到第一线性马达或第二线性马达的水平轴的垂直距离dm、力反馈传感器水平中心与执行元件末端的距离Ltool、第一线性马达和第二线性马达的直线距离h以及对第一线性马达、第二线性马达和第三线性马达进行位置标定并归零。Preferably, the mechanical arm includes a clamping part for clamping the actuator, a first linear motor and a second linear motor for driving the clamping part to rotate, and the clamping part is provided with a device for driving the actuator A third linear motor that moves linearly along the clamping part; in the first step, the parameters of the mechanical arm are measured when the mechanical arm is initialized, including the third linear motor to the first linear motor or the second linear motor The vertical distance dm of the horizontal axis of the motor, the distance Ltool between the horizontal center of the force feedback sensor and the end of the actuator, the linear distance h between the first linear motor and the second linear motor, and the distance between the first linear motor, the second linear motor and the third linear motor The motor is position calibrated and zeroed.
优选的，在所述步骤三中，给定旋转范围为所述第一线性马达和所述第二线性马达行程内达到的正负最大旋转角度[Θ，-Θ]；其中，执行元件末端的垂直水平面状态定义为角度零点，一个旋转周期定义为执行元件以角度零点为初始位置开始旋转至Θ角度，反向旋转经过角度零点到达-Θ角度，再反向旋转回到角度零点的运动周期。Preferably, in the step 3, the given rotation range is the positive and negative maximum rotation angles [Θ, -Θ] reached within the strokes of the first linear motor and the second linear motor; wherein, the end of the actuator The state of the vertical horizontal plane is defined as the zero point of the angle, and one rotation period is defined as the movement period in which the actuator starts to rotate to the angle Θ with the zero point of the angle as the initial position, reverses through the zero point of the angle to reach the -Θ angle, and then rotates back to the zero point of the angle in reverse.
优选的，远程运动中心点义为一个3*k的矩阵，k为中心点分组总数，其中每一列表示一组中心点(CL1
i,CL2
i,CL3
i)，其中CL1、CL2、CL3分别为第一线性马达、第二线性马达和第三线性马达的标记位置；远程运动中心点的选取 方法为，记录步骤三中执行元件运动过程中第一线性马达、第二线性马达和第三线性马达的实际行程范围，具体为：[minL1,maxL1],[minL2,maxL2],[minL3,maxL3]，则k个中心点可表示为:
Preferably, the remote motion center point is defined as a 3*k matrix, k is the total number of center point groups, and each column represents a group of center points (CL1 i , CL2 i , CL3 i ), where CL1, CL2, and CL3 are respectively The marking positions of the first linear motor, the second linear motor and the third linear motor; the method of selecting the remote motion center point is to record the first linear motor, the second linear motor and the third linear motor during the movement of the actuator in step 3 The actual range of travel, specifically: [minL1, maxL1], [minL2, maxL2], [minL3, maxL3], then the k center points can be expressed as:
优选的，在所述步骤四中更新学习网络的具体步骤如下：Preferably, the specific steps of updating the learning network in said step 4 are as follows:
S1：计算当前执行元件末端与垂直面所成角度θS1: Calculate the angle θ formed between the end of the current actuator and the vertical plane
其中，L1、L2分别为第一线性马达、第二线性马达的当前位置；Wherein, L1 and L2 are the current positions of the first linear motor and the second linear motor respectively;
S2：计算当前位置下的雅可比矩阵，雅可比矩阵计算公式如下：S2: Calculate the Jacobian matrix at the current position, the calculation formula of the Jacobian matrix is as follows:
其中，L3为第三线性马达3的当前位置；Wherein, L3 is the current position of the third linear motor 3;
S3：根据雅可比矩阵计算获得第一线性马达、第二线性马达和第三线性马达在本次训练中的预估相对位移，计算公式如下：S3: Calculate and obtain the estimated relative displacements of the first linear motor, the second linear motor and the third linear motor in this training according to the Jacobian matrix, and the calculation formula is as follows:
其中，ΔL1、ΔL2、ΔL3分别为第一线性马达、第二线性马达和第三线性马达的预估位移量；x，y为RCM(Remote Center of Motion，远程运动中心)点的坐标(x，y)；Among them, ΔL1, ΔL2, and ΔL3 are the estimated displacements of the first linear motor, the second linear motor, and the third linear motor respectively; x, y are the coordinates of the RCM (Remote Center of Motion, remote center of motion) point (x, y);
S4：将预估位移量ΔL1、ΔL2、ΔL3与中心矩阵输入学习网络，根据权重矩阵W对预估位移量进行调整，获得目标位移量，计算方法如下：S4: Input the estimated displacement ΔL1, ΔL2, ΔL3 and the center matrix into the learning network, adjust the estimated displacement according to the weight matrix W, and obtain the target displacement. The calculation method is as follows:
其中，ΔL1＇、ΔL2＇、ΔL3＇分别为第一线性马达、第二线性马达和第三线性马达的目标位移量；S为远程运动中心点组成的矩阵；b为3*1矩阵，表示每个位移方向经学习网络修改的偏置值；Among them, ΔL1', ΔL2', ΔL3' are the target displacements of the first linear motor, the second linear motor and the third linear motor respectively; S is the matrix composed of the remote motion center points; b is a 3*1 matrix, indicating that each A bias value modified by the learning network in each displacement direction;
S5：第一线性马达、第二线性马达和第三线性马达根据步骤S4获得的目标位移量进行移动，使得执行元件与垂直面形成的角度形成新角度,具体为：S5: The first linear motor, the second linear motor and the third linear motor move according to the target displacement obtained in step S4, so that the angle formed by the actuator and the vertical plane forms a new angle, specifically:
S6：获取并解析力反馈传感器数据，获得远程运动中心坐标方向分量：S6: Obtain and analyze the force feedback sensor data, and obtain the coordinate direction component of the remote motion center:
τ
x＝cos(θ)*τ'；
τ x =cos(θ)*τ';
τ
y＝sin(θ)*τ'；
τ y = sin(θ)*τ';
其中，τ'为执行元件的受力反馈值；τ
x，τ
y分别为τ'在x方向和y方向的分解力；
Among them, τ' is the force feedback value of the actuator; τ x , τ y are the decomposition forces of τ' in the x direction and y direction respectively;
S7：根据力反馈传感器数据更新权重矩阵，更新公式如下：S7: Update the weight matrix according to the force feedback sensor data, the update formula is as follows:
W'＝W-a*learnrate*(τ
x*J
1+τ
y*J
2)*S；
W'＝Wa*learnrate*(τ x *J 1 +τ y *J 2 )*S;
b＝a*learnrate*(τ
x*J
1+τ
y*J
2)
b＝a*learnrate*(τ x *J 1 +τ y *J 2 )
其中，W＇为更新后的权重矩阵；W为原权重矩阵；a为当前旋转方向标记(a∈{1，-1})；J
1为雅克比矩阵第一列；J
2为雅克比矩阵第二列；
Among them, W' is the updated weight matrix; W is the original weight matrix; a is the current rotation direction mark (a∈{1,-1}); J 1 is the first column of the Jacobian matrix; J 2 is the Jacobian matrix The second column;
S8：步骤S1-S7为学习网络的一个更新循环，下一个更新循环重新执行S1-S7，并以上一个更新循环所更新的参数作为学习网络的参数。S8: Steps S1-S7 are an update cycle of the learning network, and S1-S7 is re-executed in the next update cycle, and the parameters updated in the previous update cycle are used as parameters of the learning network.
优选的，所述执行元件在进行远程运动中心运动时，每步驱动前根据所述步骤四计算得到的执行元件末端位置的雅可比矩阵计算第一线性马达、第二线性马达和第三线性马达的目标位移量，并依据给定目标位移量完成执行元件的运动。Preferably, when the actuator performs remote motion center movement, the first linear motor, the second linear motor and the third linear motor are calculated according to the Jacobian matrix of the end position of the actuator calculated in step 4 before each step of driving The target displacement amount, and complete the movement of the actuator according to the given target displacement amount.
优选的，力反馈传感器的数据设置有阈值T，当本次运动后执行元件的受力反馈值超过阈值T，则触发学习网络权重W的更新，否则不进行权重更新。从而通过针对性更新学习网络达到不断减小受力，降低对周围挤压程度的目的。Preferably, the data of the force feedback sensor is set with a threshold T. When the force feedback value of the actuator exceeds the threshold T after this movement, the update of the learning network weight W is triggered; otherwise, no weight update is performed. In this way, the purpose of continuously reducing the force and reducing the degree of extrusion to the surroundings can be achieved through targeted updating of the learning network.
优选的，在连续一个旋转周期内未出现执行元件的受力反馈值超过阈值T 的情况，则认为学习网络的训练过程已完成。即经过学习已能够确保针尖在受力小于T的范围内进行远程运动中心运动，从而动态地提高远程运动中心运动的精准度。Preferably, if the force feedback value of the actuator does not exceed the threshold T within one continuous rotation cycle, it is considered that the training process of the learning network has been completed. That is, after learning, it has been able to ensure that the needle tip performs remote motion center movement within the range of force less than T, thereby dynamically improving the accuracy of remote motion center movement.
一种机械臂控制方法的自主学习***，用于实现上述的机械臂控制方法的自主学习方法，包括机械臂和控制器模块，所述控制器模块用于与机械臂电连接，获取力反馈传感器数据和控制所述机械臂的运动；所述控制器模块包括有用于执行运算的处理器。An autonomous learning system of a mechanical arm control method, used to realize the above-mentioned autonomous learning method of a mechanical arm control method, comprising a mechanical arm and a controller module, the controller module is used to electrically connect with a mechanical arm to obtain a force feedback sensor data and control the movement of the robotic arm; the controller module includes a processor for performing calculations.
与现有技术相比，有益效果是：对操作对象进行实际环境下的实时学习，通过以执行元件的受力反馈为依据进行远程运动中心的运动校准，此方法令执行元件与实际运行环境高度适配，能够将环境的形变对运动造成的偏差纳入考虑，动态地指导远程运动中心运动的调节，使机械臂在执行运动的精度大幅提高，减小机械臂实际运动与控制指令输出的标准运动之间的误差。Compared with the existing technology, the beneficial effect is: the real-time learning of the operating object in the actual environment, and the motion calibration of the remote motion center based on the force feedback of the actuator, this method makes the actuator highly consistent with the actual operating environment. Adaptation can take into account the deviation caused by the deformation of the environment, and dynamically guide the adjustment of the remote motion center motion, so that the accuracy of the robotic arm's motion execution can be greatly improved, and the standard motion between the actual motion of the robotic arm and the output of the control command can be reduced. error between.
图1是本发明的机械臂的结构示意图；Fig. 1 is the structural representation of mechanical arm of the present invention;
图2是本发明一种机械臂控制方法的自主学习方法的流程图。Fig. 2 is a flow chart of an autonomous learning method of a mechanical arm control method in the present invention.
附图仅用于示例性说明，不能理解为对本专利的限制；为了更好说明本实施例，附图某些部件会有省略、放大或缩小，并不代表实际产品的尺寸；对于本领域技术人员来说，附图中某些公知结构及其说明可能省略是可以理解的。附图中描述位置关系仅用于示例性说明，不能理解为对本专利的限制。The accompanying drawings are for illustrative purposes only, and should not be construed as limitations on this patent; in order to better illustrate this embodiment, certain components in the accompanying drawings will be omitted, enlarged or reduced, and do not represent the size of the actual product; for those skilled in the art It is understandable that some well-known structures and descriptions thereof may be omitted in the drawings. The positional relationship described in the drawings is for illustrative purposes only, and should not be construed as a limitation on this patent.
本发明实施例的附图中相同或相似的标号对应相同或相似的部件；在本发明的描述中，需要理解的是，若有术语“上”、“下”、“左”、“右”“长”“短”等指示的方位或位置关系为基于附图所示的方位或位置关系，仅是为了便于描述本发明和简化描述，而不是指示或暗示所指的装置或元件必须具有特定的方位、以特定的方位构造和操作，因此附图中描述位置关系的用语仅用于示例性说明，不能理解为对本专利的限制，对于本领域的普通技术人员而言，可以根据具体情况理解上述术语的具体含义。In the drawings of the embodiments of the present invention, the same or similar symbols correspond to the same or similar components; The orientation or positional relationship indicated by "long" and "short" are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the referred device or element must have a specific Orientation, construction and operation in a specific orientation, so the terms describing the positional relationship in the drawings are for illustrative purposes only, and should not be construed as limitations on this patent. For those of ordinary skill in the art, it can be understood according to specific circumstances The specific meaning of the above terms.
下面通过具体实施例，并结合附图，对本发明的技术方案作进一步的具体描述：Below by specific embodiment, in conjunction with accompanying drawing, the technical solution of the present invention is described in further detail:
实施例1Example 1
如图1-2所示为一种机械臂控制方法的自主学***轴的垂直距离dm、力反馈传感器201水平中心与执行元件2末端的距离Ltool、第一线性马达4和第二线性马达5的直线距离h以及对第一线性马达4、第二线性马达5和第三线性马达进行位置标定并归零。As shown in Figure 1-2, it is an embodiment of an autonomous learning method of a mechanical arm control method, which includes a mechanical arm 1 and an actuator 2 installed on the mechanical arm 1, and the mechanical arm 1 includes several driving actuators. The driving motor for the movement of the element 2; the actuator 2 also includes a force feedback sensor 201; the mechanical arm 1 includes a clamping part 3 for clamping the actuator 2, and a first linear motor for driving the clamping part 3 to rotate The motor 4 and the second linear motor 5, the first linear motor 4 and the second linear motor 5 are respectively hinged to the clamping part 3, and the clamping part 3 is provided with a device for driving the actuator 2 along the clamping part. Part 3 is a third linear motor 6 that performs linear motion; in the first step, the parameters of the mechanical arm 1 are measured when the mechanical arm 1 is initialized, including the third linear motor to the first linear motor 4 or the second linear motor The vertical distance dm of the horizontal axis of the motor 5, the distance Ltool between the horizontal center of the force feedback sensor 201 and the end of the actuator 2, the linear distance h between the first linear motor 4 and the second linear motor 5, and the distance between the first linear motor 4 and the second linear motor 4 The linear motor 5 and the third linear motor perform position calibration and return to zero.
机械臂1控制方法的自主学习方法包括如下步骤：The autonomous learning method of the control method of the mechanical arm 1 comprises the following steps:
步骤一：对机械臂1和力反馈传感器201进行初始化；Step 1: Initialize the mechanical arm 1 and the force feedback sensor 201;
步骤二：设定执行元件2的移动路径；Step 2: setting the movement path of the actuator 2;
步骤三：操作执行元件2沿着移动路径运动至远程运动中心点，并令执行元件2在给定旋转范围内完成一个旋转周期运动，获得远程运动中心点数据；给定旋转范围为所述第一线性马达4和所述第二线性马达5行程内达到的正负最大旋转角度[Θ，-Θ]；其中，执行元件2末端的垂直水平面状态定义为角度零点，一个旋转周期定义为执行元件2以角度零点为初始位置开始旋转至Θ角度，反向旋转经过角度零点到达-Θ角度，再反向旋转回到角度零点的运动周期。远程运动中心点义为一个3*k的矩阵，k为中心点分组总数，其中每一列表示一组中心点CL1
i,CL2
i,CL3
i，其中CL1、CL2、CL3分别为第一线性马达4、第二线性马达5和第三线性马达的标记位置；远程运动中心点的选取方法为，记录步骤三中执行元件2运动过程中第一线性马达4、第二线性马达5和第三线性马达的实际行程范围，具体为：[minL1,maxL1],[minL2,maxL2],[minL3,maxL3]，则k个远程运动中心点可表示为:
Step 3: Operate the actuator 2 to move to the remote motion center point along the movement path, and make the actuator 2 complete a rotation cycle movement within a given rotation range to obtain the remote motion center point data; the given rotation range is the first The positive and negative maximum rotation angle [Θ, -Θ] reached within the stroke of a linear motor 4 and the second linear motor 5; wherein, the vertical horizontal plane state at the end of the actuator 2 is defined as the zero point of the angle, and one rotation period is defined as the actuator 2 Start to rotate to the Θ angle with the zero point of the angle as the initial position, reverse the rotation through the zero point of the angle to reach the -Θ angle, and then reverse the motion cycle of returning to the zero point of the angle. The remote motion center point is defined as a 3*k matrix, k is the total number of center point groups, and each column represents a group of center points CL1 i , CL2 i , CL3 i , where CL1, CL2, and CL3 are the first linear motor 4 , the marked positions of the second linear motor 5 and the third linear motor; the method for selecting the remote motion center point is to record the first linear motor 4, the second linear motor 5 and the third linear motor during the movement of the actuator 2 in step three The actual travel range of , specifically: [minL1, maxL1], [minL2, maxL2], [minL3, maxL3], then the k remote motion center points can be expressed as:
步骤四：通过建立学习网络，给定学习步数n，执行元件2在移动路径上 开始进行n次旋转运动，每步运动后根据力反馈传感器201数据对远程运动中心点和学习网络进行更新。具体步骤如下：Step 4: By establishing a learning network and given the number of learning steps n, the actuator 2 starts to rotate n times on the moving path. After each step, the remote motion center point and the learning network are updated according to the data of the force feedback sensor 201. Specific steps are as follows:
S1：计算当前执行元件2末端与垂直面所成角度θS1: Calculate the angle θ formed between the end of the current actuator 2 and the vertical plane
其中，L1、L2分别为第一线性马达4、第二线性马达5的当前位置；Wherein, L1 and L2 are the current positions of the first linear motor 4 and the second linear motor 5 respectively;
S2：计算当前位置下的雅可比矩阵，雅可比矩阵计算公式如下：S2: Calculate the Jacobian matrix at the current position, the calculation formula of the Jacobian matrix is as follows:
其中，L3为第三线性马达3的当前位置；Wherein, L3 is the current position of the third linear motor 3;
S3：根据雅可比矩阵计算获得第一线性马达4、第二线性马达5和第三线性马达在本次训练中的预估相对位移，计算公式如下：S3: Calculate and obtain the estimated relative displacements of the first linear motor 4, the second linear motor 5, and the third linear motor in this training according to the Jacobian matrix, and the calculation formula is as follows:
其中，ΔL1、ΔL2、ΔL3分别为第一线性马达4、第二线性马达5和第三线性马达的预估位移量；x，y为RCM点的坐标x，y；Among them, ΔL1, ΔL2, ΔL3 are the estimated displacements of the first linear motor 4, the second linear motor 5 and the third linear motor respectively; x, y are the coordinates x, y of the RCM point;
S4：将预估位移量ΔL1、ΔL2、ΔL3与中心矩阵输入学习网络，根据权重矩阵W对预估位移量进行调整，获得目标位移量，计算方法如下：S4: Input the estimated displacement ΔL1, ΔL2, ΔL3 and the center matrix into the learning network, adjust the estimated displacement according to the weight matrix W, and obtain the target displacement. The calculation method is as follows:
其中，ΔL1＇、ΔL2＇、ΔL3＇分别为第一线性马达4、第二线性马达5和第三线性马达的目标位移量；S为远程运动中心点组成的矩阵；b为3*1矩阵，表示每个位移方向经学习网络修改的偏置值；Among them, ΔL1', ΔL2', ΔL3' are the target displacements of the first linear motor 4, the second linear motor 5 and the third linear motor respectively; S is a matrix composed of remote motion center points; b is a 3*1 matrix, Indicates the bias value modified by the learning network for each displacement direction;
S5：第一线性马达4、第二线性马达5和第三线性马达根据步骤S4获得的目标位移量进行移动，使得执行元件2与垂直面形成的角度形成新角度,具体为：S5: The first linear motor 4, the second linear motor 5 and the third linear motor move according to the target displacement obtained in step S4, so that the angle formed by the actuator 2 and the vertical plane forms a new angle, specifically:
S6：获取并解析力反馈传感器201数据，获得远程运动中心坐标方向分量：S6: Obtain and analyze the data of the force feedback sensor 201, and obtain the coordinate direction component of the remote motion center:
τ
x＝cos(θ)*τ'；
τ x =cos(θ)*τ';
τ
y＝sin(θ)*τ'；
τ y = sin(θ)*τ';
其中，τ'为执行元件2的受力反馈值；τ
x，τ
y分别为τ'在x方向和y方向的分解力；
Among them, τ' is the force feedback value of actuator 2; τ x , τ y are the decomposition forces of τ' in the x direction and y direction respectively;
S7：根据力反馈传感器201数据更新权重矩阵，更新公式如下：S7: update the weight matrix according to the data of the force feedback sensor 201, the update formula is as follows:
W'＝W-a*learnrate*(τ
x*J
1+τ
y*J
2)*S；
W'＝Wa*learnrate*(τ x *J 1 +τ y *J 2 )*S;
b＝a*learnrate*(τ
x*J
1+τ
y*J
2)
b＝a*learnrate*(τ x *J 1 +τ y *J 2 )
其中，W＇为更新后的权重矩阵；W为原权重矩阵；a为当前旋转方向标记a∈{1，-1}；J
1为雅克比矩阵第一列；J
2为雅克比矩阵第二列；
Among them, W' is the updated weight matrix; W is the original weight matrix; a is the current rotation direction mark a ∈ {1, -1}; J 1 is the first column of the Jacobian matrix; J 2 is the second column of the Jacobian matrix List;
S8：步骤S1-S7为学习网络的一个更新循环，下一个更新循环重新执行S1-S7，并以上一个更新循环所更新的参数作为学习网络的参数。即第n次循环中的S4的权重矩阵W是第n-1次循环中S7所得到的更新好的权重矩阵W＇。S8: Steps S1-S7 are an update cycle of the learning network, and S1-S7 is re-executed in the next update cycle, and the parameters updated in the previous update cycle are used as parameters of the learning network. That is, the weight matrix W of S4 in the nth cycle is the updated weight matrix W' obtained by S7 in the n-1th cycle.
进一步的，力反馈传感器201的数据设置有阈值T，当本次运动后执行元件2的受力反馈值超过阈值T，则触发学习网络权重W的更新，否则不进行权重更新。从而通过针对性更新学习网络达到不断减小受力，降低对周围挤压程度的目的。在连续一个旋转周期内未出现执行元件2的受力反馈值超过阈值T的情况，则认为学习网络的训练过程已完成。即经过学习已能够确保针尖在受力小于T的范围内进行远程运动中心运动，从而动态地提高远程运动中心运动的精准度。Further, the data of the force feedback sensor 201 is set with a threshold T. When the force feedback value of the actuator 2 exceeds the threshold T after this exercise, the update of the learning network weight W is triggered; otherwise, no weight update is performed. In this way, the purpose of continuously reducing the force and reducing the degree of extrusion to the surroundings can be achieved through targeted updating of the learning network. If the force feedback value of the actuator 2 does not exceed the threshold T within one continuous rotation cycle, it is considered that the training process of the learning network has been completed. That is, after learning, it has been able to ensure that the needle tip performs remote motion center movement within the range of force less than T, thereby dynamically improving the accuracy of remote motion center movement.
本实施例的工作原理：在机械臂1带动执行元件2沿着设定路径执行远程运动中心点运动的时候，通过学习网络输出执行元件2的运动位移量，然后根据执行元件2运动过程中力反馈传感器201的受力反馈来更新学习网络来重新调整执行元件2的运动位移量，使得执行元件2在运动的过程中不断的根据外部环境调整自身的位移，从而令执行元件2实际发生的运动与控制指令输出的运动尽量的保持一致。The working principle of this embodiment: when the mechanical arm 1 drives the actuator 2 to execute the remote motion center point movement along the set path, the movement displacement of the actuator 2 is output through the learning network, and then according to the force of the actuator 2 during the movement Feedback the force feedback of the sensor 201 to update the learning network to readjust the movement displacement of the actuator 2, so that the actuator 2 continuously adjusts its own displacement according to the external environment during the movement, so that the actual movement of the actuator 2 Keep as consistent as possible with the motion output by the control command.
在本实施例的有益效果：对操作对象进行实际环境下的实时学习，通过以执行元件2的受力反馈为依据进行远程运动中心的运动校准，此方法令执行元 件2的运动与实际运行环境高度适配，能够将环境的形变对运动造成的偏差纳入考虑，动态地指导远程运动中心运动的调节，使机械臂1在执行运动的精度大幅提高，减小机械臂1实际运动与控制指令输出的标准运动之间的误差。Beneficial effects of this embodiment: Real-time learning of the operating object in the actual environment, by performing motion calibration of the remote motion center based on the force feedback of the actuator 2, this method makes the motion of the actuator 2 consistent with the actual operating environment Highly adaptable, it can take into account the deviation caused by the deformation of the environment, and dynamically guide the adjustment of the remote motion center movement, so that the accuracy of the movement of the robot arm 1 is greatly improved, and the actual movement and control command output of the robot arm 1 are reduced The error between the standard movements.
实施例2Example 2
一种机械臂控制方法的自主学习***，用于实现实施例1的机械臂控制方法的自主学习方法，包括机械臂和控制器模块，所述控制器模块用于与机械臂电连接，获取力反馈传感器数据和控制所述机械臂的运动；所述控制器模块包括有用于执行运算的处理器。An autonomous learning system of a mechanical arm control method, used to realize the autonomous learning method of the mechanical arm control method of embodiment 1, comprising a mechanical arm and a controller module, the controller module is used to electrically connect with the mechanical arm to obtain force Feedback sensor data and control the movement of the robotic arm; the controller module includes a processor for performing calculations.
显然，本发明的上述实施例仅仅是为清楚地说明本发明所作的举例，而并非是对本发明的实施方式的限定。对于所属领域的普通技术人员来说，在上述说明的基础上还可以做出其它不同形式的变化或变动。这里无需也无法对所有的实施方式予以穷举。凡在本发明的精神和原则之内所作的任何修改、等同替换和改进等，均应包含在本发明权利要求的保护范围之内。Apparently, the above-mentioned embodiments of the present invention are only examples for clearly illustrating the present invention, rather than limiting the implementation of the present invention. For those of ordinary skill in the art, other changes or changes in different forms can be made on the basis of the above description. It is not necessary and impossible to exhaustively list all the implementation manners here. All modifications, equivalent replacements and improvements made within the spirit and principles of the present invention shall be included within the protection scope of the claims of the present invention.
Claims (10)
- 一种机械臂控制方法的自主学习方法，包括机械臂(1)和安装于所述机械臂(1)上的执行元件(2)，所述机械臂(1)包括若干个驱动执行元件(2)运动的驱动电机；其特征在于，所述执行元件(2)还包括力反馈传感器(201)，机械臂(1)控制方法的自主学习方法包括如下步骤：An autonomous learning method of a mechanical arm control method, comprising a mechanical arm (1) and an actuator (2) mounted on the robotic arm (1), the mechanical arm (1) including several driving actuators (2) ) a driving motor for motion; it is characterized in that, the actuator (2) also includes a force feedback sensor (201), and the self-learning method of the control method of the mechanical arm (1) includes the following steps:步骤一：对机械臂(1)和力反馈传感器(201)进行初始化；Step 1: Initializing the mechanical arm (1) and the force feedback sensor (201);步骤二：设定执行元件(2)的移动路径；Step 2: setting the moving path of the actuator (2);步骤三：操作执行元件(2)沿着移动路径运动至远程运动中心点，并令执行元件(2)在给定旋转范围内完成一个旋转周期运动，获得远程运动中心点数据；Step 3: Operate the actuator (2) to move to the remote motion center point along the moving path, and make the actuator (2) complete a rotation cycle within a given rotation range to obtain the remote motion center point data;步骤四：通过建立学习网络，给定学习步数n，执行元件(2)在移动路径上开始进行n次旋转运动，每步运动后根据力反馈传感器(201)数据对远程运动中心点和学习网络进行更新。Step 4: By establishing a learning network and given the number of learning steps n, the actuator (2) starts to perform n rotations on the moving path, and after each step moves, the remote motion center point and learning The network is updated.
- 根据权利要求1所述的一种机械臂(1)控制方法的自主学习方法，其特征在于，力反馈传感器(201)进行初始化的方法为，在无设定外力作用时，读取力反馈信号并定义为零点；在线学习过程中对力反馈信号进行差值转换，将与零点的差值作为输出。The self-learning method of a control method of a mechanical arm (1) according to claim 1, characterized in that the force feedback sensor (201) is initialized by reading the force feedback signal when there is no set external force And defined as the zero point; during the online learning process, the force feedback signal is converted into a difference, and the difference with the zero point is used as the output.
- 根据权利要求1所述的一种机械臂(1)控制方法的自主学***轴的垂直距离dm、力反馈传感器(201)水平中心与执行元件(2)末端的距离Ltool、第一线性马达(4)和第二线性马达(5)的直线距离h以及对第一线性马达(4)、第二线性马达(5)和第三线性马达进行位置标定并归零。The self-learning method of a control method of a mechanical arm (1) according to claim 1, characterized in that the mechanical arm (1) includes a clamping part (3) for clamping the actuator (2), The first linear motor (4) and the second linear motor (5) are used to drive the rotation of the clamping part (3). The third linear motor (6) of the holding portion (3) for linear motion; in the step one, the parameters of the mechanical arm (1) are measured when the mechanical arm (1) is initialized, including the third linear motor to The vertical distance dm of the horizontal axis of the first linear motor (4) or the second linear motor (5), the distance Ltool between the horizontal center of the force feedback sensor (201) and the end of the actuator (2), the first linear motor (4) and The linear distance h of the second linear motor (5) and the positions of the first linear motor (4), the second linear motor (5) and the third linear motor are calibrated and reset to zero.
- 根据权利要求3所述的一种机械臂(1)控制方法的自主学***面状态定义为角度零点，一个旋转周期定义为执行元件(2)以角度零点为初始位置开始旋转至Θ角度，反向旋转经过角度零点到达-Θ角度，再反向旋转回到角度零点的运动周期。The self-learning method of a control method of a mechanical arm (1) according to claim 3, characterized in that, in the third step, the given rotation range is the first linear motor (4) and the second linear motor (4) The positive and negative maximum rotation angle [Θ, -Θ] reached within the stroke of the two-linear motor (5); wherein, the vertical horizontal plane state at the end of the actuator (2) is defined as the zero point of the angle, and one rotation cycle is defined as the actuator (2) to The angle zero point is the movement period in which the initial position starts to rotate to the Θ angle, reversely rotates through the angle zero point to reach the -Θ angle, and then reversely rotates back to the angle zero point.
- 根据权利要求3所述的一种机械臂(1)控制方法的自主学习方法，其特征在于，远程运动中心点义为一个3*k的矩阵，k为中心点分组总数，其中每一列表示一组中心点(CL1 i,CL2 i,CL3 i)，其中CL1、CL2、CL3分别为第一线性马达(4)、第二线性马达(5)和第三线性马达的标记位置；远程运动中心点的选取方法为，记录步骤三中执行元件(2)运动过程中第一线性马达(4)、第二线性马达(5)和第三线性马达的实际行程范围，具体为：[minL1,maxL1],[minL2,maxL2],[minL3,maxL3]，则k个中心点可表示为: The self-learning method of a kind of mechanical arm (1) control method according to claim 3, it is characterized in that, remote motion center point is defined as a matrix of 3*k, and k is the center point group total number, and wherein each column represents a Group center points (CL1 i , CL2 i , CL3 i ), where CL1, CL2, and CL3 are the marker positions of the first linear motor (4), the second linear motor (5) and the third linear motor, respectively; remote motion center points The selection method is to record the actual stroke ranges of the first linear motor (4), the second linear motor (5) and the third linear motor during the movement of the actuator (2) in step 3, specifically: [minL1,maxL1] ,[minL2,maxL2],[minL3,maxL3], then k center points can be expressed as:
- 根据权利要求5所述的一种机械臂(1)控制方法的自主学习方法，其特征在于，所述步骤四的具体步骤如下：The self-learning method of a kind of mechanical arm (1) control method according to claim 5, is characterized in that, the concrete steps of described step four are as follows:S1：计算当前执行元件(2)末端与垂直面所成角度θS1: Calculate the angle θ formed between the end of the current actuator (2) and the vertical plane
其中，L1、L2分别为第一线性马达(4)、第二线性马达(5)的当前位置；Wherein, L1 and L2 are respectively the current positions of the first linear motor (4) and the second linear motor (5);S2：计算当前位置下的雅可比矩阵，雅可比矩阵计算公式如下：S2: Calculate the Jacobian matrix at the current position, the calculation formula of the Jacobian matrix is as follows:
其中，L3为第三线性马达的当前位置；Wherein, L3 is the current position of the third linear motor;S3：根据雅可比矩阵计算获得第一线性马达(4)、第二线性马达(5)和第三线性马达在本次训练中的预估相对位移，计算公式如下：S3: Calculate and obtain the estimated relative displacements of the first linear motor (4), the second linear motor (5) and the third linear motor in this training according to the Jacobian matrix, and the calculation formula is as follows:
其中，ΔL1、ΔL2、ΔL3分别为第一线性马达(4)、第二线性马达(5)和第三线性马达的预估位移量；x，y为RCM点的坐标(x，y)；Among them, ΔL1, ΔL2, and ΔL3 are the estimated displacements of the first linear motor (4), the second linear motor (5) and the third linear motor respectively; x, y are the coordinates (x, y) of the RCM point;S4：将预估位移量ΔL1、ΔL2、ΔL3与中心矩阵输入学习网络，根据权重矩阵W对预估位移量进行调整，获得目标位移量，计算方法如下：S4: Input the estimated displacement ΔL1, ΔL2, ΔL3 and the center matrix into the learning network, adjust the estimated displacement according to the weight matrix W, and obtain the target displacement. The calculation method is as follows:
其中，ΔL1＇、ΔL2＇、ΔL3＇分别为第一线性马达(4)、第二线性马达(5)和第三线性马达的目标位移量；S为远程运动中心点组成的矩阵；b为3*1矩阵，表示每个位移方向经学习网络修改的偏置值；Among them, ΔL1', ΔL2', ΔL3' are the target displacements of the first linear motor (4), the second linear motor (5) and the third linear motor respectively; S is a matrix composed of remote motion center points; b is 3 *1 matrix, representing the bias value modified by the learning network for each displacement direction;S5：第一线性马达(4)、第二线性马达(5)和第三线性马达根据步骤S4获得的目标位移量进行移动，使得执行元件(2)与垂直面形成的角度形成新角度,具体为：S5: The first linear motor (4), the second linear motor (5) and the third linear motor move according to the target displacement obtained in step S4, so that the angle formed by the actuator (2) and the vertical plane forms a new angle, specifically for:
S6：获取并解析力反馈传感器(201)数据，获得远程运动中心坐标方向分量：S6: Obtain and analyze the data of the force feedback sensor (201), and obtain the coordinate direction component of the remote motion center:τ x＝cos(θ)*τ'； τ x =cos(θ)*τ';τ y＝sin(θ)*τ'； τ y = sin(θ)*τ';其中，τ'为执行元件(2)的受力反馈值；τ x，τ y分别为τ'在x方向和y方向的分解力； Among them, τ' is the force feedback value of the actuator (2); τ x , τ y are the decomposition forces of τ' in the x direction and y direction respectively;S7：根据力反馈传感器(201)数据更新权重矩阵，更新公式如下：S7: update the weight matrix according to the data of the force feedback sensor (201), the update formula is as follows:W'＝W-a*learnrate*(τ x*J 1+τ y*J 2)*S； W'＝Wa*learnrate*(τ x *J 1 +τ y *J 2 )*S;b＝a*learnrate*(τ x*J 1+τ y*J 2) b＝a*learnrate*(τ x *J 1 +τ y *J 2 )其中，W＇为更新后的权重矩阵；W为原权重矩阵；a为当前旋转方向标记(a∈{1，-1})；J 1为雅克比矩阵第一列；J 2为雅克比矩阵第二列；learnrate意思为学习速率，该值一般设置为一个默认值，取决于实际使用中电机的行程 和速度单位； Among them, W' is the updated weight matrix; W is the original weight matrix; a is the current rotation direction mark (a∈{1,-1}); J 1 is the first column of the Jacobian matrix; J 2 is the Jacobian matrix The second column; learnrate means the learning rate, which is generally set to a default value, depending on the stroke and speed unit of the motor in actual use;S8：步骤S1-S7为学习网络的一个更新循环，下一个更新循环重新执行S1-S7，并以上一个更新循环所更新的参数作为学习网络的参数。S8: Steps S1-S7 are an update cycle of the learning network, and S1-S7 is re-executed in the next update cycle, and the parameters updated in the previous update cycle are used as parameters of the learning network. - 根据权利要求6所述的一种机械臂(1)控制方法的自主学习方法，其特征在于，所述执行元件(2)在进行远程运动中心运动时，每步驱动前根据所述步骤四计算得到的执行元件(2)末端位置的雅可比矩阵计算第一线性马达(4)、第二线性马达(5)和第三线性马达的目标位移量，并依据给定目标位移量完成执行元件(2)的运动。The self-learning method of a control method of a mechanical arm (1) according to claim 6, characterized in that, when the actuator (2) performs remote motion center movement, it is calculated according to the step 4 before driving each step Calculate the target displacements of the first linear motor (4), the second linear motor (5) and the third linear motor from the obtained Jacobian matrix of the end position of the actuator (2), and complete the actuator ( 2) Movement.
- 根据权利要求6所述的一种机械臂(1)控制方法的自主学习方法，其特征在于，力反馈传感器(201)的数据设置有阈值T，当本次运动后执行元件(2)的受力反馈值超过阈值T，则触发学习网络权重W的更新，否则不进行权重更新。The self-learning method of a control method of a mechanical arm (1) according to claim 6, characterized in that, the data of the force feedback sensor (201) is set with a threshold T, when the actuator (2) is affected by the force after this movement If the force feedback value exceeds the threshold T, the update of the learning network weight W is triggered, otherwise no weight update is performed.
- 根据权利要求8所述的一种机械臂(1)控制方法的自主学习方法，其特征在于，在连续一个旋转周期内未出现执行元件(2)的受力反馈值超过阈值T的情况，则认为学习网络的训练过程已完成。According to claim 8, an autonomous learning method of a control method for a mechanical arm (1), characterized in that, if the force feedback value of the actuator (2) does not exceed the threshold T within one continuous rotation cycle, then The training process of the learning network is considered complete.
- 一种机械臂控制方法的自主学习***，其特征在于，用于实现权利要求1-9任一所述的机械臂(1)控制方法的自主学习方法，包括机械臂(1)和控制器模块，所述控制器模块用于与机械臂(1)电连接，获取力反馈传感器(201)数据和控制所述机械臂(1)的运动；所述控制器模块包括有用于执行运算的处理器。An autonomous learning system of a mechanical arm control method, characterized in that, it is used to realize the autonomous learning method of the mechanical arm (1) control method described in any one of claims 1-9, comprising a mechanical arm (1) and a controller module , the controller module is used to electrically connect with the mechanical arm (1), acquire force feedback sensor (201) data and control the movement of the mechanical arm (1); the controller module includes a processor for performing calculations .
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| ZA2023/04394A ZA202304394B (en) | 2021-09-26 | 2023-04-13 | Autonomous learning method for mechanical arm control method |
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111130379.4A CN113715029B (en) | 2021-09-26 | 2021-09-26 | Autonomous learning method of mechanical arm control method |
| CN202111130379.4 | 2021-09-26 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2023046074A1 true WO2023046074A1 (en) | 2023-03-30 |
Family
ID=78684982
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/CN2022/120894 WO2023046074A1 (en) | 2021-09-26 | 2022-09-23 | Autonomous learning method for mechanical arm control method |
Country Status (3)
| Country | Link |
|---|---|
| CN (1) | CN113715029B (en) |
| WO (1) | WO2023046074A1 (en) |
| ZA (1) | ZA202304394B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116512254A (en) * | 2023-04-11 | 2023-08-01 | 中国人民解放军军事科学院国防科技创新研究院 | Direction-based intelligent control method and system for mechanical arm, equipment and storage medium |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113715029B (en) * | 2021-09-26 | 2022-10-11 | 广州市微眸医疗器械有限公司 | Autonomous learning method of mechanical arm control method |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2019013445A (en) * | 2017-07-06 | 2019-01-31 | リバーフィールド株式会社 | Manipulator system |
| CN110524544A (en) * | 2019-10-08 | 2019-12-03 | 深圳前海达闼云端智能科技有限公司 | A kind of control method of manipulator motion, terminal and readable storage medium storing program for executing |
| CN113334393A (en) * | 2021-08-06 | 2021-09-03 | 成都博恩思医学机器人有限公司 | Mechanical arm control method and system, robot and storage medium |
| US20210276187A1 (en) * | 2020-03-06 | 2021-09-09 | Embodied Intelligence Inc. | Trajectory optimization using neural networks |
| CN113715029A (en) * | 2021-09-26 | 2021-11-30 | 中山大学 | Autonomous learning method of mechanical arm control method |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101901580B1 (en) * | 2011-12-23 | 2018-09-28 | 삼성전자주식회사 | Surgical robot and control method thereof |
| NL2013369B1 (en) * | 2014-08-26 | 2016-09-26 | Univ Eindhoven Tech | Surgical robotic system and control of surgical robotic system. |
| CN108289719B (en) * | 2016-01-20 | 2021-09-14 | 直观外科手术操作公司 | System and method for rapidly pausing and resuming motion deviation in a medical device repositionable arm |
| EP3579780B1 (en) * | 2017-02-08 | 2024-07-17 | Intuitive Surgical Operations, Inc. | Repositioning system for a remotely controllable manipulator |
| CN110215339B (en) * | 2019-07-03 | 2022-01-07 | 中山大学 | Method for realizing path planning of automatic mechanical operation arm |
| CN112891058A (en) * | 2021-01-18 | 2021-06-04 | 中山大学 | Robot remote fixed-point control method for human eye subretinal injection |
-
2021
- 2021-09-26 CN CN202111130379.4A patent/CN113715029B/en active Active
-
2022
- 2022-09-23 WO PCT/CN2022/120894 patent/WO2023046074A1/en unknown
-
2023
- 2023-04-13 ZA ZA2023/04394A patent/ZA202304394B/en unknown
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2019013445A (en) * | 2017-07-06 | 2019-01-31 | リバーフィールド株式会社 | Manipulator system |
| CN110524544A (en) * | 2019-10-08 | 2019-12-03 | 深圳前海达闼云端智能科技有限公司 | A kind of control method of manipulator motion, terminal and readable storage medium storing program for executing |
| US20210276187A1 (en) * | 2020-03-06 | 2021-09-09 | Embodied Intelligence Inc. | Trajectory optimization using neural networks |
| CN113334393A (en) * | 2021-08-06 | 2021-09-03 | 成都博恩思医学机器人有限公司 | Mechanical arm control method and system, robot and storage medium |
| CN113715029A (en) * | 2021-09-26 | 2021-11-30 | 中山大学 | Autonomous learning method of mechanical arm control method |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116512254A (en) * | 2023-04-11 | 2023-08-01 | 中国人民解放军军事科学院国防科技创新研究院 | Direction-based intelligent control method and system for mechanical arm, equipment and storage medium |
| CN116512254B (en) * | 2023-04-11 | 2024-01-23 | 中国人民解放军军事科学院国防科技创新研究院 | Direction-based intelligent control method and system for mechanical arm, equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113715029A (en) | 2021-11-30 |
| CN113715029B (en) | 2022-10-11 |
| ZA202304394B (en) | 2023-12-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2023046074A1 (en) | Autonomous learning method for mechanical arm control method | |
| US9221176B2 (en) | Robot system and method for controlling the same | |
| US8909376B2 (en) | Robot hand and method of controlling the same | |
| US6393340B2 (en) | Robotic mechanisms with active and passive joints | |
| US20200206961A1 (en) | Backdrivable and haptic feedback capable robotic forceps, control system and method | |
| Afrisal et al. | Performance Analysis of 4-DOF RPRR Robot Manipulator Actuation Strategy for Pick and Place Application in Healthcare Environment | |
| CN115446847A (en) | System and method for improving 3D eye-hand coordination accuracy of a robotic system | |
| Cursi et al. | Optimization of surgical robotic instrument mounting in a macro–micro manipulator setup for improving task execution | |
| CN114176902B (en) | Robot automatic compensation motion control method | |
| Wang et al. | Task space compliant control and six-dimensional force regulation toward automated robotic ultrasound imaging | |
| Zhang et al. | An autonomous robotic flexible endoscope system with a DNA-inspired continuum mechanism | |
| CN113334381B (en) | Control method of movable decoupling continuum robot | |
| JP2020171989A (en) | Robot teaching system | |
| CN113742992A (en) | Master-slave control method based on deep learning and application | |
| CN110900608B (en) | Robot kinematics calibration method based on optimal measurement configuration selection | |
| CN115366115B (en) | Mechanical arm control method and control system thereof and computer readable storage medium | |
| Chen et al. | Model-based estimation of the gravity-loaded shape and scene depth for a slim 3-actuator continuum robot with monocular visual feedback | |
| Muñoz et al. | Pivoting motion control for a laparoscopic assistant robot and human clinical trials | |
| CN114452004B (en) | Control method for tail end position and posture of surgical robot | |
| WO2022198615A1 (en) | Calibration method and system for dual-arm robot puncture system | |
| Muñoz et al. | Adaptive cartesian motion control approach for a surgical robotic cameraman | |
| CN113298881B (en) | Spatial joint calibration method for monocular camera-IMU-mechanical arm | |
| CN110561426A (en) | Path planning method, device and control system of super-redundant mechanical arm | |
| CN113878583B (en) | Underwater mechanical arm control method and system | |
| Kuo et al. | A design for remote center of motion control using a software computing approach |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 22872118 Country of ref document: EP Kind code of ref document: A1 |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |