A high-precision force perception control system and method based on a large-load robot

Abstract

The present invention belongs to the technical field of precise robot assembly, and discloses a high-precision force perception control system and method based on a large-load robot. The high-precision force perception control system based on a large-load robot includes a precision assembly platform unit for providing an assembly reference plane and two-dimensional attitude adjustment, providing multi-angle installation under the fixed installation position of the robot; testing and guiding the robot adjustment requirements; an industrial robot unit for fine-tuning using a force control method during the assembly process, equipped with a high-precision force sensor, and designing an assisting module; a safety component for physically isolating the workstation, leaving a safety door at the maintenance position, and using a safety door lock for safety signal control to ensure the safety of humans and machines. The present invention can perform precise assembly at the 0.02 mm level under the control of high-performance force perception through an optical auxiliary device when the assembly accuracy is lower than the traditional manual resolution.

Description

Technical Field

[0001] The present invention belongs to the technical field of precise robot assembly, and particularly relates to a high-precision force perception control system and method based on a large-load robot. Background Technique

[0002] At present, the assembly accuracy of robots is limited by the accuracy of the robots themselves on the one hand. In particular, the accuracy of large-load robots is lower than that of miniaturized industrial robots, but large-load application scenarios are widespread in many fields. By performing higher-dimensional control constraints on the robots through force perception control, the operation accuracy of the robots can be effectively improved. The integration of force control technology and robot technology has achieved breakthroughs in many fields. Among them, it is widely used in relatively common fields such as laser detection and industrial control, aerospace, etc., especially in the integration with robot control systems. Two control technologies based on joint force and end-effector force play an important role in robot operations and are more sensitive to the basic control of robots than traditional teaching programming. With the continuous development of industrial robot technology, people's requirements for the operation accuracy of robots are also constantly increasing. In particular, in the fields of medical treatment, military industry, and precision assembly, further performance improvement requirements are put forward for the control of robots. Therefore, it is of practical significance to add force perception technology to the robotic arm to expand the constraints at the dynamic level and propose to compensate for the accuracy of the robot's motion process based on an improved dynamic model. In the process of traditional robot control and optical instrument detection, optical detection can analyze and detect the trajectory of robot motion control, but it cannot provide support for robot control in the reverse direction, resulting in multiple detections that require workers to intervene and modify control parameters. This belongs to an open-loop detection method. Due to the precision of optical detection, manual intervention often causes problems such as overshoot and hidden damage.

[0003] Through the above analysis, the problems and defects of the existing technology are as follows: The existing technology cannot provide support for robot control in the reverse direction, resulting in multiple detections that require workers to intervene and modify control parameters. This belongs to an open-loop detection method. Due to the precision of optical detection, manual intervention often causes problems such as overshoot and hidden damage. Summary of the Invention

[0004] Aiming at the problems existing in the existing technology, the present invention provides a high-precision force perception control system and method based on a large-load robot.

[0005] The present invention is implemented as follows. A high-precision force perception control system based on a large-load robot, the high-precision force perception control system based on a large-load robot includes:

[0006] A precision assembly platform unit for providing an assembly reference plane and two-dimensional attitude adjustment, and providing multi-angle installation under the fixed installation position of the robot; testing and guiding the robot adjustment requirements;

[0007] An industrial robot unit, which uses a force control method for fine-tuning during the assembly process, is equipped with a high-precision force sensor, and a boosting module is designed.

[0008] A safety component, which physically isolates the workstation, leaves a safety door at the maintenance position, and uses a safety door lock for safety signal control to ensure the safety of humans and machines.

[0009] Furthermore, the precision assembly platform unit consists of a vibration isolation platform, an integrated fixed tooling, a special calibration instrument, and auxiliary materials; the precision assembly turntable provides an assembly reference surface and two-dimensional attitude adjustment, provides multi-angle installation under the fixed installation position of the robot, and ensures compliance with the requirements of two points and one plane through the mechanism positioning function, and determines the installation reference pose through two positioning points and one plane; the special calibration instrument, which is composed of a laser, a prism, and a grating, tests and guides the robot's adjustment requirements.

[0010] Furthermore, the industrial robot unit consists of an industrial robot, a base, an end gripper tool, a robot control system, and a teaching box; the industrial robot is a large-load robot, the load of the large-load robot is greater than 500 kg, the motion radius is more than 1.5 m, and it has a repeat positioning accuracy of 0.02 mm; during the assembly process, a force control method is used for fine-tuning, it is equipped with a high-precision force sensor, and a boosting module is designed.

[0011] Furthermore, the safety component physically isolates the workstation by using a safety fence, leaves a safety door at the maintenance position, and uses a safety door lock for safety signal control to ensure the safety of humans and machines.

[0012] Another object of the present invention is to provide a high-precision force perception control method based on a large-load robot for the high-precision force perception control system based on a large-load robot, and the high-precision force perception control method based on a large-load robot includes:

[0013] Step 1, after the robot mounts the equipment object, perform inertial gravity compensation for the large-load robot; during the assembly process of the large-load robot, perform manual range traction teaching and positioning.

[0014] Step 2, during the movement process, combine the operator's instructions and environmental factors, turn on the online variable speed mode to adjust the speed of task execution, and realize online trajectory planning.

[0015] Step 3, during the assembly angle and attitude process, perform reverse supplementary guiding traction; at the same time, turn on the motion safety constraint; detect the force or torque normally applied during the assembly and adjustment process, and feedback the detected contact force to the control system for assembly position and attitude adjustment to realize precision force control assembly.

[0016] Step 4, Assembly process annotation: Set the robot motion control parameters, force control parameters, and force closed-loop control parameters, record the feedback of the robot's state, and perform control through a high-precision control algorithm.

[0017] Furthermore, in the first step, the specific process of inertial gravity compensation for the large-load robot is as follows:

[0018] After the robot mounts the equipment object, first perform gravity cancellation, compensate for the gravity influence brought by the robot's end tool and load during the movement process, so that the force sensor can more truly detect the action of other forces, and realize robot collision detection and impedance admittance control; support online zero calibration of sensor data, and shield the force deviation caused by sensor drift and gravity factors before assembly; the robot force control process package adapts to the robot body and the six-axis force sensor torque accuracy can reach 2 Nm; use an optical instrument to assist in calibrating the spatial errors △x, △y, △z in the robot motion model through the deformation of the heavy object after hanging.

[0019] The specific process of the manual range traction teaching and positioning is as follows:

[0020] During the assembly process of the large-load robot, the teaching method is used for batch replication. Since precision assembly is for small-batch and incoming material uncertain scenarios, the execution time of the robot's equipment task through the teaching method is more than 2 weeks. It is necessary to improve flexibility through manual traction positioning and attitude adjustment; detect and feedback the traction teaching force and torque of the operator through the force sensor; convert the detected force and torque into the desired end motion position of the robot, and then use inverse kinematics to convert the detected force and torque into the desired robot joint position, so that the robot moves in the direction of the force applied by the operator.

[0021] Furthermore, in the second step, the specific process of enabling the online variable speed mode to adjust the task execution speed and realizing online trajectory planning is as follows:

[0022] The robot performs speed planning online, adjusts the task execution speed during the movement process in combination with the operator's instructions and environmental factors, and realizes online trajectory planning; adopts hierarchical speed regulation technology, supports the macro-micro motion of the robot, allows the robot to move at a higher speed when far from the operation target, and move at a lower speed when approaching the operation target to meet the assembly requirements.

[0023] Furthermore, in the third step, the specific process of reverse supplementary guiding traction is as follows:

[0024] During assembly, perform assembly angle and attitude adjustment. Adopt the principle of equivalent optical path, perform assembly attitude adjustment in the non-assembly area and then migrate to the assembly area; manually control at a low speed, and map the traction force to the minimum assembly resolution of the moving position to achieve a precise assembly process.

[0025] In Step 3, the specific process of activating the motion safety constraint is as follows:

[0026] The safety constraint includes force constraint: when the force sensor senses that the force or torque on the carried module or component exceeds a given threshold, the industrial robot automatically stops or retracts its motion as required;

[0027] The safety constraint includes speed constraint. When the robot's motion speed exceeds the limit or when the speed exceeds the limit when the Cartesian space motion passes through a singularity, the robot automatically stops or retracts its motion as required;

[0028] The safety constraint includes position constraint. The robot system creates virtual walls, sets virtual wall parameters according to the actual environmental space limitations, including cube, cylinder, spherical, and custom virtual wall types, and divides the constrained space of the virtual wall into three levels, including free space, low-speed space, and no-go space; in different spaces, the robot gives corresponding feedback information. When the robot is in the no-go space, it automatically stops moving and gives an alarm message at the same time.

[0029] Furthermore, in Step 3, the specific process of precision force control assembly is as follows:

[0030] After the industrial robot carries the load, torque sensing compensates for the load gravity, senses the external force suffered by the industrial robot during the motion after compensating for the torque generated by the load weight, and manually sets the initial calibration value of the force sensor after changing the load to achieve the operation of clearing the force sensing data; during the automatic operation of the robot, collision detection is carried out. The robot detects unexpected collisions during the motion according to the installed six-axis force sensor, pauses the planned motion task, and returns to the original path manually or automatically according to the actual situation; integrating the optical detection device to achieve precise ranging and pose guidance, matching the assembly point with the assembled object, realizing dynamic sensor calibration to detect the normal force or torque applied during the assembly and adjustment process, and feeding back the detected contact force to the control system for assembly position and pose adjustment.

[0031] Furthermore, in Step 4, the specific process of assembly process annotation is as follows:

[0032] The human-machine interaction of the robot control system is responsible for setting the robot motion control parameters, motion targets, and motion speeds; setting the force control parameters, traction force constraint, and traction damping; setting the force closed-loop control parameters, control force constraint, and control position constraint; in addition, it supports the recording of the robot's status feedback, joint position, Cartesian position and pose, and contact force data;

[0033] In Step 4, the high-precision control algorithms include:

[0034] 1) High-precision motion control based on multi-level motion error compensation;

[0035] Compensate the positioning errors of industrial robots in the manufacturing site environment, classify the error sources into two categories: static positioning errors and dynamic tracking errors, and respectively construct the action mechanisms and action regions of different errors;

[0036] Use a pose tracking sensor to conduct single-axis repetitive trajectory motion and numerical simulation experiments, analyze the angular errors of each joint by integrating structural mechanics analysis methods, obtain the action laws of specific joint errors, establish specific error models, and perform specific error compensation;

[0037] According to various geometric error factors such as the parameters of each link of the industrial robot, coordinate system parameter errors, combine with external calibration equipment to establish a geometric error calibration method based on absolute position accuracy. Measure the end positioning errors of the robot in different poses by a laser tracker, and calculate the parameter errors of the robot kinematic model through iterative least squares solution; at the same time, facing the requirement of simplicity and rapidity of on-line calibration of industrial field robots, establish a geometric error self-calibration method based on fixed-point constraints. Combine with vision measurement technology, establish a geometric error self-calibration model based on fixed-point constraints, and identify the errors of geometric parameters through parameter identification algorithms to realize on-line self-calibration of the geometric errors of industrial robots;

[0038] The distribution of the end positioning errors of industrial robots in the joint space has obvious regularity. Establish a non-model error calibration method and a spatial interpolation compensation method for joint space grid division based on non-geometric errors such as gear wear and bearing clearance. Combine each grid node to establish a corresponding library of positioning errors in the joint space of the robot, and according to the distribution law and spatial correlation of the end positioning errors in the joint space, establish an optimal grid division method to achieve the unity of grid array accuracy and efficiency;

[0039] 2) Dynamics parameter identification and compensation based on the observation model;

[0040] Through the establishment of the motor system model, use the unconstrained nonlinear programming method to achieve the best identification of the dynamics model parameters to complete the mapping from motor drive current to driving torque; based on the modeling and inverse solution of robot dynamics, combine the calculation of driving torque and the random tree algorithm, randomly distribute tree nodes in the load parameter space, and through the optimal selection of parent nodes and the rapid reconstruction of subtrees, realize the generation of a fast random tree; based on the theoretical research results of Brownian motion and stochastic processes, establish a candidate set of motion states, and through machine learning and the random forest algorithm, realize the optimal estimation of the dynamics parameters of the robot body and the load;

[0041] 3) High-precision motion control based on inertia matching - torque feedforward;

[0042] Inverse kinematic solution of the whole robotic arm dynamics model: Based on the multi-body dynamics model of the whole robotic arm and the DH identification parameters in the task space, for the case of a finite solution set, an accurate inverse kinematic model is established through spatial geometric analysis; for the case of an infinite solution set, a method based on numerical analysis is used. Through research in several major planning algorithm branches such as the optimal step method, trust region method, and mixed-integer convex optimization method, accurate, fast, and robust joint motion solution is achieved; Joint feedforward torque and motion trajectory dynamic programming under time-varying non-linear constraints: Based on the numerical optimal algorithm, through research on the large-scale sequential convex optimization algorithm, optimal gradient selection within the local convex domain is achieved, fast iterative solution of the optimal motion trajectory is realized, and decoupled compensation calculation of the feedforward torque of each joint is carried out; Through the establishment of a trajectory optimization objective function based on generalized energy and solved by a path planning algorithm, smooth dynamic programming of the trajectory and dynamic feedforward compensation of the joint torque under time-varying non-linear constraints are achieved.

[0043] Combined with the above technical solutions and the technical problems solved, please analyze the advantages and positive effects of the technical solution to be protected by the present invention from the following aspects:

[0044] First, in view of the technical problems existing in the above prior art and the difficulty of solving the problem, closely combined with the technical solution to be protected by the present invention and the results and data in the R & D process, etc., analyze in detail and profoundly how the technical solution of the present invention solves the technical problems and the creative technical effects brought after solving the problems. The specific description is as follows:

[0045] The present invention can perform precise autonomous control during the assembly process of large-load robots. Especially when the assembly accuracy is lower than the traditional manual resolution, it can perform precise assembly at the 0.02 mm level under the control of high-performance force perception with the help of optical auxiliary equipment. The optical inspection and assembly process incorporates the sensitive data of optical instruments into the control of the robot dynamics model to achieve precise motion fine-tuning of large-load robots. During the optical alignment and adjustment process of the present invention, the control parameters of the robot itself and the measurement of the object to be assembled are comprehensively analyzed through control algorithms. The present invention overcomes the error accumulation scheme with the robot accuracy as the single control source, reduces the random errors brought by manual parameter adjustment in traditional methods, and thus realizes efficient flexible assembly. Especially during the assembly process of precious instruments, flexible precise assembly can effectively reduce errors by breaking through the precise assembly level.

[0046] Second, regarding the technical solution as a whole or from the perspective of the product, the technical effects and advantages of the technical solution to be protected by the present invention are specifically described as follows:

[0047] In response to the control requirements in the precise assembly scenario, the present invention proposes an optical inspection system and control method for large-load robots based on dynamics, which solves the problem of reverse compensation control for the object to be assembled. Brief Description of the Drawings

[0048] Figure 1 FIG. 1 is a schematic structural diagram of a high-precision force perception control system based on a large-load robot provided by an embodiment of the present invention;

[0049] Figure 2 FIG. 2 is a flowchart of a high-precision force perception control method based on a large-load robot provided by an embodiment of the present invention;

[0050] Figure 3 FIG. 3 is a schematic diagram of a high-precision motion error compensation process provided by an embodiment of the present invention;

[0051] Figure 4 FIG. 4 is a schematic diagram of a dynamic parameter identification and compensation process based on an observation model provided by an embodiment of the present invention;

[0052] In the figures: 1, precision assembly platform unit; 2, industrial robot unit; 3, safety component. Detailed Embodiment

[0053] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0054] This part is an explanatory embodiment that expands and explains the technical solutions of the claims in order to enable those skilled in the art to fully understand how the present invention is specifically implemented.

[0055] As Figure 1 shown, the high-precision force perception control system based on a large-load robot provided by an embodiment of the present invention includes:

[0056] A precision assembly platform unit 1 for providing an assembly reference plane and two-dimensional attitude adjustment, providing multi-angle installation under the fixed installation position of the robot; testing and guiding the robot adjustment requirements.

[0057] An industrial robot unit 2 for fine-tuning using a force control method during the assembly process, equipped with a high-precision force sensor, and a boosting module is designed.

[0058] A safety component 3 for physically isolating the workstation, leaving a safety door at the maintenance position, and using a safety door lock for safety signal control to ensure the safety of humans and machines.

[0059] The precision assembly platform unit provided by the embodiment of the present invention is composed of a vibration isolation platform, an integrated fixing tooling, a special calibration instrument, and auxiliary materials. The precision assembly turntable provides an assembly reference surface and a two-dimensional attitude adjustment function, and can provide multi-angle installation under the fixed installation position of the robot. It is required that the mechanism positioning function meets the requirements of "two points and one plane", that is, the installation reference pose is determined by two positioning points and one plane. The special calibration instrument is composed of a laser, a prism, and a grating. Based on this set of equipment, the adjustment requirements of the robot can be tested and guided.

[0060] The industrial robot unit provided by the embodiment of the present invention is composed of an industrial robot, a base, an end gripper tool, a robot control system, and a teaching box. The industrial robot is a large-load robot. The large-load robot is a general heavy-duty robot in the industry, with a load greater than 500 kg, a motion radius of more than 1.5 m, and a repeat positioning accuracy of 0.02 mm. The system standard recommends using the KR1000titan industrial robot of KUKA. During the assembly process, a force control method needs to be used for fine adjustment, a high-precision force sensor needs to be equipped, and a boosting module needs to be designed.

[0061] The safety component provided by the embodiment of the present invention physically isolates the workstation by using a safety fence, leaves a safety door at the maintenance position, and uses a safety door lock to control the safety signal to ensure the safety of humans and machines.

[0062] As Figure 2 shown, the high-precision force perception control method based on a large-load robot provided by the embodiment of the present invention includes:

[0063] S101: After the robot mounts the equipment object, perform inertial gravity compensation for the large-load robot; during the assembly process of the large-load robot, perform manual range traction teaching and positioning.

[0064] S102: During the movement process, combine the operator's instructions and environmental factors, and turn on the online variable speed mode to adjust the speed of task execution to achieve online trajectory planning.

[0065] S103: During the assembly angle and attitude process, perform reverse supplementary guiding traction; at the same time, turn on the motion safety constraint; detect the force or torque normally applied during the assembly and adjustment process, and feed the detected contact force back to the control system for assembly position and attitude adjustment to achieve precise force control assembly.

[0066] S104: Assembly process marking, set the robot motion control parameters, force control parameters, and force closed-loop control parameters, feedback and record the state of the robot, and perform control through a high-precision control algorithm.

[0067] In S101 provided by the embodiment of the present invention, the specific process of inertial gravity compensation for the large-load robot is:

[0068] After the robot mounts the equipment object, gravity cancellation is first performed to compensate for the gravity influence brought by the end tool and load of the robot during the movement process, so that the force sensor can more accurately detect the action of other forces, and functions such as robot collision detection and impedance admittance control can be realized. It supports the online zero calibration function of sensor data to shield the force deviation caused by factors such as sensor drift and gravity before assembly. The force control process package of the robot is adapted to the robot body and the six-axis force sensor, and the torque accuracy can reach 2 Nm. The spatial errors △x, △y, and △z are secondarily calibrated in the robot motion model by using the deformation of the heavy object assisted by the optical instrument after hanging.

[0069] In S101 provided by the embodiment of the present invention, the specific process of manual range traction teaching and positioning is as follows:

[0070] During the assembly process of large-load robots, teaching is usually used for batch replication. However, since precision assembly often targets small batches and scenarios with uncertain incoming materials, the execution time of the robot's equipment tasks in this mode through the teaching method is more than 2 weeks. However, manual traction positioning and attitude adjustment can improve flexibility requirements. Specifically, the traction teaching force and torque (movement direction) of the operator are detected and fed back through the force sensor; and the detected force and torque are converted into the desired end movement position of the robot, and then the detected force and torque are converted into the desired robot joint position by using inverse kinematics to enable the robot to move in the direction of the force applied by the operator.

[0071] In S102 provided by the embodiment of the present invention, the specific process of enabling the online variable speed mode to adjust the task execution speed and realizing online trajectory planning is as follows:

[0072] The robot has an online speed planning function, which adjusts the task execution speed during the movement process in combination with the operator's instructions and environmental factors to realize online trajectory planning. It adopts hierarchical speed regulation technology, supports the macro-micro movement of the robot, allows the robot to move at a higher speed when far from the operation target, and move at a lower speed when approaching the operation target to meet the assembly requirements.

[0073] In S103 provided by the embodiment of the present invention, the specific process of reverse supplementary guiding traction is as follows:

[0074] In addition to the position, the core in assembly is the assembly angle and attitude. Therefore, the principle of equivalent optical path is adopted to adjust the assembly attitude in the non-assembly area and then migrate it to the assembly area. During this process, manual control is performed at a low speed, and the traction force is mapped to the minimum assembly resolution of the moving position, thereby realizing a precise assembly process.

[0075] In S103 provided by the embodiment of the present invention, the specific process of enabling motion safety constraints is as follows:

[0076] Safety constraints include force constraints: when the force sensor senses that the force or torque received by the carried module or component exceeds a given threshold, the industrial robot can automatically stop or withdraw its movement as required (when ensuring that the withdrawal movement is unobstructed);

[0077] Safety constraints include speed constraints. When the movement speed of the robot exceeds the limit, such as the speed limit situation that occurs when the Cartesian space movement passes through a singularity, the robot can automatically stop or withdraw its movement as required (when ensuring that the withdrawal movement is unobstructed);

[0078] Safety constraints include position constraints. The robot system has a virtual wall function. The operator can set virtual wall parameters according to the actual environmental space limitations, such as cube, cylinder, sphere, and custom virtual wall types, and divide the constrained space of the virtual wall into three levels, including free space, low-speed space, and no-entry space. The robot gives corresponding feedback information in different spaces. For example, when the robot is in the no-entry space, the robot automatically stops moving and gives an alarm message at the same time.

[0079] In S103 provided by the embodiment of the present invention, the specific process of precise force control assembly is as follows:

[0080] After the industrial robot carries the load, the torque sensing has a load gravity compensation function. After compensating the torque generated by the load weight, it can sense the external force suffered by the industrial robot during the movement process. It is also possible to manually set the initial value of the force sensor calibration after changing the load to achieve the "zeroing" operation of the force sensing data; during the automatic operation process of the robot, it has a collision detection function. The robot can detect accidental collisions during the movement process according to the installed six-dimensional force sensor, pause the currently planned movement task, and can return to the original path manually or automatically according to the actual situation. Integrating the optical detection device, it can achieve precise ranging and pose guidance, match the assembly point with the assembly object, and can achieve the dynamic sensor calibration function to detect the force or torque normally applied during the assembly and adjustment process, and feedback the detected contact force to the control system for the adjustment of the assembly position and pose.

[0081] During the assembly process marking, set the robot motion control parameters, force control parameters, and force closed-loop control parameters, record the feedback of the robot state, and perform control through a high-precision control algorithm.

[0082] In S104 provided by the embodiment of the present invention, the specific process of assembly process marking is as follows:

[0083] The human-computer interaction part of the robot control system is mainly responsible for setting the motion control parameters of the robot, such as motion targets, motion speeds, etc.; setting the force control parameters, such as traction constraints, traction damping, etc.; and setting the force closed-loop control parameters, such as control force constraints, control position constraints, etc. In addition, it supports the recording of the robot's state feedback, such as data on joint positions, Cartesian positions and postures, contact forces, etc.

[0084] In S104 provided by the embodiment of the present invention, the high-precision control algorithm includes:

[0085] 1) High-precision motion control based on multi-level motion error compensation

[0086] Aiming at the problem of industrial robot positioning error compensation in the manufacturing site environment, the error sources are divided into two categories: static positioning errors and dynamic tracking errors, and the action mechanisms and action regions of different errors are respectively constructed.

[0087] Using a pose tracking sensor to perform single-axis repetitive trajectory motion and numerical simulation experiments, and combining structural mechanics analysis methods to analyze the angular errors of each joint, obtaining the action law of specific joint errors and establishing a specific error model for specific error compensation.

[0088] Comprehensively considering various geometric error factors such as the parameters of each link of the industrial robot and the errors of coordinate system parameters, combining external calibration equipment to study the geometric error calibration method based on absolute position accuracy, measuring the end positioning errors of the robot at different poses through a laser tracker, and calculating the parameter errors of the robot kinematic model through iterative least squares solution.

[0089] At the same time, aiming at the requirement of simplicity and rapidity of on-line calibration of industrial field robots, a geometric error self-calibration method based on fixed-point constraints is studied. By combining with vision measurement technology, a geometric error self-calibration model is established based on fixed-point constraints, and the errors of geometric parameters are identified through a parameter identification algorithm to realize on-line self-calibration of industrial robot geometric errors.

[0090] Aiming at the obvious regularity of the distribution of the end positioning error of the industrial robot in its joint space, a non-model error calibration method and a space interpolation compensation method for joint space grid division are studied based on non-geometric errors such as gear wear and bearing clearance. Combining each grid node to establish a corresponding library of positioning errors in the robot joint space, and according to the distribution law and spatial correlation of the end positioning error in the joint space, studying the optimal grid division method to achieve the unity of grid array accuracy and efficiency.

[0091] 2) Dynamics parameter identification and compensation based on the observation model

[0092] By establishing a motor system model and adopting a nonlinear programming method without constraints, the optimal identification of the dynamic model parameters is achieved to complete the mapping from the motor drive current to the driving torque.

[0093] Based on the modeling and inverse kinematic solution of robot dynamics, combined with the solution of the driving torque and the random tree algorithm, by randomly distributing tree nodes in the load parameter space and through the optimal selection of parent nodes and the rapid reconstruction of subtrees, the generation of a rapid random tree is realized.

[0094] Based on the theoretical research results of Brownian motion and stochastic processes, a candidate set of motion states is established, and through machine learning and the random forest algorithm, the optimal estimation of the dynamic parameters of the robot body and the load is realized.

[0095] 3) High-precision motion control based on inertia matching - torque feedforward

[0096] Inverse kinematic solution of the whole-arm dynamics model of the robotic arm: Based on the multi-body dynamics model of the whole robotic arm and the DH identification parameters in the task space, for the case of a finite solution set, through the space geometry analysis method, the establishment of an accurate inverse kinematic model is realized; for the case of an infinite solution set, based on the numerical analysis method, through the research of several major planning algorithm branches such as the optimal step method, the trust region method, and the mixed-integer convex optimization method, the accurate, fast, and robust solution of the joint motion is realized.

[0097] Dynamic programming of joint feedforward torque and motion trajectory under time-varying non-linear constraints: Based on the numerical optimal algorithm, through the research of the large-scale sequential convex optimization algorithm, the optimal gradient selection within the local convex domain is realized, so as to realize the rapid iterative solution of the optimal motion trajectory and the decoupling calculation of the compensation of the feedforward torque of each joint; through the establishment of a trajectory optimization objective function based on the generalized energy and the solution by the path planning algorithm, the smooth dynamic programming of the trajectory and the dynamic feedforward compensation of the joint torque under time-varying non-linear constraints are finally realized.

[0098] The technical solutions of the present invention will be described in detail below in conjunction with specific embodiments.

[0099] The high-precision force perception control method for large-load robots provided by the embodiments of the present invention includes:

[0100] High-precision motion control technology based on dynamic and static kinematic error compensation, adopting methods for classifying and compensating joint errors of industrial robots based on error classification research, geometric error analysis and modeling, and non-geometric error non-model calibration, constructing a multi-level hierarchical error compensation mechanism to achieve effective error compensation and eliminate or reduce the robot positioning error.

[0101] High-precision motion control technology based on inertia matching - torque feedforward. According to the decoupled low-frequency dynamic model of the whole machine system, it adopts a motor execution torque observation method based on neural network and an adaptive feedforward compensation method based on torque observer, combines torque feedforward control, realizes torque feedforward control with dynamic inertia matching, solves the problem of jitter during the movement of the robot, and improves the accuracy and efficiency of the robot system.

[0102] Research on high-precision motion control technology based on joint flexibility and stiffness compensation. According to the dynamic model of the rigid-flexible coupling of the robot joints, it adopts a method of precise compensation of joint stiffness and rapid error correction under external disturbance, combines Bayesian filtering model and predictive control, and rolls out the stiffness compensation values of each joint of the robot to achieve high-precision compliant compensation control under disturbance conditions and improve the absolute accuracy and trajectory accuracy of the robot.

[0103] Research on high-precision motion control technology for various dynamic unknown environments. It adopts a comprehensive control strategy of a neural network full approximation control strategy without relying on dynamic information and a control framework based on the nominal dynamic model and its improved full neural network approximation framework to improve the response speed and adaptability of the control.

[0104] Adopt high-sensitivity force perception control based on dynamics. Through the force sensor at the end of the robot, the contact surface force of precision assembly is sensed and analyzed. At the same time, the motion accuracy is processed by an assembly and adjustment process algorithm for precision assembly, specifically including a standard force control process package (calibration, stop, traction). By detecting the load end of the large-load robot, it improves the online variable-speed process of the robot at low speed, reduces the jitter during the start and stop of the robot, and through cooperation with the optical measurement of the precision assembly, the position information is transmitted to the robot control system in real time during the movement. The robot performs force control closed-loop guided assembly, and the human-machine system provides low-speed settings and movement direction adjustment for the robot, etc.

[0105] It has the function of coordinate transformation based on a large-load robot, supports calibrating the coordinate system of the industrial robot according to the sensor measurement data to obtain the accurate assembly position required. Through the industrial robot having the function of planning and moving according to the given target space pose, that is, given the motion target pose (X, Y, Z, Rx, Ry, Rz) and the corresponding coordinate system, the industrial robot can realize automatic planning and motion control according to the given target coordinates; the motion accuracy of the robot can be simply determined by 3 groups of micrometers for absolute positioning accuracy. This accuracy supports measurement using a micrometer, and the positioning accuracy is at the order of 0.01mm.

[0106] It has a task trajectory speed adaptation function. When the robot starts to move while carrying a load, it can perform trajectory planning according to the startup trajectory requirements (such as position, speed, time limitations, etc.) to enable the robot to start safely. When the robot approaches the target position of the operation task, it reduces the running trajectory speed according to the operation requirements to achieve fine movement at low speed and smoothly complete the specified task.

[0107] It has an online measurement function for optical instruments to improve the control accuracy of the robot. During the robot calibration process, after loading, it can perform workpiece equivalent coordinate system calibration through optical path guidance to determine the robot's position and movement feed direction. During the micro-adjustment of the robot's position and posture, it can utilize the principle of optical path reflection and refraction to provide optical path equivalent test analysis support, give priority to completing the robot's posture adjustment, and perform position fine-tuning according to the calibrated robot position feed direction. It provides the function of motion planning and control for the target position and posture in the specified coordinate system, and supports online compensation and modification of the planned results.

[0108] The traction control is divided into two stages. The first stage is range traction, which is applicable to the scenario of quickly approaching the assembly object to the assembly point in the initial assembly stage. The second stage is reverse supplementary guiding traction. In the assembly, information such as the assembly angle and posture is determined by optical detection technology. Therefore, by adopting the principle of equivalent optical path, after adjusting the assembly posture in the non-assembly area and then migrating to the assembly area, a high-precision assembly task can be achieved.

[0109] As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention shall be covered by the protection scope of the present invention.

Claims

1. A control method for a high-precision force perception control system based on a large-load robot, characterized in that, The high-precision force sensing control system based on a large-load robot includes: A precision assembly platform unit, which is used to provide an assembly reference plane and two-dimensional attitude adjustment, and provide multi-angle installation under the fixed installation position of the robot; test and guide the robot's adjustment requirements; An industrial robot unit, which uses force control mode for fine adjustment during the assembly process, is equipped with a high-precision force sensor, and a boosting module is designed; A safety component, which physically isolates the workstation, leaves a safety door at the maintenance position, and uses a safety door lock to control safety signals to ensure the safety of humans and machines; The control method includes: Step 1, after the robot mounts the equipment object, perform inertial gravity compensation for the large-load robot; during the assembly process of the large-load robot, perform manual range traction teaching and positioning; Step 2, during the movement process, combine the operator's instructions and environmental factors, and turn on the online variable speed mode to adjust the speed of task execution to achieve online trajectory planning; Step 3, during the assembly angle and attitude adjustment process, perform reverse supplementary guiding traction; at the same time, turn on the motion safety constraint; detect the force or torque normally applied during the assembly and adjustment process, and feed the detected contact force back to the control system for assembly position and attitude adjustment to achieve precision force control assembly; Step 4, assembly process annotation, set the robot motion control parameters, force control parameters, force closed-loop control parameters, feedback and record the state of the robot, and perform control through a high-precision control algorithm; In the third step, the specific process of reverse supplementary guiding traction is as follows: During the assembly angle and attitude adjustment in the assembly, adopt the principle of equivalent optical path, perform the assembly attitude adjustment in the non-assembly area and then migrate it to the assembly area; manually control at a low speed, and map the traction force to the minimum assembly resolution of the moving position to achieve a precise assembly process; In the third step, the specific process of turning on the motion safety constraint is as follows: The safety constraints include force constraint: when the force sensor senses that the force or torque received by the carried module or component exceeds the given threshold, the industrial robot automatically stops or withdraws the motion according to the requirements; The safety constraints include speed constraint: when the robot's motion speed exceeds the limit or the speed exceeds the limit when the Cartesian space motion passes through the singularity point, the robot automatically stops or withdraws the motion according to the requirements; The safety constraints include position constraint: the robot system sets virtual walls, sets virtual wall parameters according to the actual environmental space limitations, including cube, cylinder, spherical and custom virtual wall types, and divides the constraint space of the virtual wall into three levels, including free space, low-speed space and non-permissible space; the robot gives corresponding feedback information in different spaces. When the robot is in the non-permissible space, the robot automatically stops moving and gives an alarm message at the same time; In the third step, the specific process of precision force control assembly is as follows: After the industrial robot holds a load, torque sensing is performed to compensate for the load gravity. After compensating for the torque generated by the load weight, the external forces suffered by the industrial robot during movement are sensed. After changing the load, the initial calibration value of the force sensor is manually set to achieve the operation of clearing the force sensing data. During the automatic operation of the robot, collision detection is performed. The robot detects unexpected collisions during movement according to the installed six-axis force sensor, pauses the planned movement task, and returns to the original path manually or automatically according to the actual situation. By integrating an optical detection device, precise distance measurement and attitude guidance are achieved. The assembly point is matched with the assembled object to achieve dynamic sensor calibration, so as to detect the force or torque normally applied during the assembly and adjustment process, and feedback the detected contact force to the control system for assembly position and attitude adjustment.

2. The control method of the high-precision force perception control system based on the large-load robot according to claim 1, characterized in that In step 1, the specific process of inertial gravity compensation for the large-load robot is as follows: After the robot mounts the equipment object, gravity cancellation is first performed to compensate for the gravity influence brought by the robot's end tool and load during movement, so that the force sensor can more truly detect the action of other forces, and realize robot collision detection and impedance admittance control; support online zeroing of sensor data to shield the force deviation caused by sensor drift and gravity factors before assembly; The force control process package of the robot adapts to the robot body and the six-axis force sensor, and the torque accuracy is 2 Nm; an optical instrument is used to assist in measuring the deformation of the heavy object after mounting, and the spatial errors △x, △y, and △z are calibrated twice in the robot motion model; The specific process of the manual range traction teaching and positioning is as follows: During the assembly process of the large-load robot, teaching is used for batch replication. Since precision assembly is for small-batch and uncertain incoming material scenarios, the execution time of the robot's equipment task by teaching is more than 2 weeks. It is necessary to improve flexibility through manual traction positioning and attitude adjustment; the traction teaching force and torque of the operator are detected and fed back through the force sensor; the detected force and torque are converted into the desired end movement position of the robot, and then the detected force and torque are converted into the desired robot joint position by using inverse kinematics to make the robot move in the direction of the force applied by the operator.

3. The control method of the high-precision force perception control system based on a large-load robot according to claim 1, characterized in that, In step 2, the specific process of adjusting the speed of task execution by enabling the online variable speed mode and realizing online trajectory planning is as follows: The robot performs speed planning online and adjusts the speed of task execution in combination with the operator's instructions and environmental factors during movement to achieve online trajectory planning; the hierarchical speed regulation technology is adopted to support the macro and micro movement of the robot, allowing the robot to move at a higher speed when far from the operation target and at a lower speed when approaching the operation target to meet the assembly requirements.

4. The control method of the high-precision force perception control system based on a large-load robot according to claim 1, characterized in that, In step 4, the specific process of assembly process marking is as follows: The human-machine interaction of the robot control system is responsible for setting the robot motion control parameters, motion target, and motion speed; setting the force control parameters, traction force constraint, and traction damping; setting the force closed-loop control parameters, control force constraint, and control position constraint; In addition, it supports the recording of the robot's status feedback, including joint position, Cartesian position and attitude, and contact force data; In the fourth step, the high-precision control algorithm includes: 1) High-precision motion control based on multi-level motion error compensation; Compensate the positioning error of the industrial robot in the manufacturing site environment. Divide the error sources into two categories: static positioning error and dynamic tracking error, and respectively construct the action mechanisms and action regions of different errors; Use a pose tracking sensor to conduct single-axis repeated trajectory motion and numerical simulation experiments. Combine the structural mechanics analysis method to analyze the angular error of each joint, obtain the action law of the joint error and establish an error model for error compensation; According to various geometric error factors such as the parameters of each link of the industrial robot and the coordinate system parameter error, establish a geometric error calibration method based on absolute position accuracy in combination with an external calibration device. Measure the end positioning error of the robot in different poses through a laser tracker, and calculate the kinematic model parameter error of the robot through the iterative least squares method; At the same time, in response to the requirement of simplicity and rapidity of on-line calibration of industrial field robots, establish a geometric error self-calibration method based on fixed-point constraints. Combine it with vision measurement technology, establish a geometric error self-calibration model based on fixed-point constraints, and identify the error of geometric parameters through a parameter identification algorithm to realize on-line self-calibration of the geometric error of industrial robots; The distribution of the end positioning error of the industrial robot in the joint space has obvious regularity. Establish a non-model error calibration method and a space interpolation compensation method for joint space grid division based on non-geometric errors such as gear wear and bearing clearance. Combine each grid node to establish a corresponding library of positioning errors in the robot joint space, and establish an optimal grid division method according to the distribution law and spatial correlation of the end positioning error in the joint space to achieve the unity of grid array accuracy and efficiency; 2) Dynamics parameter identification and compensation based on the observation model; Through the establishment of the motor system model, use the unconstrained nonlinear programming method to achieve the best identification of the dynamics model parameters to complete the mapping from the motor drive current to the driving torque; Based on the modeling and inverse solution of the robot dynamics, combine the calculation of the driving torque and the random tree algorithm. Randomly distribute tree nodes in the load parameter space, and through the optimal selection of the parent node and the rapid reconstruction of the subtree, realize the generation of a fast random tree; Based on the theoretical research results of Brownian motion and stochastic processes, establish a candidate set of motion states, and through machine learning and the random forest algorithm, realize the optimal estimation of the dynamics parameters of the robot body and the load; 3) High-precision motion control based on inertia matching - torque feedforward; Inverse solution of the whole-arm dynamics model: Based on the multi-body dynamics model of the whole robotic arm and the DH identification parameters in the task space, for the case of a finite solution set, an accurate inverse kinematics model is established through spatial geometric analysis; for the case of an infinite solution set, a method based on numerical analysis is used. Through research in several major branches of programming algorithms such as the optimal step size method, trust region method, and mixed-integer convex optimization method, accurate, fast, and robust joint motion solution is achieved; Joint feedforward torque and motion trajectory dynamic programming under time-varying non-linear constraints: Based on numerical optimal algorithms, through research on large-scale sequential convex optimization algorithms, optimal gradient selection within the local convex domain is achieved, fast iterative solution of the optimal motion trajectory is realized, and decoupled compensation calculation of the feedforward torque of each joint is performed; Through the establishment of a trajectory optimization objective function based on generalized energy and solved by a path planning algorithm, smooth dynamic programming of the trajectory and dynamic feedforward compensation of joint torque under time-varying non-linear constraints are realized.

5. The control method of the high-precision force perception control system based on the large-load robot according to claim 1, characterized in that, The precision assembly platform unit consists of a vibration isolation platform, an integrated fixed tooling, a special calibration instrument, and auxiliary materials; The precision assembly platform provides an assembly reference plane and two-dimensional attitude adjustment, provides multi-angle installation under the fixed installation position of the robot, and ensures compliance with the requirements of two points and one plane through the mechanism positioning function. The installation reference pose is determined by two positioning points and one plane. The special calibration instrument, composed of a laser, a prism, and a grating, tests and guides the adjustment requirements of the robot.

6. The control method of the high-precision force perception control system based on the large-load robot according to claim 1, characterized in that, The industrial robot unit consists of an industrial robot, a base, an end effector tool, a robot control system, and a teaching pendant; The industrial robot is a large-load robot. The large-load robot has a load greater than 500 kg, a motion radius of more than 1.5 m, and a repeat positioning accuracy of 0.02 mm; During the assembly process, force control is used for fine adjustment, a high-precision force sensor is equipped, and a boosting module is designed.

7. The control method of the high-precision force sensing control system based on a large-load robot according to claim 1, characterized in that, The safety component physically isolates the workstation in the form of a safety fence, leaves a safety door at the maintenance position, and uses a safety door lock for safety signal control to ensure the safety of humans and machines.

Images