WO2020133880A1 - 一种工业机器人振动抑制方法 - Google Patents
一种工业机器人振动抑制方法 Download PDFInfo
- Publication number
- WO2020133880A1 WO2020133880A1 PCT/CN2019/086711 CN2019086711W WO2020133880A1 WO 2020133880 A1 WO2020133880 A1 WO 2020133880A1 CN 2019086711 W CN2019086711 W CN 2019086711W WO 2020133880 A1 WO2020133880 A1 WO 2020133880A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- robot
- vibration
- compensation amount
- vibration suppression
- speed
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
Definitions
- the invention belongs to the technical field of robots, relates to industrial robots, and is a vibration suppression method for industrial robots.
- Vibration suppression of industrial robots refers to the method of controlling vibrations during robot motions.
- the vibration suppression of the robot can be carried out from two aspects: mechanical structure and control algorithm.
- the mechanical structure of the robot is optimized by increasing the rigidity of the mechanical structure and the damping of the mechanical system, but this will increase the overall quality of the mechanical system, increase the energy consumption of the robot system, and easily affect the response speed of the system.
- the method cannot fundamentally solve the vibration problem of the robot.
- Robot control algorithms generally use the robot's kinematic model or dynamic model to improve the vibration suppression effect by selecting appropriate feedback parameters or control rates. Compared with optimizing the mechanical structure design, this type of method is easier to implement and has more Research and application.
- the method based on the dynamic model has the problem of inaccurate dynamic model. It is also difficult to ensure the accuracy of the model through the parameter identification method, especially for the flexible tandem industrial robot, so the dynamic model is used.
- the kinematics-based method is to add sensors at the end of the robot, such as acceleration sensors, laser trackers, etc. Such methods invisibly increase the cost of the robot, and the operation is complicated, especially for the use of acceleration sensors, the actual field application requires higher Calibration accuracy, otherwise it is counterproductive.
- the present invention provides a new method of vibration suppression for industrial robots.
- the problem to be solved by the present invention is that: in the existing vibration suppression method for robots, the use of peripheral devices increases the burden of robot energy consumption and affects the robot's motion.
- the use of control algorithms requires additional detection sensors, which increases costs and is complicated .
- the technical solution of the present invention is: an industrial robot vibration suppression method, which records a given position issued by the control system during the operation of the robot, and the actual position information fed back by the encoder, and calculates position compensation according to the deviation of the given position and the feedback position Volume and speed compensation, add position compensation to a given position, use speed feed-forward interface to add speed compensation to suppress vibration of the robot.
- the robot runs according to the preset motion trajectory, collects the given position and feedback position of the robot, calculates the position deviation, and uses the vibration signal filter to obtain the vibration signal for the position deviation , And then iteratively learn the vibration signal to obtain the learned position compensation amount, and then obtain the vibration suppression position compensation amount by multiplying by the coefficient function; at the same time, use a low-pass filter to filter the position deviation amount to remove the burr, and then obtain the speed by differential processing The amount of deviation, and multiplied by the gain to obtain the amount of speed compensation;
- the iterative learning is: add the last position compensation amount and the current given position to the robot servo system, and send the speed compensation amount to the servo speed loop through the servo feedforward interface, the robot runs to get the feedback position, The compensation amount is calculated again to form iterative learning;
- the robot repeats the operation several times until the robot feedback position indicates that the robot does not vibrate or stops within the acceptable range.
- the calculation and learning of the position compensation amount is specifically:
- the given position of each axis of the robot is ⁇ set
- the feedback position is ⁇ act
- the positional deviation ⁇ offs of the two is:
- ⁇ offs ⁇ set - ⁇ act * ⁇ (tt delay )
- t delay is the action delay
- ⁇ is the step function
- vibration signal filter uses the vibration signal filter to filter the position deviation ⁇ offs , extract the robot's vibration signal ⁇ vib , establish a PI-type iterative learning method, and learn and calculate the vibration signal and the previous vibration suppression compensation amount, first based on the filtered vibration signal ⁇ vib vibration signals calculated differential ⁇ 'vib:
- vibration suppression is not performed, otherwise vibration suppression is performed.
- the calculation of the speed compensation amount is:
- the given position of each axis of the robot is ⁇ set
- the feedback position is ⁇ act
- the positional deviation ⁇ offs of the two is:
- ⁇ offs ⁇ set - ⁇ act * ⁇ (tt delay )
- t delay is the action delay
- ⁇ is the step function
- Low-pass filter for filtering the positional deviation amount, the position deviation amount obtained after deburring ⁇ 'offs, the position deviation amount ⁇ ' offs differentiating, and multiplied by the speed gain K v, to give the amount of compensation velocity v comp And store the speed compensation amount for the next vibration learning or vibration suppression.
- the form of the speed gain K v includes a constant, a linear expression, and a non-linear expression.
- the implementation of the vibration signal filter includes a band-pass filter and a wavelet filter, and when the filter is used, it includes a time-domain filter and a frequency-domain filter.
- iterative learning includes P-type iterative learning, PI-type iterative learning, adaptive iterative learning, iterative learning based on frequency domain analysis, iterative learning based on 2-D theory, and optimal iterative learning.
- a combination of three methods of position control, speed feedforward, and torque feedforward can be used to suppress vibration of the robot.
- the present invention directly uses the position encoder of the industrial robot without adding an external sensor, and avoids the inaccuracy in calculating the position compensation amount and the speed compensation amount caused by the calibration error of the external sensor.
- the method of the present invention can not only solve the jitter during the robot motion, but also solve the positioning jitter.
- the invention repeatedly learns many times, and can calculate the learning compensation amount of each interpolation point of the full trajectory through the position deviation of each interpolation point of the robot trajectory to suppress the jitter during the movement and improve the trajectory accuracy of the robot. Calculate the amount of learning compensation for position deviation to solve the jitter during positioning.
- the present invention does not rely on the kinematics and dynamics models of the robot, but directly calculates the compensation through the deviation of the robot's movement position.
- the method is versatile and easy to implement.
- the present invention does not need to optimize the design of the robot mechanical structure, and can avoid increasing costs.
- the present invention reduces or even avoids the continuous effect of jitter by increasing the coefficient function, and optimizes the learning effect of vibration suppression.
- the present invention not only performs position compensation, but also effectively optimizes the learning effect of vibration suppression by increasing the speed feedforward and speed gain.
- the present invention is easy to implement, without changing the robot, and can improve the working efficiency of the robot, reduce the cost of the robot and the cost of the production line system.
- the vibration suppression compensation amount is obtained through learning in the present invention, if the motion trajectory of the robot does not change, the position compensation amount and the speed compensation amount can be used repeatedly without vibration suppression learning.
- FIG. 1 is a structural diagram of a robot vibration suppression system of the present invention.
- FIG. 3 is a schematic diagram of an industrial robot in an embodiment of the present invention.
- FIG. 4 is a schematic diagram of a given position and a feedback position of the industrial robot axis 1 in the embodiment of the present invention.
- FIG. 5 is a schematic diagram of the positional deviation between the given position of the industrial robot axis and the feedback position in the embodiment of the present invention.
- FIG. 6 is a schematic diagram of an industrial robot vibration signal in an embodiment of the present invention.
- FIG. 7 is a schematic diagram of the vibration compensation amount of an industrial robot in an embodiment of the present invention.
- FIG. 8 is a schematic diagram of a coefficient function in an embodiment of the present invention.
- FIG. 9 is a schematic diagram of the result of the speed compensation amount according to an embodiment of the present invention.
- the purpose of the present invention is to provide a method for suppressing vibration of industrial robots, by recording the actual position information given by the control system and the encoder feedback during the operation of the robot, and calculating the vibration suppression according to the designed vibration suppression system Position compensation amount and speed compensation amount. Finally, add the position compensation amount to the given position and add the speed compensation amount using the speed feedforward interface to achieve the vibration suppression of the robot.
- FIG. 1 shows a configuration diagram of a robot vibration suppression system according to an embodiment of the present invention.
- the motion kernel performs kinematics planning according to the expected position of the robot motion to obtain the given position ⁇ set of each axis of the robot.
- the expected position of the robot motion is set by online teaching or offline programming. This type of method is to set the expected pose (X, Y, Z, A, B, C) in the robot base coordinate space, where (X , Y, Z) is the position where the robot is expected to reach, (A, B, C) is the posture that the end of the robot is expected to reach.
- the servo system uses the position control regulator to obtain the speed command according to the given position ⁇ set issued by the motion controller.
- the speed control regulator obtains the current command according to the speed command.
- the current control regulator controls the power converter to output a certain voltage according to the current command. 3.
- the current signal is given to the servo motor to drive the axis of the robot.
- the servo motor acts according to the given position, and the servo system obtains the actual change angle of each axis by collecting the position encoder, that is, the feedback position ⁇ act .
- the vibration signal filter filters the deviation between the given position of the robot and the feedback position to obtain the vibration signal.
- the low-pass filter filters the deviation between the given position of the robot and the feedback position, which is used to remove the glitch signal in the deviation amount.
- the invention adopts an iterative learning method in vibration suppression to realize the tracking of vibration compensation, and the coefficient function is used to optimize the processing position compensation amount to obtain an accurate vibration suppression position compensation amount.
- learning is based on the vibration signal and the previous position compensation amount to obtain the learned position compensation amount.
- a position compensation memory can be set in the controller of the robot to store the position compensation amount of the robot.
- the compensation amount in the compensation amount file is read and stored in the position compensation memory.
- the position compensation memory temporarily stores the position compensation amount during the learning process.
- the vibration suppression learning is completed, the position compensation amount is backed up from the position compensation memory to a file, which is ready for the next system startup reading.
- a speed compensation memory can be set in the controller of the robot to store the speed compensation amount of the robot.
- the compensation amount in the compensation amount file is read and stored in the speed compensation memory.
- the speed compensation memory temporarily stores the speed compensation amount. After the vibration suppression learning is completed, the speed compensation amount is backed up to a file, which is ready for the next system startup reading.
- Fig. 2 is a flowchart of vibration suppression.
- the robot runs according to the preset motion trajectory, collects the given position and feedback position of the robot, calculates the position deviation, uses the vibration signal filter to obtain the vibration signal, and then iteratively learns through the learning system to obtain the learned position compensation, and then Through the coefficient function processing, the position compensation amount of vibration suppression is obtained, and the low-pass filter is used to filter the position deviation amount to remove the burr, and then the speed deviation amount is obtained by differential processing, and multiplied by the gain K v to obtain the speed compensation amount, the next learning The position compensation amount and the given position are added to the servo system, and the speed compensation amount is sent to the servo speed loop through the servo feedforward interface.
- the present invention uses position given and speed feed-forward to suppress the vibration of the robot, but the present invention is not limited to this mode. According to the system or actual application, a combination of position control, speed feed-forward, and torque feed-forward can be performed.
- the robot runs in accordance with a preset motion trajectory. As shown in FIG. 3, during the motion, the given position ⁇ set and feedback position information ⁇ act of each axis of the robot are collected. In this example, the data example of the robot axis 1 is given. The given position and feedback position of axis 1 are shown in Fig. 4. Unless otherwise specified, the following are the experimental data of axis 1.
- ⁇ offs ⁇ set - ⁇ act * ⁇ (tt delay )
- t delay is the action time lag
- ⁇ is the step function
- the vibration signal filter is used to filter the position deviation ⁇ offs to extract the vibration signal ⁇ vib of the robot.
- the parameter selection of the vibration signal filter is determined according to the output characteristics of the industrial robot. The example results are shown in Figure 6.
- ⁇ and ⁇ are the iterative learning gains and k is the number of learnings.
- the learning amount ⁇ k after iterative learning is multiplied by the coefficient function A to obtain the vibration compensation amount ⁇ comp .
- the vibration compensation amount is shown in FIG. 7.
- i is the number of robot joints
- i 1,...,m
- m is the total number of sampled data
- the coefficient function A is established based on the total number of sampled data.
- the composition of the coefficient function A needs to consider two types of motion acceleration and deceleration and uniform motion speed.
- the value of the coefficient function A is between [0,1], and the construction method can use the linear method and the nonlinear method.
- Figure 8 is an example. the way.
- the position deviation amount is filtered by a low-pass filter to obtain the position deviation amount ⁇ 'offs after removing the burr.
- Fig. 9 is an example result of the speed compensation amount.
- the position of the compensation amount of the robot ⁇ comp ⁇ set for a given position of the addition operation, the position of the next operation as the robot is given ⁇ 'set, ⁇ ' set ⁇ comp + ⁇ set.
- the speed compensation amount v comp is sent to the speed feedforward interface.
- the robot moves with the position given ⁇ 'set and the given speed feedforward compensation amount v comp , it will acquire a new feedback position ⁇ 'act , and then according to the above The process repeats the action learning until the robot does not vibrate or the vibration is within an acceptable range.
- the present invention provides a novel vibration suppression method.
- technical means can be replaced during the implementation process, including the following:
- the invention uses a position encoder of an industrial robot for vibration suppression.
- the robot's own encoder is used to extract an effective vibration signal, which is implemented using a band-pass filter, but it is not limited to the use of a band-pass filter, but also includes other methods.
- Filters for extracting vibration signals include time-domain filters and frequency-domain filters.
- the method of the present invention achieves vibration suppression through iterative learning.
- the present invention uses PI-type iterative learning control, but it is not limited to this method, and other methods such as adaptive iterative learning control and iterative learning control based on frequency domain analysis can also be used. , Iterative learning control based on 2-D theory, optimal iterative learning control, etc.
- the coefficient function is between [0,1], and the construction method can use the linear method and the nonlinear method.
- the invention optimizes the learning effect of vibration suppression by increasing the speed loop of the speed feedforward operation servo system and increasing the speed gain.
- Speed gain The example of the present invention provides a constant, but it is not limited to a constant, and other linear and nonlinear expressions can also be used.
- the suppression method of the present invention is applicable to robots of various position control systems.
- the present invention uses position given and speed feed-forward to suppress the vibration of the robot, but it is not limited to this mode.
- Position control and speed feed-forward can also be used. 3.
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Manipulator (AREA)
- Feedback Control In General (AREA)
Abstract
一种工业机器人振动抑制方法,记录机器人动作过程中控制系统下发的给定位置,及编码器反馈的实际位置信息,根据给定位置和反馈位置的偏差计算位置补偿量及速度补偿量,对给定位置加入位置补偿量,利用速度前馈接口加入速度补偿量,进行机器人的振动抑制。直接使用工业机器人的位置编码器,无需增加外部传感器,避免增加优化成本,并避免了额外增加外设所带来的数据误差,该方法既可以解决机器人动作过程中的抖动,也可以解决定位抖动。通过学习得到振动抑制补偿量后,若机器人的动作轨迹没有发生变化,位置补偿量及速度补偿量可以反复使用,无需再进行振动抑制学习。
Description
本发明属于机器人技术领域,涉及工业机器人,为一种工业机器人振动抑制方法。
工业机器人的振动抑制是指对机器人动作过程中的振动的控制方法。机器人的振动抑制可以从机械结构、控制算法两方面进行。通过增加机械结构的刚度、机械系统的阻尼以实现对机器人的机械结构优化,但这会增加机械系统的整体质量,使得机器人系统的能耗增大,也容易影响系统的响应速度,而且这种方法不能根本解决机器人的振动问题。机器人控制算法一般利用机器人的运动学模型或动力学模型,通过选择合适的反馈参数或控制率来提高振动抑制的效果,相比于优化机械结构设计,这类方法更容易实现,得到了更多的研究和应用。
在机器人控制算法中,基于动力学模型的方法存在动力学模型不精确的问题,通过参数辨识的方法也很难保证模型的精确性,尤其是对柔性的串联型工业机器人,因此利用动力学模型进行振动抑制的方式存在一定的问题。基于运动学的方法是在机器人的末端增加传感器,如加速度传感器、激光跟踪仪等,该类方法无形地增加了机器人成本,而且操作复杂,尤其对于使用加速度传感器,实际现场应用时需要更高的标定精度,否则适得其反。针对以上存在的问题,本发明提供了一种新型的工业机器人振动抑制的方法。
发明内容
本发明要解决的问题是:现有对机器人的振动抑制方法中,采用外设的方式增加机器人能耗负担,影响机器人运动,采用控制算法的方式需要另外设置检测传感器,增加成本,且设置复杂。
本发明的技术方案为:一种工业机器人振动抑制方法,记录机器人动作过程中控制系统下发的给定位置,及编码器反馈的实际位置信息,根据给定位置和反馈位置的偏差计算位置补偿量及速度补偿量,对给定位置加入位置补偿量,利用速度前馈接口加入速度补偿量,进行机器人的振动抑制。
进一步的,对振动抑制的过程进行学习,首先,机器人按照预设的动作轨迹运行,采集机器人的给定位置及反馈位置,通过计算位置偏差量,对位置偏差量使用振动信号 滤波器得到振动信号,再对振动信号进行迭代学习得到学习位置补偿量,然后通过乘以系数函数得到振动抑制的位置补偿量;同时,利用低通滤波器对位置偏差量进行滤波去除毛刺,再通过微分处理得到速度偏差量,并乘以增益得到速度补偿量;
所述迭代学习为:将上一次的位置补偿量与当前给定位置相加下发至机器人伺服系统,并将速度补偿量通过伺服前馈接口下发伺服速度环中,机器人运行得到反馈位置,再进行补偿量计算,形成迭代学习;
机器人重复多次运行,直至机器人反馈位置表明机器人不振动或振动达到接受范围内停止。
作为优选方式,位置补偿量的计算学习具体为:
机器人的各轴给定位置为θ
set,反馈位置为θ
act,两者的位置偏差量θ
offs为:
θ
offs=θ
set-θ
act*δ(t-t
delay)
其中,t
delay为动作时滞,δ为阶跃函数;
利用振动信号滤波器对位置偏差θ
offs进行滤波处理,提取机器人的振动信号θ
vib,建立PI型迭代学习方式,对振动信号及上次振动抑制补偿量进行学习计算,首先依据滤波后的振动信号θ
vib计算振动信号的微分θ'
vib:
θ
k=θ
comp-Φθ
vib-Γθ'
vib
其中Φ、Γ为迭代学习增益,k为学习次数;
对迭代学习后的学习量θ
k乘以系数函数A,得到振动补偿量θ
comp:
θ
comp,i=θ
k,i*A
i
其中,i为机器人关节数,i=1,…,m,m总数目,系数函数A依据采样数据的总数目建立,系数函数A的构成考虑机器人动作加速及减速、动作匀速两种类型段,取值在[0,1]之间,构建方式包括使用线性方式和非线性方式,同时保证系数函数的元素单调不递增。
进一步的,得到位置偏差量后,对位置偏差量进行判断处理,若没有发生振动或者振动小于设定的阈值,则不进行振动抑制,否则进行振动抑制。
作为优选方式,速度补偿量的计算学习为:
机器人的各轴给定位置为θ
set,反馈位置为θ
act,两者的位置偏差量θ
offs为:
θ
offs=θ
set-θ
act*δ(t-t
delay)
其中,t
delay为动作时滞,δ为阶跃函数;
利用低通滤波器对位置偏差量进行滤波处理,得到去除毛刺后的位置偏差量θ'
offs,将位置偏差量θ'
offs进行微分处理,并乘以速度增益K
v,得到速度补偿量v
comp,并将速度补偿量进行存储,供下次振动学习或者振动抑制使用。
本发明方法中,速度增益K
v的形式包括常数、线性表达式及非线性表达式。
本发明方法中,振动信号滤波器的实现方式包括带通滤波器和小波滤波,使用滤波器时,包括时域滤波器和频域滤波器。
本发明方法中,迭代学习包括P型迭代学习、PI型迭代学习、自适应迭代学习、基于频域分析的迭代学习、基于2-D理论的迭代学习、和最优化迭代学习。
本发明方法中,作为替换方式,还可根据位置补偿量及速度补偿量,使用位置控制、速度前馈、力矩前馈三种方式组合,对机器人进行振动抑制。
与现有技术相比,本发明方法的有益效果如下:
(1)本发明直接使用工业机器人的位置编码器,无需增加外部传感器,避免了外部传感器标定误差带来计算位置补偿量和速度补偿量的不精确性。
(2)本发明方法既可以解决机器人动作过程中的抖动,也可以解决定位抖动。本发明多次反复学习,可以通过机器人运动轨迹各个插补点的位置偏差计算全轨迹各个插补点的学习补偿量抑制运动过程中的抖动,提高机器人的轨迹精度,也可以通过机器人定位时的位置偏差计算学习补偿量,解决定位时的抖动。
(3)本发明没有依据机器人的运动学和动力学模型,而是直接通过机器人运动位置偏差来计算补偿,方法通用性强,易于实现。
(4)本发明无需优化机器人机械结构的设计,可以避免增加成本。
(5)本发明通过增加系数函数,减少甚至避免了抖动的连续影响,优化振动抑制的学习效果。
(6)对于振动产生的位置偏差,本发明不仅进行位置补偿,还通过增加速度前馈及速度增益,有效地优化振动抑制的学习效果。
(7)本发明易于实现,无需对机器人进行改动,能够提高机器人的工作效率、降低机器人成本及生产线系统成本。
(8)本发明通过学习得到振动抑制补偿量后,若机器人的动作轨迹没有发生变化,位置补偿量及速度补偿量可以反复使用,无需再进行振动抑制学习。
图1为本发明机器人振动抑制系统的结构图。
图2为本发明振动抑制方法的流程图。
图3为本发明实施例中工业机器人的示意图。
图4为本发明实施例中工业机器人轴1的给定位置及反馈位置示意图。
图5为本发明实施例中工业机器人轴给定位置与反馈位置的位置偏差示意图。
图6为本发明实施例中工业机器人振动信号示意图。
图7为本发明实施例中工业机器人振动补偿量示意图。
图8为本发明实施例中系数函数的示意图。
图9为本发明实施例速度补偿量的结果示意图。
本发明的目的是提供一种工业机器人振动抑制的方法,通过记录机器人动作过程中的控制系统下发给定位置及编码器反馈的实际位置信息,并依据设计的振动抑制系统计算得到振动抑制的位置补偿量及速度补偿量,最后通过对给定位置加入位置补偿量及利用速度前馈接口加入速度补偿量,以实现机器人的振动抑制。
以下,参照附图,说明本发明的实施例所涉及的机器人振动抑制系统。图1表示本发明的实施例所涉及的机器人振动抑制系统的结构图。
运动内核根据机器人动作的预期位置进行运动学规划,以得到机器人的各轴给定位置θ
set。机器人动作的预期位置通过在线示教或离线编程的方式设定,该类方式是在机器人基坐标空间设定预期的位姿(X、Y、Z、A、B、C),其中,(X、Y、Z)为机器人预期到达的位置,(A、B、C)为机器人末端预期到达的姿态。
伺服系统根据运动控制器下发的给定位置θ
set,利用位置控制调节器得到速度指令,速度控制调节器依据速度指令得到电流指令,电流控制调节器依据电流指令控制功率变换器输出一定的电压、电流信号给伺服电机以驱动机器人的轴动作。在位置控制模式下,伺服电机依据给定位置进行动作,伺服系统通过采集位置编码器得到各轴的实际变化角度,即反馈位置θ
act。
在计算位置补偿量时,振动信号滤波器对机器人的给定与反馈位置的偏差进行滤波处理,以得到振动信号。
在计算速度补偿量时,由低通滤波器对机器人的给定与反馈位置的偏差进行滤波处 理,用于去除偏差量中的毛刺信号。
本发明在振动抑制中采用迭代学习方式,实现对振动补偿的跟踪,系数函数用于对学习位置补偿量进行优化处理,以得到精准的振动抑制位置补偿量。通过迭代学习方式,依据振动信号及上一次的位置补偿量进行学习,得到学习位置补偿量。
本发明实施时可以在机器人的控制器中设置位置补偿存储器,用于储存机器人的位置补偿量,机器人上电启动时将补偿量文件中的补偿量读取并存放至位置补偿存储器中,当需要进行振动抑制学习时,位置补偿存储器暂存学习过程中的位置补偿量,当振动抑制学习结束后,将位置补偿量从位置补偿存储器中备份至文件,已备下次系统启动读取。
本发明实施时可以在机器人的控制器中设置速度补偿存储器,用于储存机器人的速度补偿量,机器人上电启动时将补偿量文件中的补偿量读取并存放至速度补偿存储器中,当需要进行振动抑制学习时,速度补偿存储器暂存速度补偿量,当振动抑制学习结束后,将速度补偿量备份至文件,已备下次系统启动读取。
接下来说明机器人利用本发明方法进行振动抑制的流程。
图2为振动抑制的流程图。首先机器人按照预先设定的动作轨迹运行,采集机器人的给定位置及反馈位置,通过计算位置偏差量,使用振动信号滤波器得到振动信号,再通过学习系统进行迭代学习得到学习位置补偿量,然后通过系数函数处理得到振动抑制的位置补偿量,并利用低通滤波器对位置偏差量进行滤波去除毛刺,再通过微分处理得到速度偏差量,并乘以增益K
v得到速度补偿量,下次学习时将位置补偿量与给定位置相加下发至伺服系统,以及将速度补偿量通过伺服前馈接口下发伺服速度环中,重复多次运行直至机器人不振动或振动达到接受范围内停止。本发明使用位置给定及速度前馈的方式进行机器人的振动抑制,但本发明不限于该模式的情况,根据系统或实际应用,可进行位置控制、速度前馈、力矩前馈方式的组合。
机器人按照预先设定的动作轨迹运行,如图3所示,在动作过程中,采集机器人各轴的给定位置θ
set及反馈位置信息θ
act,本实例中给出机器人轴1的数据实例,轴1的给定位置及反馈位置如图4所示,以下如不进行特殊说明,均为轴1的实验数据。
计算机器人轴给定位置θ
set与反馈位置θ
act的位置偏差量θ
offs,如图5所示。
θ
offs=θ
set-θ
act*δ(t-t
delay)
其中,t
delay为动作时滞,δ为阶跃函数。
对位置偏差量进行判断处理,若没有发生振动或者振动小于设定的阈值,则不进行振动抑制的学习,否则继续以下流程处理。
利用振动信号滤波器对位置偏差θ
offs进行滤波处理,以提取机器人的振动信号θ
vib。振动信号滤波器的参数选择依据工业机器人的输出特性确定,实例结果如图6所示。
建立PI型迭代学习方式,对振动信号及上次振动抑制补偿量进行学习计算,具体流程如下所示。
首先依据滤波后的振动信号θ
vib计算振动信号的微分θ'
vib,具体学习的运算方式如下:
其中Φ、Γ为迭代学习增益,k为学习次数。
对迭代学习后的学习量θ
k乘以系数函数A,以得到振动补偿量θ
comp,振动补偿量如图7所示。
θ
comp,i=θ
k,i*A
i
其中,i为机器人关节数,i=1,…,m,m为采样数据的总数目,系数函数A依据采样数据的总数目建立。系数函数A的构成需要考虑动作加速及减速、动作匀速两种类型段。系数函数A的取值在[0,1]之间,构建方式可以使用线性方式、非线性方式,但建立两种类型段时,需要保证系数函数的元素单调不递增,图8为一种实例方式。
利用低通滤波器对位置偏差量进行滤波处理,得到去除毛刺后的位置偏差量θ'
offs。
将位置偏差量θ'
offs进行微分处理,并乘以速度增益K
v,得到速度补偿量v
comp,并将速度补偿量存储于速度补偿存储器中,供下次学习或者振动抑制使用。图9为速度补偿量的实例结果。
将位置补偿量θ
comp与机器人的给定位置θ
set进行相加运算,作为机器人的下次运行的位置给定θ'
set,θ'
set=θ
comp+θ
set。将速度补偿量v
comp下发至速度前馈接口,当机器人以位置给定θ'
set及给定速度前馈补偿量v
comp动作后,会采集得到新的反馈位置θ'
act,再依据以上的流程重复动作学习,直至机器人不振动或振动在可接受的范围内。
本发明提供了一种新型振动抑制方法,除了上述实施例的具体方式,在实施过程中还可进行技术手段的替换,包括以下:
本发明使用工业机器人的位置编码器进行振动抑制,流程处理过程中通过机器人自身的编码器提取有效的振动信号,使用带通滤波器实现,但不仅限于使用带通滤波器,还包含其他方式,如小波处理方法。提取振动信号的滤波器包含时域滤波器和频域滤波器。
本发明方法通过迭代学习的方式实现振动抑制,本发明使用的是PI型迭代学习控制,但不限于该方式,还可使用其他方式,如自适应迭代学习控制、基于频域分析的迭代学习控制、基于2-D理论的迭代学习控制、最优化迭代学习控制等。
本发明在位置补偿时,通过增加系数函数,减少甚至避免了抖动的连续影响,优化振动抑制的学习效果。系数函数的取值在[0,1]之间,构建方式可以使用线性方式、非线性方式,但建立两种类型段时,需要保证系数函数的元素单调不递增。
本发明通过增加速度前馈操作伺服系统的速度环,并增加速度增益,优化振动抑制的学习效果。速度增益本发明实例提供的是一个常数,但不仅限于常数,还可使用其他线性及非线性表达式。
本发明的抑制方法适用于各种位置控制系统的机器人,本发明使用位置给定及速度前馈的方式进行机器人的振动抑制,但不限于该模式的情况,还可使用位置控制、速度前馈、力矩前馈方式的组合。
Claims (9)
- 一种工业机器人振动抑制方法,其特征是记录机器人动作过程中控制系统下发的给定位置,及编码器反馈的实际位置信息,根据给定位置和反馈位置的偏差计算位置补偿量及速度补偿量,对给定位置加入位置补偿量,利用速度前馈接口加入速度补偿量,进行机器人的振动抑制。
- 根据权利要求1所述的一种工业机器人振动抑制方法,其特征是对振动抑制的过程进行学习,首先,机器人按照预设的动作轨迹运行,采集机器人的给定位置及反馈位置,通过计算位置偏差量,对位置偏差量使用振动信号滤波器得到振动信号,再对振动信号进行迭代学习得到学习位置补偿量,然后通过乘以系数函数得到振动抑制的位置补偿量;同时,利用低通滤波器对位置偏差量进行滤波去除毛刺,再通过微分处理得到速度偏差量,并乘以增益得到速度补偿量;所述迭代学习为:将上一次的位置补偿量与当前给定位置相加下发至机器人伺服系统,并将速度补偿量通过伺服前馈接口下发伺服速度环中,机器人运行得到反馈位置,再进行补偿量计算,形成迭代学习;机器人重复多次运行,直至机器人反馈位置表明机器人不振动或振动达到接受范围内停止。
- 根据权利要求1或2所述的一种工业机器人振动抑制方法,其特征是位置补偿量的计算学习具体为:机器人的各轴给定位置为θ set,反馈位置为θ act,两者的位置偏差量θ offs为:θ offs=θ set-θ act*δ(t-t delay)其中,t delay为动作时滞,δ为阶跃函数;利用振动信号滤波器对位置偏差θ offs进行滤波处理,提取机器人的振动信号θ vib,建立PI型迭代学习方式,对振动信号及上次振动抑制补偿量进行学习计算,首先依据滤波后的振动信号θ vib计算振动信号的微分θ' vib:θ k=θ comp-Φθ vib-Γθ' vib其中Φ、Γ为迭代学习增益,k为学习次数;对迭代学习后的学习量θ k乘以系数函数A,得到振动补偿量θ comp:θ comp,i=θ k,i*A i其中,i为机器人关节数,i=1,…,m,m总数目,系数函数A依据采样数据的总数目建立,系数函数A的构成考虑机器人动作加速及减速、动作匀速两种类型段,取值在 [0,1]之间,构建方式包括使用线性方式和非线性方式,同时保证系数函数的元素单调不递增。
- 根据权利要求2所述的一种工业机器人振动抑制方法,其特征是得到位置偏差量后,对位置偏差量进行判断处理,若没有发生振动或者振动小于设定的阈值,则不进行振动抑制,否则进行振动抑制。
- 根据权利要求1或2所述的一种工业机器人振动抑制方法,其特征是速度补偿量的计算学习为:机器人的各轴给定位置为θ set,反馈位置为θ act,两者的位置偏差量θ offs为:θ offs=θ set-θ act*δ(t-t delay)其中,t delay为动作时滞,δ为阶跃函数;利用低通滤波器对位置偏差量进行滤波处理,得到去除毛刺后的位置偏差量θ' offs,将位置偏差量θ' offs进行微分处理,并乘以速度增益K v,得到速度补偿量v comp,并将速度补偿量进行存储,供下次振动学习或者振动抑制使用。
- 根据权利要求5所述的一种工业机器人振动抑制方法,其特征是速度增益K v的形式包括常数、线性表达式及非线性表达式。
- 根据权利要求1或2所述的一种工业机器人振动抑制方法,其特征是振动信号滤波器的实现方式包括带通滤波器和小波滤波,使用滤波器时,包括时域滤波器和频域滤波器。
- 根据权利要求1或2所述的一种工业机器人振动抑制方法,其特征是迭代学习包括P型迭代学习、PI型迭代学习、自适应迭代学习、基于频域分析的迭代学习、基于2-D理论的迭代学习、和最优化迭代学习等迭代学习控制方法。
- 根据权利要求1或2所述的一种工业机器人振动抑制方法,其特征是根据位置补偿量及速度补偿量,使用位置控制、速度前馈、力矩前馈三种方式组合,对机器人进行振动抑制。
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811642227 | 2018-12-29 | ||
CN201811642227.0 | 2018-12-29 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2020133880A1 true WO2020133880A1 (zh) | 2020-07-02 |
Family
ID=66563531
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2019/086711 WO2020133880A1 (zh) | 2018-12-29 | 2019-05-13 | 一种工业机器人振动抑制方法 |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN109799701B (zh) |
WO (1) | WO2020133880A1 (zh) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114185274A (zh) * | 2021-12-06 | 2022-03-15 | 东北大学 | 基于迭代学习的钢铁生产过程重复性误差补偿控制方法 |
CN114185370A (zh) * | 2020-08-24 | 2022-03-15 | 广东博智林机器人有限公司 | 一种伺服系统及其转速补偿方法 |
CN114505844A (zh) * | 2022-01-24 | 2022-05-17 | 华南理工大学 | 一种面向工业机器人的残余振动抑制系统及方法 |
CN114800498A (zh) * | 2022-04-20 | 2022-07-29 | 上海捷勃特机器人有限公司 | 一种scara机器人力矩前馈补偿方法 |
CN117681252A (zh) * | 2024-01-12 | 2024-03-12 | 华中科技大学 | 一种柔顺末端力控执行器振动抑制方法、系统及设备 |
Families Citing this family (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110465765B (zh) * | 2019-08-19 | 2021-02-12 | 南京埃斯顿机器人工程有限公司 | 一种焊接机器人正弦摆焊的启停抖动削弱方法 |
CN111152213B (zh) * | 2019-12-05 | 2021-03-16 | 北京蒂斯科技有限公司 | 一种基于混合控制的机械臂振动补偿方法及装置 |
CN114619483A (zh) * | 2020-12-08 | 2022-06-14 | 山东新松工业软件研究院股份有限公司 | 基于轨迹校正的柔性机器人振动抑制方法及其系统 |
CN114460838A (zh) * | 2021-12-31 | 2022-05-10 | 上海新时达机器人有限公司 | 一种机械末端抖动抑制方法、位置环和驱动装置 |
CN116690595A (zh) * | 2022-01-04 | 2023-09-05 | 陈威 | 一种使用多峰输入整形法的抑振机器人 |
CN115674208B (zh) * | 2022-12-30 | 2023-05-26 | 广东隆崎机器人有限公司 | 一种机器人振动抑制装置、控制方法及其机器人 |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2004071717A1 (en) * | 2003-02-13 | 2004-08-26 | Abb Ab | A method and a system for programming an industrial robot to move relative to defined positions on an object, including generation of a surface scanning program |
CN105773623A (zh) * | 2016-04-29 | 2016-07-20 | 江南大学 | 基于预测型间接迭代学习的scara机器人轨迹跟踪控制方法 |
CN105773622A (zh) * | 2016-04-29 | 2016-07-20 | 江南大学 | 一种基于iekf的工业机器人绝对精度校准方法 |
CN107433589A (zh) * | 2017-07-28 | 2017-12-05 | 珞石(山东)智能科技有限公司 | 基于加速度传感器的机器人振动抑制方法 |
CN107942680A (zh) * | 2017-12-20 | 2018-04-20 | 南京埃斯顿自动控制技术有限公司 | 一种机器人抖动抑制方法 |
Family Cites Families (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0268491A3 (en) * | 1986-11-20 | 1988-08-03 | Unimation Inc. | Multiaxis robot having improved motion control |
JP2005301508A (ja) * | 2004-04-08 | 2005-10-27 | Fanuc Ltd | 制御装置 |
CN101623867B (zh) * | 2008-07-11 | 2010-12-01 | 中国科学院沈阳自动化研究所 | 一种使机器人高精度跟踪指定路径的设备和方法 |
JP6347595B2 (ja) * | 2013-11-25 | 2018-06-27 | キヤノン株式会社 | ロボット制御方法、及びロボット制御装置 |
CN104400783B (zh) * | 2014-09-24 | 2016-08-24 | 哈尔滨工业大学 | 张力和关节位置反馈机器人关节用挠性驱动单元控制方法 |
CN104898427B (zh) * | 2015-05-19 | 2017-06-23 | 浙江工业大学 | 电机伺服系统的分段滤波迭代学习控制方法 |
JP6717768B2 (ja) * | 2017-03-09 | 2020-07-01 | ファナック株式会社 | 生産ラインにおける運用を考慮した学習制御を行うロボット及びその制御方法 |
CN107070341B (zh) * | 2017-03-24 | 2019-05-10 | 中国科学院长春光学精密机械与物理研究所 | 基于鲁棒迭代学习控制的永磁同步电机转矩脉动抑制方法 |
JP6499720B2 (ja) * | 2017-06-22 | 2019-04-10 | ファナック株式会社 | 機械学習装置、サーボ制御装置、サーボ制御システム、及び機械学習方法 |
CN107370433B (zh) * | 2017-09-05 | 2019-11-08 | 南京工程学院 | 基于pid自学习时滞补偿器的电流谐波补偿方法 |
CN108267959B (zh) * | 2018-01-31 | 2021-06-08 | 珞石(北京)科技有限公司 | 基于迭代学习控制和输入整形技术的联合抑制振动的方法 |
-
2019
- 2019-03-14 CN CN201910194468.1A patent/CN109799701B/zh active Active
- 2019-05-13 WO PCT/CN2019/086711 patent/WO2020133880A1/zh active Application Filing
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2004071717A1 (en) * | 2003-02-13 | 2004-08-26 | Abb Ab | A method and a system for programming an industrial robot to move relative to defined positions on an object, including generation of a surface scanning program |
CN105773623A (zh) * | 2016-04-29 | 2016-07-20 | 江南大学 | 基于预测型间接迭代学习的scara机器人轨迹跟踪控制方法 |
CN105773622A (zh) * | 2016-04-29 | 2016-07-20 | 江南大学 | 一种基于iekf的工业机器人绝对精度校准方法 |
CN107433589A (zh) * | 2017-07-28 | 2017-12-05 | 珞石(山东)智能科技有限公司 | 基于加速度传感器的机器人振动抑制方法 |
CN107942680A (zh) * | 2017-12-20 | 2018-04-20 | 南京埃斯顿自动控制技术有限公司 | 一种机器人抖动抑制方法 |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114185370A (zh) * | 2020-08-24 | 2022-03-15 | 广东博智林机器人有限公司 | 一种伺服系统及其转速补偿方法 |
CN114185274A (zh) * | 2021-12-06 | 2022-03-15 | 东北大学 | 基于迭代学习的钢铁生产过程重复性误差补偿控制方法 |
CN114185274B (zh) * | 2021-12-06 | 2023-07-04 | 东北大学 | 基于迭代学习的钢铁生产过程重复性误差补偿控制方法 |
CN114505844A (zh) * | 2022-01-24 | 2022-05-17 | 华南理工大学 | 一种面向工业机器人的残余振动抑制系统及方法 |
CN114505844B (zh) * | 2022-01-24 | 2023-06-20 | 华南理工大学 | 一种面向工业机器人的残余振动抑制系统及方法 |
CN114800498A (zh) * | 2022-04-20 | 2022-07-29 | 上海捷勃特机器人有限公司 | 一种scara机器人力矩前馈补偿方法 |
CN117681252A (zh) * | 2024-01-12 | 2024-03-12 | 华中科技大学 | 一种柔顺末端力控执行器振动抑制方法、系统及设备 |
CN117681252B (zh) * | 2024-01-12 | 2024-05-28 | 华中科技大学 | 一种柔顺末端力控执行器振动抑制方法、系统及设备 |
Also Published As
Publication number | Publication date |
---|---|
CN109799701A (zh) | 2019-05-24 |
CN109799701B (zh) | 2022-04-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2020133880A1 (zh) | 一种工业机器人振动抑制方法 | |
JP6774637B2 (ja) | 制御装置及び制御方法 | |
CN109274314B (zh) | 机器学习装置、伺服电动机控制装置、伺服电动机控制系统以及机器学习方法 | |
JP6490127B2 (ja) | 機械学習装置、サーボ制御装置、サーボ制御システム、及び機械学習方法 | |
US10824121B2 (en) | Machine learning device, servo motor controller, servo motor control system, and machine learning method | |
CN105773623B (zh) | 基于预测型间接迭代学习的scara机器人轨迹跟踪控制方法 | |
CN105772917B (zh) | 一种三关节点焊机器人轨迹跟踪控制方法 | |
JP4673326B2 (ja) | 回転軸の位置制御装置 | |
JP6499720B2 (ja) | 機械学習装置、サーボ制御装置、サーボ制御システム、及び機械学習方法 | |
CN109927032A (zh) | 一种基于高阶滑模观测器的机械臂轨迹跟踪控制方法 | |
CN109669345B (zh) | 基于eso的水下机器人模糊pid运动控制方法 | |
CN110948504B (zh) | 机器人加工作业法向恒力跟踪方法和装置 | |
US11029650B2 (en) | Machine learning device, control system, and machine learning method | |
US11087509B2 (en) | Output device, control device, and evaluation function value output method | |
CN105549383B (zh) | 直线电机精密轨迹跟踪装置及方法 | |
JP2014136260A (ja) | 制御装置 | |
CN113765453B (zh) | 宽-窄极特征的磁悬浮开关磁阻电机悬浮控制系统 | |
CN110881107A (zh) | 一种基于神经网络的增稳云台控制方法 | |
CN107490958A (zh) | 一种五自由度混联机器人的模糊自适应控制方法 | |
WO2018209860A1 (zh) | 解耦控制方法、装置、机器人及存储介质 | |
US11243501B2 (en) | Machine learning device, control system, and machine learning | |
CN111152213B (zh) | 一种基于混合控制的机械臂振动补偿方法及装置 | |
JP3936242B2 (ja) | サーボモータ制御用コントローラにおけるゲイン設定法、コントローラの有効性検証法およびロボット制御法 | |
WO2020133881A1 (zh) | 机械设备的控制学习方法和具备学习功能的机械设备控制学习系统 | |
CN116079740A (zh) | 一种基于观测器的机器人变阻抗加工方法 |
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: 19904662 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 19904662 Country of ref document: EP Kind code of ref document: A1 |