A robot stable contact force control method for assembly scene

Abstract

The application relates to the technical field of robot automatic assembly, and provides a robot stable contact force control method for an assembly scene. The method comprises the following steps: performing zero-point calibration operation on a six-dimensional force sensor to obtain sensor zero values, and then performing load gravity compensation processing on sensor detection data; a robot end is controlled to drive a to-be-assembled shaft piece to move to a target hole position adjacent area, and the to-be-assembled shaft piece is driven to perform feeding movement along an axial direction towards a workpiece fitting surface; contact force data of the robot end is collected in real time through the six-dimensional force sensor, and real-time values of the axial contact force are continuously extracted; according to the real-time axial contact force values, the axial position and the feeding speed of the robot end are dynamically adjusted through the mobility control logic, the axial contact force is stably maintained in a preset value range, a stable axial force interaction environment is formed, and a reliable contact basis is provided for hole position searching and shaft piece insertion operation. The method effectively improves the accuracy of contact force detection.

Description

Technical Field

[0001] This application relates to the field of automated robot assembly technology, and in particular to a method for controlling stable contact force of a robot for assembly scenarios. Background Technology

[0002] In current automated precision shaft and hole assembly operations, force feedback-guided robot compliant assembly schemes are commonly used. In existing assembly processes, after the robot completes initial positioning, it directly enters the hole search and insertion stage without stable control over the initial axial contact state between the shaft and the mating surface.

[0003] The existing technology has two core defects: First, the end force detection data is easily affected by sensor zero drift, load gravity and robot pose changes, resulting in insufficient contact force detection accuracy and inability to provide reliable feedback for assembly control. Second, the axial contact force fluctuates greatly during the initial contact stage, making it impossible to form a stable force interaction environment. This can easily cause damage to the workpiece or end effector due to excessive contact force, and the unstable contact state can also reduce the success rate of subsequent hole search and shaft insertion operations, making it difficult to meet the requirements of high-precision assembly. Summary of the Invention

[0004] This application provides a robot stable contact force control method for assembly scenarios, which aims to solve the problems that end-effector force detection data is easily affected by sensor zero-point drift, load gravity and robot pose changes, and the contact force detection accuracy is insufficient, which cannot provide reliable feedback for assembly control.

[0005] In a first aspect, embodiments of this application provide a method for controlling stable contact force of a robot in an assembly scenario, the method comprising: Zero-point calibration is performed on the six-dimensional force sensor to obtain the sensor's zero-drift value. Then, the sensor detection data is processed for load gravity compensation to eliminate the interference of gravity components caused by changes in robot pose on the force detection results. The robot's end effector moves the shaft to be assembled to the vicinity of the target hole, driving the shaft to feed axially toward the workpiece mating surface. A six-dimensional force sensor collects contact force data from the robot's end effector in real time, continuously extracting the real-time value of the axial contact force. This process includes: reading the raw detection data from the six-dimensional force sensor at a fixed sampling frequency; filtering the raw detection data; establishing a robot end effector coordinate system; separating the force component corresponding to the axial direction from the filtered detection data; and using the force component corresponding to the axial direction as the real-time value of the axial contact force. Based on the real-time acquired axial contact force values, the axial position and feed speed of the robot end effector are dynamically adjusted through admittance control logic to stably maintain the axial contact force within the preset value range, forming a stable axial force interaction environment and providing a reliable contact basis for hole position search and shaft insertion operations.

[0006] In some embodiments, performing zero-point calibration on the six-dimensional force sensor to obtain the sensor zero-drift value includes: controlling the robot end effector to remain in an unloaded and stationary state; continuously collecting multiple sets of output data from the six-dimensional force sensor at a preset sampling frequency; calculating the average of the collected multiple sets of output data; and using the calculated average as the sensor zero-drift value.

[0007] In some embodiments, the process of performing load gravity compensation processing on sensor detection data to eliminate the interference of gravity components caused by changes in robot pose on the force detection results includes: pre-collecting sensor output data of the end-effector load in multiple different poses; calculating the force component parameters corresponding to the load gravity in different poses by combining the sensor zero drift values; reading the current pose information during robot operation; calling the stress component parameters according to the current pose information; subtracting the stress component parameters from the original sensor detection data; and obtaining the true external contact force data.

[0008] In some embodiments, the control robot end effector moves the shaft to be assembled to a region adjacent to the target hole, and drives the shaft to be assembled to feed axially toward the workpiece mating surface, including: receiving pre-taught and stored position parameters of the adjacent point of the target hole; controlling the robot end effector to move the shaft to be assembled to the adjacent point position; identifying the axial direction of the shaft to be assembled; driving the shaft to be assembled to move axially toward the workpiece mating surface according to a preset feed speed; detecting the numerical change of the axial contact force in real time; and stopping the feed motion when the axial contact force reaches a preset contact threshold.

[0009] In some embodiments, the step of using the real-time acquired axial contact force value includes: smoothing the axial contact force values ​​for multiple consecutive sampling periods; removing abnormal data whose values ​​jump beyond a preset range; and using the smoothed axial contact force value as an input parameter for admittance control logic.

[0010] In some embodiments, dynamically adjusting the axial position and feed rate of the robot end effector through admittance control logic includes: pre-setting virtual inertia parameters, virtual damping parameters, and virtual stiffness parameters corresponding to the admittance control logic; calculating the difference between the real-time value of the axial contact force and the preset target contact force; calculating the axial position correction amount by combining the virtual inertia parameters, virtual damping parameters, and virtual stiffness parameters; and adjusting the axial position and corresponding feed rate of the robot end effector according to the position correction amount.

[0011] In some embodiments, maintaining the axial contact force stably within a preset value range to form a stable axial force interaction environment includes: real-time monitoring of the fluctuation amplitude of the axial contact force; adjusting the axial feed speed amplitude when the axial contact force exceeds the preset value range; maintaining the current axial position unchanged when the axial contact force falls within the preset value range; continuously maintaining the stable state of the axial contact force; and establishing a constant axial force interaction basis.

[0012] In some embodiments, the method further includes: during the process of maintaining the axial contact force stably, collecting the workpiece material parameters and shaft size parameters corresponding to the current assembly; retrieving pre-stored historical assembly data under the same working conditions; extracting the assembly success rate index from the historical assembly data; adaptively adjusting the virtual damping parameters and virtual stiffness parameters corresponding to the admittance control logic according to the assembly success rate index; and matching to obtain the optimal control parameters for the current assembly working condition.

[0013] In some embodiments, the method further includes: continuously collecting trend data of axial contact force changes; calculating the rate of change of axial contact force for multiple consecutive sampling periods; determining whether there is an abnormality in the current contact state based on the rate of change of axial contact force; when it is determined that there is an abnormality in the contact state, immediately controlling the robot end effector to retreat a preset distance in the axial direction; and re-executing the axial feed and contact force stabilization control process.

[0014] This application has the following beneficial effects: 1. By performing zero-point calibration and load gravity compensation on the six-dimensional force sensor, the interference of sensor zero drift, end-effector load gravity and robot pose changes on the force detection results is eliminated, effectively improving the accuracy of contact force detection and providing accurate feedback data support for axial contact force control.

[0015] 2. By collecting axial contact force data in real time, the axial position and feed speed of the robot end effector are dynamically adjusted to keep the axial contact force stable within the preset range, establish a stable and reliable axial force interaction environment, avoid damage to the workpiece or actuator caused by contact force fluctuations, and provide a stable contact basis for subsequent hole position search and shaft insertion operations.

[0016] 3. Stable axial contact can be established without relying on a vision positioning system, simplifying the hardware configuration of the assembly system, reducing deployment costs, and effectively improving the overall operational stability and assembly success rate of precision shaft and hole assembly.

[0017] 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

[0018] 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.

[0019] Figure 1 This is a schematic flowchart illustrating the steps of a robot stable contact force control method for assembly scenarios provided in an embodiment of this application; Figure 2 This is a schematic diagram illustrating the principle of a robot stable contact force control method for assembly scenarios provided in an embodiment of this application; Figure 3 This is a schematic block diagram of a robot stable contact force control system for assembly scenarios 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.

[0020] 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

[0021] 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.

[0022] 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.

[0023] 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.

[0024] 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.

[0025] 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.

[0026] In current automated precision shaft and hole assembly operations, force feedback-guided robot compliant assembly schemes are commonly used. In existing assembly processes, after the robot completes initial positioning, it directly enters the hole search and insertion stage without stable control over the initial axial contact state between the shaft and the mating surface.

[0027] The existing technology has two core defects: First, the end force detection data is easily affected by sensor zero drift, load gravity and robot pose changes, resulting in insufficient contact force detection accuracy and inability to provide reliable feedback for assembly control. Second, the axial contact force fluctuates greatly during the initial contact stage, making it impossible to form a stable force interaction environment. This can easily cause damage to the workpiece or end effector due to excessive contact force, and the unstable contact state can also reduce the success rate of subsequent hole search and shaft insertion operations, making it difficult to meet the requirements of high-precision assembly.

[0028] Please refer to Figure 1 This application provides a method for controlling stable contact force of a robot in assembly scenarios, applied to computer equipment. The computer equipment can be deployed on a single server or a server cluster. It can also be deployed on handheld terminals, laptops, wearable devices, or robots, etc.

[0029] It should be noted that the robot stable contact force control method for assembly scenarios provided in this embodiment runs on a computer control device, which can be deployed on an independent industrial controller, server, server cluster, or in a mobile operating terminal or robot built-in control unit. All data acquisition, storage, and processing operations involved in this method are performed with the legal authorization of the relevant parties and comply with relevant data security and privacy protection regulations throughout the process.

[0030] This method is applied to a robot shaft hole assembly system with a six-dimensional force sensor at the end, and can be adapted to various high-precision assembly scenarios such as precision electronic component assembly, automotive precision component assembly, and aerospace small-size part assembly.

[0031] The provided robot stable contact force control method for assembly scenarios includes steps S101 to S103. Details are as follows: Step S101. Perform zero-point calibration on the six-dimensional force sensor to obtain the sensor zero-drift value, and then perform load gravity compensation processing on the sensor detection data to eliminate the interference of gravity component generated by the robot's pose change on the force detection result.

[0032] Specifically, the core function of this step is to eliminate systematic errors in the force detection process, providing a high-precision feedback data foundation for subsequent force feedback control throughout the entire process. The output value of the six-dimensional force sensor when the robot is running at low speed or stationary is composed of three parts: the sensor's own zero-point drift, the end-effector load gravity, and the external actual contact force. Without pre-correction, the force detection result will continuously deviate with changes in the robot's pose, failing to accurately reflect the actual contact state of the end-effector, thus leading to insufficient force control accuracy.

[0033] This step consists of two core execution stages: The first step is zero-point calibration. With the robot unloaded, multiple sets of continuous data are collected under a fixed posture to calculate the zero-drift values ​​of the sensor in six dimensions. These values ​​are used to correct the sensor's basic output error and eliminate detection deviations caused by the sensor's own electrical drift.

[0034] The second step is load gravity compensation processing. Based on the real-time end-effector pose obtained from the robot's kinematics calculation, the gravity component generated by the end-effector load under the current posture is matched and calculated. The influence of the gravity component is subtracted from the original sensor output data, so that the end-effector force detection results are not affected by changes in robot posture, and only the real force generated by external contact is retained.

[0035] Step S102. Control the robot end effector to move the shaft to be assembled to the vicinity of the target hole, and drive the shaft to be assembled to feed along the axial direction toward the workpiece mating surface; collect the contact force data of the robot end effector in real time through a six-dimensional force sensor, and continuously extract the real-time value of the axial contact force, including: reading the raw detection data of the six-dimensional force sensor at a fixed sampling frequency; filtering the raw detection data; establishing the robot end effector coordinate system; separating the force component corresponding to the axial direction from the filtered detection data; and using the force component corresponding to the axial direction as the real-time value of the axial contact force.

[0036] Specifically, the core function of this step is to complete the coarse positioning before assembly, establish the initial contact between the shaft and the workpiece mating surface, and continuously extract real-time contact force data in the axial direction to provide stable input parameters for subsequent closed-loop force control.

[0037] This step consists of two core execution stages: The first step is shaft positioning and axial feeding. First, using pre-stored positioning parameters, the robot end effector moves the shaft to be assembled to the vicinity of the target hole, completing the coarse positioning for the assembly operation. Then, the shaft is driven to feed along its own central axis at a low speed toward the workpiece mating surface. During the feeding process, the contact force changes are monitored in real time. Feeding stops after initial contact is detected to avoid impact damage to the workpiece.

[0038] The second step is the real-time extraction of axial contact force. A six-dimensional force sensor collects full-dimensional contact force data at a fixed frequency from the end effector. After filtering, noise reduction, and coordinate system transformation, the axial contact force component is separated from the multi-dimensional force data and used as the real-time input for subsequent closed-loop control.

[0039] The axial contact force acquisition and extraction operation in step S102 is further refined, and the specific implementation process is as follows: The sampling frequency of the six-dimensional force sensor is matched with the hardware output frequency, typically set to 100Hz to 500Hz, to ensure the real-time performance of the force feedback data and meet the delay requirements of closed-loop control. After reading the raw six-dimensional force and torque data from the sensor in each sampling cycle, a first-order Butterworth low-pass filter algorithm is first used to denoise the raw data. The filter cutoff frequency is set to 10Hz to remove high-frequency noise components caused by robot joint vibration and environmental electromagnetic interference.

[0040] Establish a robot end effector coordinate system with the origin set at the center of the end flange. The Z-axis is defined as the axis pointing from the center of the component to be assembled to the workpiece. The transformation relationship between the end effector coordinate system and the robot base coordinate system is updated in real time through robot forward kinematics calculation.

[0041] From the filtered six-dimensional force data, the force component in the Z-axis direction is extracted. After coordinate system transformation and calibration, the contact force value in the axial direction is obtained. The corresponding value is used as the real-time value of the axial contact force and output to the subsequent control module.

[0042] Step S103. Based on the real-time acquired axial contact force value, dynamically adjust the axial position and feed speed of the robot end effector through admittance control logic to stably maintain the axial contact force within the preset value range, forming a stable axial force interaction environment, and providing a reliable contact basis for hole position search and shaft insertion operations.

[0043] Specifically, the core function of this step is to establish a stable and reliable axial force interaction environment through force feedback closed-loop control, providing a consistent contact basis for subsequent hole location search and shaft insertion operations.

[0044] The specific execution logic of this step is as follows: The real-time value of the axial contact force obtained from step S102 is used as the control input and imported into the admittance control logic for calculation. The admittance control logic adopts an equivalent mass-damping-spring compliant control model, converting the deviation of the contact force into a position correction amount for the robot end effector, and dynamically adjusting the axial position and feed speed of the robot end effector. Through closed-loop continuous correction, the axial contact force is always stably maintained within the preset value range, forming a stable axial force interaction environment. This control mode can automatically adapt the end effector position according to the change of contact force, avoiding both excessive contact force that damages the workpiece or sensor and insufficient contact force that leads to contact disengagement, ensuring contact consistency in subsequent assembly operations.

[0045] In some embodiments, performing zero-point calibration on the six-dimensional force sensor to obtain the sensor zero-drift value includes: controlling the robot end effector to remain in an unloaded and stationary state; continuously collecting multiple sets of output data from the six-dimensional force sensor at a preset sampling frequency; calculating the average of the collected multiple sets of output data; and using the calculated average as the sensor zero-drift value.

[0046] This embodiment further refines the zero-point calibration operation in step S101, and the specific implementation process is as follows: Zero-point calibration is performed after the robot completes its power-on self-test and joint homing. First, confirm that no tools, shafts, or additional loads are installed on the robot's end effector, and that it is in a completely unloaded state. Then, control the robot to move to the preset calibration posture and keep the end effector stationary for at least 30 seconds to eliminate the influence of residual stress inside the robotic arm joints on the test results.

[0047] The sampling frequency of the six-dimensional force sensor was set to 100Hz, and data was continuously collected for 10 seconds to obtain a total of 1000 sets of raw output data of six-dimensional force and six-dimensional torque. The arithmetic mean of the 1000 sets of data in each of the six dimensions was calculated, and a zero drift value was obtained for each dimension.

[0048] The calculated six sets of zero drift values ​​are stored in the controller's non-volatile memory. Each time the sensor is powered on, the corresponding zero drift value is automatically loaded to perform basic zero-point correction on all raw output data, eliminating the sensor's own electrical zero-point drift error.

[0049] In some embodiments, the process of performing load gravity compensation processing on sensor detection data to eliminate the interference of gravity components caused by changes in robot pose on the force detection results includes: pre-collecting sensor output data of the end-effector load in multiple different poses; calculating the force component parameters corresponding to the load gravity in different poses by combining the sensor zero drift values; reading the current pose information during robot operation; calling the stress component parameters according to the current pose information; subtracting the stress component parameters from the original sensor detection data; and obtaining the true external contact force data.

[0050] This embodiment further refines the load gravity compensation process in step S101, and the specific implementation process is as follows: After zero-point calibration, the execution tools and the shaft to be assembled are installed on the robot's end effector, and the two are used as a whole load for gravity compensation parameter identification. The robot is controlled to move to at least 6 different end effector postures, all of which cover the common workspace of robot assembly operations, and the posture angle differences are sufficient to support gravity parameter fitting. The end effector is kept still for 5 seconds in each posture, and the filtered sensor output data is collected. After subtracting the pre-stored zero drift value, the load force data for the corresponding posture is obtained.

[0051] By combining the end-effector rotation matrix corresponding to each posture obtained from the robot's forward kinematics calculation, the total gravity of the overall load and the three-dimensional coordinate parameters of the load's center of mass in the sensor coordinate system are identified through the least squares fitting algorithm. A gravity compensation calculation model is then constructed, and the parameters are stored in the system's compensation parameter library.

[0052] During the robot's actual operation, the angle data of each joint of the robot is read in real time during each control cycle. The attitude matrix of the current end flange is obtained through forward kinematics calculation. Based on the attitude matrix, the load gravity vector is transformed into the current sensor coordinate system, and the six-dimensional force and torque components generated by gravity under the current attitude are calculated. The zero drift value and the gravity component corresponding to the current attitude are deducted sequentially from the original sensor detection data to finally obtain the real force data generated only by external contact, completely eliminating the interference of gravity components generated by robot pose changes on the force detection results.

[0053] After compensation is completed, the accuracy can be confirmed through verification: keep the end effector free from external contact, change the robot's pose arbitrarily, and if the deviation of the force data in each dimension after compensation does not exceed 0.5% of the sensor's rated range, the compensation is considered qualified.

[0054] In some embodiments, the control robot end effector moves the shaft to be assembled to a region adjacent to the target hole, and drives the shaft to be assembled to feed axially toward the workpiece mating surface, including: receiving pre-taught and stored position parameters of the adjacent point of the target hole; controlling the robot end effector to move the shaft to be assembled to the adjacent point position; identifying the axial direction of the shaft to be assembled; driving the shaft to be assembled to move axially toward the workpiece mating surface according to a preset feed speed; detecting the numerical change of the axial contact force in real time; and stopping the feed motion when the axial contact force reaches a preset contact threshold.

[0055] This embodiment further refines the shaft positioning and feeding operation in step S102. The specific implementation process is as follows: The position parameters of the adjacent points of the target hole are generated in advance by drag-and-drop teaching or offline programming. The corresponding adjacent points are located directly above the center of the target hole. The axial distance from the upper surface of the workpiece is set to 10mm to 20mm. During teaching, the coaxiality deviation between the axis of the shaft to be assembled and the theoretical axis of the hole is controlled within 2mm.

[0056] After assembly is started, the robot end effector drives the shaft to be assembled to move to the nearest point at a normal operating speed to complete the coarse positioning of the assembly operation; after confirming that the axial parallelism between the shaft and the normal parallelism of the workpiece mating surface meets the assembly requirements, it switches to low-speed feed mode and moves along the central axis of the shaft toward the workpiece mating surface at a constant speed of 0.5 mm / s to 1 mm / s.

[0057] During the feeding process, the axial contact force value is read in real time for each sampling cycle. When the axial contact force value of three consecutive sampling cycles exceeds the preset initial contact threshold, the axial feed stop command is immediately triggered to complete the initial contact establishment between the shaft and the workpiece mating surface. The initial contact threshold is usually set to 10% to 20% of the target stable contact force to avoid false triggering and prevent contact impact caused by excessive feed speed.

[0058] In some embodiments, the step of using the real-time acquired axial contact force value includes: smoothing the axial contact force values ​​for multiple consecutive sampling periods; removing abnormal data whose values ​​jump beyond a preset range; and using the smoothed axial contact force value as an input parameter for admittance control logic.

[0059] This embodiment further refines the preprocessing operation of the axial contact force value in step S103. The specific implementation process is as follows: After obtaining the real-time value of the axial contact force, the values ​​of multiple consecutive sampling periods are first smoothed. In this embodiment, the sliding window averaging method is used, and the sliding window length is set to 5 to 10 consecutive sampling periods. When each new sampling period arrives, the axial force data in the window is updated, the earliest set of data is removed, the latest set of data is added, and the arithmetic mean of all data in the window is recalculated to obtain the smoothed axial contact force value.

[0060] Synchronous execution of abnormal data removal operation: If the absolute value of the deviation between the original value of the current sampling period and the smoothed value of the previous period exceeds 30% of the preset target contact force, the corresponding data is determined to be abnormal jump data, and the data set is directly removed. The smoothed value of the previous period is used as the replacement value of the current period to avoid abnormal data caused by external accidental shocks and signal interference from affecting control stability.

[0061] The axial contact force value, after smoothing and anomaly removal, is used as the input parameter of the admittance control logic to improve the smoothness of the closed-loop control process.

[0062] In some embodiments, dynamically adjusting the axial position and feed rate of the robot end effector through admittance control logic includes: pre-setting virtual inertia parameters, virtual damping parameters, and virtual stiffness parameters corresponding to the admittance control logic; calculating the difference between the real-time value of the axial contact force and the preset target contact force; calculating the axial position correction amount by combining the virtual inertia parameters, virtual damping parameters, and virtual stiffness parameters; and adjusting the axial position and corresponding feed rate of the robot end effector according to the position correction amount.

[0063] This embodiment further refines the admittance control adjustment operation in step S103, and the specific implementation process is as follows: The three sets of parameters corresponding to the admittance control logic are pre-tuned according to the assembly conditions: virtual inertia parameter, virtual damping parameter, and virtual stiffness parameter. Among them, the virtual inertia parameter determines the response inertia of the system. The larger the value, the smoother the system response and the smaller the impact. The virtual damping parameter is used to suppress the oscillation in the force control process. The larger the value, the lower the oscillation amplitude. The virtual stiffness parameter determines the ratio between force deviation and position correction. The smaller the value, the larger the position correction for the same force deviation, and the stronger the system compliance.

[0064] The execution cycle of admittance control is synchronized with the sensor sampling cycle. The following operations are performed in each control cycle: First, the difference between the preset target contact force and the current smoothed axial contact force value is calculated to obtain the force deviation. The force deviation is input into the discretized admittance control model and iteratively calculated using three sets of virtual parameters to obtain the axial position correction increment for the current cycle. The position correction increment is superimposed on the axial position command of the previous cycle to generate the target axial position for the current cycle. At the same time, the corresponding axial feed speed command is obtained based on the ratio of the position correction increment to the control cycle. The position and speed commands are sent to the robot motion controller to drive the robot end effector to perform the corresponding axial adjustment action.

[0065] In some embodiments, maintaining the axial contact force stably within a preset value range to form a stable axial force interaction environment includes: real-time monitoring of the fluctuation amplitude of the axial contact force; adjusting the axial feed speed amplitude when the axial contact force exceeds the preset value range; maintaining the current axial position unchanged when the axial contact force falls within the preset value range; continuously maintaining the stable state of the axial contact force; and establishing a constant axial force interaction basis.

[0066] This embodiment further refines the axial force stabilization maintenance operation in step S103, and the specific implementation process is as follows: A stable range of axial contact force is preset. The range is centered on the target contact force, with a fluctuation of 5% above and below as the stability judgment interval. For example, when the target contact force is set to 10N, the stable range is 9.5N to 10.5N.

[0067] Real-time monitoring of the relationship between the axial contact force value and the stable range: When the axial contact force is higher than the upper limit of the range, it indicates that the feed depth of the shaft is too large and the contact pressure is too high. Control the robot end to retract the tool at a low speed in the opposite direction along the axial direction. The retraction speed is positively correlated with the magnitude of the force deviation. The larger the deviation, the faster the retraction speed. When the axial contact force is lower than the lower limit of the range, it indicates that the contact pressure is insufficient and there is a risk of disengagement. Control the robot end to feed at a low speed along the axial direction. The feed speed is positively correlated with the magnitude of the force deviation. The larger the deviation, the faster the feed speed.

[0068] When the axial contact force falls within the preset stable range for 10 consecutive sampling cycles, the stable contact state is determined to be formally established. Thereafter, the robot maintains its current axial position and only makes minor positional corrections to offset the minor fluctuations in the contact force, thus maintaining the stable state of the axial contact force and establishing a constant axial force interaction basis. This ensures that the shaft always maintains a uniform contact pressure with the workpiece surface during subsequent operations such as hole position search and attitude adjustment.

[0069] In some embodiments, the method further includes: during the process of maintaining the axial contact force stably, collecting the workpiece material parameters and shaft size parameters corresponding to the current assembly; retrieving pre-stored historical assembly data under the same working conditions; extracting the assembly success rate index from the historical assembly data; adaptively adjusting the virtual damping parameters and virtual stiffness parameters corresponding to the admittance control logic according to the assembly success rate index; and matching to obtain the optimal control parameters for the current assembly working condition.

[0070] This embodiment adds an adaptive optimization function for admittance control parameters to the basic method flow. The specific implementation process is as follows: During the process of maintaining stable axial contact force, the working condition parameters corresponding to the current assembly task are first collected, including workpiece material parameters and shaft size parameters. Among them, the workpiece material parameters include workpiece surface hardness and surface friction coefficient, and the shaft size parameters include shaft diameter, shaft length, and shaft hole fit tolerance. All parameters are pre-entered into the system's working condition parameter library.

[0071] The system accesses a locally stored historical assembly database. This database contains multiple sets of historical assembly records for different material and size combinations. Each record corresponds to a set of admittance control parameters, as well as the corresponding assembly success rate and contact force fluctuation amplitude. Based on the currently collected operating parameters, the system matches historical data with the same or similar operating conditions from the database and extracts the set of admittance parameters with the highest assembly success rate as the initial parameters for the current control.

[0072] During the stable contact phase, the system calculates the root mean square value of the axial contact force fluctuation in real time. If the fluctuation value is greater than the preset qualified threshold, the virtual damping parameter and virtual stiffness parameter are iteratively adjusted step by step with the minimum force fluctuation as the optimization goal until the contact force fluctuation meets the threshold requirement. Finally, the optimal control parameters under the current assembly condition are obtained, and the corresponding set of parameters, corresponding working conditions, and assembly results are synchronously updated to the historical database to achieve adaptive adaptation of control parameters under different assembly scenarios.

[0073] In some embodiments, the method further includes: continuously collecting trend data of axial contact force changes; calculating the rate of change of axial contact force for multiple consecutive sampling periods; determining whether there is an abnormality in the current contact state based on the rate of change of axial contact force; when it is determined that there is an abnormality in the contact state, immediately controlling the robot end effector to retreat a preset distance in the axial direction; and re-executing the axial feed and contact force stabilization control process.

[0074] This embodiment adds contact anomaly detection and automatic recovery functions to the basic method flow. The specific implementation process is as follows: Throughout the entire process of axial feed and stable contact force maintenance, the trend data of axial contact force change is continuously collected. The axial contact force change rate is calculated by taking the difference between the force value of the current cycle and the force values ​​of the previous 5 cycles, dividing it by the corresponding time interval, and obtaining the force change rate per unit time.

[0075] An abnormal change rate threshold is preset. When the force change rate exceeds the preset threshold for three consecutive sampling cycles, the current contact state is determined to be abnormal. Abnormal situations include protrusions on the workpiece surface, jamming of shafts, and excessive initial positioning deviation. If feeding continues, it may cause damage to the workpiece or sensor.

[0076] Upon triggering the abnormal protection logic, the robot immediately stops all current feed actions and retreats axially at a constant low speed to a preset safety distance, typically set to 1mm to 3mm, to completely eliminate the abnormal contact state. Once the contact force data stabilizes, the system re-executes the axial feed, initial contact detection, and contact force stabilization control process. If a stable contact state cannot be established after three consecutive retries, the system outputs an assembly abnormality alarm signal, stops the current operation, and awaits manual intervention to prevent damage to the equipment or workpiece.

[0077] In some embodiments, this example is a complete application solution for precision shaft-hole compliant assembly, and the overall system architecture is as follows: Figure 2 As shown, it covers the entire chain of functions including sensor data processing, admittance control, assembly strategy execution, and environmental interaction status recognition, forming a complete closed-loop force control assembly system. The specific implementation process is as follows: This corresponds to the sensor calibration and gravity compensation algorithm module in the system architecture. Before formal assembly, the six-dimensional force sensor at the robot's end effector is first zero-point calibrated to obtain the zero drift values ​​of each dimension of the sensor. Then, the assembly tools and the shaft to be assembled are installed, and the sensor output data of the end effector load under multiple different poses are collected to identify all the parameters of the load gravity compensation model and build a gravity compensation parameter library.

[0078] During assembly and operation, the six-dimensional force sensor collects raw data of the feedback force generated by the contact between the end and the environment in real time. It then performs zero-point correction, gravity compensation, and filtering and noise reduction in sequence to obtain high-precision real external contact force data. The processed contact force data is used as the feedback input of the entire control system.

[0079] This corresponds to the admittance controller and assembly strategy module in the system architecture. The system pre-tunes the virtual inertial parameters, virtual damping parameters, and virtual stiffness parameters corresponding to the admittance control, and stores the pre-planned desired assembly trajectory. During control, the position command of the desired trajectory is superimposed with the position correction amount output by the admittance calculation to generate the robot's actual control position command; simultaneously, the real-time feedback contact force is compared with the preset desired contact force, and the position correction amount is obtained through admittance control logic calculation to perform real-time smooth correction of the robot's motion trajectory.

[0080] The assembly strategy employs a three-level progressive execution process: The first stage is the coarse positioning and stable contact establishment stage. Through manual teaching or offline programming, the shaft to be assembled is guided to the vicinity of the target hole, and the shaft is controlled to approach the workpiece mating surface axially. Through the stable contact force control method described above in this application, the axial contact force is stably maintained within a preset range, and a stable axial force interaction environment is established.

[0081] The second stage is the hole location search stage. After stable contact is established, the control shaft performs a spiral trajectory search in the horizontal plane. The search trajectory adopts an Archimedean spiral with equal step size, with the step size set to 0.1 mm. The maximum search radius is set according to the initial positioning deviation. During the search, a stable axial contact force is maintained, and the changes in contact force in the horizontal direction are collected in real time. When a sudden change in the force characteristics in the horizontal direction is detected, it is determined that the shaft has contacted the edge of the hole, and the spiral search stops.

[0082] The third stage is the attitude fine-tuning and insertion phase. After the attitude adjustment mechanism is triggered, the shaft is controlled to swing slightly to correct the tilt angle of the axis and gradually reduce the insertion resistance. During the adjustment process, the axial contact force is continuously controlled to remain stable, pushing the shaft to gradually insert into the hole.

[0083] This corresponds to the environment interaction module in the system architecture. A two-point contact force analysis model is used to calculate the contact state of the shaft-hole assembly. When the shaft enters the hole at an angle, the shaft end and the edge of the hole form two contact points, generating corresponding force and torque distributions.

[0084] Based on the three-dimensional force and three-dimensional torque data fed back by the six-dimensional force sensor, the system calculates the tilt angle and offset direction of the current shaft through the superposition and decomposition of force and torque. It transforms the contact state, which cannot be directly observed, into a quantifiable pose deviation, and then derives the corresponding adjustment movement direction. The system controls the shaft to rotate around the line connecting the two contact points, while coordinating with translation adjustment in the horizontal plane, so that the shaft gradually stands upright and aligns with the center of the hole. High-precision autonomous alignment can be achieved without visual feedback.

[0085] The assembly quality evaluation is carried out using dual criteria of insertion time and contact state. The system records the duration of the insertion operation and sets a preset time window; within the time window, the insertion depth of the shaft part and the value of the axial contact force are detected in real time. When the insertion depth reaches the preset qualified depth and the axial contact force remains stably within the preset theoretical interval, the current assembly is determined to be qualified; if the insertion depth requirement is not met after exceeding the preset time window, or the axial contact force has an abnormal surge, the assembly is determined to have failed, the operation is terminated, and a failure prompt is output.

[0086] The entire system forms a complete closed-loop control link: the robot assembly control system outputs motion commands to control the robot end to perform corresponding actions; the robot end comes into contact with the workpiece environment to generate contact force, and the six-dimensional force sensor collects the contact force and feeds it back to the front-end data processing module; the processed force data is input into the admittance controller to generate corresponding position correction commands, which adjust the motion state of the robot again, forming a fully closed-loop force-controlled compliant assembly system, realizing highly reliable automatic precision shaft-hole assembly without visual dependence.

[0087] In some embodiments, the method further includes an adaptive tuning process for admittance parameters based on deep reinforcement learning, which solves the technical problem that fixed admittance parameters cannot adapt to the contact characteristics of workpieces of different materials in different assembly stages, and it is difficult to balance the response speed and stability of force control. It is applicable to complex precision shaft-hole assembly scenarios with mixed production of multiple varieties and multiple assembly stages. By constructing a deep deterministic policy gradient reinforcement learning agent, the dynamic adjustment of admittance control parameters is taken as the control target. The input features defining the state space include: current axial contact force value, force deviation, force change rate, current insertion depth, workpiece material and surface stiffness grade, current assembly stage identifier, and position correction amounts of the three historical control cycles; the action space is the adjustment increments of the three core parameters of admittance control, that is, the virtual inertia adjustment amount, the virtual damping adjustment amount, and the virtual stiffness adjustment amount, and each parameter is set with upper and lower limit constraints to ensure control stability. The reward function adopts a multi-objective weighted design: the negative value of the contact force tracking error is taken as the main penalty term, the negative value of the force fluctuation amplitude is taken as the secondary penalty term, the smoothness of position adjustment is taken as the positive reward term, and a penalty term for force overshoot is set at the same time to guide the agent to improve the response speed on the premise of ensuring no force overshoot.

[0088] Before formal production, the agent is pre-trained offline in a simulation environment based on a large amount of assembly data under different working conditions, covering typical scenarios of different materials, different fitting clearances, and different assembly stages, so that the agent can learn the optimal parameter adjustment strategies under different working conditions. After the pre-training is completed, the agent is deployed to the robot controller and runs in the inference mode.

[0089] During actual assembly, the intelligent agent collects current state characteristics in each control cycle and outputs real-time adjustment increments for admittance parameters, dynamically correcting virtual inertia, damping, and stiffness values: In the initial contact approach phase, it automatically increases damping and decreases stiffness to avoid contact impact; in the stable contact hole-finding phase, it balances response speed and stability to adapt to the contact characteristics of planar sliding; in the shaft insertion phase, it adjusts stiffness and damping in real-time based on the insertion depth and changes in contact resistance to adapt to the resistance characteristics of the hole fit. The parameter adjustment process employs smooth interpolation to avoid system oscillations caused by sudden parameter changes.

[0090] After each assembly, the system performs online incremental learning on the agent based on the force control effect of that assembly, continuously optimizing the parameter tuning strategy. By adaptively tuning admittance parameters through reinforcement learning, it can replace the fixed parameter scheme of manual tuning, enabling the force control system to autonomously adapt to the contact characteristics of different working conditions and different assembly stages. This significantly reduces the process debugging cost of multi-variety products, while improving the force control stability and tracking accuracy of the entire assembly process.

[0091] In some embodiments, the method further includes a contact force feedforward compensation process based on a digital twin model, addressing the technical problems of control lag, large force control fluctuations, and easy force overshoot in pure feedback force control when facing workpiece surface morphology undulations or sudden changes in local stiffness. It is suitable for high-precision assembly scenarios with complex surface morphology and containing local soft coatings or brittle materials. By constructing a digital twin of the assembly station, it includes a robot kinematic twin model, an end-effector load dynamics model, a workpiece 3D morphology and stiffness distribution twin model, and a contact mechanics simulation model. The workpiece morphology and stiffness model can be calibrated using 3D scanning and microhardness testing data to accurately reproduce the microscopic undulations, coating distribution, and stiffness differences in different regions of the workpiece surface, simulating the force response characteristics of shafts when in contact at different positions on the workpiece surface.

[0092] Before assembly, based on the theoretical hole positions and surface topography data of the workpiece, the entire approach, contact, and hole-searching motion process is pre-simulated in a digital twin environment. The expected contact force change curves at each point on the motion path are calculated in advance through contact mechanics simulation, and the corresponding feedforward position correction sequence is generated.

[0093] In the actual assembly process, a composite force control architecture of "feedforward prediction + feedback correction" is adopted: on the one hand, the real-time collected robot pose and contact force data are synchronously mapped to the digital twin, driving the twin model to run synchronously with the physical assembly, and correcting the physical parameters such as contact stiffness and friction coefficient of the twin model online according to the measured force data, thereby continuously improving the simulation accuracy of the twin model; on the other hand, based on the twin model, the contact force change trend of the next stage is predicted, and the feedforward position correction amount is generated in advance and superimposed on the feedback correction amount of the admittance control output, so as to adjust the end position in advance before the contact force fluctuates.

[0094] For example, when a shaft is about to move to a raised area on the workpiece surface, the twin model predicts in advance that the contact force will increase and outputs a feedforward correction amount for axial retraction to offset the force impact caused by the protrusion. When it moves to a recessed area, it outputs a feedforward amount in advance to prevent the contact force from decreasing and causing it to lose contact. By offsetting the force disturbance caused by changes in shape through feedforward compensation, and then eliminating residual deviations through feedback control, a high-precision and stable force control effect is achieved.

[0095] By combining digital twin feedforward and force feedback control, the hysteresis of pure feedback control can be effectively reduced, the contact force fluctuations caused by workpiece surface morphology undulations and stiffness abrupt changes can be significantly suppressed, and stable contact pressure can be maintained under complex surface conditions. At the same time, the risk of contact impact can be reduced, protecting brittle workpieces and precision sensors.

[0096] In some embodiments, the method further includes a federated transfer learning-based force control model optimization process for multi-robot assembly stations. This addresses the technical problems of data silos across multiple workstations, limited data volume at a single workstation making it difficult to train high-precision force control models, and the inconvenience of centralized uploading of raw production data, posing data security risks. It is suitable for large-scale precision assembly lines with parallel multi-workstation operations. Each assembly robot in the production line acts as a local computing node, storing its own local assembly force perception data and process parameters. All raw production data is stored locally on the controller and not transmitted externally, ensuring data privacy and security. A federated coordination node is set up to be responsible for the aggregation, distribution, and iterative management of the global model, without ever accessing the raw production data.

[0097] A horizontal federated transfer learning framework is adopted to achieve multi-workstation collaborative optimization: Each local node trains a force control parameter optimization model and a contact state recognition model locally based on its own historical assembly data. After training, only the parameter gradients of the model are encrypted and uploaded to the coordination node. The coordination node performs weighted aggregation of the gradients uploaded by all workstations. The weights are positively correlated with the effective data volume of each workstation to generate an updated global basic model. The global model is then encrypted and distributed to each local workstation.

[0098] After receiving the global model, each workstation performs migration and fine-tuning based on its own equipment characteristics and workpiece type, adapting the global model to the mechanical characteristics and working conditions of its local workstation, thus obtaining a locally optimized force control model. After multiple rounds of federated iterations, the global model integrates the data features of all workstations, possessing stronger generalization ability and control accuracy than single-workstation models. At the same time, all original data is always stored locally, complying with data security and privacy compliance requirements.

[0099] Furthermore, a unified benchmark for force control parameters across the entire production line can be achieved based on a global model, synchronizing validated optimal control parameters to all workstations and improving the consistency of force control and the stability of assembly quality throughout the line. Through distributed collaborative optimization using federated transfer learning, the data value of the entire production line can be aggregated while protecting data privacy, overcoming the bottleneck of model accuracy caused by insufficient data at individual workstations and continuously improving the intelligence level of force control across the entire line.

[0100] In some embodiments, the method further includes a contact state intelligent recognition and multi-mode force control switching process based on temporal force perception characteristics, which solves the technical problems that traditional threshold judgment methods cannot accurately distinguish complex contact states and different force control strategies need to be matched for different contact stages. It is applicable to the entire assembly process including multiple stages such as approach, surface contact, hole edge contact, and hole jamming. By constructing a multi-category contact state dataset, six-dimensional force and temporal data of various typical contact states in the entire assembly process are collected and corresponding state labels are labeled, including more than ten contact states such as free space approach, workpiece surface contact, hole edge single-sided contact, hole two-point jamming, hole bottom contact, foreign object obstruction, and contact slippage. The dataset covers working conditions with different materials, different fit gaps, and different movement speeds to ensure sample diversity.

[0101] A temporal classification model is constructed using a one-dimensional convolutional network combined with a bidirectional long short-term memory network. The model takes six-dimensional force and torque temporal time-series data with a fixed time window as input and outputs the category and confidence level of the current contact state. The model is trained offline using a labeled dataset, autonomously learning the force perception temporal features corresponding to different contact states. After training, the model is deployed to the control unit.

[0102] Throughout the assembly process, the system continuously inputs the current force sensing timing data into the classification model using a sliding time window to identify the current contact state and confidence level in real time. It automatically switches the corresponding force control mode based on different contact states: when identified as a free-space approach state, it switches to high-speed position control mode, retaining only the force monitoring and anti-collision function; when identified as a workpiece surface contact state, it switches to constant force control mode to maintain a stable axial contact force; when identified as a hole edge contact state, it switches to low-stiffness, high-damping compliant mode to adapt to sudden force changes at the hole edge; when identified as a stuck state inside the hole, it switches to variable impedance adjustment mode, automatically increasing compliance and triggering attitude correction logic; when identified as a foreign object obstruction state, it triggers abnormal protection logic, stopping the feed and retracting.

[0103] When the confidence level of the model's output state falls below a set threshold, the current force control mode is conservatively maintained and the motion speed is reduced. Switching is only initiated once the confidence level recovers, avoiding control oscillations caused by erroneous switching. Through intelligent recognition and mode switching, the force control system can accurately match the contact characteristics at different stages, replacing manually set fixed parameter switching logic and improving the force control adaptability and operational stability of the entire assembly process.

[0104] In some embodiments, the method further includes an overshoot-free adaptive force control and contact impact suppression process for brittle workpieces, addressing the technical problems of force overshoot, damage to brittle workpieces or precision coatings caused by traditional force control in the initial contact stage. It is applicable to assembly scenarios of brittle / soft precision workpieces such as ceramics, glass, and semiconductor wafers. During the axial feed approach to the workpiece, a three-stage variable speed approach strategy is employed: high-speed feed at a long distance to improve efficiency; switching to low-speed feed when approaching a preset distance before the predicted contact position; and switching to micro-step feed mode when extremely close to the contact position, gradually reducing the feed speed to minimize contact impact. Simultaneously, a contact position prediction model is constructed based on historical contact data, predicting the contact critical point based on minute force changes during the feed process, and initiating force control intervention in advance.

[0105] During the contact establishment phase, an adaptive impedance-based soft-landing control logic is introduced: at the instant initial contact is detected, the virtual damping and virtual inertia parameters are automatically and temporarily increased to significantly enhance the system's damping characteristics, absorb the contact impact kinetic energy, and suppress force overshoot; simultaneously, the maximum amplitude of the position correction is limited to avoid force oscillation caused by excessive correction. Once the contact force gradually approaches the target value and the fluctuation amplitude is below the threshold, the system smoothly transitions to normal constant force control parameters, ensuring that the force curve is free of spikes and overshoot throughout the entire contact process.

[0106] During assembly, the equivalent contact stiffness of the workpiece surface is identified in real time online: based on the correspondence between contact force and position correction, the equivalent stiffness value of the current contact point is recursively calculated; the virtual stiffness parameter is dynamically adjusted according to the identified stiffness value, with smaller virtual stiffness settings corresponding to areas with lower stiffness, further reducing the risk of force overshoot. Simultaneously, a contact force peak protection mechanism is set; once the force value is detected to be close to the safety threshold, reverse retreat protection is immediately triggered to prevent force overshoot from damaging the workpiece.

[0107] Through multiple mechanisms including soft landing control, online stiffness identification, and peak protection, the force overshoot of initial contact can be controlled within 5% of the target value, completely avoiding contact impact damage to brittle workpieces, while ensuring the stability of steady-state contact force and meeting the stringent force control requirements of high-precision brittle component assembly.

[0108] Please see Figure 3 As shown, Figure 3This is a schematic diagram of a robot stable contact force control system 200 for assembly scenarios provided in this application embodiment. The robot stable contact force control system 200 for assembly scenarios is used to execute the steps of the robot stable contact force control method for assembly scenarios shown in the above embodiments. The robot stable contact force control system 200 for assembly scenarios 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.

[0109] like Figure 3 As shown, the robot stable contact force control system 200 for assembly scenarios includes: The calibration operation unit 201 is used to perform zero-point calibration operation on the six-dimensional force sensor, obtain the zero drift value of the sensor, and then perform load gravity compensation processing on the sensor detection data to eliminate the interference of gravity component generated by the robot's pose change on the force detection result. The numerical extraction unit 202 is used to control the robot end effector to move the shaft to be assembled to the vicinity of the target hole, and drive the shaft to be assembled to feed along the axial direction toward the workpiece mating surface; it continuously extracts the real-time value of the axial contact force by acquiring the contact force data of the robot end effector in real time through a six-dimensional force sensor, including: reading the raw detection data of the six-dimensional force sensor at a fixed sampling frequency; filtering the raw detection data; establishing the robot end effector coordinate system; separating the force component corresponding to the axial direction from the filtered detection data; and using the force component corresponding to the axial direction as the real-time value of the axial contact force. The basic providing unit 203 is used to dynamically adjust the axial position and feed speed of the robot end effector through admittance control logic based on the real-time acquired axial contact force value, so as to stably maintain the axial contact force within the preset value range, form a stable axial force interaction environment, and provide a reliable contact basis for hole position search and shaft insertion operation.

[0110] In some embodiments, performing zero-point calibration on the six-dimensional force sensor to obtain the sensor zero-drift value includes: controlling the robot end effector to remain in an unloaded and stationary state; continuously collecting multiple sets of output data from the six-dimensional force sensor at a preset sampling frequency; calculating the average of the collected multiple sets of output data; and using the calculated average as the sensor zero-drift value.

[0111] In some embodiments, the process of performing load gravity compensation processing on sensor detection data to eliminate the interference of gravity components caused by changes in robot pose on the force detection results includes: pre-collecting sensor output data of the end-effector load in multiple different poses; calculating the force component parameters corresponding to the load gravity in different poses by combining the sensor zero drift values; reading the current pose information during robot operation; calling the stress component parameters according to the current pose information; subtracting the stress component parameters from the original sensor detection data; and obtaining the true external contact force data.

[0112] In some embodiments, the control robot end effector moves the shaft to be assembled to a region adjacent to the target hole, and drives the shaft to be assembled to feed axially toward the workpiece mating surface, including: receiving pre-taught and stored position parameters of the adjacent point of the target hole; controlling the robot end effector to move the shaft to be assembled to the adjacent point position; identifying the axial direction of the shaft to be assembled; driving the shaft to be assembled to move axially toward the workpiece mating surface according to a preset feed speed; detecting the numerical change of the axial contact force in real time; and stopping the feed motion when the axial contact force reaches a preset contact threshold.

[0113] In some embodiments, the step of using the real-time acquired axial contact force value includes: smoothing the axial contact force values ​​for multiple consecutive sampling periods; removing abnormal data whose values ​​jump beyond a preset range; and using the smoothed axial contact force value as an input parameter for admittance control logic.

[0114] In some embodiments, dynamically adjusting the axial position and feed rate of the robot end effector through admittance control logic includes: pre-setting virtual inertia parameters, virtual damping parameters, and virtual stiffness parameters corresponding to the admittance control logic; calculating the difference between the real-time value of the axial contact force and the preset target contact force; calculating the axial position correction amount by combining the virtual inertia parameters, virtual damping parameters, and virtual stiffness parameters; and adjusting the axial position and corresponding feed rate of the robot end effector according to the position correction amount.

[0115] In some embodiments, maintaining the axial contact force stably within a preset value range to form a stable axial force interaction environment includes: real-time monitoring of the fluctuation amplitude of the axial contact force; adjusting the axial feed speed amplitude when the axial contact force exceeds the preset value range; maintaining the current axial position unchanged when the axial contact force falls within the preset value range; continuously maintaining the stable state of the axial contact force; and establishing a constant axial force interaction basis.

[0116] In some embodiments, the method further includes: during the process of maintaining the axial contact force stably, collecting the workpiece material parameters and shaft size parameters corresponding to the current assembly; retrieving pre-stored historical assembly data under the same working conditions; extracting the assembly success rate index from the historical assembly data; adaptively adjusting the virtual damping parameters and virtual stiffness parameters corresponding to the admittance control logic according to the assembly success rate index; and matching to obtain the optimal control parameters for the current assembly working condition.

[0117] In some embodiments, the method further includes: continuously collecting trend data of axial contact force changes; calculating the rate of change of axial contact force for multiple consecutive sampling periods; determining whether there is an abnormality in the current contact state based on the rate of change of axial contact force; when it is determined that there is an abnormality in the contact state, immediately controlling the robot end effector to retreat a preset distance in the axial direction; and re-executing the axial feed and contact force stabilization control process.

[0118] 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 robot stable contact force control system and its modules for assembly scenarios described above can be found in the corresponding embodiments of the robot stable contact force control method for assembly scenarios, and will not be repeated here.

[0119] The aforementioned robot stable contact force control method for assembly scenarios can be implemented as a computer program, which can be used in, for example... Figure 3 It runs on the system shown.

[0120] 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.

[0121] 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 robot-stabilized contact force control method for assembly scenarios.

[0122] The processor provides computing and control capabilities, supporting the operation of the entire computer device.

[0123] The internal memory provides an environment for the execution of computer programs in non-volatile storage media. When the computer program is executed by the processor, it enables the processor to execute any robot stable contact force control method for assembly scenarios.

[0124] This network interface is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that... Figure 4The 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.

[0125] 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.

[0126] In one embodiment, the processor is configured to run a computer program stored in memory to perform the following steps: Zero-point calibration is performed on the six-dimensional force sensor to obtain the sensor's zero-drift value. Then, the sensor detection data is processed for load gravity compensation to eliminate the interference of gravity components caused by changes in robot pose on the force detection results. The robot's end effector moves the shaft to be assembled to the vicinity of the target hole, driving the shaft to feed axially toward the workpiece mating surface. A six-dimensional force sensor collects contact force data from the robot's end effector in real time, continuously extracting the real-time value of the axial contact force. This process includes: reading the raw detection data from the six-dimensional force sensor at a fixed sampling frequency; filtering the raw detection data; establishing a robot end effector coordinate system; separating the force component corresponding to the axial direction from the filtered detection data; and using the force component corresponding to the axial direction as the real-time value of the axial contact force. Based on the real-time acquired axial contact force values, the axial position and feed speed of the robot end effector are dynamically adjusted through admittance control logic to stably maintain the axial contact force within the preset value range, forming a stable axial force interaction environment and providing a reliable contact basis for hole position search and shaft insertion operations.

[0127] In some embodiments, performing zero-point calibration on the six-dimensional force sensor to obtain the sensor zero-drift value includes: controlling the robot end effector to remain in an unloaded and stationary state; continuously collecting multiple sets of output data from the six-dimensional force sensor at a preset sampling frequency; calculating the average of the collected multiple sets of output data; and using the calculated average as the sensor zero-drift value.

[0128] In some embodiments, the process of performing load gravity compensation processing on sensor detection data to eliminate the interference of gravity components caused by changes in robot pose on the force detection results includes: pre-collecting sensor output data of the end-effector load in multiple different poses; calculating the force component parameters corresponding to the load gravity in different poses by combining the sensor zero drift values; reading the current pose information during robot operation; calling the stress component parameters according to the current pose information; subtracting the stress component parameters from the original sensor detection data; and obtaining the true external contact force data.

[0129] In some embodiments, the control robot end effector moves the shaft to be assembled to a region adjacent to the target hole, and drives the shaft to be assembled to feed axially toward the workpiece mating surface, including: receiving pre-taught and stored position parameters of the adjacent point of the target hole; controlling the robot end effector to move the shaft to be assembled to the adjacent point position; identifying the axial direction of the shaft to be assembled; driving the shaft to be assembled to move axially toward the workpiece mating surface according to a preset feed speed; detecting the numerical change of the axial contact force in real time; and stopping the feed motion when the axial contact force reaches a preset contact threshold.

[0130] In some embodiments, the step of using the real-time acquired axial contact force value includes: smoothing the axial contact force values ​​for multiple consecutive sampling periods; removing abnormal data whose values ​​jump beyond a preset range; and using the smoothed axial contact force value as an input parameter for admittance control logic.

[0131] In some embodiments, dynamically adjusting the axial position and feed rate of the robot end effector through admittance control logic includes: pre-setting virtual inertia parameters, virtual damping parameters, and virtual stiffness parameters corresponding to the admittance control logic; calculating the difference between the real-time value of the axial contact force and the preset target contact force; calculating the axial position correction amount by combining the virtual inertia parameters, virtual damping parameters, and virtual stiffness parameters; and adjusting the axial position and corresponding feed rate of the robot end effector according to the position correction amount.

[0132] In some embodiments, maintaining the axial contact force stably within a preset value range to form a stable axial force interaction environment includes: real-time monitoring of the fluctuation amplitude of the axial contact force; adjusting the axial feed speed amplitude when the axial contact force exceeds the preset value range; maintaining the current axial position unchanged when the axial contact force falls within the preset value range; continuously maintaining the stable state of the axial contact force; and establishing a constant axial force interaction basis.

[0133] In some embodiments, the method further includes: during the process of maintaining the axial contact force stably, collecting the workpiece material parameters and shaft size parameters corresponding to the current assembly; retrieving pre-stored historical assembly data under the same working conditions; extracting the assembly success rate index from the historical assembly data; adaptively adjusting the virtual damping parameters and virtual stiffness parameters corresponding to the admittance control logic according to the assembly success rate index; and matching to obtain the optimal control parameters for the current assembly working condition.

[0134] In some embodiments, the method further includes: continuously collecting trend data of axial contact force changes; calculating the rate of change of axial contact force for multiple consecutive sampling periods; determining whether there is an abnormality in the current contact state based on the rate of change of axial contact force; when it is determined that there is an abnormality in the contact state, immediately controlling the robot end effector to retreat a preset distance in the axial direction; and re-executing the axial feed and contact force stabilization control process.

[0135] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to implement the steps of the robot stable contact force control method for assembly scenarios provided in any embodiment of this application.

[0136] 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.

[0137] 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 robot stable contact force control method for assembly scenarios, applied to a robot shaft hole assembly system with a six-dimensional force sensor at the end effector, characterized in that, include: Zero-point calibration is performed on the six-dimensional force sensor to obtain the sensor's zero-drift value. Then, the sensor detection data is processed for load gravity compensation to eliminate the interference of gravity components caused by changes in robot pose on the force detection results. The robot end effector moves the shaft to be assembled to the vicinity of the target hole, and drives the shaft to be assembled to feed along the axial direction toward the workpiece mating surface. The robot end effector acquires contact force data in real time using a six-dimensional force sensor and continuously extracts the real-time value of the axial contact force. This process includes: reading the raw detection data from the six-dimensional force sensor at a fixed sampling frequency; filtering the raw detection data; establishing the robot end effector coordinate system; separating the force component corresponding to the axial direction from the filtered detection data; and using the force component corresponding to the axial direction as the real-time value of the axial contact force. Based on the real-time acquired axial contact force values, the axial position and feed speed of the robot end effector are dynamically adjusted through admittance control logic to stably maintain the axial contact force within the preset value range, forming a stable axial force interaction environment and providing a reliable contact basis for hole position search and shaft insertion operations.

2. The method according to claim 1, characterized in that, The zero-point calibration operation of the six-dimensional force sensor to obtain the sensor's zero-drift value includes: Control the robot's end effector to remain stationary and unloaded; Multiple sets of output data from the six-dimensional force sensor are continuously collected according to a preset sampling frequency. Calculate the mean of the multiple sets of output data collected; The calculated mean value is used as the zero drift value of the sensor.

3. The method according to claim 1, characterized in that, The process of performing load gravity compensation processing on sensor detection data to eliminate the interference of gravity components caused by changes in robot pose on the force detection results includes: Pre-collect sensor output data of the end-effector load under multiple different poses; Calculate the force component parameters of the load gravity under different poses by combining the zero-drift values ​​of the sensor; The robot reads its current pose information during operation; The stress component parameters are retrieved based on the current pose information; the stress component parameters are subtracted from the original sensor detection data. Obtain real external contact force data.

4. The method according to claim 1, characterized in that, The control robot end effector moves the shaft to be assembled to the vicinity of the target hole, and drives the shaft to be assembled to feed axially toward the workpiece mating surface, including: Receive the pre-taught and stored location parameters of the target hole's neighboring points; The robot's end effector drives the shaft to be assembled to a nearby position. Identify the axial direction of the shaft to be assembled; Drive the shaft to be assembled to move axially toward the workpiece mating surface according to the preset feed speed; Real-time detection of numerical changes in axial contact force; The feed motion stops when the axial contact force reaches the preset contact threshold.

5. The method according to claim 1, characterized in that, The step of using the axial contact force value acquired in real time includes: The axial contact force values ​​from multiple consecutive sampling periods are smoothed. Remove abnormal data whose values ​​jump beyond the preset range; The smoothed axial contact force value is used as the input parameter of the admittance control logic.

6. The method according to claim 1, characterized in that, The method of dynamically adjusting the axial position and feed rate of the robot end effector through admittance control logic includes: The virtual inertial parameters, virtual damping parameters, and virtual stiffness parameters corresponding to the admittance control logic are preset. Calculate the difference between the real-time value of the axial contact force and the preset target contact force; The axial position correction is calculated by combining virtual inertia parameters, virtual damping parameters, and virtual stiffness parameters. Adjust the axial position of the robot's end effector and the corresponding feed rate based on the position correction amount.

7. The method according to claim 1, characterized in that, Maintaining the axial contact force stably within a preset value range to form a stable axial force interaction environment includes: Real-time monitoring of fluctuations in axial contact force; When the axial contact force exceeds the preset value range, the axial feed speed amplitude should be adjusted accordingly. Once the axial contact force falls within the preset value range, the current axial position remains unchanged. Maintain a stable state of axial contact force; establish a constant basis for axial force interaction.

8. The method according to claim 1, characterized in that, The method further includes: During the process of maintaining stable axial contact force, collect the workpiece material parameters and shaft dimension parameters corresponding to the current assembly; Retrieve pre-stored historical assembly data under the same working conditions; extract assembly success rate indicators from the historical assembly data. The virtual damping parameters and virtual stiffness parameters corresponding to the admittance control logic are adaptively adjusted based on the assembly success rate index. The optimal control parameters for the current assembly condition are obtained through matching.

9. The method according to claim 1, characterized in that, The method further includes: Continuously collect data on the changing trend of axial contact force; calculate the rate of change of axial contact force over multiple consecutive sampling periods; Determine whether there is an abnormality in the current contact state based on the rate of change of axial contact force; When an abnormality is detected in the contact state, the robot end effector is immediately controlled to retreat a preset distance in the axial direction. Re-execute the axial feed and contact force stabilization control process.

Images