WO2021068334A1

WO2021068334A1 - Drive-control integrated control system

Info

Publication number: WO2021068334A1
Application number: PCT/CN2019/117615
Authority: WO
Inventors: 冯伟; 吴新宇; 张艳辉; 陈清朋; 徐天添; 马跃; 孙建铨; 王大帅; 郭师峰
Original assignee: 深圳先进技术研究院
Priority date: 2019-10-12
Filing date: 2019-11-12
Publication date: 2021-04-15

Abstract

An intelligent drive-control integrated control system, comprising: a control module (1), an intelligent dynamic parameter identification module (2), a sensorless active compliance control module (3), a force feedback human-machine collaboration anti-collision control module (4) and a multi-shaft driving module (5). The control module (1) is configured to perform real-time control on the multi-shaft driving module (5); the intelligent dynamic parameter identification module (2) is configured to feed back, according to the motion state of a collaborative robot (6), a mechanical parameter identification signal of the collaborative robot (6) to the control module (1) in time; the sensorless active compliance control module (3) is configured to feed back, according to the motion state of the collaborative robot (6), a position signal, a force signal and an environmental signal of the collaborative robot (6) to the control module (1) in time; the force feedback human-machine collaboration anti-collision control module (4) is configured to feed back, according to the motion state of the collaborative robot (6), a safety state signal of the collaborative robot (6) to the control module (1) in time; and the multi-shaft driving module (5) is configured to control the motion of the collaborative robot in real time according to an instruction of the control module. By these means, the capability for controlling the robot can be improved.

Description

An integrated control system for driving and controlling

【Technical Field】

The invention belongs to the field of collaborative robots, in particular to an integrated drive and control control system.

【Background technique】

With the development of industrial automation technology, industrial robots have played an important role in more and more production tasks. However, due to technical maturity and implementation costs, some complex operation tasks still need to be completed manually by humans. This gave birth to collaborative robots that can operate in a human-machine communicative environment.

Compared with traditional industrial robots, collaborative robots do not require a separate isolation space, and can cooperate with humans to complete production tasks at close range. For example, on the assembly line of 3C products, humans can complete complex assembly tasks, while collaborative robots can quickly and accurately The task of picking and placing parts is completed. This collaborative division of labor greatly improves production efficiency and reduces production costs. In order to achieve this collaborative goal, it is necessary to ensure a safe human-computer interaction environment, which puts forward requirements for the control of collaborative robots far higher than traditional robots in terms of accuracy and flexibility.

At present, industrial robots generally adopt a distributed control mode of "central motion controller + multiple servo drives", which is convenient in layout and simple in application. Most of the traditional industrial robots work in the position control mode. Each joint uses the driver to achieve precise position loop PID control, and receives the command requirements of the motion controller through the bus. This mode has simple control algorithms and small calculations. The data communication volume is not large. For collaborative robots, it is necessary to implement complex feedforward control, compliance control and other algorithms. However, the distributed architecture has limited signal transmission rate and synchronization mechanism problems. Its real-time and rapidity are difficult to meet the requirements of collaborative robots. In order to solve this problem, a controller with integrated drive and control for collaborative robots has appeared, which has the characteristics of compact structure, fast response speed, high control accuracy, and low cost. However, the existing integrated drive and control controllers still have the following problems in application: First, the algorithm implementation still needs a higher-level controller to implement, which will cause data transmission, real-time and synchronization problems between different systems; Second, most of the existing dynamic model parameter identification algorithms are based on the traditional excitation trajectory and least squares method for iterative estimation. The modeling is complicated, the estimation accuracy is not high, and the parameters that cannot be modeled can be identified; Third, the active compliance force control information of the existing robots is mainly obtained through the force/torque sensors on the joints, but the force/torque sensors are large in size and not suitable for use on collaborative robots. If a smaller-volume force/torque sensor is used, Then there is the problem of high price; fourth, there is the problem that the anti-collision detection ability is limited, and the safety protection strategy is not suitable for use on collaborative robots.

[Summary of the invention]

In order to solve the above problems, the present invention provides the society with a method that can identify parameters that cannot be modeled, can perform active compliance control without sensors, and has good anti-collision detection capabilities and safety protection strategies suitable for use on collaborative robots. Integrated intelligent drive and control control system.

The technical solution of the present invention is to provide a drive-control integrated control system, which is used to control a collaborative robot (6), and the drive-control integrated control system includes: a control module (1), intelligent power Learning parameter identification module (2), sensorless active compliance control module (3), force feedback human-machine cooperation anti-collision control module (4) and multi-axis drive module (5); the control module (1) is used to Set instructions, or/and by the intelligent dynamic parameter identification module (2), the sensorless active compliance control module (3) and the force feedback human-machine cooperation anti-collision control module (4) according to the collaborative robot (6) The signal fed back from the state controls the multi-axis drive module (5) in real time, so that the multi-axis drive module (5) controls the motion of the collaborative robot (6); among them, the intelligent dynamic parameter identification module (2) ) Is used to feed back the mechanical parameter identification signal of the collaborative robot (6) to the control module (1) according to the motion state of the collaborative robot (6); the sensorless active compliance control module (3) is used to respond to the motion state of the collaborative robot (6) The position signal, force signal and environment signal of the collaborative robot (6) are fed back to the control module (1); the force feedback human-machine cooperation anti-collision control module (4) is used to feed the control module (1) according to the motion state of the collaborative robot (6) The safety status signal of the cooperative robot (6) is fed back.

Among them, the intelligent dynamic parameter identification module (2) includes: the nominal model (21), the nominal model (21) is based on the Lagrangian dynamic model, used for: according to the motion state of the collaborative robot (6), Obtain the movement speed of any point on each link of the collaborative robot, calculate the kinetic energy of each link of the collaborative robot during the movement, and the total kinetic energy of the movement of the collaborative robot; calculate the potential energy of each link of the collaborative robot during the movement, and the collaborative robot The total potential energy relative to the reference potential energy surface during the movement; construct the Lagrangian function of the collaborative robot according to the total kinetic energy and total potential energy of the collaborative robot; perform the derivative operation on the Lagrangian function to obtain the standard of the collaborative robot It is called the dynamic equation; the actual dynamic model (22) is used to obtain the actual dynamic equation of the actual dynamic model of the collaborative robot according to the preset parameters; the parameter identification neural network (23) is used to set the collaborative robot to In the torque working mode, a smooth torque curve is selected as the input of the collaborative robot from the minimum to the maximum joint torque, and the code disc of each joint is used to obtain the angular displacement, angular velocity and angular acceleration of each joint; in a sampling period (T) Set the sampling time (t) inside, and take N groups of data containing torque, angular displacement, angular velocity and angular acceleration as a training sample data; the learning optimization module (24) is used to combine the torque τ(k) in the sample data ) Get the theoretical output value through the nominal model

Combine the torque τ(k) with the actual output value in the sample

Input to the parameter identification neural network to obtain the output correction value

Combine the theoretical output value with the output correction value to get the identification output value

Make the difference between the actual output value and the identification output value to obtain the output error

The output error is used to establish the loss function of the parameter identification neural network, and the parameter identification neural network is trained to complete the correction of the dynamic model.

Among them, the nominal kinetic equation is:

Among them, D(q)∈R ^n×n is a symmetric positive definite inertia matrix;

Is the Coriolis force and centrifugal force matrix; G(q)∈R ^n×1 is the center of gravity term matrix;

q is the mechanical joint angular displacement vector,

Is the angular velocity vector of the robotic arm

Is the angular acceleration vector of the manipulator; τ∈R ⁿ is the control torque vector of each joint of the manipulator.

Among them, the actual dynamic equation is:

Among them, F(q) represents the friction of joint movement,

Represents the disturbance in the motion of the robotic arm.

Among them, the sensorless active compliance control module (3) includes a position loop (31), a force loop (32) and a force position hybrid control law output module (33); wherein the position loop (31) includes an end position input terminal (311) ), position selection matrix (312) and position control law module (313); the end position input terminal (311) is used to input the end position signal to the position selection matrix (312), and the end position signal passes through the position selection matrix (312) and The position signal processed by the position control law module (313) is input to the force position mixed control law output module (33); wherein the force loop (32) includes an end force input terminal (321), a force selection matrix (322), and a force control law The module (313) and the joint torque estimation module (324) based on motor current. The end force input terminal (321) is used to input the end force signal to the force selection matrix (322). The end force signal depends on the passing force selection matrix (322) and the force The force signal processed by the control law module (323) is input to the force position hybrid control law output module (33), and the joint torque estimation module (324) feeds back the real-time current of the joint motor (35) to the end force input terminal (321); The position mixing control law output module (33) inputs the force position mixing control law output signal (G) to the joint motor (35).

The sensorless active compliance control module (3) also includes a robot kinematics model (314). The robot kinematics model (314) feeds back the joint angle and angular velocity of the collaborative robot (6) to the end position input terminal (311).

Among them, the sensorless active compliance control module (3) further includes a compensation module (34), and the compensation module (34) is interposed between the force position hybrid control law output module (33) and the joint motor (35).

Among them, the structure of the joint torque estimation module is a complete robot dynamics equation:

Among them, M∈R _n×n is the joint space inertia matrix; C∈R _n×n is the Coriolis force and centripetal force calculation matrix; g∈R _n×1 is the gravity term vector; q∈R _n×1 is the driving joint angle vector ；Τ∈R _n×1 is the driving joint torque; the motor torque τ _m driving equation is: τ=Nτ _m ; where N∈R _n×n is the diagonal matrix of the reduction ratio of each joint, and set J _m as The inertia of the motor rotor; in the derivation process, the friction term at the motor rotor

Substituting the relationship between the motor torque module and the joint torque, the joint torque estimation module based on the motor current is obtained, and the force detection output is obtained:

Among them, the force feedback human-machine cooperation anti-collision detection module (4) includes: the dynamic equation establishment module (41), which is used to establish the linkage coordinate system by the DH parameter method on the predetermined robot platform, and according to the Lagrangian dynamics The robot dynamics equation is established by the scientific formula; the collision detection operator and the disturbance observer establishment module (42), according to the robot dynamics equation and momentum equation, construct the collision detection operator based on the constant energy of the robot and the disturbance based on the generalized momentum change Observer; data analysis module (43), based on the real-time feedback of the robot system current, determines the relationship between the torque of each joint and the collision force, and gives the robot Jacobian matrix solution method, and analyzes its effectiveness in detecting collisions; safety protection The strategy formulation module (44), based on the detection results of the collision detection model, formulates different safety protection strategies for different collision situations and combined with actual working conditions; simulation verification and optimization module (45), based on the ADAMS-Simulink joint simulation platform for robots The effectiveness of the collision detection operator and the rationality of the safety protection strategy are simulated and optimized; the actual effect verification module (46), based on the predetermined robot platform, verifies and evaluates the actual effect of the obstacle avoidance protection safety strategy based on force feedback.

Among them, the force feedback human-machine cooperation anti-collision detection module (4) also includes: a monocular dual-view stereo matching module (47), used to construct a SVS-based monocular dual-view stereo matching model, and optimize geometric constraints on the loss function , Through the left and right view synthesis process and dual-view stereo matching, the accurate estimation of the detection target depth in the monocular image is realized; the convolution feature extraction module (48), based on the RGB image collected by the monocular camera, uses the ResNet model to perform deep convolution features Extraction; Human bone key point processing module (49), according to the prior knowledge of human bone joint geometry and the correlation between joints, optimize the design of the dual-branch deep convolutional neural network structure, and realize the synchronization processing of joint points and their joint relations. One branch uses the combination of probabilistic heat map and offset to perform human bone key point regression. The other branch detects the joint association information of multiple people in the image, and forms human bone sequence data through bipartite map matching; human skeleton image data processing module (410), used to reconstruct the human skeleton image data set combined with the characteristics of the industrial human-machine collaboration scene, and perform joint point data annotation, combined with manual adjustment to obtain the posture data set for the industrial collaboration scene.

The invention has the advantages of identifying parameters that cannot be modeled, performing active compliance control without sensors, and having good anti-collision detection capabilities and safety protection strategies, which are suitable for use on collaborative robots.

【Explanation of the drawings】

Figure 1 is a schematic block diagram of an embodiment of the method of the present invention;

2 is a block diagram of the intelligent dynamic parameter identification module of the present invention;

Figure 3 is a block diagram of the sensorless active compliance control module of the present invention;

4 is a schematic block diagram of the force feedback human-machine cooperation anti-collision control module of the present invention;

Fig. 5 is a schematic structural diagram of another expression form of Fig. 4.

【Detailed ways】

Please refer to Figure 1. Figure 1 discloses an intelligent drive and control integrated control system for multi-axis collaborative industrial robots, including: control module 1, intelligent dynamic parameter identification module 2, sensorless active compliance Control module 3, force feedback man-machine cooperation anti-collision control module 4 and multi-axis drive module 5.

The control module 1 is used to follow the pre-set instructions, or/and the intelligent dynamic parameter identification module 2, the sensorless active compliance control module 3 and the force feedback human-machine cooperation anti-collision control module 4 according to the cooperation The signal fed back from the state of the robot 6 controls the multi-axis drive module 5 in real time.

The intelligent dynamic parameter identification module 2 provides the mechanical parameter identification signal of the collaborative robot 6 to the control module 1 in time according to the motion state of the collaborative robot 6.

The sensorless active compliance control module 3 provides the position signal, force signal and environment signal of the collaborative robot 6 to the control module 1 in time according to the motion state of the collaborative robot 6.

The force feedback human-machine cooperation anti-collision control module 4 provides the control module 1 with a safety status signal of the collaborative robot 6 in time according to the motion state of the collaborative robot 6.

The multi-axis drive module 5 controls the movement of the collaborative robot 6 in real time according to the instructions of the control module 1.

The mains input interface 7 in FIG. 1 provides power for the control module 1 and the multi-axis drive module 5. Since the control module 1 requires low-voltage direct current, a power adapter 8 is provided between the mains input interface 7 and the control module 1.

Please refer to Figure 2. The intelligent dynamic parameter identification module 2 of the present invention includes the nominal model 21 based on the Lagrangian dynamic model. According to the motion state of the collaborative robot 6, each link of the collaborative robot is calculated The movement speed of one point; calculate the kinetic energy of each link of the collaborative robot during the movement, and the total kinetic energy of the entire collaborative robot movement; calculate the potential energy of each link of the collaborative robot during the movement, and the relative Refer to the total potential energy of the potential energy surface; construct the Lagrangian function of the collaborative robot system based on the total kinetic energy and total potential energy of the collaborative robot obtained above; perform the derivative operation on the Lagrangian function obtained in the above process to obtain The nominal dynamic equation of the collaborative robot system.

The actual dynamics model 22 is used to establish an actual dynamics model based on the nominal model. Starting from the actual use of the collaborative robot system, adding preset parameters that are difficult to model to obtain the actual dynamics model of the collaborative robot Kinetic equation.

The neural network training sample acquisition module 23, acquires neural network training sample data, sets the collaborative robot to the torque working mode, selects a smooth torque curve as the input of the collaborative robot within the range of the minimum to maximum joint torque, and uses the code of each joint The disk obtains the angular displacement, angular velocity and angular acceleration of each joint; the sampling time is set to t within a sampling period T, and N sets of data including torque, angular displacement, angular velocity and angular acceleration are taken as a training sample data.

The parameter identification neural network training module 24 uses the torque τ(k) in the sample data to obtain the theoretical output value through the nominal model

Combine the torque τ(k) with the actual output value in the sample

Input to the parameter identification neural network; get the output correction value

The theoretical output value is combined with the output correction value to obtain the identification output value

The difference between the actual output value and the identification output value to obtain the output error

Use the output error to establish the loss function of the parameter identification neural network; adopt the optimization strategy of self-learning evolution; train the neural network to complete the correction of the dynamic model.

Preferably, the nominal kinetic equation is:

Among them, D(q)∈R ^n×n is a symmetric positive definite inertia matrix;

q is the mechanical joint angular displacement vector,

Is the angular velocity vector of the robotic arm

Preferably, the actual dynamic equation is:

In the above formula, F(q) represents the friction of joint movement

Represents the disturbance in the motion of the robotic arm.

Preferably, the disturbance includes load changes, modeling errors or/and electrical interference.

Preferably, the parameters that are difficult to model include friction parameters, clearance parameters, or/and deformation parameters of the collaborative robot.

Please refer to FIG. 3, which shows the sensorless active compliance control module 3, which includes a position loop 31 and a force loop 32. The position loop 31 includes an end position input terminal 311, a position selection matrix 312, and a position control law module 313. The end position input terminal 311 is used to input an end position signal to the position selection matrix 312, and the end position signal is input to the force position hybrid control law output module according to the position signal processed by the position selection matrix 312 and the position control law module 313 33; The force loop 32 includes an end force input terminal 321, a force selection matrix 322, a force control law module 323, and a joint torque estimation module 324 based on motor current. The end force input terminal 321 is used to input an end force signal to all According to the force selection matrix 322, the terminal force signal is input to the force position hybrid control law output module 33 according to the force signal processed by the force selection matrix 322 and the force control law module 323. The joint torque estimation module 324 calculates the real-time current of the joint motor 35 The feedback is fed to the terminal force input terminal 321; the force position hybrid control law output module 33 inputs the force position hybrid control law output signal G to the joint motor 35, and the joint motor 35 controls the action of the collaborative robot 6 through the transmission mechanism 36.

In Figure 3, X _d ,

And f _d are the ideal end position and contact force, respectively.

Preferably, the present invention further includes a robot kinematics model 314 that feeds back the joint angle and angular velocity of the collaborative robot 6 to the end position input terminal 311.

Preferably, the present invention further includes a compensation module 34 interposed between the force-position hybrid control law output module 33 and the joint motor 35.

Preferably, the position selection matrix 312 and the force selection matrix 322 are combined into one, which is called the compliance selection matrix S.

Preferably, the expression of the compliance selection matrix S is:

S=diag(s ₁ ,s ₂ ,...,s _n );

When the i-th degree of freedom at the end of the robotic arm is position control, s _i =1, and when it is force control, s _i =0; the compliance selection matrix S is adjusted online based on force information.

Preferably, the structure of the joint torque estimation module 324 based on the motor current is that the complete robot dynamics equation is Equation 1:

Among them, M∈R _n×n is the joint space inertia matrix; C∈R _n×n is the Coriolis force and centripetal force calculation matrix; g∈R _n×1 is the gravity term vector; q∈R _n×1 is the driving joint angle vector ；Τ∈R _n×1 is the driving joint torque;

The equation driven by the motor torque τ _{m is:}

τ=Nτ _m ;

Where N ∈ _{R n×n} is the diagonal matrix of the reduction ratio of each joint, and J _m is the inertia of the motor rotor; in the derivation process, the friction term at the motor rotor

Among them, Ψ(i)=τ _m is the mapping module between the motor output torque and current.

4 and 5, in the present invention, the force feedback human-machine cooperation anti-collision detection module 4 includes:

The dynamic equation establishment module 41 is used to establish the linkage coordinate system by the D-H parameter method on the predetermined robot platform, and establish the robot dynamic equation according to the Lagrangian dynamic formula;

The collision detection operator and disturbance observer establishment module 42, according to the robot dynamics equation and momentum equation, constructs the collision detection operator based on the robot's constant energy and the disturbance observer based on the generalized momentum change;

The data analysis module 43, based on the real-time feedback of the robot system current, determines the relationship between the torque of each joint and the collision force, and provides a method for solving the Jacobian matrix of the robot, and analyzes its effectiveness in detecting collisions;

The safety protection strategy formulation module 44, based on the detection results of the collision detection model, formulates different safety protection strategies for different collision situations and in combination with actual working conditions;

Simulation verification and optimization module 45, based on the ADAMS-Simulink joint simulation platform, performs simulation verification and optimization on the effectiveness of the robot collision detection operator and the rationality of the safety protection strategy;

The actual effect verification module 46, based on a predetermined robot platform, verifies and evaluates the actual effect of the obstacle avoidance protection safety strategy based on force feedback.

As an improvement to the present invention, the safety protection strategy includes: 1. Stop after collision, that is, after the robot control system detects the collision signal, the control system will immediately turn off the servo drive and enable; or, 2. After the collision, the robot control system switches control Mode, convert the position mode to the torque mode; or, 3. After the collision, the robot changes its original motion trajectory and leaves the collision area.

Preferably, the present invention also includes:

Monocular and dual-view stereo matching module 47, used to construct a SVS-based monocular and dual-view stereo matching model, optimize geometric constraints on the loss function, and achieve detection targets in monocular images through the left and right view synthesis process and dual-view stereo matching Accurate estimation of depth;

The convolution feature extraction module 48 uses the ResNet model to extract deep convolution features based on the RGB images collected by the monocular camera;

The human bone key point processing module 49 optimizes the structure design of the dual-branch deep convolutional neural network based on the prior geometric knowledge of the human bone joints and the correlation between the joints, and realizes the synchronization processing of the joint points and the joint correlation. One of the branches passes Probabilistic heat map and offset are combined to perform human bone key point regression, one branch detects the joint related information of multiple people in the image, and the human bone sequence data is formed through bipartite map matching;

The human skeleton image data processing module 410, based on the Microsoft COCO data set, combines the characteristics of industrial human-machine collaboration scenes to reconstruct the human skeleton image data set, uses the open source alphapose of Shanghai Jiaotong University for joint point data annotation, and combines manual adjustment to obtain industrial-oriented collaboration The pose dataset of the scene.

Claims

A drive and control integrated control system, characterized in that the drive and control integrated control system is used to control a collaborative robot (6), and the drive and control integrated control system includes: a control module (1), intelligent dynamics Parameter identification module (2), sensorless active compliance control module (3), force feedback man-machine cooperation anti-collision control module (4) and multi-axis drive module (5);

The control module (1) is used for pre-set instructions, or/and by the intelligent dynamic parameter identification module (2), the sensorless active compliance control module (3) and the force feedback man-machine cooperation defense The collision control module (4) controls the multi-axis drive module (5) in real time according to the signal fed back from the state of the collaborative robot (6), so that the multi-axis drive module (5) controls the collaborative robot ( 6) Exercise;

Wherein, the intelligent dynamic parameter identification module (2) is used to feed back the mechanical parameter identification signal of the collaborative robot (6) to the control module (1) according to the motion state of the collaborative robot (6); The sensory active compliance control module (3) is used to feed back the position signal, force signal and environment signal of the collaborative robot (6) to the control module (1) according to the motion state of the collaborative robot (6); the force feedback human-machine cooperation defense The collision control module (4) is used for feeding back the safety state signal of the collaborative robot (6) to the control module (1) according to the motion state of the collaborative robot (6).
The drive and control integrated control system according to claim 1, characterized in that:

The intelligent dynamic parameter identification module (2) includes:

Nominal model (21), the nominal model (21) is based on Lagrange's dynamics model, and is used to obtain the motion speed of any point on each link of the collaborative robot according to the motion state of the collaborative robot (6) Calculate the kinetic energy of each link of the collaborative robot in the movement process and the total kinetic energy of the movement of the collaborative robot; calculate the potential energy of each link of the collaborative robot in the movement process and the movement process of the collaborative robot The total potential energy relative to the reference potential energy surface; the Lagrangian function of the collaborative robot is constructed according to the total kinetic energy and total potential energy of the collaborative robot; the derivative operation is performed on the Lagrangian function to obtain the total potential energy State the nominal dynamics equation of the collaborative robot;

The actual dynamics model (22) is used to obtain the actual dynamics equation of the actual dynamics model of the collaborative robot according to preset parameters;

The parameter identification neural network (23) is used to set the collaborative robot to the torque working mode, select a smooth torque curve as the input of the collaborative robot in the range from the minimum to the maximum joint torque, and use the code wheel of each joint Obtain the angular displacement, angular velocity and angular acceleration of each joint; set the sampling time (t) within a sampling period (T), and take N groups of data including torque, angular displacement, angular velocity and angular acceleration as a training sample data ；

The learning optimization module (24) is used to obtain the theoretical output value of the torque τ(k) in the sample data through the nominal model
Combine the torque τ(k) with the actual output value in the sample
Input to the parameter identification neural network to obtain the output correction value
And combine the theoretical output value with the output correction value to obtain the identification output value
Make the difference between the actual output value and the identification output value to obtain the output error
The output error is used to establish the loss function of the parameter identification neural network, and the parameter identification neural network is trained to complete the correction of the dynamic model.
The drive and control integrated control system according to claim 2, characterized in that:

The nominal kinetic equation is:

Among them, D(q)∈R n×n is a symmetric positive definite inertia matrix;
Is the Coriolis force and centrifugal force matrix; G(q)∈R n×1 is the center of gravity term matrix;
q is the mechanical joint angular displacement vector,
Is the angular velocity vector of the robotic arm
Is the angular acceleration vector of the manipulator; τ∈R n is the control torque vector of each joint of the manipulator.
The drive and control integrated control system according to claim 2, characterized in that:

The actual dynamic equation is:

Among them, F(q) represents the friction of joint movement,
Represents the disturbance in the motion of the robotic arm.
The drive and control integrated control system according to claim 1, characterized in that:

The sensorless active compliance control module (3) includes a position loop (31), a force loop (32) and a force position hybrid control law output module (33);

Wherein, the position loop (31) includes an end position input terminal (311), a position selection matrix (312) and a position control law module (313); the end position input terminal (311) is used to input an end position signal to the The position selection matrix (312), and the end position signal is input to the force position mixing control law output module (33) according to the position signal processed by the position selection matrix (312) and the position control law module (313) ；

Wherein, the force loop (32) includes a terminal force input terminal (321), a force selection matrix (322), a force control law module (313), and a joint torque estimation module (324) based on motor current. The terminal force input The terminal (321) is used to input the terminal force signal to the force selection matrix (322), and the terminal force signal is input to the force position hybrid control law based on the force signal processed by the force selection matrix (322) and the force control law module (323) The output module (33), the joint torque estimation module (324) feeds back the real-time current of the joint motor (35) to the end force input terminal (321); the force position mixed control law output module (33) feeds the joint motor ( 35) Input the output signal of the mixed control law of force and position (G).
The drive and control integrated control system according to claim 5, characterized in that:

The sensorless active compliance control module (3) also includes a robot kinematics model (314), which feeds back the joint angle and angular velocity of the collaborative robot (6) to the end position input terminal (311). ).
The drive and control integrated control system according to claim 5 or 6, characterized in that:

The sensorless active compliance control module (3) also includes a compensation module (34), which is interposed between the force-position hybrid control law output module (33) and the joint motor (35). between.
The drive and control integrated control system according to claim 5 or 6, wherein the joint torque estimation module is configured as a complete robot dynamics equation:

Among them, M∈R n×n is the joint space inertia matrix; C∈R n×n is the Coriolis force and centripetal force calculation matrix; g∈R n×1 is the gravity term vector; q∈R n×1 is the driving joint angle vector ；Τ∈R n×1 is the driving joint torque;

The equation driven by the motor torque τ m is:

τ=Nτ m ;

Where N ∈ R n×n is the diagonal matrix of the reduction ratio of each joint, and J m is the inertia of the motor rotor; in the derivation process, the friction term at the motor rotor
Substituting the relationship between the motor torque module and the joint torque, the joint torque estimation module based on the motor current is obtained, and the force detection output is obtained:

Among them, Ψ(i)=τ m is the mapping module between the motor output torque and current.
The drive and control integrated control system according to claim 1 or 2, characterized in that:

The force feedback human-machine cooperation anti-collision detection module (4) includes:

The dynamic equation establishment module (41) is used to establish the linkage coordinate system by the D-H parameter method on the predetermined robot platform, and establish the robot dynamic equation according to the Lagrangian dynamic formula;

Collision detection operator and disturbance observer establishment module (42), according to the robot dynamic equation and momentum equation, construct a collision detection operator based on the invariable robot energy and a disturbance observer based on the generalized momentum change;

The data analysis module (43), based on the real-time feedback of the robot system current, determines the relationship between the torque of each joint and the collision force, and provides a method for solving the Jacobian matrix of the robot, and analyzes its effectiveness in detecting collisions;

The safety protection strategy formulation module (44), based on the detection results of the collision detection model, formulates different safety protection strategies for different collision situations and combined with actual working conditions;

The simulation verification and optimization module (45), based on the ADAMS-Simulink joint simulation platform, performs simulation verification and optimization on the effectiveness of the robot collision detection operator and the rationality of the safety protection strategy;

The actual effect verification module (46), based on the predetermined robot platform, verifies and evaluates the actual effect of the obstacle avoidance protection safety strategy based on force feedback.
The drive and control integrated control system according to claim 9, characterized in that:

The force feedback human-machine cooperation anti-collision detection module (4) further includes:

The monocular dual-view stereo matching module (47) is used to construct a SVS-based monocular dual-view stereo matching model, optimize geometric constraints on the loss function, and realize the monocular image through the left and right view synthesis process and dual-view stereo matching Accurate estimation of detection target depth;

Convolution feature extraction module (48), based on the RGB image collected by the monocular camera, uses the ResNet model for deep convolution feature extraction;

The human bone key point processing module (49) optimizes the design of the dual-branch deep convolutional neural network structure based on the prior geometric knowledge of human bone joints and the correlation between the joints, and realizes the synchronization processing of the joint points and their joint relations. One of them The branch uses the combination of probabilistic heat map and offset to perform human bone key point regression. One branch detects the joint information of multiple people in the image, and forms human bone sequence data through bipartite map matching;

The human body skeleton image data processing module (410) is used to reconstruct the human body skeleton image data set combined with the characteristics of the industrial human-machine collaboration scene, and perform joint point data annotation, combined with manual adjustment to obtain a posture data set oriented to the industrial collaboration scene.