CN104476544A - Self-adaptive dead zone inverse model generating device of visual servo mechanical arm system - Google Patents

Abstract

The invention discloses a device for generating an adaptive dead zone inverse model of a visual servo manipulator system. The visual servo manipulator system is composed of a visual servo controller, a vision module, a motion control module, a drive module, a six-degree-of-freedom manipulator, a torque feedback module, a speed and position acquisition module, and a detection module; the above-mentioned visual servo controller The adaptive dead zone inverse model generating device in the system is used to eliminate the dead zone nonlinear constraints, and the device includes: dead zone inverse model module, adaptive module, and operation control module; the dead zone inverse model module is used to construct a smooth dead zone non-linear Linear inverse model; the adaptive module adopts the adaptive law to adjust the estimated dead zone parameters online, and then transmits to the dead zone inverse model module to change the parameters of the inverse model; through the signal communication between the above modules, the visual servo manipulator A closed-loop control system is formed internally. The invention effectively eliminates the influence of the non-linear constraint of the dead zone and achieves higher image tracking precision.

Description

An Adaptive Dead Zone Inverse Model Generation Device for Visual Servo Manipulator System

technical field

The invention relates to an adaptive dead zone inverse model generating device of a visual servo manipulator system, which belongs to the transformation technology of the visual servo manipulator system on the input signal.

Background technique

Robot research began in the middle of the 20th century, and it developed along with the development of automation and computer technology and the development and utilization of atomic energy. Robots have been in the process of continuous development from simple to complex, from single to diverse, from low-level to high-level. With the development of electronic device technology, the computing power of computers has been greatly improved and various sensors are used in On the robot system, the working environment of most robots is pre-set, so once there is any change in the working environment, the robot system needs to be reset. As a result, the application range of industrial robots is greatly limited. In order to change this situation and make the robot adapt to various changing environments, people apply various external sensors such as touch, distance, force and vision sensors to the robot control system. To carry out contact measurement, so that the robot has a wider range of application and higher performance, people introduce visual sensors for robots. In the research of the visual servo system of the manipulator, the intelligent control strategy between the robot's visual sensor and the actuator has broad application prospects, and has become a research hotspot. The so-called visual servoing is to use the target position information obtained by the visual sensor as feedback to construct a position closed-loop robot control system P. . Visual servoing controls robot motion by automatically acquiring and analyzing images of objects. In short, visual servoing is to use the principle of machine vision to quickly process the image feedback information, and quickly give the position control signal according to the processed information to form a position closed-loop robot control.

In the early 1970s, people began to combine computer vision with robotic servo technology. Shirai and Inoue were the first scholars to apply vision to robotic systems. The visual servo method they adopt is to calculate and estimate the position of the target through the image information collected by the visual system, which is a static visual control method. The connection between the vision and the robot system is an open loop, so the positioning accuracy of the robot is related to the accuracy of the vision system and the robot. Among them, the accuracy of the vision system is mainly affected by the vision calibration error and resolution. The accuracy of the robot is mainly related to the accuracy of the joint position sensor and the inverse motion of the robot.

It is related to factors such as the accuracy of the scientific model, the joint control algorithm, the backlash and the flexibility of the robot. According to the different feedback information, visual servoing is mainly divided into image-based visual servoing and position-based visual servoing. Image-based visual servoing directly uses images to calculate errors and transmits them to the vision controller to plan the movement of the robot. This method does not need to calculate the position of the target and is not sensitive to the pose of the robot, but the design of the controller is relatively difficult. The position-based visual servoing calculates the position of the target according to the image and the pose of the robot itself, and the vision controller plans the movement of the robot according to the target position. The advantage of this method is that the signal fed back to the vision controller is a position signal, and the inner loop controller also needs a position signal, so the design and implementation of the vision controller is relatively easy. However, this method needs to calculate the three-dimensional information of the image and obtain the spatial coordinates of the target. At the same time, the position of the target depends on the pose of the robot and the camera, and is sensitive to the calibration error of the robot and the camera.

But the key issues at present are: First, the common robotic arm system with visual servo needs to calibrate the camera. However, camera calibration is a cumbersome and inefficient work, although many methods have been used in vision systems. calibration, but the cost of camera calibration is still high. The process of calculating the external parameters and internal parameters of the camera is the calibration of the camera. The working principle of the vision system is to start from the two-dimensional image information obtained by the camera, calculate the geometric information such as the shape and position of the object in the three-dimensional environment, and then reconstruct the three-dimensional object. The position of each point on the two-dimensional image is related to the geometric position of the corresponding point on the surface of the object in three-dimensional space. This interrelationship depends on the camera imaging geometric model, and the parameters of the geometric model are also called camera parameters, mainly including external parameters and internal parameters. The external parameters refer to the representation of the camera coordinate system in the reference coordinate system, and the internal parameters mainly include the magnification factor from the imaging plane coordinates to the image coordinates, the image coordinates of the center point of the optical axis, and the lens distortion coefficient. Camera calibration provides the link between professional cameras and non-surveyed cameras. The so-called non-measurement camera refers to such a type of camera whose internal parameters are completely unknown, partially unknown or in principle uncertain. Camera calibration is to obtain the internal and external parameters of the camera through calibration experiments; secondly, because the gears are difficult to fully mesh and the machinery wears down with time and times of use, there will be various nonlinear factors in the torque input of the manipulator system , will not only greatly affect the control effect but also cause the servo system to be unstable, among which the more common one is the input dead zone nonlinear constraint. When the driver provides the input torque for the motor, the non-linearity of the dead zone is manifested by the overlap of the gear meshing. When the torque can be accurately applied to the motor within a certain range, but when the torque exceeds the dead zone threshold, it appears that the input is applied to the motor. The torque on the motor presents a threshold characteristic, maintaining the maximum input value of the dead zone, causing a drastic drop in control performance and even instability.

Contents of the invention

The object of the present invention is to propose an adaptive dead zone inverse model generation device for a visual servo manipulator system considering that the above-mentioned visual servo manipulator control system is affected by the dead zone nonlinear constraint on the input signal.

The technical solution of the present invention is: an adaptive dead zone inverse model generating device of a visual servo manipulator system, including a visual servo controller (1), a vision module (2), a motion control module (3), and a drive module (4) , a six-degree-of-freedom mechanical arm (5), a torque feedback module (6), a speed (7) and a position acquisition module (8), and a detection module (9); the visual servo controller (1) is composed of a computer control unit (15 ), a control signal generating unit (11), an adaptive dead zone inverse model generating device (12), an adaptive camera calibration device (13), and a communication unit (14); it is characterized in that the visual servo controller (1) receives The actual image trajectory obtained by the image processing unit (23) and the error signal formed by the expected image trajectory, the position signal collected by the position acquisition module (7), the speed signal collected by the speed acquisition module (8), and the torque feedback module ( 6) The torque signal collected is calculated by the computer operation control unit (15), and the information is exchanged by the communication unit (14) (21) between the servo vision controller (1) and the vision module (2) and calibrated by the adaptive camera The device (13) calibrates the camera online, the dead zone inversion is constructed by the adaptive dead zone inverse model generating device (12) and acts on the control signal, and the control signal generating unit (11) sends the control signal to the control module (3); the motion control The module (3) modulates the PWM wave in the drive module (4) to drive the motor to drive the mechanical arm (5) to move; the detection module (9) detects the motor current, speed and position information in the drive module (4), and feedbacks and motion control The module (3) realizes closed-loop control; the vision module (2) collects the image coordinates of the feature points at the end of the manipulator (5) and feeds back the input of the controller (1), forming a closed-loop visual servo control system with dead zone nonlinear constraints control.

The above-mentioned adaptive dead zone inverse model generator (12) comprises an adaptive module (121), a dead zone inverse model module (122), an operation control module (123); the adaptive module (121) comprises an adaptive law memory ( 1211), dead zone parameter adjustment value memory (1212), dead zone parameter initial value storage value (1213); dead zone inverse model module (122) includes parameter storage value (1221), power amplifier (1222), piezoelectric drive ( 1223), threshold value circuit (1224), generating circuit (1225); operation control module (123) comprises operation memory (1231) (1234), integration module (1232), multiplication and division operation module (1233), addition and subtraction operation module ( 1234).

In the above-mentioned adaptive dead zone inverse model generation device (12), the parameter memory (1221) of the dead zone inverse model module (122) accepts the data transmitted by the communication unit (14) and the operation memory (1231), and obtains the dead zone inverse a vector of estimated parameters for the model Furthermore, the voltage value applied by the power amplifier circuit (1222) to the electric test drive (1223) is regulated, and the threshold value circuit (1224) uses the voltage signal to perform switch control for the generation circuit (1225), and finally the generation circuit reverses the dead zone to the model module (122 ) to send the control signal used to eliminate the non-linear influence of the dead zone to the control signal generating unit (11); at the same time, the dead zone parameter adjustment value memory (1212) and the parameter memory (1221) carry out camera estimation parameters and dead zone information exchange.

In the above-mentioned self-adaptive dead zone inverse model generating device (12), the self-adaptive law memory (1211) in the self-adaptive module (121) has stored the programming code of self-adaptive law, can be expressed as in mathematical form:

${\stackrel{<span\; class="notranslate">\; \&Center\; Dot;</span>}{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&Lambda;</span>}}_{{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; \&omega;</span>}}}_{\mathrm{<span\; class="notranslate">\; j</span>}}{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{{\mathrm{<span\; class="notranslate">\; q</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}$

in is the vector of estimated parameters for the inverse model of the dead zone derivative of is a positive definite symmetric matrix, is the estimated value of the dead zone structure parameter, is the joint velocity error, is the domain reference angular velocity associated with the vision system, is the angular velocity; the aforementioned parameters are passed to the adaptive law memory (1211) by the parameter memory (1221) through the dead zone parameter adjustment value memory (1212); at the initial stage of system operation, the dead zone parameter initial value memory (1213) will The initial information is transferred to the operation memory (1234); after the system runs, the data to be calculated is transferred to the operation memory (1234) by the adaptive law memory (1211) and the dead zone parameter adjustment value memory (1212).

In the above-mentioned self-adaptive dead zone inverse model generation device (12), the data received by the operation memory (1234) in the operation control module is simultaneously processed by the integral module (1232), the differential module (1232), the multiplication and division operation module (1233) Perform calculus, multiplication, division, addition and subtraction operations with the addition and subtraction module (1233), and then store the calculated data in the operation memory (1231), and then transfer it to the parameter memory (1221).

The above-mentioned adaptive dead zone inverse model generation device (12) needs to be connected with the communication unit (14) and the adaptive camera calibration device (13) through the bus, and the parameter variable value in the adaptive law memory (1211) is the vision module (2 ) data transmitted to the controller and the position signal and speed signal collected in the position acquisition module (8) and the speed acquisition module (7), the parameter memory (1211) in the dead zone inverse model module (122) also stores the torque feedback module (6) The transmitted signal; the torque generation signal that eliminates the nonlinear influence of the dead zone is generated by the dead zone inverse model (122).

In the above-mentioned visual servo manipulator system with dead zone nonlinear constraints, the visual servo controller (1) communicates with the vision module (2) through the communication unit (14) (21), and the adaptive camera calibration device (13) online Estimate the model parameters of the vision module (2), establish a non-calibrated independent depth vision model, and process the image captured by the camera unit (24) in real time through the image processing unit (23) and the operation control unit (22) to obtain the feature points the actual image trajectory.

In the above-mentioned visual servo manipulator system with dead zone nonlinear constraints, the visual servo controller (1) receives the image error formed by the input image trajectory signal and the actual image trajectory signal obtained by the visual system after image processing, and receives the position The acquisition module (8) and the speed acquisition module (7) obtain the joint angle, joint speed, and end position of the manipulator, and receive the torque change after the torque collected by the torque feedback module (6) passes through the dead zone nonlinear module, so as to realize the control of the mechanical arm. Collect the position information of the arm, quantify the movement trajectory of the manipulator, and directly transmit the expected position information of the manipulator to the motion control module.

In the above-mentioned visual servo manipulator system with dead zone nonlinear constraints, the motion control module (3) uses a DSP controller to realize three closed-loop control and PWM control; the outermost loop of the three closed-loop control is controlled by the position (31) The realized position control loop, the middle one is the speed control loop realized by the speed control (32), the innermost loop is the current control loop realized by the current control (33), and the DSP controller communicates with the control signal generation unit.

In the above visual servo manipulator system with dead zone nonlinear constraints, the drive module (4) receives the PWM modulation signal sent by the PWM control (34), and the driver (41) drives the motor (42) (43) with dead zone constraints , the electric power drags the transmission device (44) and thereby drags the six-degree-of-freedom mechanical arm (5) to move.

In the above-mentioned visual servo manipulator system with dead zone nonlinear constraints, the torque feedback module (6) collects the torque applied to the motor by the driver (41) without the dead zone constraint and the motor speed after the dead zone constraint, to realize The dead zone constrains the feedback to the visual servo controller (1).

In the above-mentioned visual servo manipulator system with dead zone nonlinear constraints, the detection module (9) realizes detection and provides closed-loop feedback signals for three closed-loop control, including QEP circuit (91), frequency measurement circuit (92), photoelectric encoder (93), A/D converter (94), current sensor (95); the pulse signal output by the photoelectric encoder (93) on the motor shaft is transmitted to the QEP circuit (91) and the frequency measurement circuit (92), and the pulse signal After being processed by the QEP circuit (91), the position feedback signal is obtained and sent to the position control loop (31) in the motion control module (3). The pulse signal is processed by the frequency measurement circuit to obtain a speed feedback signal and sent to the motion control module ( 3) in the speed control module (32), the current sensor (95) detects the motor winding current, and obtains its digital current signal through the A/D converter (94), and then sends it to the motion control module (3) Current control loop (33).

In the above-mentioned visual servo manipulator system with dead zone nonlinear constraints, the six-degree-of-freedom manipulator (5) marks a plurality of feature points at the end that can be photographed by the camera unit (24) and detected by the image processing unit (23), The image coordinates of the feature points are obtained by the vision module (2).

The present invention adopts a visual servo manipulator structure with hand-eye separation, that is, the camera is installed at a fixed position convenient for observing the feature points at the end of the manipulator, and the image is transmitted to the image processing unit to extract the image trajectory of the feature points by taking pictures. In addition, the problem of camera calibration is fully considered, and the online estimation of the visual model through the adaptive camera calibration device reduces the complicated workload of calibrating the camera. At the same time, the present invention also fully considers that when the manipulator system loses the nonlinear constraint input, it uses the adaptive dead zone inverse model generating device to eliminate the dead zone nonlinear constraint, and the adaptive law adopted can effectively establish the corresponding dead zone inverse model. Experiments have proved that this method has achieved very good results. The present invention is an adaptive dead zone inverse model generation device with excellent performance and easy to build on a computer system.

Description of drawings

Fig.1 Overall block diagram of visual servo manipulator system with dead zone nonlinear constraints

Figure 2 Principle block diagram of adaptive dead zone inverse model generator

Figure 3 Schematic diagram of the physical structure of hand-eye separation visual servoing

The inverse dead zone model constructed in Fig. 4 and the nonlinear schematic diagram of the dead zone through which the input torque passes

Figure 5 Schematic diagram of the tracking trajectory of the device without adaptive dead zone inverse model generation

Figure 6 Schematic diagram of tracking trajectory using adaptive dead zone inverse model generation device

Figure 7 Control block diagram of visual servo manipulator control system

Detailed ways

The invention relates to a dead zone inverse model generating device of a visual servo manipulator control system, which uses the designed adaptive law to estimate the parameters of the dead zone nonlinear constraint model online, and then constructs the corresponding dead zone inverse model, which can effectively The visual servo controller controls the manipulator to make the end feature points gradually track the desired image trajectory on the image plane under the condition that the input signal is eliminated, so as to achieve high image tracking accuracy. The dead zone inverse model generation device of the visual servo manipulator control system designed by the present invention will be described in detail below in conjunction with the accompanying drawings and specific examples.

Figure 1 is the overall block diagram of the visual servo manipulator system with nonlinear constraints in the dead zone. The purpose of designing the visual servo controller in Figure 1 is to control the movement of the manipulator so that the feature points on the end of the manipulator are in the image plane when the camera is not calibrated and the input torque of the manipulator is constrained by nonlinearity. Projection is able to track a given desired image trajectory. Controller The visual servo controller receives the error signal formed by the actual image trajectory and the expected image trajectory obtained by the image processing unit, the position signal collected by the position acquisition module, the speed signal collected by the speed acquisition module, and the torque collected by the torque feedback module The signal is calculated by the computer operation control unit, and the information is exchanged by the communication unit between the servo vision controller and the vision module. The camera is calibrated online by the adaptive camera calibration device, and the dead zone inversion is constructed by the adaptive dead zone inverse model generation device. Based on the control signal, the control signal generation unit sends the control signal to the control module; the motion control module modulates the PWM wave in the drive module to drive the motor to drive the mechanical arm; the detection module detects the motor current, speed and position information in the drive module, and feeds back Realize closed-loop control with the motion control module; the vision module collects the image coordinates of the feature points at the end of the manipulator and feeds back to the input of the controller to form a closed-loop control of the visual servo control system with dead zone nonlinear constraints. The controller can feedback according to the image, Velocity feedback adjusts controller output in time to provide optimal image tracking performance.

Fig. 7 is a control block diagram of the visual servo manipulator control system, which is the embodiment of the principle device diagram in Fig. 1 in terms of control.

Fig. 2 is a schematic block diagram of an adaptive dead zone inverse model generating device. It can be obtained from Figure 1 that Figure 2 is a part of the visual servo controller. The function of this device is to construct a dead zone inverse model that can be adjusted online to eliminate the influence of input nonlinearity and restore the designed input torque to the maximum extent. The dead zone inverse model module is based on the obtained estimated parameter vector Construct the inverse model of the dead zone, and the adaptive module transfers the designed adaptive law to the operation control module for calculation. According to the adaptive law:

${\stackrel{<span\; class="notranslate">\; \&Center\; Dot;</span>}{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&Lambda;</span>}}_{{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; \&omega;</span>}}}_{\mathrm{<span\; class="notranslate">\; j</span>}}{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{{\mathrm{<span\; class="notranslate">\; q</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}$

get the parameter vector Constantly modify the parameter vector during the operation of the system The values of , that is, the slopes k _r , k _l of the dead zone and the estimated values of the breakpoints _hr , h _l , enable the designed input torque to offset the disturbance caused by the nonlinear input to the greatest extent.

Fig. 3 is a schematic diagram of the physical structure of the hand-eye separation visual servo used in the present invention. The camera is installed at a fixed position convenient for observing the characteristic points at the end of the mechanical arm. The mechanical arm and the camera are connected to the computer through a bus to exchange information. Compared with the structure where the camera is installed at the end of the manipulator, the structure of separating the hands and eyes can effectively reduce the camera shake caused by the movement of the manipulator, and can clearly observe the global movement of the manipulator, and obtain the global information of the feature points. Through the computer The motion control card can send instructions to the motion control module to control the movement of the robotic arm.

Fig. 4 is a schematic diagram of the simulated input torque passing through the dead zone inverse dead zone model and then passing through the dead zone constraint. Dead zone nonlinearity can be described as:

${\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\left\{\begin{array}{cc}{\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; v</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}}<span\; class="notranslate">\; )</span>& {\mathrm{<span\; class="notranslate">\; v</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&Greater\; Equal;</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; l</span>}}<span\; class="notranslate">\; \&le;</span>{\mathrm{<span\; class="notranslate">\; v</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&le;</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}}\\ {\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; l</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; v</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; l</span>}}<span\; class="notranslate">\; )</span>& {\mathrm{<span\; class="notranslate">\; v</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&le;</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; l</span>}}\end{array}\right.$

Among them, u _{d and} v _d are the output and input of the dead zone, and k _r , k _l , _hr and h _l are constants. Parametrize the above model, and define β _d =[k _r , k _r h _r , k _l , k _l h _l ] ^T , The parameterized dead zone model can be obtained as: $u_d\left(t\right)=-{\mathrm{\&beta;}}_d^T{\mathrm{\&omega;}}_d.$ in

Construct a smooth dead zone inverse model as follows:

${\mathrm{<span\; class="notranslate">\; v</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}}{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}}}{{\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}}}{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\frac{{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; l</span>}}{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; l</span>}}}{{\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; l</span>}}}{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}_{\mathrm{<span\; class="notranslate">\; l</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; )</span>$

${\mathrm{<span\; class="notranslate">\; \&delta;</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; e</span>}}^{{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; /</span>{\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}}}{{\mathrm{<span\; class="notranslate">\; e</span>}}^{{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; /</span>{\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; /</span>{\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}}}$

${\mathrm{<span\; class="notranslate">\; \&delta;</span>}}_{\mathrm{<span\; class="notranslate">\; l</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; /</span>{\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}}}{{\mathrm{<span\; class="notranslate">\; e</span>}}^{{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; /</span>{\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; /</span>{\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}}}$

Among them, v _d and u _d are the output and input of the inverse model, and δ _r (u _d ) and δ _l (u _d ) are continuous indicator functions. Since the parameters of the dead zone model are unknown, only the estimated values of the parameters can be used in the design, that is, To build a smooth deadzone inverse model:

${\mathrm{<span\; class="notranslate">\; v</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}<span\; class="notranslate">\; +</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; k</span>}}}_{\mathrm{<span\; class="notranslate">\; r</span>}}{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; h</span>}}}_{\mathrm{<span\; class="notranslate">\; r</span>}}}{{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; k</span>}}}_{\mathrm{<span\; class="notranslate">\; r</span>}}}{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\frac{{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}<span\; class="notranslate">\; +</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; k</span>}}}_{\mathrm{<span\; class="notranslate">\; l</span>}}{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; h</span>}}}_{\mathrm{<span\; class="notranslate">\; l</span>}}}{{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; k</span>}}}_{\mathrm{<span\; class="notranslate">\; l</span>}}}{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}_{\mathrm{<span\; class="notranslate">\; l</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}<span\; class="notranslate">\; )</span>$

Among them, v _o and u _o do not predict the output and input of the inverse model. From the above formula, it can be seen that whether the constructed inverse model can eliminate the influence of the dead zone constraint to the greatest extent depends on the parameters The estimates are accurate enough. The following explains in detail how to apply the adaptive law to obtain the estimated value of the parameter vector:

1) According to the initialization parameters of the dead zone inverse model and the subsequent input and output measurement values, the estimated dead zone inverse model multiplication factor is defined as:

where v _o is the output of the dead zone inverse model, and is defined as above.

2) The angle q and angular velocity of the manipulator can be collected by the speed module and the position module angular acceleration And the position x of the end, the speed of the end Let the rod lengths of the first and second joints of the mechanical arm be l ₁ , l ₂ . The Jacobian matrix of the manipulator is obtained as J(q(t)):

$\mathrm{<span\; class="notranslate">\; J</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; q</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\left[\begin{array}{ccc}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; l</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; sin</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; q</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span>& <span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; l</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; sin</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; q</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span>& \mathrm{<span\; class="notranslate">\; 0</span>}\\ {\mathrm{<span\; class="notranslate">\; l</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; cos</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; q</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span>& {\mathrm{<span\; class="notranslate">\; l</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; sin</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; q</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span>& \mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]$

Among them, q(1), q(2), and q(3) are the first, second, and third elements in q. Similarly, according to the collected image coordinates y=[u, v] ^T , the expected image trajectory y _d =[u _d , v _d ] ^T , and the estimated parameter matrix delivered by the adaptive camera calibration device The image depth-independent interaction matrix can be constructed for the following form:

$\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; A</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\left(\begin{array}{c}{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; m</span>}}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; *</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; m</span>}}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}\\ {\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; m</span>}}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; *</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; m</span>}}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}\end{array}\right)$

in for the matrix The first, second and third lines of the Then the following parameter matrix can be obtained

$\mathrm{<span\; class="notranslate">\; Q</span>}<span\; class="notranslate">\; =</span>\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; A</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; J</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; q</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span>$

According to the collected end position information, the estimated depth of the camera relative to the projection plane can be obtained as:

$\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; z</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; m</span>}}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; x</span>}$

The domain reference image speed of the image plane is defined below. According to the image error Δy=yy _d , the domain reference image speed is:

${\stackrel{<span\; class="notranslate">\; \&Center\; Dot;</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}}_{\mathrm{<span\; class="notranslate">\; r</span>}}<span\; class="notranslate">\; =</span>{\stackrel{<span\; class="notranslate">\; \&Center\; Dot;</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; \&lambda;</span>}<span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; \&Delta;y</span>}$

Then the following parameter matrix can be obtained:

${\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; f</span>}}}_{\mathrm{<span\; class="notranslate">\; yr</span>}}<span\; class="notranslate">\; =</span>_{\mathrm{<span\; class="notranslate">\; c</span>}}\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; z</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; *</span>{\stackrel{<span\; class="notranslate">\; \&Center\; Dot;</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}}_{\mathrm{<span\; class="notranslate">\; r</span>}}$

3) From the above parameter matrix, we can further define the domain reference joint velocity for:

${\stackrel{<span\; class="notranslate">\; \&CenterDot;</span>}{\mathrm{<span\; class="notranslate">\; q</span>}}}_{\mathrm{<span\; class="notranslate">\; r</span>}}<span\; class="notranslate">\; =</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; R</span>}}}^{<span\; class="notranslate">\; +</span>}<span\; class="notranslate">\; *</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; f</span>}}}_{\mathrm{<span\; class="notranslate">\; yr</span>}}$

in for The generalized inverse of , so the joint velocity error vector can be obtained

4) According to the definition of the appeal and the collected variable values, the estimated value adaptive law about the dead zone parameter vector can be constructed:

$\stackrel{<span\; class="notranslate">\; \&Center\; Dot;</span>}{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&Lambda;</span>}}_{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; \&omega;</span>}}{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; q</span>}}$

Where Λ _β is a positive definite symmetric matrix, and the adaptive law of the appeal can transfer the data to the operation control module for operation, and get value, and the dead zone inverse model module is used to construct the dead zone inverse model.

So far, we have completed the construction of the inverse model of the dead zone. It should be noted that several steps of the appeal are completed in the generation device of the inverse model of the adaptive dead zone. Through the information exchange between the modules, it can be effectively estimated The dead zone parameter ensures that the designed control torque can be accurately applied to the manipulator, eliminating the influence of dead zone nonlinearity. Next, we verify the performance of the designed adaptive dead zone inverse model generator through experiments. The image trajectory to be tracked is selected as follows:

${\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\left(\begin{array}{c}\mathrm{<span\; class="notranslate">\; 40</span>}<span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; sin</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 0.2</span>}<span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 30</span>}\\ <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 40</span>}<span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; cos</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 0.2</span>}<span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 17</span>}\end{array}\right)$

The estimated initial value of the selected dead zone parameter is β _d0 =[1, -4, 1, -4] ^T .

Fig. 5 is a schematic diagram of the tracking trajectory of the device not using the adaptive dead zone inverse model generator. Among them, the red curve is the actual image trajectory, and the blue curve is the expected image trajectory curve. It can be seen that the actual trajectory is difficult to track the desired trajectory, and there is a large image error, which reflects the performance impact of the dead zone nonlinear constraint on the visual servo manipulator system, so it is particularly useful to design an adaptive dead zone inverse model generator necessary.

Fig. 6 is a schematic diagram of tracking trajectory of the device adopting the adaptive dead zone inverse model generation device. It can be clearly seen that the actual trajectory quickly tracks the desired trajectory in the initial stage, and gradually tracks the desired trajectory in the subsequent stage, which reflects the good performance of the designed adaptive dead zone inverse model generator.

Claims (13)

1. An adaptive dead zone inverse model generating device for a visual servo manipulator system, comprising a visual servo controller (1), a visual module (2), a motion control module (3), a drive module (4), and a six-freedom A mechanical arm (5), a torque feedback module (6), a speed (7) and a position acquisition module (8), and a detection module (9); the visual servo controller (1) is controlled by a computer control unit (15), A signal generating unit (11), an adaptive dead zone inverse model generating device (12), an adaptive camera calibration device (13), and a communication unit (14); is characterized in that the visual servo controller (1) receives the image processing unit (23) The error signal formed by the actual image trajectory obtained and the desired image trajectory, the position signal collected by the position acquisition module (7), the speed signal collected by the speed acquisition module (8), and the collection by the torque feedback module (6) The torque signal is calculated by the computer operation control unit (15), and the communication unit (14) (21) between the servo vision controller (1) and the vision module (2) performs information exchange through the adaptive camera calibration device (13 ) to calibrate the camera online, the dead zone inversion is constructed by an adaptive dead zone inverse model generating device (12) and acts on the control signal, and the control signal generating unit (11) sends a control signal to the control module (3); the motion control module (3 ) modulates the PWM wave in the drive module (4) to drive the motor to drive the mechanical arm (5) to move; the detection module (9) detects the motor current, speed and position information in the drive module (4), and feeds back to the motion control module (3 ) to realize closed-loop control; the vision module (2) collects the image coordinates of the end feature points of the mechanical arm (5) and feeds back the input of the controller (1), forming a closed-loop control of a visual servo control system with dead zone nonlinear constraints. 2. adaptive dead zone inverse model generation device (12) according to claim 1, is characterized in that, device comprises adaptive module (121), dead zone inverse model module (122), operation control module (123); Adaptive module (121) comprises adaptive law memory (1211), dead zone parameter adjustment value memory (1212), dead zone parameter initial value storage value (1213); Dead zone inverse model module (122) comprises parameter storage value (1221 ), power amplifier (1222), piezoelectric drive (1223), threshold circuit (1224), generating circuit (1225); the operation control module (123) includes operation memory (1231) (1234), integral module (1232), Multiplication and division operation module (1233), addition and subtraction operation module (1234). 3. adaptive dead zone inverse model generation device (12) according to claim 1, is characterized in that the parameter memory (1221) of dead zone inverse model module (122) accepts by communication unit (14), computing memory (1231) ) to obtain the estimated parameter vector of the dead zone inverse model Furthermore, the voltage value applied by the power amplifier circuit (1222) to the electric test drive (1223) is regulated, and the threshold value circuit (1224) uses the voltage signal to perform switch control for the generation circuit (1225), and finally the generation circuit reverses the dead zone to the model module (122 ) to send the control signal used to eliminate the non-linear influence of the dead zone to the control signal generating unit (11); at the same time, the dead zone parameter adjustment value memory (1212) and the parameter memory (1221) carry out camera estimation parameters and dead zone information exchange. 4. adaptive dead zone inverse model generating device (12) according to claim 1 is characterized in that the adaptive law memory (1211) in the adaptive module (121) has stored the programming code of adaptive law, uses mathematics The form can be expressed as: ${\stackrel{<span\; class="notranslate">\; .</span>}{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&Lambda;</span>}}_{{\mathrm{<span\; class="notranslate">\; \&beta;</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; \&omega;</span>}}}_{\mathrm{<span\; class="notranslate">\; j</span>}}{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{{\mathrm{<span\; class="notranslate">\; q</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}$ in is the vector of estimated parameters for the inverse model of the dead zone derivative of is a positive definite symmetric matrix, is the estimated value of the dead zone structure parameter, is the joint velocity error, is the domain reference angular velocity associated with the vision system, is the angular velocity; the aforementioned parameters are passed to the adaptive law memory (1211) by the parameter memory (1221) through the dead zone parameter adjustment value memory (1212); at the initial stage of system operation, the dead zone parameter initial value memory (1213) will The initial information is transferred to the operation memory (1234); after the system runs, the data to be calculated is transferred to the operation memory (1234) by the adaptive law memory (1211) and the dead zone parameter adjustment value memory (1212). 5. adaptive dead zone inverse model generation device (12) according to claim 1 is characterized in that the data that the computing memory (1234) in the computing control module will receive is simultaneously by integral module (1232), differential module ( 1232), the multiplication and division module (1233) and the addition and subtraction module (1233) perform calculus, multiplication, division, addition and subtraction, and then store the calculated data in the operation memory (1231), and then transfer it to the parameter memory (1221). 6. The adaptive dead zone inverse model generating device (12) according to claim 1, characterized in that in the visual servo controller (1), the adaptive dead zone inverse model generating device (12) needs to communicate with The unit (14) is connected to the adaptive camera calibration device (13), and it can be obtained from claim 4 that the parameter variable value in the adaptive law memory (1211) is the data and the position transmitted by the vision module (2) to the controller Acquisition module (8), the position signal and speed signal that gather in the speed acquisition module (7), the parameter memory (1211) in the dead zone inverse model module (122) also stores the signal that torque feedback module (6) transmits; The zone inverse model (122) produces a torque generating signal that eliminates the non-linear effects of the dead zone. 7. The visual servo manipulator system with dead zone nonlinear constraints according to claim 1, characterized in that the above-mentioned visual servo controller (1) communicates with the visual module (2) through a communication unit (14) (21), The adaptive camera calibration device (13) estimates the model parameters of the vision module (2) online, establishes a non-calibrated independent depth vision model, and passes the image captured by the camera unit (24) through the image processing unit (23) and operation control The unit (22) performs real-time processing to obtain the actual image trajectory of the feature points. 8. The visual servo manipulator system with dead zone nonlinear constraints according to claim 1, characterized in that the above-mentioned visual servo controller (1) receives the input image trajectory signal and the actual image obtained by the visual system after image processing. The image error formed by the image trajectory signal receives the joint angle, joint speed, and end position of the mechanical arm obtained by the position acquisition module (8) and the speed acquisition module (7), and receives the torque collected by the torque feedback module (6) through the dead zone. The torque change after the linear module realizes the collection of position information of the manipulator, quantifies the movement trajectory of the manipulator, and directly transmits the desired position information of the manipulator to the motion control module. 9. The visual servo manipulator system with dead zone nonlinear constraint according to claim 1, is characterized in that above-mentioned motion control module (3) adopts DSP controller to realize three closed-loop control and PWM control; Described three closed-loop control The outermost loop is a position control loop realized by position control (31), the middle loop is a speed control loop realized by speed control (32), and the innermost loop is a current control loop realized by current control (33). The DSP controller communicates with the control signal generation unit. 10. The visual servo manipulator system with dead zone nonlinear constraints according to claim 1, characterized in that the above-mentioned drive module (4) accepts the PWM modulation signal sent by the PWM control (34), and the driver (41) drives the dead zone The electric motor (42) (43) constrained by the area, the electric drive transmission device (44) and thus drags the six-degree-of-freedom mechanical arm (5) to move. 11. The visual servo manipulator system with dead zone nonlinear constraints according to claim 1, characterized in that the torque feedback module (6) of the above-mentioned torque feedback module (6) collects the torque and The motor speed after the dead zone constraint realizes the feedback of the dead zone constraint to the visual servo controller (1). 12. The visual servo manipulator system with dead zone nonlinear constraints according to claim 1, characterized in that the above-mentioned detection module (9) realizes detection and provides a closed-loop feedback signal of three closed-loop control, including a QEP circuit (91) and Frequency measurement circuit (92), photoelectric encoder (93), A/D converter (94), current sensor (95); the pulse signal output by the photoelectric encoder (93) on the motor shaft is transmitted to the QEP circuit (91) and frequency measurement circuit (92), the pulse signal is processed by the QEP circuit (91) to obtain a position feedback signal, and sent to the position control loop (31) in the motion control module (3), the pulse signal is processed by the frequency measurement circuit to obtain the speed feedback signal, and sent to the speed control module (32) in the motion control module (3), the current sensor (95) detects the motor winding current, and obtains its digital current signal by the A/D converter (94), and then converts it Send to the current control loop (33) in the motion control module (3). 13. The visual servo manipulator system with dead zone nonlinear constraints according to claim 1, characterized in that the above-mentioned six-degree-of-freedom manipulator (5) has a plurality of markings at the end that can be photographed by the camera unit (24), image processed The feature points detected by the unit (23), the image coordinates of the feature points are obtained by the vision module (2).