A robot visual-free precision assembly method and system based on force feedback

Abstract

The application relates to the technical field of wireless communication, and provides a robot visual-free precision assembly method and system based on force feedback. The method establishes an initial positioning reference through manual teaching, guides a shaft piece at the end of a robot to the vicinity of a to-be-mated hole position, controls the shaft piece to move in a direction perpendicular to the end face of the to-be-mated hole until axial contact force is detected, controls the shaft piece to perform spiral trajectory search in a horizontal plane, and collects force and torque data of a six-dimensional force sensor in real time; when a force feature mutation is detected, it is determined that the shaft piece contacts a hole rim, the shaft piece is controlled to perform slight swinging to correct the shaft line inclination angle and reduce the insertion resistance; the shaft piece is controlled to perform an insertion operation along the axial direction, and if the insertion depth reaches a set value and the axial contact force is stable in a preset interval within a preset limited time, it is determined that the assembly is qualified, otherwise, it is determined that the assembly fails and the work is terminated. The method eliminates the influence of environmental light, shielding and other factors on the assembly process, and is more widely applicable.

Description

Technical Field

[0001] This application relates to the field of wireless communication technology, and in particular to a method and system for precision assembly of robots without vision based on force feedback. Background Technology

[0002] Currently, precision shaft and hole assembly in robots mainly relies on vision-guided technology, requiring high-precision cameras, light sources, and complex calibration systems. This results in high hardware costs and susceptibility to interference from factors such as changes in ambient lighting, workpiece surface reflection, and occlusion in the assembly space, leading to insufficient assembly stability. Existing force-controlled assembly solutions suffer from low hole-searching path efficiency, poor posture adjustment accuracy, and a single criterion for judging assembly results. They cannot achieve stable precision fit without vision assistance, making it difficult to meet the high reliability and high efficiency automated assembly requirements in industrial scenarios. Summary of the Invention

[0003] This application provides a robot vision-free precision assembly method and system based on force feedback, aiming to solve the problems that current robot precision shaft and hole assembly mainly relies on vision guidance technology, which requires high-precision cameras, light sources and complex calibration systems, resulting in high hardware investment costs and susceptibility to interference from factors such as changes in ambient light, reflections on the workpiece surface and occlusion in the assembly space, leading to insufficient assembly stability.

[0004] In a first aspect, embodiments of this application provide a robot vision-free precision assembly method based on force feedback, the method comprising: An initial positioning reference is established through manual teaching, and the shaft at the end of the robot is guided to the vicinity of the hole to be mated. The shaft is controlled to move in a direction perpendicular to the end face of the hole to be mated until an axial contact force is detected. The axial contact force is kept constant to maintain a stable force interaction environment. The control shaft performs a spiral trajectory search in the horizontal plane, and collects force and torque data from the six-dimensional force sensor in real time; when a sudden change in force characteristics is detected, the shaft is determined to be in contact with the hole edge, and the shaft is controlled to swing slightly to correct the axis tilt angle and reduce insertion resistance; The control shaft is inserted axially while continuously monitoring the insertion depth and axial contact force. If the insertion depth reaches the set value and the axial contact force is stable within the preset range within a preset time, the assembly is deemed qualified; otherwise, the assembly is deemed unsuccessful and the operation is terminated.

[0005] In some embodiments, the step of establishing an initial positioning reference through manual teaching and guiding the shaft of the robot end to the vicinity of the hole to be mated includes: dragging the teaching robot end to a preset height position directly above the hole to be mated, recording the coordinates of the preset height position as the initial reference point, and controlling the robot end to move the shaft to the initial reference point position.

[0006] In some embodiments, the control shaft moves along a direction perpendicular to the end face of the hole to be mated until an axial contact force is detected, including: setting an axial movement speed, controlling the shaft to move downward along a direction perpendicular to the end face of the hole to be mated at the set speed, collecting axial force data from a six-dimensional force sensor in real time, and stopping the movement of the shaft when the axial force reaches a preset contact threshold.

[0007] In some embodiments, maintaining a constant axial contact force to maintain a stable force interaction environment includes: acquiring axial force data from a six-dimensional force sensor in real time, comparing the acquired axial force with a preset target force, and adjusting the axial position of the robot end effector based on the comparison result to keep the axial force within the allowable error range of the preset target force.

[0008] In some embodiments, the control shaft performs a spiral trajectory search in the horizontal plane and collects force and torque data from the six-dimensional force sensor in real time, including: setting the initial radius, pitch and rotation speed of the spiral trajectory, controlling the shaft to perform a spiral motion from the inside out in the horizontal plane with the set parameters, and collecting three-dimensional force and three-dimensional torque data from the six-dimensional force sensor at a fixed frequency during the motion.

[0009] In some embodiments, when a sudden change in force characteristics is detected, determining that the shaft is in contact with the hole edge and controlling the shaft to perform a slight oscillation to correct the axis tilt angle and reduce the insertion resistance includes: calculating the difference between the real-time collected force and torque data and the corresponding data at the previous moment; determining that the shaft is in contact with the hole edge when the difference exceeds a preset sudden change threshold; controlling the shaft to perform a slight oscillation around two orthogonal axes in the horizontal plane in sequence; collecting axial force data in real time during the oscillation; recording the shaft posture corresponding to the minimum axial force; and controlling the shaft to adjust to that posture.

[0010] In some embodiments, the control shaft is inserted axially while continuously monitoring the insertion depth and axial contact force, including: setting an axial insertion speed, controlling the shaft to be inserted downwards at a set speed in a direction perpendicular to the end face of the hole to be mated, reading robot joint encoder data in real time to calculate the current insertion depth, and simultaneously acquiring axial force data from a six-dimensional force sensor at a fixed frequency.

[0011] In some embodiments, the step of determining that the assembly is qualified if the insertion depth reaches a set value and the axial contact force is stable within a preset range within a preset time limit includes: starting the timing module to start timing, continuously judging whether the insertion depth reaches the set value within a preset time limit, if the insertion depth reaches the set value, continuing to judge whether the axial contact force is always within a preset range within a preset stable time period, and judging whether the assembly is qualified if the insertion depth reaches the set value and the axial contact force is always within a preset range within a preset stable time period.

[0012] In some embodiments, the step of otherwise determining assembly failure and terminating the operation includes: if the insertion depth does not reach the set value within a preset time limit, or the axial contact force exceeds the preset range, immediately stopping all movement of the shaft, controlling the shaft to move axially upward to the initial reference point position, and outputting an assembly failure signal.

[0013] Secondly, this application provides a vision-free precision assembly system for robots based on force feedback, the system comprising: The reference holding unit is used to establish an initial positioning reference through manual teaching, guide the shaft of the robot end to the vicinity of the hole to be mated, control the shaft to move in a direction perpendicular to the end face of the hole to be mated until an axial contact force is detected, and maintain the axial contact force constant to maintain a stable force interaction environment. The resistance reduction unit is used to control the shaft to perform a spiral trajectory search in the horizontal plane and to collect force and torque data from the six-dimensional force sensor in real time. When a sudden change in force characteristics is detected, the unit determines the contact hole edge of the shaft and controls the shaft to swing slightly to correct the axis tilt angle and reduce insertion resistance. The termination unit is used to control the insertion of the shaft along the axial direction, while continuously monitoring the insertion depth and axial contact force. If the insertion depth reaches the set value and the axial contact force is stable within the preset range within a preset time, the assembly is deemed qualified; otherwise, the assembly is deemed unsuccessful and the operation is terminated.

[0014] This application requires no vision system, significantly reducing hardware costs and completely eliminating the impact of ambient light, occlusion, and other factors on the assembly process, making it applicable to a wider range of scenarios. It employs a spiral trajectory search combined with force feature mutation detection, enabling rapid and accurate hole positioning and improving hole-finding efficiency and accuracy. Micro-oscillation correction of the axis tilt angle effectively reduces insertion resistance, preventing damage to the workpiece or equipment. The use of both insertion depth and axial contact force as dual criteria for determining assembly results significantly improves the reliability of assembly quality assessment. Full-process closed-loop force control enables high-precision automated shaft and hole assembly without vision, improving industrial production efficiency and product qualification rate.

[0015] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1This is a schematic flowchart illustrating the steps of a robot vision-free precision assembly method based on force feedback, provided in one embodiment of this application. Figure 2 This is a schematic diagram illustrating the principle of a robot vision-free precision assembly method based on force feedback, according to an embodiment of this application. Figure 3 This is a schematic block diagram of a visionless precision assembly system for robots based on force feedback, provided in one embodiment of this application. Figure 4 This is a schematic block diagram of the structure of a computer device provided in an embodiment of this application.

[0018] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0021] It should be understood that, in order to clearly describe the technical solutions of the embodiments of the present invention, the terms "first" and "second" are used in the embodiments of the present invention to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and the terms "first" and "second" are not necessarily different.

[0022] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0023] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0024] Currently, precision shaft and hole assembly in robots mainly relies on vision-guided technology, requiring high-precision cameras, light sources, and complex calibration systems. This results in high hardware costs and susceptibility to interference from factors such as changes in ambient lighting, workpiece surface reflection, and occlusion in the assembly space, leading to insufficient assembly stability. Existing force-controlled assembly solutions suffer from low hole-searching path efficiency, poor posture adjustment accuracy, and a single criterion for judging assembly results. They cannot achieve stable precision fit without vision assistance, making it difficult to meet the high reliability and high efficiency automated assembly requirements in industrial scenarios.

[0025] Please refer to Figure 1 and Figure 2 This application provides a vision-free precision assembly method for robots based on force feedback, applicable to computer equipment. The computer equipment can be deployed on a single server or server cluster. It can also be deployed on handheld terminals, laptops, wearable devices, or robots, etc. It should be noted that all information involved in the method provided in this application is extracted with the authorization of the relevant users and in accordance with relevant regulations, and will not infringe on user privacy.

[0026] Please refer to Figure 1 and Figure 2 , Figure 1 This is a flowchart of a vision-free precision assembly method for robots based on force feedback, provided in an embodiment of the present invention. Figure 2 This is a schematic diagram illustrating the three core states of the assembly process provided in this embodiment of the invention. Area A represents the initial positioning state, area B represents the horizontal spiral hole-finding state, and area C represents the attitude adjustment and insertion state. The robot-based, vision-free precision assembly method based on force feedback provided in this embodiment of the invention comprises three core steps: step S101, initial positioning and force control approximation; step S102, spiral hole-finding and attitude correction; and step S103, insertion operation and result determination. The technical details and specific implementation methods of each step are described in detail below: Step S101. Establish an initial positioning reference through manual teaching, guide the shaft of the robot end to the vicinity of the hole to be mated; control the shaft to move in a direction perpendicular to the end face of the hole to be mated until an axial contact force is detected, and keep the axial contact force constant to maintain a stable force interaction environment.

[0027] Specifically, step S101 is the initial preparation stage of the assembly process. Its core purpose is to establish a reliable initial positioning datum and to create a stable force interaction environment between the shaft and the end face of the hole to be mated, laying the foundation for subsequent hole searching and insertion operations. This step includes two sub-steps: the initial positioning datum establishment sub-step and the axial force control approximation sub-step.

[0028] The initial positioning reference establishment sub-step determines the approximate location of the hole to be mated through manual teaching, without requiring a high-precision vision positioning system. The manual teaching process involves the operator manually dragging the robot's end effector, moving the shaft to the area directly above the hole. The range of this area is determined based on the tolerance grade of the shaft-hole mating and the robot's repeatability accuracy, typically chosen at a certain height above the hole's end face to ensure the shaft covers all possible hole offset ranges during subsequent helical hole-finding. Once the shaft reaches the target position, the computer records its three-dimensional coordinates in the robot's base coordinate system, storing these coordinates as the initial reference point in the control system. After recording the initial reference point, the computer controls the robot's end effector to move the shaft to this initial reference point, completing the initial positioning.

[0029] The axial force control approximation sub-step involves, after initial positioning, controlling the shaft to move downwards in a direction perpendicular to the end face of the hole to be mated, until the shaft contacts the hole end face and generates a stable axial contact force. During this movement, the computer device continuously collects axial force data from a six-dimensional force sensor mounted on the robot's end effector. This six-dimensional force sensor can simultaneously detect forces and moments in three orthogonal directions in space; this step primarily utilizes its axial force detection function. When the collected axial force value reaches a preset contact threshold, the computer device immediately controls the shaft to stop moving downwards. Subsequently, the computer device enters a constant force control mode, continuously collecting axial force data from the six-dimensional force sensor. It compares the real-time collected axial force value with a preset target contact force, calculates the axial position correction of the robot's end effector based on the difference, and adjusts the axial position of the robot's end effector to ensure the axial contact force remains within the allowable error range of the preset target force. This creates a stable force interaction environment, ensuring uniform contact between the shaft and the hole end face during subsequent hole-finding processes, preventing workpiece damage due to excessive contact force or hole-finding failure due to insufficient contact force.

[0030] Step S102. Control the shaft to perform a spiral trajectory search in the horizontal plane and collect force and torque data from the six-dimensional force sensor in real time; when a sudden change in force characteristics is detected, determine that the shaft is in contact with the hole edge, and control the shaft to perform a slight swing to correct the axis tilt angle and reduce the insertion resistance.

[0031] Specifically, step S102 is the core positioning stage of the assembly process. Its core objective is to accurately locate the hole position through force feedback without visual assistance, and to correct the axis inclination of the shaft to reduce insertion resistance. This step includes two sub-steps: a spiral trajectory hole search sub-step and an attitude correction sub-step.

[0032] The helical trajectory hole-finding process involves a computer-controlled shaft moving in a helical trajectory from the inside out in a horizontal plane after a stable force interaction environment has been established. The parameters of the helical trajectory include the initial radius, pitch, and rotational speed, which are preset based on the diameter of the hole to be fitted, the clearance between the shaft and the hole, and the robot's motion accuracy. The initial radius is typically set smaller than the radius of the hole to ensure the helical motion starts from the center of the hole. The pitch is determined based on the clearance between the shaft and the hole, ensuring the distance between two adjacent helical rotations is less than the maximum clearance, preventing missed holes. The rotational speed is determined based on the data acquisition frequency of the six-dimensional force sensor and the robot's response speed, ensuring timely detection of sudden changes in force characteristics at the hole edge. During the helical motion of the shaft, the computer acquires three-dimensional force data and three-dimensional torque data output from the six-dimensional force sensor at a fixed frequency, monitoring the force interaction between the shaft and the hole end face in real time.

[0033] The attitude correction sub-step addresses the abrupt change in contact between the shaft and the hole edge during helical motion. This abrupt change causes a significant shift in the force and torque data detected by the six-dimensional force sensor. The computer calculates the difference between the real-time force and torque data and the corresponding data collected at the previous moment. When the calculated difference exceeds a pre-set abrupt change threshold, it is determined that the shaft has contacted the hole edge. At this point, the computer controls the shaft to stop helical motion and enters attitude correction mode. In attitude correction mode, the computer controls the shaft to perform micro-oscillations sequentially around two orthogonal axes in the horizontal plane. The oscillation angle range is pre-set according to the allowable tilt angle of the shaft-hole fit. During the oscillation, the computer collects the axial force data from the six-dimensional force sensor in real time and records the shaft attitude corresponding to the minimum axial force value. This attitude represents the highest degree of overlap between the shaft axis and the hole axis. Subsequently, the computer controls the shaft to adjust to this attitude, completing the correction of the axis tilt angle and effectively reducing insertion resistance during subsequent insertion.

[0034] Step S103. Control the shaft to be inserted along the axial direction, while continuously monitoring the insertion depth and axial contact force; if the insertion depth reaches the set value and the axial contact force is stable within the preset range within a preset time limit, the assembly is deemed qualified; otherwise, the assembly is deemed unsuccessful and the operation is terminated.

[0035] Specifically, step S103 is the final execution stage of the assembly process. Its core purpose is to complete the insertion operation of the shaft and accurately determine the assembly result through multiple criteria to ensure assembly quality. This step includes two sub-steps: an axial insertion sub-step and an assembly result determination sub-step.

[0036] The axial insertion sub-step involves the computer-controlled shaft being inserted downwards in a direction perpendicular to the end face of the hole to be fitted, after attitude correction. The insertion speed is pre-set based on the tolerance grade of the shaft-hole fit and the material of the workpiece, ensuring a smooth insertion process and preventing workpiece jamming or damage due to excessive insertion speed. During insertion, the computer reads encoder data from each joint of the robot in real time and calculates the current insertion depth of the shaft using the robot's kinematic model. Simultaneously, the computer collects axial force data from a six-dimensional force sensor at a fixed frequency, monitoring changes in axial contact force in real time during the insertion process.

[0037] The assembly result determination sub-step involves the computer device activating its internal timing module to begin timing simultaneously with the insertion operation. Within a pre-set time limit, the computer device continuously checks whether the insertion depth reaches the pre-set target insertion depth. If the insertion depth reaches the target depth, the computer device further checks whether the axial contact force remains within a preset acceptable range for a pre-set stabilization period. If both the insertion depth and axial contact force are simultaneously satisfied, the assembly is deemed successful, the computer device stops the shaft movement, and outputs an assembly success signal. If, within the pre-set time limit, the insertion depth does not reach the target insertion depth, or the axial contact force exceeds the preset acceptable range, the assembly is deemed a failure. The computer device immediately stops all movement of the shaft, controls the shaft to move axially upwards to the initial reference point, outputs an assembly failure signal, and awaits further processing by the operator.

[0038] In some embodiments, the step of establishing an initial positioning reference through manual teaching and guiding the shaft of the robot end to the vicinity of the hole to be mated includes: dragging the teaching robot end to a preset height position directly above the hole to be mated, recording the coordinates of the preset height position as the initial reference point, and controlling the robot end to move the shaft to the initial reference point position.

[0039] This embodiment specifically illustrates the method of establishing an initial positioning reference and guiding the shaft to the vicinity of the hole in step S101 through manual teaching. In this embodiment, the operator first switches the robot to drag teaching mode. In this mode, the robot's joints are in a compliant state, and the operator can easily drag the robot's end effector. The operator manually drags the robot's end effector, moving the shaft clamped on the end effector to directly above the hole to be mated. The height of the shaft is adjusted so that the distance between the lower end face of the shaft and the end face of the hole to be mated is a preset initial height value. This initial height value is determined according to the length of the shaft and the robot's range of motion, and is usually set between 5 mm and 20 mm. After the shaft is adjusted to the target position, the operator presses the record button on the teach pendant. The computer automatically records the X, Y, and Z three-dimensional coordinates of this position in the robot's base coordinate system and stores these coordinates as the initial reference point in the control system's memory. After recording, the operator switches the robot to automatic operation mode. The computer controls the robot's end effector to move the shaft to the stored initial reference point position, completing the initial positioning.

[0040] This embodiment establishes the initial positioning reference by dragging and teaching, which is simple and convenient to operate, requires no complicated calibration process, and can quickly complete the pre-assembly preparation work, making it suitable for small-batch, multi-variety production scenarios.

[0041] In some embodiments, the control shaft moves along a direction perpendicular to the end face of the hole to be mated until an axial contact force is detected, including: setting an axial movement speed, controlling the shaft to move downward along a direction perpendicular to the end face of the hole to be mated at the set speed, collecting axial force data from a six-dimensional force sensor in real time, and stopping the movement of the shaft when the axial force reaches a preset contact threshold.

[0042] In this embodiment, after initial positioning, the computer device first sets the axial movement speed. This speed is determined based on the robot's motion accuracy and the response speed of the six-dimensional force sensor, typically set between 1 mm / s and 5 mm / s. Subsequently, the computer device sends motion commands to the robot controller, controlling the shaft to move downwards at the set speed along a direction perpendicular to the end face of the hole to be mated. During the movement, the computer device collects axial force data output by the six-dimensional force sensor at a frequency of 100 Hz, comparing the real-time collected axial force value with a preset contact threshold. This contact threshold is determined based on the workpiece material and the weight of the shaft, typically set between 1 Newton and 5 Newtons. When the collected axial force value is greater than or equal to the preset contact threshold, the computer device immediately sends a stop command to the robot controller, controlling the shaft to stop moving downwards. This embodiment determines the contact between the shaft and the hole end face through force feedback. Compared to position control, this effectively avoids collision damage caused by workpiece position deviations, improving the safety of the assembly process.

[0043] In some embodiments, maintaining a constant axial contact force to maintain a stable force interaction environment includes: acquiring axial force data from a six-dimensional force sensor in real time, comparing the acquired axial force with a preset target force, and adjusting the axial position of the robot end effector based on the comparison result to keep the axial force within the allowable error range of the preset target force.

[0044] In this embodiment, after the shaft stops moving, the computer device enters a constant force control mode. In this mode, the computer device continuously acquires axial force data output from the six-dimensional force sensor at a frequency of 100 Hz, compares the real-time acquired axial force value with a pre-set target contact force, and calculates the difference between the two. Based on this difference, the computer device uses a proportional-integral-derivative (PID) control algorithm to calculate the axial position correction of the robot's end effector, converts the position correction into joint motion commands for the robot, and sends them to the robot controller. This controller then moves the shaft along the axial direction using the robot's end effector, adjusting its axial position. Through this closed-loop control process, the axial contact force is always maintained within the allowable error range of the preset target force, typically set to ±10% of the target contact force. This embodiment maintains a constant axial contact force through a closed-loop force control algorithm, ensuring that the shaft and the hole end face maintain uniform contact throughout the hole-finding process, avoiding decreased hole-finding accuracy due to contact force fluctuations, and improving the reliability of the hole-finding process.

[0045] In some embodiments, the control shaft performs a spiral trajectory search in the horizontal plane and collects force and torque data from the six-dimensional force sensor in real time, including: setting the initial radius, pitch and rotation speed of the spiral trajectory, controlling the shaft to perform a spiral motion from the inside out in the horizontal plane with the set parameters, and collecting three-dimensional force and three-dimensional torque data from the six-dimensional force sensor at a fixed frequency during the motion.

[0046] In this embodiment, after entering the hole-finding stage, the computer device first sets the parameters of the helical trajectory: the initial radius is set to 0.5 mm, the pitch is set to 0.2 mm, and the rotation speed is set to 30 degrees per second. These parameters are preset based on the working condition that the diameter of the hole to be fitted is 5 mm and the shaft-hole fit clearance is 0.1 mm. Subsequently, the computer device generates a sequence of trajectory points for the helical motion according to the set parameters, with each trajectory point containing X and Y coordinate values ​​in the horizontal plane. The computer device sends motion commands to the robot controller sequentially according to the trajectory point sequence, controlling the shaft to perform an inward-outward helical motion in the horizontal plane. During the motion, the computer device acquires three-dimensional force data and three-dimensional torque data output by the six-dimensional force sensor at a frequency of 200 Hz, storing the acquired data in an internal buffer for subsequent force characteristic abrupt change detection. This embodiment uses an inward-outward helical hole-finding trajectory, which, compared to linear or circular hole-finding trajectories, has the advantages of comprehensive hole-finding coverage and high efficiency, enabling the hole to be located in the shortest possible time.

[0047] In some embodiments, when a sudden change in force characteristics is detected, determining that the shaft is in contact with the hole edge and controlling the shaft to perform a slight oscillation to correct the axis tilt angle and reduce the insertion resistance includes: calculating the difference between the real-time collected force and torque data and the corresponding data at the previous moment; determining that the shaft is in contact with the hole edge when the difference exceeds a preset sudden change threshold; controlling the shaft to perform a slight oscillation around two orthogonal axes in the horizontal plane in sequence; collecting axial force data in real time during the oscillation; recording the shaft posture corresponding to the minimum axial force; and controlling the shaft to adjust to that posture.

[0048] In this embodiment, the computer device collects force and torque data while simultaneously calculating the difference between the current data and the previous data in real time. When the shaft is not in contact with the hole edge, the shaft and the hole end face are in surface contact, and the force and torque data change gradually with a small difference. When the shaft contacts the hole edge, the contact state changes to line contact or point contact, and the force and torque data change abruptly, with the difference increasing significantly. When the calculated force difference or torque difference exceeds a preset abrupt change threshold, the computer device determines that the shaft has contacted the hole edge and immediately stops the helical motion. Subsequently, the computer device controls the shaft to perform a slight oscillation around the X-axis in the horizontal plane, with an oscillation angle range of ±1 degree and an oscillation speed of 1 degree per second. During the oscillation, axial force data is collected in real time, and the X-axis angle corresponding to the minimum axial force is recorded. Next, the computer device controls the shaft to perform the same slight oscillation around the Y-axis in the horizontal plane, and the Y-axis angle corresponding to the minimum axial force is recorded. Finally, the computer device controls the shaft to adjust to the posture corresponding to the recorded X-axis and Y-axis angles, completing the correction of the axis tilt angle. This embodiment detects hole edge contact through force characteristic abrupt change and optimizes shaft posture through dual-axis micro-oscillation, which can effectively correct the shaft axis tilt angle, significantly reduce insertion resistance, and avoid jamming and workpiece damage caused by axis tilt.

[0049] In some embodiments, the control shaft is inserted axially while continuously monitoring the insertion depth and axial contact force, including: setting an axial insertion speed, controlling the shaft to be inserted downwards at a set speed in a direction perpendicular to the end face of the hole to be mated, reading robot joint encoder data in real time to calculate the current insertion depth, and simultaneously acquiring axial force data from a six-dimensional force sensor at a fixed frequency.

[0050] In this embodiment, after attitude correction, the computer device sets the axial insertion speed to 0.5 mm / s, which is preset based on the working condition of a shaft-hole fit tolerance grade of H7 / h6. Subsequently, the computer device sends an insertion command to the robot controller, controlling the shaft to be inserted downwards at the set speed in a direction perpendicular to the end face of the hole to be fitted. During insertion, the computer device reads encoder data from each joint of the robot at a frequency of 100 Hz, calculates the Z-coordinate value of the shaft end in the base coordinate system using the robot's forward kinematics model, and subtracts the current Z-coordinate value from the initial reference point to obtain the current insertion depth. Simultaneously, the computer device acquires axial force data from a six-dimensional force sensor at a frequency of 200 Hz to monitor changes in axial contact force during insertion in real time. This embodiment calculates the insertion depth using robot joint encoder data, eliminating the need for additional displacement sensors, reducing hardware costs, and ensuring the accuracy of insertion depth detection.

[0051] In some embodiments, the step of determining that the assembly is qualified if the insertion depth reaches a set value and the axial contact force is stable within a preset range within a preset time limit includes: starting the timing module to start timing, continuously judging whether the insertion depth reaches the set value within a preset time limit, if the insertion depth reaches the set value, continuing to judge whether the axial contact force is always within a preset range within a preset stable time period, and judging whether the assembly is qualified if the insertion depth reaches the set value and the axial contact force is always within a preset range within a preset stable time period.

[0052] In this embodiment, at the same time when the insertion operation starts, the computer device activates the timing module to start timing, and the preset limited time is 10 seconds. Within the limited time of 10 seconds, the computer device continuously determines whether the current insertion depth reaches the preset target insertion depth, which is set to 10 mm according to the requirements of shaft-hole fit. When the insertion depth reaches 10 mm, the computer device further determines whether the axial contact force is always within the preset qualified interval within a stable duration of 1 consecutive second, and the qualified interval is set to be between 2 Newtons and 8 Newtons. If the axial contact force is always between 2 Newtons and 8 Newtons within 1 second, it is determined that the current assembly is qualified, the computer device controls the shaft part to stop moving, and outputs an assembly qualified signal. This embodiment adopts three conditions of time, insertion depth and axial force stability to determine assembly qualification. Compared with the single insertion depth determination method, it can reflect the assembly quality more accurately and avoid the false qualification phenomenon caused by jamming.

[0053] In some embodiments, the step of otherwise determining that the assembly fails and terminating the operation includes: if the insertion depth does not reach the set value within the preset limited time, or the axial contact force exceeds the preset interval, all movements of the shaft part are stopped immediately, the shaft part is controlled to move axially upward to the position of the initial reference point, and an assembly failure signal is output.

[0054] In this embodiment, if the insertion depth does not reach 10 mm within the preset limited time of 10 seconds, or the axial contact force exceeds the qualified interval of 2 Newtons to 8 Newtons during the insertion process, it is determined that the current assembly fails. At this time, the computer device immediately sends an emergency stop instruction to the robot controller to stop all movements of the shaft part. Subsequently, the computer device controls the shaft part to move axially upward until it returns to the position of the initial reference point. Finally, the computer device outputs an assembly failure signal to the external production management system through the communication interface, and displays prompt information of assembly failure on the teach pendant, waiting for the operator to troubleshoot and handle the fault. This embodiment provides a complete assembly failure exception handling process, which can stop the equipment operation and reset in time when the assembly fails, avoid further damage to the equipment and the workpiece, and improve the safety and maintainability of the system.

[0055] In some embodiments, by solving the technical problems in the prior art that manual teaching is required for each hole in batch multi-hole assembly and the assembly efficiency is low, the present disclosure is applicable to multi-hole continuous operation scenarios such as circuit board connector insertion and box multi-bolt assembly.

[0056] The establishment of the multi-hole reference matrix is performed by the operator dragging and teaching the robot end to sequentially move to the preset height positions directly above all the to-be-mated hole positions on the workpiece, recording the three-dimensional coordinate values corresponding to each hole position respectively, and the computer device stores all the coordinate values as the multi-hole reference matrix according to the row and column sequence of the hole positions. At the same time, the computer device records the relative position deviation between adjacent hole positions, and generates a hole position deviation compensation table.

[0057] The first hole assembly is performed by controlling the end effector of a robot via computer equipment to move the shaft to the initial reference point position corresponding to the first hole position in the multi-hole reference matrix, and completing the assembly of the first hole position according to steps S101 to S103 as described in Examples 1-8.

[0058] Automatic hole position switching works by having the computer read the coordinate value of the next hole position from the multi-hole reference matrix after the first hole position is assembled, correcting the coordinate value in conjunction with the hole position deviation compensation table, and controlling the robot end effector to automatically move the shaft to the initial reference point position of the next hole position.

[0059] The continuous assembly cycle completes the assembly of the current hole position by repeatedly executing steps S101 to S103 until all holes in the multi-hole reference matrix are assembled. During the continuous assembly process, the computer equipment continuously monitors the assembly time, hole search time, and insertion resistance data for each hole position, and generates an assembly quality statistical report.

[0060] This embodiment achieves one-time teaching and automatic continuous assembly of batch hole positions by establishing a multi-hole reference matrix, eliminating the need for manual intervention for each hole and greatly improving the efficiency of batch assembly. At the same time, the hole position deviation compensation table corrects the hole position offset caused by processing errors, ensuring the consistency and pass rate of batch assembly.

[0061] In some embodiments, by solving the technical problems of fixed-parameter spiral hole searching in the prior art being unable to adapt to different shaft-hole fit clearances and having difficulty in balancing hole searching efficiency and accuracy, it is suitable for small-batch production scenarios with large fluctuations in shaft-hole fit tolerances.

[0062] The initial hole-finding parameters are set by computer equipment based on the nominal diameter and nominal clearance of the hole to be fitted, setting the initial radius, initial pitch and initial rotation speed of the helical trajectory as the starting parameters for the hole-finding process.

[0063] Force feedback fluctuation monitoring uses a computer to calculate the standard deviation of axial force data over 10 consecutive acquisition cycles as the force feedback fluctuation value during the helical motion of the shaft. When the force feedback fluctuation value is less than a preset stability threshold, the shaft is determined to be in planar contact with the hole end face; when the force feedback fluctuation value is greater than the preset stability threshold, the shaft is determined to be close to the hole edge area.

[0064] The hole-searching parameters are adaptively adjusted by increasing the pitch and rotation speed when the shaft is determined to be in planar contact. The pitch is adjusted to 1.5 times the initial pitch, and the rotation speed is adjusted to 2 times the initial rotation speed to improve hole-searching efficiency. When the shaft is determined to be close to the hole edge area, the pitch and rotation speed are decreased. The pitch is adjusted to 0.5 times the initial pitch, and the rotation speed is adjusted to 0.5 times the initial rotation speed to improve the accuracy of hole edge detection.

[0065] The hole edge contact determination is based on the detection of a force characteristic change exceeding a preset change threshold. When this occurs, the shaft is determined to be in contact with the hole edge, the spiral motion is stopped, and the shaft enters the attitude correction mode.

[0066] This embodiment uses force feedback fluctuation values ​​to determine the contact state of the shaft in real time and dynamically adjusts the parameters of the spiral hole search. While ensuring the accuracy of hole edge detection, it significantly shortens the hole search time. Compared with the fixed parameter hole search method, the hole search efficiency can be improved by more than 40%.

[0067] In some embodiments, by solving the technical problems of jamming and workpiece scratching that easily occur when inserting deep holes with a single force control parameter in the prior art, it is applicable to deep hole assembly scenarios with a depth greater than 5 times the shaft diameter.

[0068] The insertion process is divided into three stages—initial insertion, intermediate insertion, and final insertion—based on the depth of the hole to be fitted, using computer equipment. The initial insertion stage is from 0 to 20% of the hole depth, the intermediate insertion stage is from 20% to 80% of the hole depth, and the final insertion stage is from 80% to 100% of the hole depth.

[0069] The graded force control parameter settings involve setting different target contact forces and insertion speeds for different insertion stages. The initial insertion stage sets a higher target contact force and a faster insertion speed, with the target contact force being 1.2 times the normal insertion force and the insertion speed 1.5 times the normal insertion speed. The intermediate insertion stage sets a medium target contact force and insertion speed, consistent with the normal insertion parameters. The final insertion stage sets a lower target contact force and a slower insertion speed, with the target contact force being 0.7 times the normal insertion force and the insertion speed 0.3 times the normal insertion speed.

[0070] The phase switching and force control execution involve the computer device calculating the current insertion depth in real time during insertion and automatically switching the corresponding force control parameters based on the insertion depth. Within each phase, the axial contact force is maintained stably according to the constant force control method described in Example 3.

[0071] The jamming warning and handling system monitors the rate of change of axial contact force in real time during the insertion process. When the rate of change of axial contact force exceeds the preset jamming threshold, it is determined that the shaft is jamming. The insertion speed is immediately reduced and the target contact force is increased. If the jamming trend is not eliminated, the shaft is controlled to retract 1 to 2 mm axially upwards, and then the insertion operation is re-executed.

[0072] This embodiment adapts to the resistance characteristics at different stages of deep hole insertion through a graded force control strategy, effectively avoiding jamming and workpiece scratches during deep hole assembly; at the same time, the jamming warning and automatic processing mechanism further improves the success rate of deep hole assembly.

[0073] In some embodiments, by addressing the technical problem in the prior art that assembly failures can only be addressed through manual intervention and cannot be automatically adjusted for retrying, this method is applicable to unattended automated production line scenarios.

[0074] The failure cause classification and determination system classifies the cause of failure based on the force data and insertion depth data at the time of failure when assembly failure is determined. If the insertion depth does not reach the set value and the axial contact force continues to increase, it is determined that the hole position is not found; if the insertion depth does not reach the set value and the axial contact force suddenly increases, it is determined that the axis tilt angle is too large, causing jamming; if the insertion depth reaches the set value but the axial contact force exceeds the preset range, it is determined that the fit is too tight.

[0075] Adaptive parameter adjustment automatically adjusts the corresponding assembly parameters based on different failure reasons. If the failure is determined to be due to a missing hole, the maximum radius of the spiral hole search and the search time limit are increased; if the failure is determined to be due to excessive axis tilt causing jamming, the angle range and number of swings of the micro-oscillation are increased; if the failure is determined to be due to excessive tightness, the insertion speed is reduced and the target contact force is increased.

[0076] After the automatic retry execution is completed and the parameters are adjusted, the computer equipment controls the shaft to return to the initial reference point position and re-executes the assembly process from steps S101 to S103 according to the adjusted parameters. A maximum of 3 retries are allowed for a single assembly failure. If all 3 retries fail, a serious fault signal is output and the operation is stopped.

[0077] The parameter self-learning storage mechanism stores the adjusted parameters, along with the corresponding workpiece number, shaft-hole fit tolerances, and other information, in a parameter database upon successful re-assembly. Subsequent assembly of workpieces of the same specifications will automatically utilize the optimal parameters stored in the database.

[0078] This embodiment achieves automatic retry after assembly failure by classifying and judging the reasons for failure and adjusting adaptive parameters, reducing the number of manual interventions; at the same time, through the parameter self-learning mechanism, the assembly parameters are continuously optimized, so that the assembly success rate gradually increases with the number of production runs.

[0079] In some embodiments, by solving the technical problem that the prior art can only realize the assembly of circular holes and cannot adapt to irregular holes such as square holes, it is applicable to the assembly scenarios of irregular shaft holes such as square pins and spline shafts.

[0080] Initial positioning and force control approximation are completed by following the steps described in Examples 1-3, so that the shaft and the end face of the square hole form a stable force interaction environment.

[0081] The square spiral trajectory hole search is generated by computer equipment, forming a square spiral trajectory from the inside out. Each turn of the trajectory is a square, and the difference in side length between two adjacent squares is the preset square pitch. The control shaft moves along the square spiral trajectory, and force and torque data from a six-dimensional force sensor are collected in real time during the movement.

[0082] The square hole edge and corner identification method works by detecting a unidirectional sudden change in torque detected by the force sensor when the shaft contacts the edge of the square hole; and a bidirectional sudden change in torque detected when the shaft contacts the corner of the square hole. The computer equipment identifies whether the shaft is contacting the edge or the corner of the hole based on the characteristics of the torque sudden change and records the location of the contact point.

[0083] Attitude correction is achieved by the computer calculating the direction of the line connecting two adjacent hole edges after identifying the contact points. This direction corresponds to the direction of one side of the square hole. The control shaft rotates around its axis to align the corresponding side of the square shaft with the side of the square hole. Subsequently, the control shaft makes slight oscillations around two orthogonal axes in the horizontal plane to correct the axis tilt angle.

[0084] Insertion and result determination are performed by following the steps described in Examples 6-8.

[0085] This embodiment achieves visual-free precision assembly of square holes through square spiral trajectory hole searching and torque feature recognition technology, expanding the scope of application of the present invention. Compared with circular hole assembly, only the hole searching trajectory and attitude correction logic need to be modified, without adding additional hardware costs.

[0086] In some embodiments, this invention addresses the technical challenges of large workpiece shaft and hole assembly being impossible with a single robot in the prior art, and the high precision requirements for workpiece positioning. It is applicable to assembly scenarios for heavy workpieces such as automobile chassis and large mechanical equipment.

[0087] Dual-robot coordinate system calibration is performed using calibration tools to calibrate the coordinate systems of the master and slave robots, establishing the transformation relationship between the two robot coordinate systems. The master robot is used to clamp shafts, and the slave robot is used to clamp workpieces with holes.

[0088] Initial relative positioning is achieved by the operator dragging the shaft at the end of the master robot and the workpiece with holes at the end of the slave robot to roughly align the shaft with the holes, and then recording the position coordinates of the master robot and the slave robot at this point as the initial relative position.

[0089] The collaborative force control approximation method involves the main robot controlling the shaft to remain stationary, while the slave robot controls the perforated workpiece to move towards the shaft along a direction perpendicular to its axis until an axial contact force is detected. The slave robot then enters a constant force control mode to maintain a constant axial contact force.

[0090] The collaborative spiral hole-finding and attitude correction system involves a slave robot controlling a workpiece with a hole to perform a spiral trajectory search in a horizontal plane, while the master robot collects force and torque data from a six-dimensional force sensor in real time. When a sudden change in force characteristics is detected, the master and slave robots work together to adjust the attitude. The master robot controls the shaft to make a slight oscillation, while the slave robot controls the workpiece with the hole to perform corresponding position compensation and correct the axis tilt angle.

[0091] Collaborative insertion and result determination are achieved by moving the perforated workpiece axially toward the shaft under robot control. The main robot continuously monitors the axial contact force and insertion depth, and determines the assembly result according to the criteria described in Examples 7-8.

[0092] This embodiment achieves vision-free precision assembly of large workpieces through dual-robot collaborative control, reducing the requirements for workpiece positioning accuracy; at the same time, the force sensing unit is deployed on the main robot end that holds the shaft, ensuring the accuracy and response speed of force feedback.

[0093] In some embodiments, the method further includes a hole type and contact state intelligent recognition process based on deep temporal networks, which solves the technical problems of low accuracy, easy omission and misjudgment, and inability to automatically adapt to multiple hole types in traditional threshold judgment methods for complex hole types and complex contact states. It is suitable for flexible assembly scenarios with multiple product lines and multiple shaft hole matings including round, square, spline, and irregular hole types. The specific implementation is as follows.

[0094] A pre-constructed force-sensor time-series dataset was used to collect six-dimensional force and torque data for the entire assembly process under different hole types, mating clearances, and posture deviations. Each time segment was labeled with a corresponding contact state label, including more than ten states such as planar sliding, hole edge contact, single-sided jamming, two-point contact, hole bottom contact, and foreign object obstruction. Hole type category labels were also added. The dataset covers different materials, surface roughnesses, and assembly speeds, ensuring sample diversity and scenario coverage.

[0095] A multi-task recognition model is constructed using a one-dimensional convolutional neural network combined with a long short-term memory network. The model takes six-dimensional force and torque time-series data with a fixed time window as input and outputs the prediction results of two branches: one branch outputs the current contact state category and confidence level, and the other branch outputs the current aperture type category and confidence level. The model is trained offline using a labeled dataset, autonomously learning the force perception time-series features corresponding to different contact states and aperture types. After training, the model is deployed to the control processing unit.

[0096] During actual assembly, the control processing unit continuously inputs the six-dimensional force and torque data within the current window into the recognition model in the form of a sliding time window, and outputs the current contact state and hole shape recognition results in real time. When different hole shapes are identified, the corresponding hole search trajectory and attitude correction strategy are automatically matched: when a circular hole is identified, Archimedes spiral hole search and dual-axis attitude swing correction are used; when a square hole is identified, it automatically switches to square spiral hole search and spiral alignment correction; when a spline hole is identified, it switches to tooth groove detection trajectory and circumferential alignment strategy, without requiring manual switching of process parameters throughout the process.

[0097] When an abnormal contact state is detected, corresponding differentiated processing logic is triggered: if it is detected as unilateral jamming, a small-amplitude reciprocating retraction and attitude correction are automatically performed; if it is detected as bottoming out of the hole, insertion is terminated and the depth is determined to be in place; if it is detected as foreign object obstruction, insertion is immediately stopped and an anomaly is reported. When the recognition confidence level is lower than a set threshold, the movement speed is automatically reduced and the sampling density is increased, and the normal speed is restored after the confidence level recovers.

[0098] Intelligent recognition through deep learning can replace traditional multi-threshold judgment logic, accurately distinguish complex and diverse contact states and hole types, greatly improve the adaptability and anomaly recognition accuracy of multi-variety mixed-line assembly, and reduce assembly failures caused by misjudgment.

[0099] In some embodiments, the method further includes an assembly trajectory pre-simulation and online dynamic correction process based on a digital twin model, which solves the technical problems of traditional process parameters relying on manual trial and error debugging, long new product introduction cycles, and the inability to predict risks during the assembly process. It is suitable for high-precision, small-batch, multi-category precision assembly production lines. The specific implementation is as follows.

[0100] A digital twin model of the assembly station is constructed, comprising a robot kinematics digital twin, a dynamic model of the end effector load and force sensor, geometric models of shafts and holes, a contact mechanics simulation model, and an environmental disturbance parameter model. This model can simulate force response under different poses, feed rates, and attitude deviations, and predict force characteristic changes during hole searching and jamming risks during insertion. Model parameters can be calibrated using historical assembly data to gradually approximate real physical characteristics.

[0101] Before assembling any new product specification, a virtual assembly simulation is performed in a digital twin environment. Theoretical hole positions, shaft and hole dimensions, fit tolerances, and initial process parameters are input. The complete spiral hole-finding, attitude correction, and insertion process is simulated in the twin model, outputting the predicted force curves for each stage, estimated hole-finding time, potential jamming locations, and assembly risk points. Based on the simulation results, the pitch, speed, attitude adjustment step size, and insertion speed curves of the spiral hole-finding process are automatically optimized to proactively avoid jamming risks, forming an optimized assembly process package, which is then sent to the physical robot for actual assembly.

[0102] During actual assembly, real-time robot pose data and six-dimensional force data are synchronously mapped to the digital twin model, driving the twin model to run synchronously with the physical assembly. Based on the real-world force data, physical parameters such as contact stiffness and friction coefficient in the twin model are corrected online, making the twin model's output gradually approach the real working conditions. At the same time, based on the twin model's prediction results, the force change trend in the next stage is predicted in advance, and the posture and feed speed of the shaft are finely adjusted in advance. For example, when a jamming trend is predicted, the axis tilt angle is slightly corrected in advance to avoid jamming.

[0103] Each time an assembly is completed, the parameters of the digital twin model are iteratively optimized using the full-process data from that assembly. The model's accuracy continuously improves as production data accumulates. Through digital twin virtual-physical mapping and pre-simulation optimization, the number of physical trial and error attempts in new product assembly can be significantly reduced, the process debugging cycle can be shortened, and assembly risks can be predicted and intervened in advance, further improving the success rate of precision assembly on the first attempt.

[0104] In some embodiments, the method further includes a rapid transfer and adaptation process for process parameters based on transfer learning, which solves the technical problems of requiring a large number of trial and error samples to form a stable process for the assembly of new products and new specifications of shafts and holes, and the low efficiency of new product introduction. It is suitable for flexible assembly scenarios with multiple varieties, small batches, and rapid product iteration. The specific implementation is as follows.

[0105] Mature shaft and hole products with large production volume and sufficient data accumulation are selected as the source domain. Based on their massive historical assembly data, the source domain process parameter optimization model is trained. The model input includes product attributes such as shaft diameter, hole diameter, fit tolerance, material, and surface roughness. The output includes the optimal hole search parameters, force control parameters, posture adjustment parameters, and insertion parameters. The source domain model has mature parameter mapping capabilities and assembly effect prediction capabilities.

[0106] When introducing a new target product specification, a transfer learning strategy is used to transfer the source domain model to the target domain: First, the product feature spaces of the source domain and the target domain are aligned through feature transfer, preserving the general knowledge related to assembly mechanical properties in the model; then, only a small number of trial assembly samples of the target product are needed to fine-tune the top-level parameters of the model, quickly obtaining a process parameter prediction model adapted to the target product, without having to accumulate data from scratch.

[0107] Before formal production, the effectiveness of the parameters was verified through a small-scale trial assembly. Based on the trial assembly results, the model parameters were further fine-tuned to quickly achieve a stable assembly success rate. During production, the migration model was continuously iterated and optimized using actual assembly data of the target product, and the assembly effect continued to improve as the production volume increased.

[0108] For products of different specifications within the same series, initial process parameters can be directly output based on the transfer model, skipping the manual debugging stage and significantly shortening the process introduction cycle for new products. Simultaneously, a product family parameter transfer spectrum is established, allowing products within the same series to share the transferred model parameters, further improving the efficiency of process development for multiple product varieties. Through transfer learning, process knowledge can be reused across products, enabling rapid process adaptation of new product specifications with minimal sample size, significantly reducing the debugging costs and introduction cycle for new product assembly.

[0109] In some embodiments, the method further includes a federated learning process optimization process for multi-robot assembly stations, which solves the technical problems of data silos among multiple assembly robots on the production line, limited data volume per station making it difficult to train high-precision models, and inconvenience in centralized uploading of production data, which poses data security risks. It is suitable for large-scale assembly production lines with parallel operations of multiple stations, and the specific implementation is as follows.

[0110] Each assembly robot on the production line acts as a local node, maintaining its own local assembly force and process parameter data. All raw production data is stored on the local controller and is not uploaded to the central server. A federated coordination node is set up to be responsible for the aggregation and distribution of the global model, without ever touching the raw production data, thus ensuring data privacy and security.

[0111] The model training employs a horizontal federated learning framework: each local node trains a contact state recognition model and a process parameter optimization model locally based on its own local assembly data. After training, the model's parameter gradients are uploaded to the federated coordination node. The coordination node performs weighted aggregation of the gradients from all nodes to generate an updated global model, which is then distributed to each local node. Each node updates its local model with the distributed global model, completing one round of training. After multiple iterations, the global model integrates the data features from all workstations, exhibiting superior generalization ability and recognition accuracy compared to single-workstation models.

[0112] Each workstation uses a global model locally to perform contact state identification and process parameter optimization, while continuously fine-tuning the model with local production data to adapt to the equipment characteristics and operating conditions of its specific workstation. During model updates, all original production data remains locally, with only model parameter gradients being transmitted, complying with data security and privacy protection requirements.

[0113] Simultaneously, it can achieve unified optimization of process parameters across the entire production line based on a global model, synchronizing the optimal parameters verified across the entire production line to all workstations, thereby improving the assembly success rate and quality consistency of the entire line. Through the distributed optimization mode of federated learning, it can aggregate the data value of multiple workstations while protecting data privacy, continuously improving the assembly intelligence level of the entire line and solving the problems of insufficient data at single workstations and limited model accuracy.

[0114] In some embodiments, the method further includes an impedance parameter adaptive tuning process based on deep reinforcement learning, which solves the technical problems of high insertion resistance, easy jamming, and the inability of fixed impedance parameters to adapt to the dynamically changing contact state during the insertion process, resulting in low assembly success rate. It is suitable for press-fitting scenarios of small-gap interference fits and high-precision shaft holes. The specific implementation is as follows.

[0115] A reinforcement learning-based compliant control agent is constructed, using real-time observations of six-dimensional force and torque, current insertion depth, and insertion velocity as the state space, and adjustments to virtual mass, virtual damping, and virtual stiffness parameters of the admittance control model as the action space. The reward function uses assembly success rate, insertion time, maximum contact force, and number of jams as comprehensive indicators: successful assembly with a smooth insertion process and low maximum contact force results in a positive reward; jams, force exceeding the safety threshold, or assembly failure result in a negative reward.

[0116] The agent is pre-trained in a simulation environment to simulate interference fit scenarios with different interference amounts, initial posture deviations, and surface friction coefficients. This allows the agent to learn a strategy of dynamically adjusting impedance parameters based on real-time contact states, automatically improving compliance when insertion resistance increases and ensuring insertion efficiency when resistance is stable. After pre-training, the agent is deployed to the robot controller.

[0117] During actual insertion, the agent collects the current force and pose states in real time, dynamically outputs the impedance parameter adjustment amount, and corrects the parameters of the admittance control model in real time: in the initial alignment stage of insertion, the damping is appropriately reduced to improve the response speed and ensure alignment efficiency; after entering the interference fit stage, the damping and stiffness are gradually increased with the increase of insertion depth to adapt to the gradually increasing insertion resistance; when the lateral force increases and a jamming trend is detected, the compliance in the corresponding direction is quickly adjusted, and the feed rate is slightly reduced to resolve the jamming risk.

[0118] After assembly, the agent undergoes online incremental learning based on the assembly results data to continuously optimize the control strategy. By using reinforcement learning to adaptively tune impedance parameters, it can replace fixed-parameter compliant control, precisely adapting to dynamic force changes throughout the interference fit process. This effectively reduces the jamming rate and workpiece scratch rate of the interference fit, improving the quality and success rate of precision interference fit assembly.

[0119] Please see Figure 3 As shown, Figure 3 This is a schematic diagram of the structure of a force feedback-based robot vision-free precision assembly system 200 provided in this application embodiment. The force feedback-based robot vision-free precision assembly system 200 is used to execute the steps of the force feedback-based robot vision-free precision assembly method shown in the above embodiments. The force feedback-based robot vision-free precision assembly system 200 can be a single server or a server cluster, or it can be a terminal, such as a handheld terminal, a laptop computer, a wearable device, or a robot.

[0120] like Figure 3 As shown, the visionless precision assembly system 200 for robots based on force feedback includes: The reference holding unit is used to establish an initial positioning reference through manual teaching, guide the shaft of the robot end to the vicinity of the hole to be mated, control the shaft to move in a direction perpendicular to the end face of the hole to be mated until an axial contact force is detected, and maintain the axial contact force constant to maintain a stable force interaction environment. The resistance reduction unit is used to control the shaft to perform a spiral trajectory search in the horizontal plane and to collect force and torque data from the six-dimensional force sensor in real time. When a sudden change in force characteristics is detected, the unit determines the contact hole edge of the shaft and controls the shaft to swing slightly to correct the axis tilt angle and reduce insertion resistance. The termination unit is used to control the insertion of the shaft along the axial direction, while continuously monitoring the insertion depth and axial contact force. If the insertion depth reaches the set value and the axial contact force is stable within the preset range within a preset time, the assembly is deemed qualified; otherwise, the assembly is deemed unsuccessful and the operation is terminated.

[0121] In some embodiments, the step of establishing an initial positioning reference through manual teaching and guiding the shaft of the robot end to the vicinity of the hole to be mated includes: dragging the teaching robot end to a preset height position directly above the hole to be mated, recording the coordinates of the preset height position as the initial reference point, and controlling the robot end to move the shaft to the initial reference point position.

[0122] In some embodiments, the control shaft moves along a direction perpendicular to the end face of the hole to be mated until an axial contact force is detected, including: setting an axial movement speed, controlling the shaft to move downward along a direction perpendicular to the end face of the hole to be mated at the set speed, collecting axial force data from a six-dimensional force sensor in real time, and stopping the movement of the shaft when the axial force reaches a preset contact threshold.

[0123] In some embodiments, maintaining a constant axial contact force to maintain a stable force interaction environment includes: acquiring axial force data from a six-dimensional force sensor in real time, comparing the acquired axial force with a preset target force, and adjusting the axial position of the robot end effector based on the comparison result to keep the axial force within the allowable error range of the preset target force.

[0124] In some embodiments, the control shaft performs a spiral trajectory search in the horizontal plane and collects force and torque data from the six-dimensional force sensor in real time, including: setting the initial radius, pitch and rotation speed of the spiral trajectory, controlling the shaft to perform a spiral motion from the inside out in the horizontal plane with the set parameters, and collecting three-dimensional force and three-dimensional torque data from the six-dimensional force sensor at a fixed frequency during the motion.

[0125] In some embodiments, when a sudden change in force characteristics is detected, determining that the shaft is in contact with the hole edge and controlling the shaft to perform a slight oscillation to correct the axis tilt angle and reduce the insertion resistance includes: calculating the difference between the real-time collected force and torque data and the corresponding data at the previous moment; determining that the shaft is in contact with the hole edge when the difference exceeds a preset sudden change threshold; controlling the shaft to perform a slight oscillation around two orthogonal axes in the horizontal plane in sequence; collecting axial force data in real time during the oscillation; recording the shaft posture corresponding to the minimum axial force; and controlling the shaft to adjust to that posture.

[0126] In some embodiments, the control shaft is inserted axially while continuously monitoring the insertion depth and axial contact force, including: setting an axial insertion speed, controlling the shaft to be inserted downwards at a set speed in a direction perpendicular to the end face of the hole to be mated, reading robot joint encoder data in real time to calculate the current insertion depth, and simultaneously acquiring axial force data from a six-dimensional force sensor at a fixed frequency.

[0127] In some embodiments, the step of determining that the assembly is qualified if the insertion depth reaches a set value and the axial contact force is stable within a preset range within a preset time limit includes: starting the timing module to start timing, continuously judging whether the insertion depth reaches the set value within a preset time limit, if the insertion depth reaches the set value, continuing to judge whether the axial contact force is always within a preset range within a preset stable time period, and judging whether the assembly is qualified if the insertion depth reaches the set value and the axial contact force is always within a preset range within a preset stable time period.

[0128] In some embodiments, the step of otherwise determining assembly failure and terminating the operation includes: if the insertion depth does not reach the set value within a preset time limit, or the axial contact force exceeds the preset range, immediately stopping all movement of the shaft, controlling the shaft to move axially upward to the initial reference point position, and outputting an assembly failure signal.

[0129] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the above-described force feedback-based robot vision-free precision assembly system and its modules can be found in the corresponding embodiments of the force feedback-based robot vision-free precision assembly method, and will not be repeated here.

[0130] The aforementioned method for visionless precision assembly of robots based on force feedback can be implemented as a computer program, which can be used in various applications such as... Figure 3 It runs on the system shown.

[0131] Please see Figure 4 , Figure 4 This is a schematic block diagram of the structure of a computer device provided in an embodiment of this application. The computer device includes a processor, a memory, and a network interface connected via a device bus, wherein the memory may include a storage medium and internal memory.

[0132] The storage medium can store operating devices and computer programs. The computer program includes program instructions that, when executed, cause the processor to perform any force feedback-based, vision-free precision assembly method for robots.

[0133] The processor provides computing and control capabilities, supporting the operation of the entire computer device.

[0134] Internal memory provides an environment for the execution of computer programs in non-volatile storage media. When executed by a processor, the computer program enables the processor to perform any vision-free precision assembly method for robots based on force feedback.

[0135] This network interface is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that... Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the terminal to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0136] It should be understood that the processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, a general-purpose processor can be a microprocessor or any conventional processor.

[0137] In one embodiment, the processor is configured to run a computer program stored in memory to perform the following steps: An initial positioning reference is established through manual teaching, and the shaft at the end of the robot is guided to the vicinity of the hole to be mated. The shaft is controlled to move in a direction perpendicular to the end face of the hole to be mated until an axial contact force is detected. The axial contact force is kept constant to maintain a stable force interaction environment. The control shaft performs a spiral trajectory search in the horizontal plane, and collects force and torque data from the six-dimensional force sensor in real time; when a sudden change in force characteristics is detected, the shaft is determined to be in contact with the hole edge, and the shaft is controlled to swing slightly to correct the axis tilt angle and reduce insertion resistance; The control shaft is inserted axially while continuously monitoring the insertion depth and axial contact force. If the insertion depth reaches the set value and the axial contact force is stable within the preset range within a preset time, the assembly is deemed qualified; otherwise, the assembly is deemed unsuccessful and the operation is terminated.

[0138] In some embodiments, the step of establishing an initial positioning reference through manual teaching and guiding the shaft of the robot end to the vicinity of the hole to be mated includes: dragging the teaching robot end to a preset height position directly above the hole to be mated, recording the coordinates of the preset height position as the initial reference point, and controlling the robot end to move the shaft to the initial reference point position.

[0139] In some embodiments, the control shaft moves along a direction perpendicular to the end face of the hole to be mated until an axial contact force is detected, including: setting an axial movement speed, controlling the shaft to move downward along a direction perpendicular to the end face of the hole to be mated at the set speed, collecting axial force data from a six-dimensional force sensor in real time, and stopping the movement of the shaft when the axial force reaches a preset contact threshold.

[0140] In some embodiments, maintaining a constant axial contact force to maintain a stable force interaction environment includes: acquiring axial force data from a six-dimensional force sensor in real time, comparing the acquired axial force with a preset target force, and adjusting the axial position of the robot end effector based on the comparison result to keep the axial force within the allowable error range of the preset target force.

[0141] In some embodiments, the control shaft performs a spiral trajectory search in the horizontal plane and collects force and torque data from the six-dimensional force sensor in real time, including: setting the initial radius, pitch and rotation speed of the spiral trajectory, controlling the shaft to perform a spiral motion from the inside out in the horizontal plane with the set parameters, and collecting three-dimensional force and three-dimensional torque data from the six-dimensional force sensor at a fixed frequency during the motion.

[0142] In some embodiments, when a sudden change in force characteristics is detected, determining that the shaft is in contact with the hole edge and controlling the shaft to perform a slight oscillation to correct the axis tilt angle and reduce the insertion resistance includes: calculating the difference between the real-time collected force and torque data and the corresponding data at the previous moment; determining that the shaft is in contact with the hole edge when the difference exceeds a preset sudden change threshold; controlling the shaft to perform a slight oscillation around two orthogonal axes in the horizontal plane in sequence; collecting axial force data in real time during the oscillation; recording the shaft posture corresponding to the minimum axial force; and controlling the shaft to adjust to that posture.

[0143] In some embodiments, the control shaft is inserted axially while continuously monitoring the insertion depth and axial contact force, including: setting an axial insertion speed, controlling the shaft to be inserted downwards at a set speed in a direction perpendicular to the end face of the hole to be mated, reading robot joint encoder data in real time to calculate the current insertion depth, and simultaneously acquiring axial force data from a six-dimensional force sensor at a fixed frequency.

[0144] In some embodiments, the step of determining that the assembly is qualified if the insertion depth reaches a set value and the axial contact force is stable within a preset range within a preset time limit includes: starting the timing module to start timing, continuously judging whether the insertion depth reaches the set value within a preset time limit, if the insertion depth reaches the set value, continuing to judge whether the axial contact force is always within a preset range within a preset stable time period, and judging whether the assembly is qualified if the insertion depth reaches the set value and the axial contact force is always within a preset range within a preset stable time period.

[0145] In some embodiments, the step of otherwise determining assembly failure and terminating the operation includes: if the insertion depth does not reach the set value within a preset time limit, or the axial contact force exceeds the preset range, immediately stopping all movement of the shaft, controlling the shaft to move axially upward to the initial reference point position, and outputting an assembly failure signal.

[0146] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to perform the steps of the visionless precision assembly method for robots based on force feedback as provided in any embodiment of this application.

[0147] The computer-readable storage medium may be an internal storage unit of the computer device described in the foregoing embodiments, such as the hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, SmartMedia Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the computer device.

[0148] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A vision-free precision assembly method for robots based on force feedback, characterized in that, include: An initial positioning reference is established through manual teaching, and the shaft at the end of the robot is guided to the vicinity of the hole to be mated. The control shaft moves along a direction perpendicular to the end face of the hole to be fitted until an axial contact force is detected. The axial contact force is kept constant to maintain a stable force interaction environment. The control shaft performs a spiral trajectory search in the horizontal plane, and collects force and torque data from the six-dimensional force sensor in real time; when a sudden change in force characteristics is detected, the shaft is determined to be in contact with the hole edge, and the shaft is controlled to swing slightly to correct the axis tilt angle and reduce insertion resistance; The control shaft is inserted axially while continuously monitoring the insertion depth and axial contact force. If the insertion depth reaches the set value and the axial contact force is stable within the preset range within a preset time, the assembly is deemed qualified; otherwise, the assembly is deemed unsuccessful and the operation is terminated.

2. The method according to claim 1, characterized in that, The step of establishing an initial positioning reference through manual teaching and guiding the robot's end effector to the vicinity of the hole to be mated includes: Drag the end effector of the teaching robot to a preset height position directly above the hole to be fitted, record the coordinates of the preset height position as the initial reference point, and control the end effector of the robot to move the shaft to the initial reference point position.

3. The method according to claim 1, characterized in that, The control shaft moves along a direction perpendicular to the end face of the hole to be fitted until an axial contact force is detected, including: Set the axial movement speed and control the shaft to move downwards at the set speed in a direction perpendicular to the end face of the hole to be mated. Collect the axial force data of the six-dimensional force sensor in real time. Stop the movement of the shaft when the axial force reaches the preset contact threshold.

4. The method according to claim 3, characterized in that, Maintaining a constant axial contact force to preserve a stable force interaction environment includes: The robot collects axial force data from a six-dimensional force sensor in real time, compares the collected axial force with a preset target force, and adjusts the axial position of the robot end effector based on the comparison results to keep the axial force within the allowable error range of the preset target force.

5. The method according to claim 1, characterized in that, The control shaft performs a spiral trajectory search in the horizontal plane, and collects force and torque data from the six-dimensional force sensor in real time, including: The initial radius, pitch, and rotation speed of the helical trajectory are set, and the shaft is controlled to perform an inward-outward helical motion in the horizontal plane according to the set parameters. During the motion, three-dimensional force and three-dimensional torque data of the six-dimensional force sensor are collected at a fixed frequency.

6. The method according to claim 1, characterized in that, When a sudden change in force characteristics is detected, the contact hole edge of the shaft is determined, and the shaft is controlled to make a slight oscillation to correct the axis tilt angle and reduce the insertion resistance, including: The force and torque data collected in real time are compared with the corresponding data at the previous moment. When the difference exceeds the preset sudden change threshold, the shaft is determined to be in contact with the hole edge. The shaft is controlled to swing slightly around two orthogonal axes in the horizontal plane. Axial force data is collected in real time during the swing. The shaft posture corresponding to the minimum axial force is recorded and the shaft is controlled to adjust to that posture.

7. The method according to claim 1, characterized in that, The control shaft is inserted axially while continuously monitoring the insertion depth and axial contact force, including: Set the axial insertion speed, control the shaft to be inserted downwards at the set speed in a direction perpendicular to the end face of the hole to be mated, read the robot joint encoder data in real time to calculate the current insertion depth, and at the same time collect the axial force data of the six-dimensional force sensor at a fixed frequency.

8. The method according to claim 1, characterized in that, Within the preset limited time, if the insertion depth reaches the set value and the axial contact force is stable within the preset interval, it is determined that the assembly is qualified, comprising: Start the timing module to start timing, and continuously judge whether the insertion depth reaches the set value within the preset limited time. If the insertion depth reaches the set value, further judge whether the axial contact force is always within the preset interval within the preset stable duration. If the insertion depth reaches the set value and the axial contact force is always within the preset interval within the preset stable duration, it is determined that the assembly is qualified.

9. The method according to claim 8, characterized in that, Otherwise, it is determined that the assembly fails and the operation is terminated, comprising: If the insertion depth does not reach the set value within the preset limited time, or the axial contact force exceeds the preset interval, all movements of the shaft part are immediately stopped, the shaft part is controlled to move axially upward to the initial reference point position, and an assembly failure signal is output.

10. A vision-free precision assembly system for robots based on force feedback, used to implement the method as described in any one of claims 1-9, characterized in that, comprising: A reference maintaining unit, configured to establish an initial positioning reference through manual teaching, and guide the shaft part at the end of the robot to the area near the hole to be mated; Control the shaft part to move in the direction perpendicular to the end face of the hole to be mated until the axial contact force is detected, and keep the axial contact force constant to maintain a stable force interaction environment; A resistance reduction unit, configured to control the shaft part to perform spiral trajectory search in the horizontal plane, and collect force and torque data of the six-dimensional force sensor in real time; when a sudden change in the force characteristic is detected, it is determined that the shaft part contacts the hole edge, and the shaft part is controlled to perform small-amplitude swing to correct the axis inclination angle and reduce the insertion resistance; An operation termination unit, configured to control the shaft part to perform an insertion operation along the axial direction, and continuously monitor the insertion depth and the axial contact force at the same time; within the preset limited time, if the insertion depth reaches the set value and the axial contact force is stable within the preset interval, it is determined that the assembly is qualified, otherwise it is determined that the assembly fails and the operation is terminated.

Images