A method for automatically subdividing and scheduling a humanoid robot loading and unloading task

Abstract

The application provides a humanoid robot loading and unloading task automatic subdivision and scheduling method, comprising: according to the shear force time sequence record, using Kalman filtering to suppress the noise and false adhesion components superimposed in the record, separating out the potential adhesion disturbance, and determining the interference level of the false adhesion component; through the adhesion effect classification conclusion, evaluating the degree of false adhesion component covering the real slip characteristics, and updating the slip identification threshold accordingly; identifying the mutation point in the shear force time sequence record of the humanoid robot loading and unloading clamping process which crosses the updated slip identification threshold, and determining the real mutation position; according to the real mutation position, analyzing the shear force direction and intensity change at the mutation point, confirming whether it meets the slip signal characteristics, and obtaining the slip signal confirmation conclusion; after evaluating the positive slip signal confirmation conclusion, immediately sending a trigger command to the humanoid robot control system to activate the emergency subtask to execute the stop and then re-clamp operation.

Description

Technical Field

[0001] This invention relates to the field of robot control technology, and in particular to a method for automatic subdivision and scheduling of loading and unloading tasks for humanoid robots. Background Technology

[0002] In modern industrial manufacturing, the loading and unloading tasks of humanoid robots are considered a crucial pillar for improving production efficiency and flexible manufacturing. Especially under complex working conditions, their automated operation has irreplaceable value in ensuring production safety and reducing labor costs. With the deepening of intelligent manufacturing, how to enable robots to more accurately grasp and transport workpieces has become a focus of industry attention. However, current methods in this field mostly focus on simply increasing mechanical gripping force or optimizing the gripping angle, failing to fully consider the dynamic effects of liquid properties on the gripping effect. This leads to difficulties in accurately determining whether the workpiece is truly stable during actual operation, thus affecting the reliability of the entire task. The viscosity of the cutting fluid film plays a dual and contradictory role in the gripping process. On the one hand, it establishes an adhesive bond between the workpiece and the robot's gripping surface in the initial stage of gripping, making the workpiece appear to be firmly held, leading the robot to mistakenly believe that the gripping is in place. On the other hand, it continuously adds a false adhesive component to the output of the shear force sensor, submerging the mechanical abrupt changes that should appear when the workpiece begins to slip slightly under the viscous drag effect. This makes conventional slip detection methods almost ineffective under high-viscosity liquid film conditions, and the system cannot promptly trigger emergency subtasks such as "stop and then re-grip". The viscosity of the cutting fluid film refers to the viscosity of the thin film formed between the workpiece and the robot's gripping surface, which directly determines the magnitude of friction and adhesion during gripping. If the liquid film viscosity is high, the workpiece may appear stable in the initial stage of gripping, but this adhesion may mask the subtle signals of actual workpiece slippage; conversely, when the viscosity is low, the workpiece is more likely to slip directly, but the signals may be more obvious. This contradiction makes it difficult for robots to accurately distinguish whether a workpiece is slipping in different liquid environments, thus hindering their ability to take timely countermeasures. Specifically, in a typical loading / unloading scenario, suppose a robot picks up a freshly machined metal part covered in high-viscosity sulfurized cutting oil used for deep-hole drilling. Due to the adhesive effect of the liquid film, the workpiece appears firmly clamped, but in reality, it may have already begun to slip slightly during transport. The robot, however, fails to detect this due to signal interference, ultimately causing the workpiece to fall, resulting in production interruption or equipment damage. Therefore, accurately identifying the actual slipping signals during workpiece picking under different liquid film viscosities and triggering timely emergency measures has become a key issue in improving the reliability of humanoid robots in loading / unloading tasks. Summary of the Invention

[0003] This invention provides a method for automatic subdivision and scheduling of loading and unloading tasks for humanoid robots, mainly including: Shear force data of the contact area between the end-effector gripping surface of the humanoid robot and the workpiece is acquired and fused to obtain a shear force time-series record reflecting the force evolution during the gripping process; Based on the shear force time-series record, Kalman filtering is used to suppress noise and spurious adhesion components, separate potential adhesion disturbances, and determine the interference level. Based on the interference level, the strength of the adhesion effect is analyzed, and a classification conclusion on the adhesion effect is obtained. Based on the classification conclusions of adhesion, assess the degree to which false adhesion components obscure real slip features and update the slip recognition threshold; Based on the updated slip recognition threshold, identify the abrupt change points in the shear force time series record and determine the actual abrupt change location; Based on the actual abrupt change location, analyze the shear force change characteristics to confirm whether it conforms to the characteristics of a slip signal, and obtain a slip signal confirmation conclusion; If the sliding signal confirmation is positive, a trigger command is sent to the humanoid robot control system to activate the emergency subtask to perform the stop and re-clamping operation.

[0004] Furthermore, the acquisition of shear force data in the contact area between the end-effector of the humanoid robot and the workpiece, and the fusion of this data to obtain a shear force time-series record reflecting the force evolution during the clamping process, includes: The shear force intensity value and vector direction of each measuring point in the contact area between the clamping surface and the workpiece are collected, and the measurement data are obtained according to a fixed sampling period to obtain shear force distribution data covering the entire contact area. Based on the shear force distribution data, the adhesion force reading between the clamping surface and the workpiece surface is read in the initial stage of gripping, the thickness and viscosity of the residual liquid film at the contact interface before the workpiece is unloaded are identified, and the liquid film viscosity level is divided according to the preset range. The shear force distribution data, the adhesion force readings, and the liquid film viscosity level are fused using a unified time reference to generate a shear force time series record that includes changes in shear force intensity, direction deflection, and adhesion characteristics.

[0005] Furthermore, the step of using Kalman filtering to suppress noise and spurious adhesion components, separating potential adhesion disturbances, and determining the interference level based on the shear force time-series record includes: Based on the shear force time series record, the shear force intensity value and direction deflection data at each sampling time are extracted, a state space representation is established, and the measured values ​​are used as observation inputs to obtain the state observation sequence; Kalman filtering is used to recursively process the state observation sequence. The shear force state at the current moment is predicted based on the state estimate of the previous moment. The predicted value is compared with the actual observed value to obtain the observation residual. The predicted value is corrected and updated based on the observation residual. After iteration, a smooth shear force sequence is obtained. The difference between the smooth shear force sequence and the original record is calculated to obtain the filtered residual. The measurement noise component and the adhesion disturbance component are identified based on the changing frequency, and the potential adhesion disturbance sequence is separated. For the adhesion perturbation sequence, the average perturbation amplitude and the number of continuous sampling times are statistically analyzed. If the preset conditions are met, the interference level of the false adhesion component is determined to be high; otherwise, it is determined to be low.

[0006] Furthermore, based on the interference level, the strength of the adhesion is analyzed to obtain a classification conclusion on the adhesion, including: The interference level is compared with a preset threshold. If the interference level exceeds the preset threshold, the current clamping condition is marked as a high viscosity liquid film state. The adhesion effect is divided into levels according to the magnitude of the exceedance. If the exceedance is greater than the preset threshold, it is classified as a strong adhesion level; otherwise, it is classified as a weak adhesion level, thus obtaining the adhesion effect classification conclusion. If the interference level does not exceed the preset threshold, the cutting fluid dragging effect is determined to be weak, and the conventional slip recognition threshold is used to obtain the adhesion classification conclusion. Based on the aforementioned classification conclusions of adhesion, the strength of adhesion under high-viscosity liquid films is distinguished, thus forming a classification result of adhesion.

[0007] Furthermore, based on the interference level, the strength of the adhesion is analyzed to obtain a classification conclusion on the adhesion, including: When the interference level exceeds a preset threshold, the thickness and viscosity of the liquid film at the contact interface are obtained; Based on the thickness and viscosity of the liquid film, the spread range of the viscous liquid on the clamping surface is evaluated. The spread range value is calculated by weighted summation of the number of measuring points and the distribution area. At the same time, the duration of workpiece pulling and stopping is recorded. Based on the spreading range and the duration of the pulling pause, the motion characteristics of the workpiece when it leaves the clamping surface are identified. If the spreading range and the duration of the pulling pause both meet the preset range threshold and duration threshold respectively, it is determined to be a delayed detachment and classified into the strong adhesion category; otherwise, it is determined to be an instantaneous detachment and classified into the weak adhesion category, thus obtaining the adhesion effect classification conclusion.

[0008] Furthermore, the step of evaluating the degree to which false adhesion components obscure real slip features based on the adhesion classification conclusion, and updating the slip recognition threshold, includes: Based on the classification conclusion of adhesion, the judgment content of strong adhesion and weak adhesion is retrieved. For the performance of delayed detachment and instantaneous detachment, the degree of occlusion of false adhesion components on real slip characteristics is evaluated. The occlusion degree value is calculated by the ratio of the adhesion disturbance amplitude to the normal fluctuation amplitude. Based on the stated degree of obscuration, a preset adjustment coefficient is matched, and the increase in the original slip recognition threshold by the high viscosity liquid film is calculated, wherein the adjustment coefficient corresponding to the strong adhesion level is greater than the adjustment coefficient corresponding to the weak adhesion level. By superimposing the aforementioned lifting amplitude on the original sliding recognition threshold, an updated sliding recognition threshold is obtained.

[0009] Furthermore, the step of evaluating the degree to which false adhesion components obscure real slip features based on the adhesion classification conclusion, and updating the slip recognition threshold, includes: Retrieve the strong adhesion and weak adhesion determination content from the adhesion classification conclusion, and obtain the spreading range value and pulling pause time corresponding to the two determination contents of strong adhesion and weak adhesion. Based on the spread range value and the duration of slack, the degree of occlusion of the false adhesion component is evaluated to cover the real slip change point. The occlusion value is determined by comparing the adhesion disturbance amplitude with the normal fluctuation amplitude. The occlusion degree of the strong adhesion is higher than that of the weak adhesion. Based on the stated degree of obscuration, a preset lifting coefficient is matched to calculate the additional lifting amplitude of the high-viscosity liquid film on the original slip recognition threshold. The additional lifting amplitude is superimposed on the original slip recognition threshold to obtain the updated slip recognition threshold.

[0010] Furthermore, the step of identifying abrupt change points in the shear force time series record and determining the actual abrupt change location based on the updated slip recognition threshold includes: Based on the shear force time sequence record, the shear force intensity value is compared with the updated slip recognition threshold at each time step, and the sampling time that exceeds the threshold is marked to obtain a set of candidate mutation times; For each sampling time in the candidate mutation time set, the amplitude of the shear force intensity jump and the deflection angle are identified. If both the jump amplitude and the deflection angle exceed a preset threshold, the sampling time is determined to be the actual mutation location.

[0011] Furthermore, the step of analyzing the shear force change characteristics based on the actual abrupt change location to confirm whether it conforms to the characteristics of a slip signal and obtaining a slip signal confirmation conclusion includes: Extract the shear force intensity values ​​and direction angles at the actual mutation location and the sampling times before and after it, obtain the shear force intensity jump and direction deflection at the mutation point, and obtain the shear force change characteristics; The shear force variation characteristics are matched and compared with the preset slip signal characteristics, wherein the slip signal characteristics include a combination of a decreasing trend in shear force intensity and a significant deflection of the direction angle. If the shear force change characteristics match the slip signal characteristics, it is determined to be a real workpiece slip event and a slip signal confirmation conclusion is output as positive; otherwise, it is determined to be an interference signal and a confirmation conclusion is output as negative, thus obtaining a slip signal confirmation conclusion.

[0012] Furthermore, if the sliding signal confirmation is positive, a trigger command is sent to the humanoid robot control system to activate the emergency subtask to perform a stop and re-clamping operation, including: If the confirmation conclusion of the slip signal is positive, a trigger command is sent to the humanoid robot control system. The trigger command carries the actual change position of the current clamping condition and the slip time mark. According to the trigger command, the pre-configured emergency sub-task is activated, controlling the humanoid robot to stop the current handling action and maintain the workpiece position; According to the emergency sub-task execution process, the clamping force is increased to re-clamp the workpiece, completing the stop and re-clamping operation, and obtaining the emergency adjustment result of the clamping process.

[0013] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: This invention discloses an automatic subdivision and scheduling method for humanoid robot loading and unloading tasks. It proposes a complete solution to address the challenges of adhesion interference and slippage recognition caused by high-viscosity liquid films during workpiece loading and unloading operations. The invention collects shear force, adhesion force, and liquid film viscosity data of the contact area between the clamping surface and the workpiece, fuses them to form a shear force time-series record, and uses Kalman filtering to suppress noise and separate false adhesion components. This allows for the assessment of interference levels and adhesion strength, classifying adhesion strength into strong and weak conclusions. Based on this, the invention dynamically updates the slippage recognition threshold to adapt to the detection requirements of the clamping process in high-viscosity environments, monitors shear force abrupt changes in real time, accurately confirms slippage signals, and finally triggers an emergency subtask when slippage risk is detected, performing a stop and re-clamping operation. The core innovation of this invention lies in its significantly improved accuracy and reliability of slippage recognition in complex environments through adhesion interference analysis and adaptive threshold adjustment, ensuring the safety and efficiency of loading and unloading operations. Attached Figure Description

[0014] Figure 1 This is a flowchart of an automatic subdivision and scheduling method for loading and unloading tasks of a humanoid robot according to the present invention.

[0015] Figure 2 This is a schematic diagram of an automatic subdivision and scheduling method for loading and unloading tasks of a humanoid robot according to the present invention.

[0016] Figure 3 This is another schematic diagram of an automatic subdivision and scheduling method for loading and unloading tasks of a humanoid robot according to the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0018] like Figures 1-3 This embodiment of the method for automatic subdivision and scheduling of loading and unloading tasks for humanoid robots may specifically include: S101. During the workpiece loading and unloading operation of the humanoid robot, the shear force intensity and direction information of the contact area between the end clamping surface and the workpiece are collected. At the same time, the adhesion force reading at the initial stage of gripping and the viscosity level of the liquid film formed at the contact interface before the workpiece is unloaded are obtained. The results are fused to obtain a shear force time sequence record that reflects the force evolution during the loading and unloading clamping process.

[0019] During the workpiece loading and unloading operation of the humanoid robot, multi-point force-sensitive elements arranged on the end gripping surface collect the shear force intensity values ​​at each measuring point in the contact area between the gripping surface and the workpiece, and simultaneously record the vector direction of the shear force at each measuring point. Shear force intensity and direction measurement data are obtained according to a fixed sampling period, resulting in shear force distribution data covering the entire contact area. Based on the shear force distribution data, the adhesion force reading generated between the gripping surface and the workpiece surface is read in the initial stage of gripping. The thickness and viscosity of the liquid film formed by residual cutting fluid at the contact interface before workpiece unloading are identified. Based on preset liquid film thickness and viscosity ranges, the liquid film viscosity level is divided, obtaining the adhesion force reading and liquid film viscosity level reflecting the current gripping condition. Using a unified time reference, the shear force distribution data, the adhesion force reading, and the liquid film viscosity level are fused and processed. The force evolution state at each moment is arranged according to the time sequence of the loading and unloading gripping process, generating a shear force time series record including changes in shear force intensity, direction deflection, and adhesion characteristics.

[0020] During the workpiece loading and unloading operation of the humanoid robot, the end-effector gripping surface is in direct contact with the workpiece surface, and the force state of this contact area directly reflects the gripping stability. The multi-point force-sensitive elements are distributed in an array in the contact area of ​​the gripping surface, with each element corresponding to an independent measuring point to sense the shear force information at that measuring point.

[0021] Specifically, the multi-point force-sensitive element adopts a piezoresistive or capacitive thin-film sensing structure. When the workpiece contacts the clamping surface, the shear force at each measuring point in the contact area causes the sensing element to generate an electrical signal change. Each measuring point simultaneously senses the components of the shear force in two orthogonal directions, and obtains the shear force intensity value and vector direction of the measuring point through vector synthesis. The fixed sampling period is set according to the movement speed of the loading and unloading operation, and the shear force information of all measuring points is read synchronously at each sampling moment to form shear force distribution data covering the entire contact area. In one embodiment, the acquisition of the adhesion force reading occurs in the initial stage of gripping, that is, within a short period of time after the clamping surface has just established stable contact with the workpiece surface. At this time, the workpiece has not yet started to be moved, and the cutting fluid between the clamping surface and the workpiece forms a thin liquid film under the action of contact pressure. The adhesion effect generated by this liquid film is manifested as a pulling force perpendicular to the contact surface direction, and the adhesion force value is read by the normal force-sensitive element arranged on the clamping surface.

[0022] It should be noted that the classification of the liquid film viscosity grade is based on two dimensions: liquid film thickness and viscosity. Liquid film thickness refers to the size of the thin layer gap formed by the cutting fluid between the workpiece surface and the clamping surface, which can be indirectly obtained by the change in normal displacement of the clamping surface during the initial contact. Viscosity reflects the flow resistance characteristics of the cutting fluid, is related to the type of cutting fluid and temperature, and can be determined based on the decay rate of the initial adhesion force reading over time.

[0023] For example, if the adhesion force reading rapidly decreases to a stable value within a short period, it indicates that the liquid film viscosity is low; if the adhesion force reading remains at a high level for a long time and then slowly decreases, it indicates that the liquid film viscosity is high. Based on a preset lookup table of liquid film thickness and viscosity ranges, the current clamping condition is categorized into the corresponding liquid film viscosity level, which is used to indicate the strength of the adhesive characteristics at the current contact interface. After acquiring the shear force distribution data, adhesion force reading, and liquid film viscosity level, a unified time reference is used to align the aforementioned multi-source data. The unified time reference uses the start time of the robot control cycle as a reference, converting the acquisition time of various data types to the same time coordinate system. During the fusion processing, according to the time sequence of the loading and unloading clamping process, the changes in shear force intensity, direction deflection, and corresponding adhesion characteristics at each moment are integrated into a continuous time-series data structure, generating the shear force time-series record. This record fully reflects the evolution trajectory of the force state of the clamping surface from the initial gripping stage to the handling process.

[0024] S102. Based on the shear force time-series record, Kalman filtering is used to suppress the noise and spurious adhesion components superimposed within the record, separate potential adhesion disturbances, and determine the interference level of spurious adhesion components.

[0025] Based on the shear force time series record, the shear force intensity and direction deflection data at each sampling moment are extracted from the record. A state-space representation with shear force intensity and direction as state variables is established. The shear force measurement value at each sampling moment is used as the observation input to obtain a state observation sequence for filtering. Kalman filtering is used to recursively process the state observation sequence. The shear force state at the current moment is predicted based on the state estimate of the previous moment. The predicted value is compared with the actual observed value at the current moment to obtain the observation residual. The predicted value is corrected and updated based on the observation residual. After iterating step by step, a smooth shear force sequence with suppressed noise is obtained. The smooth shear force sequence and the original shear force time series record are subjected to step-by-step difference calculation to obtain the filtered residual at each sampling moment. The components are identified based on the changing frequency of the filtered residual. The residual with a changing frequency higher than a preset frequency threshold is classified as measurement noise component, and the residual with a changing frequency lower than the preset frequency threshold and showing a slow decay trend is classified as adhesion disturbance component, thus separating the potential adhesion disturbance sequence. For the adhesion perturbation sequence, the average value of the perturbation amplitude and the number of sampling moments during which the perturbation lasts are statistically analyzed. If the average value exceeds a preset perturbation intensity threshold and the number of sampling moments exceeds a preset duration threshold, the interference level of the false adhesion component is determined to be high; otherwise, the interference level of the false adhesion component is determined to be low.

[0026] During the loading and unloading operation of the humanoid robot, the shear force time-series record contains complete force evolution information of the contact area between the clamping surface and the workpiece. This record not only contains random noise generated by sensor measurements, but also superimposed false mechanical signals caused by the adhesion of the cutting fluid film. The state space representation combines the intensity value and direction angle of the shear force into a multi-dimensional state vector. The change of this state vector with the sampling time constitutes a complete state evolution trajectory.

[0027] Specifically, the state variables include the components of the shear force in two orthogonal directions within the clamping plane, and the combined shear force intensity and pointing angle calculated from these two components. At each sampling moment, the shear force measurement output by the sensor serves as an observation of the true state. Due to uncertainties in the measurement process, there is a deviation between the observed value and the true state. Arranging the observations from multiple consecutive sampling moments in chronological order forms the state observation sequence, which is the input data for Kalman filtering. In one embodiment, the recursive processing of the Kalman filter follows an iterative mechanism of alternating prediction and update. In the prediction phase, based on the state estimate corrected at the previous moment and combined with the variation pattern of the shear force over a short period, the predicted value of the shear force state at the current moment is calculated. This predicted value reflects the prior judgment of the current state without considering new observation information. In the update phase, the predicted value is compared with the actual observation value output by the sensor at the current moment. The difference between the two is the observation residual, which contains a mixture of prediction bias and measurement noise. Based on the magnitude of the observation residuals and the relative proportion of prediction uncertainty to observation uncertainty, the predicted values ​​are weighted and corrected. The corrected result is used as the final state estimate for the current moment. This prediction and update process is repeated at each sampling moment. After iterating through the entire time series record, the fluctuation amplitude of random noise in the output smooth shear force sequence is significantly reduced.

[0028] It should be noted that the filtered residual is the difference between the original observation and the smoothed estimate at the same time. This difference comprehensively reflects the various interference components suppressed by the filtering process. Measurement noise typically exhibits high-frequency random fluctuations, with no regularity in the direction and amplitude of its changes between adjacent sampling times, showing a rapid alternation of positive and negative values ​​in the filtered residual sequence. The spurious adhesion component generated by liquid film adhesion has different time-domain characteristics. Since the adhesion force, after being established in the initial stage of grasping, slowly decays as the liquid film is gradually squeezed out, the adhesion disturbance in the filtered residual appears as a low-frequency component that changes slowly and shows a monotonically decreasing trend.

[0029] For example, the preset frequency threshold is set based on the typical motion cycle of loading and unloading operations and the viscosity characteristics of the cutting fluid. In scenarios involving gripping and transporting metal workpieces, the shear force on the robot's gripping surface fluctuates periodically due to the workpiece's inertia and acceleration, and this fluctuation frequency is related to the transport cycle. The fluctuation frequency of the measured noise is typically several times higher than the frequency of mechanical changes caused by motion, while the decay period of adhesion disturbances is much longer than the duration of a single transport action. By setting the frequency threshold in the boundary region between the motion fluctuation frequency and the noise fluctuation frequency, the high-frequency components and low-frequency components in the filtered residual can be separated.

[0030] In one embodiment, the adhesion perturbation sequence obtained after frequency separation records the amplitude of the spurious adhesion component at each sampling moment. When statistically processing this sequence, the average perturbation amplitude reflects the overall intensity level of the adhesion effect, while the number of sampling moments during which the perturbation persists reflects the duration of the adhesion effect over time. High-viscosity cutting fluids form strong and long-lasting liquid film adhesion, corresponding to higher average perturbation amplitudes and a larger number of durations; low-viscosity cutting fluids form weak and rapidly dissipating liquid film adhesion, resulting in relatively mild perturbation characteristics.

[0031] Preferably, the disturbance intensity threshold and duration threshold are pre-calibrated based on the viscosity parameters of different types of cutting fluids. When the statistical characteristics of the adhesion disturbance sequence simultaneously exceed both thresholds, it indicates that the degree of interference of the spurious adhesion components on the shear force measurement is high under the current clamping condition, and the interference level is determined to be high. When any statistical characteristic does not reach the corresponding threshold, it indicates that the influence of the spurious adhesion components is relatively limited, and the interference level is determined to be low. The determination result of this interference level directly reflects the severity of the masking effect of liquid film adhesion on subsequent slip recognition.

[0032] S103. Analyze the situation where the interference level of false adhesion components exceeds the preset threshold, distinguish the strength of adhesion under high viscosity liquid film, and obtain the classification conclusion of adhesion. If it does not exceed the preset threshold, it is determined that the drag effect of cutting fluid is weak and the conventional slip recognition threshold is used.

[0033] Based on the interference level of the false adhesive components, this interference level is compared with a preset threshold. If the interference level exceeds the preset threshold, the current clamping condition is marked as a high-viscosity liquid film state. The adhesion effect is then classified according to the magnitude of the interference level exceeding the preset threshold: if the exceedance is greater than a preset threshold, it is classified as strong adhesion; if the exceedance is less than or equal to the preset threshold, it is classified as weak adhesion, thus obtaining the adhesion effect classification conclusion. If the interference level does not exceed the preset threshold, the current cutting fluid drag effect is determined to be weak, and the conventional slip recognition threshold is used to obtain the adhesion effect classification conclusion.

[0034] After obtaining the interference level of the false adhesion component, this interference level is compared with a preset threshold. The preset threshold is pre-calibrated based on the viscosity characteristics of different types of cutting fluids and is used to distinguish whether the current clamping condition is significantly affected by a high-viscosity liquid film. In one embodiment, if the interference level exceeds the preset threshold, it indicates that the liquid film adhesion has significantly interfered with the shear force measurement. In this case, the amplitude value of the interference level exceeding the preset threshold is further calculated. This amplitude value reflects the severity of the adhesion interference. This amplitude value is compared with a preset amplitude threshold. When the amplitude value is greater than the preset amplitude threshold, the adhesion effect is determined to be strong adhesion; when the amplitude value is less than or equal to the preset amplitude threshold, the adhesion effect is determined to be weak adhesion.

[0035] It should be noted that if the interference level does not exceed the preset threshold, it indicates that the current cutting fluid drag effect has a relatively weak impact on the clamping process. In this case, the pre-established conventional slip recognition threshold is used, and there is no need to adjust the slip detection judgment criteria. Through the differentiation and processing of the above two situations, a classification conclusion on adhesion effect adapted to the current clamping conditions is obtained.

[0036] Another way to classify strong and weak adhesion is to analyze the thickness and viscosity of the liquid film at the contact interface when the interference level of the false adhesion component exceeds the threshold, evaluate the spread of the viscous liquid on the clamping surface and the duration of the workpiece being pulled and stopped, identify the different manifestations of delayed detachment of the workpiece under strong adhesion and instantaneous detachment under weak adhesion, and obtain a classification conclusion of adhesion covering strong and weak adhesion.

[0037] When the interference level of the false adhesive component exceeds a preset threshold, the thickness and viscosity of the liquid film at the contact interface are obtained. The thickness of the liquid film is determined based on the change in the contact gap between the clamping surface and the workpiece in the initial stage of gripping. The viscosity is determined based on the rate at which the adhesive force reading decays over time, thus obtaining a thickness and viscosity label that reflects the characteristics of the liquid film at the current contact interface. Based on the thickness and viscosity markings, the spreading range of the cutting fluid viscous on the clamping surface is evaluated. The spreading range is determined through the following process: input the number N of measuring points where the adhesion force reading in the contact area exceeds the preset adhesion threshold of 5N and the area A of the measuring points. The number of measuring points is dimensionless. Calculate the spreading range value S = w1*N + w2*A, where w1 is the quantity weight, set to 0.4, and w2 is the area weight, set to 0.6. For example, when N=20 and A=0.5, S=0.4*20+0.6*0.5. At the same time, record the duration of the workpiece being pulled and stopped by the viscous on the clamping surface to obtain the spreading range value and the duration of the pulling stop. By using the spread range value and the pulling pause duration, the motion characteristics of the workpiece when it detaches from the clamping surface are identified. If the spread range value is greater than a preset threshold and the pulling pause duration exceeds a preset duration threshold, the workpiece is determined to exhibit delayed detachment and is classified as a strong adhesion class. If the spread range value is less than or equal to the preset threshold or the pulling pause duration does not exceed the preset duration threshold, the workpiece is determined to exhibit instantaneous detachment and is classified as a weak adhesion class. Based on the classification results of the strong and weak adhesion classes, and combined with the different performance characteristics of delayed detachment and instantaneous detachment, a classification conclusion on adhesion effects covering both strong and weak adhesion classes is obtained.

[0038] When the interference level of the false adhesion component exceeds a preset threshold, it indicates that the current clamping condition is significantly affected by a high-viscosity liquid film. At this point, the thickness and viscosity of the liquid film at the contact interface are further measured. The liquid film thickness reflects the size of the thin gap formed by the cutting fluid between the workpiece surface and the clamping surface. This gap size is indirectly obtained by the displacement change of the clamping surface during the initial gripping process as it approaches the workpiece.

[0039] Specifically, at the instant the clamping surface establishes contact with the workpiece surface, the cutting fluid is squeezed to form a liquid film. The difference between the displacement of the clamping surface and the theoretical contact position is a characteristic of the liquid film thickness. The viscosity is determined by observing the change in adhesion force readings over time. When the adhesion force reading rapidly decreases and stabilizes within a short period, it indicates a low viscosity of the liquid film; when the adhesion force reading slowly decays over a longer period, it indicates a high viscosity. Combining the liquid film thickness and viscosity creates a labeling information reflecting the characteristics of the liquid film at the current contact interface. In one embodiment, the evaluation of the spreading range is based on the distribution of adhesion force readings at various measuring points within the contact area. Multiple force-sensitive measuring points are arranged on the clamping surface. Each measuring point records an adhesion force reading at the initial stage of gripping. When the adhesion force reading at a measuring point exceeds a preset adhesion threshold, it indicates that there is significant slime coverage at that measuring point. The number of all measurement points exceeding the preset adhesion threshold is counted, and the distribution area of ​​these measurement points on the clamping surface is calculated. This distribution area is a quantitative representation of the viscous fluid spreading range. The larger the spreading range, the more extensive the coverage of the cutting fluid viscous fluid on the contact interface, and the more significant its interference with clamping stability.

[0040] It should be noted that the aforementioned pull-and-hold duration refers to the time interval during which the workpiece's movement is slowed due to the adhesion of viscous liquid during the process of detaching from the clamping surface. Under normal clamping conditions, the workpiece should immediately separate from the clamping surface after the clamping force is released; however, under high-viscosity liquid film conditions, the adhesive force generated by the viscous liquid will pull the workpiece, hindering its movement, resulting in a brief adhesion between the workpiece and the clamping surface. By monitoring changes in the workpiece position or the force state of the clamping surface, and recording the time interval from the release of the clamping force to the complete detachment of the workpiece, the pull-and-hold duration is obtained by subtracting the reference time required for normal detachment from this time interval.

[0041] For example, when a humanoid robot grasps a metal part whose surface is coated with sulfurized cutting oil, if the viscous liquid spreads to cover most of the measuring points in the contact area of ​​the clamping surface, and the workpiece exhibits obvious adhesion and stagnation during release, then this condition exhibits typical characteristics of strong adhesion. Conversely, if the viscous liquid only forms a cover at local measuring points, and the workpiece quickly separates from the clamping surface after release, then this condition exhibits characteristics of weak adhesion.

[0042] In one embodiment, the distinction between delayed release and instantaneous release is based on a combined determination of the spreading range value and the duration of the pulling pause. Delayed release is characterized by a slow process of the workpiece leaving the clamping surface accompanied by significant adhesion and dragging, corresponding to the combination of a spreading range value exceeding a preset threshold and a pulling pause duration exceeding a preset duration threshold. Instantaneous release is characterized by a rapid process of the workpiece leaving the clamping surface without significant pause, corresponding to the condition that either the spreading range value or the pulling pause duration has not reached the corresponding threshold. Based on the above determination, the current clamping condition is categorized into a strong adhesion level or a weak adhesion level.

[0043] The strong adhesion level corresponds to the condition where a thick liquid film is formed by high-viscosity cutting fluid. Under this condition, the spurious adhesion components have a strong masking effect on the real slip signal. The weak adhesion level corresponds to the condition where a thinner liquid film is formed by medium-viscosity cutting fluid. Under this condition, the masking effect of spurious adhesion components is relatively limited. In another embodiment, based on the classification results of the strong adhesion and weak adhesion levels, combined with the different motion performance characteristics of delayed detachment and instantaneous detachment, an adhesion effect classification conclusion covering both levels is output. This conclusion provides a basis for subsequent adjustment of the slip recognition threshold.

[0044] S104. Based on the classification conclusions of adhesion, assess the degree to which false adhesion components mask true slip characteristics, and update the slip recognition threshold accordingly.

[0045] Based on the classification conclusions of adhesion, the judgment content of strong adhesion or weak adhesion is retrieved. For the delayed detachment performance corresponding to the strong adhesion level and the instantaneous detachment performance corresponding to the weak adhesion level, the degree of masking of the true slip characteristics by the spurious adhesion component is evaluated. The masking degree is determined by the ratio of the adhesion disturbance amplitude to the normal fluctuation amplitude in the shear force time-series record. The masking degree is used to determine the increase in the original slip recognition threshold caused by the high-viscosity liquid film. The increase is obtained by multiplying the masking degree value by a preset threshold adjustment coefficient. The preset threshold adjustment coefficient for the strong adhesion level is 1.5, which is greater than the preset threshold adjustment coefficient for the weak adhesion level (0.8), thus obtaining the threshold increase. Based on this threshold increase, the updated slip recognition threshold is obtained by superimposing this increase on the original slip recognition threshold.

[0046] After obtaining the classification conclusion of adhesion, the degree to which spurious adhesion components mask the true slip characteristics is further evaluated based on the determination of strong or weak adhesion. The degree of masking reflects the strength of the masking effect of the adhesion perturbation signal on the slip abrupt change signal in the shear force time-series record. The stronger the masking effect, the more difficult it is for the slip signal to be identified by conventional thresholds.

[0047] Specifically, the degree of obscuring is determined by comparing the ratio of the adhesion disturbance amplitude to the normal fluctuation amplitude in the shear force time-series record. The normal fluctuation amplitude refers to the periodic change in shear force due to workpiece inertia and acceleration when no slippage occurs, while the adhesion disturbance amplitude refers to the spurious component amplitude superimposed on the shear force measurement by the liquid film adhesion effect. When the ratio of the adhesion disturbance amplitude to the normal fluctuation amplitude is large, the obscuring value is high, indicating a greater risk that the true slippage signal is masked by the adhesive component; when the ratio is small, the obscuring value is low, indicating that the slippage signal is relatively easy to identify. In one embodiment, the preset gear adjustment coefficient is set according to the characteristic differences between the strong adhesion gear and the weak adhesion gear. The hysteresis detachment performance corresponding to the strong adhesion gear means that the adhesion effect lasts for a long time and is strong, resulting in more severe obscuring of the slippage signal; therefore, the corresponding preset gear adjustment coefficient is larger. The instantaneous detachment performance corresponding to the weak adhesion gear means that the adhesion effect fades quickly and is limited in strength, resulting in relatively weak obscuring; therefore, the corresponding preset gear adjustment coefficient is smaller.

[0048] It should be noted that the threshold increase is obtained by multiplying the occlusion level value by the corresponding preset adjustment coefficient. This increase represents the amount by which the original slip recognition threshold needs to be adjusted upward under the current adhesion condition. Based on the above threshold increase, this increase is superimposed on the original slip recognition threshold to obtain the updated slip recognition threshold. The updated slip recognition threshold is higher than the original threshold, which can filter out interference signals generated by false adhesion components under high viscosity liquid film conditions, thereby accurately detecting real slip abrupt changes in the loading and unloading clamping process.

[0049] The strong adhesion and weak adhesion criteria in the adhesion classification conclusions were retrieved, and the degree to which false adhesion components obscured the real slip change points under the two criteria was analyzed. The additional increase of the original slip recognition threshold by the high viscosity liquid film was evaluated, and the updated slip recognition threshold value was obtained to adapt to the recognition requirements of the loading and unloading clamping process.

[0050] The strong adhesion and weak adhesion levels are retrieved from the adhesion classification conclusions. The spreading range and drag-and-hold duration corresponding to the strong adhesion level are obtained, as are the spreading range and drag-and-hold duration corresponding to the weak adhesion level, thus obtaining adhesion characteristic parameters for both levels. Based on these adhesion characteristic parameters, the degree of occlusion of the true slip abrupt change point by the spurious adhesion component is evaluated for both the strong and weak adhesion levels. The occlusion degree is determined by comparing the relative magnitude of the adhesion disturbance amplitude and the normal fluctuation amplitude in the shear force time series record. The occlusion degree is higher in the strong adhesion level than in the weak adhesion level, thus obtaining the occlusion degree value corresponding to the current level. Using this occlusion degree value, a preset level increase coefficient is matched. The preset level increase coefficient corresponding to the strong adhesion level is greater than that corresponding to the weak adhesion level. The occlusion degree value is multiplied by the corresponding preset level increase coefficient to obtain the additional increase in the original slip recognition threshold by the high-viscosity liquid film. By adding the additional lifting amplitude to the original sliding recognition threshold, an updated sliding recognition threshold value is obtained. The updated sliding recognition threshold value is adapted to the recognition requirements of the loading and unloading clamping process.

[0051] After obtaining the classification conclusion of adhesion, the adhesion characteristic parameters for the corresponding level are retrieved based on the classification results of strong or weak adhesion. The adhesion characteristic parameters include the spreading range value and the duration of pausing during pulling. These two parameters have been determined in the aforementioned adhesion classification process and are directly used here as the input basis for subsequent occlusion assessment.

[0052] In one embodiment, the strong adhesion setting corresponds to a larger spreading range and a longer dragging stagnation time, indicating that the high-viscosity cutting fluid has formed a widely covering and continuously acting viscous layer on the clamping surface; the weak adhesion setting corresponds to a smaller spreading range and a shorter dragging stagnation time, indicating that the coverage of the cutting fluid viscous layer is limited and the adhesion effect dissipates rapidly. The difference in characteristic parameters between these two settings directly affects the masking strength of the false adhesion components on the real slip signal. In another embodiment, the masking degree is evaluated by comparing the relative magnitude of the adhesion disturbance amplitude with the normal fluctuation amplitude in the shear force time series record. The normal fluctuation amplitude refers to the background fluctuation amplitude of the shear force caused by the workpiece motion inertia and clamping force fine-tuning under the condition of no slippage and slight adhesion interference. This amplitude has a relatively stable range in the same type of loading and unloading operations. The adhesion disturbance amplitude refers to the additional fluctuation amplitude generated by the adhesion of the high-viscosity liquid film in the shear force measurement. This amplitude is extracted from the adhesion disturbance sequence separated by the aforementioned Kalman filter. When the ratio of the adhesion disturbance amplitude to the normal fluctuation amplitude is large, it indicates that the adhesion component occupies a significant proportion in the shear force signal, and the real slip change signal is easily submerged by the adhesion fluctuation, resulting in a correspondingly high degree of obscuration. When the ratio is small, the slip signal is relatively easy to identify, and the degree of obscuration is low.

[0053] It should be noted that the occlusion level value under the strong adhesion setting is usually higher than that under the weak adhesion setting. This is because the strong adhesion setting corresponds to a larger spreading range and a longer dwell time, meaning that the adhesion disturbance signal has stronger coverage capabilities in both time and space dimensions. When evaluating the occlusion level, the corresponding occlusion level value is obtained according to the current clamping condition.

[0054] For example, the preset level lifting coefficient is calibrated according to the different degrees of influence of the strong adhesion level and the weak adhesion level on slip recognition. The preset level lifting coefficient corresponding to the strong adhesion level has a larger value, which is used to compensate for the severe masking of the slip signal by the false adhesion components at this level; the preset level lifting coefficient corresponding to the weak adhesion level has a smaller value, which is used to moderately adjust the threshold to deal with the relatively mild adhesion interference at this level. These two coefficients are predetermined and stored based on the actual test data of cutting fluids of different viscosity grades before being put into use in the loading and unloading operation.

[0055] In one embodiment, the additional lifting amplitude is obtained by multiplying the shielding degree value by the corresponding preset lifting coefficient. This multiplication means that the shielding degree value reflects the actual intensity of adhesion interference under the current clamping condition, and the preset lifting coefficient reflects the baseline proportion of the threshold adjustment at the corresponding level. Multiplying the two yields the specific threshold adjustment value for the current condition. Based on the above additional lifting amplitude, this lifting amplitude is superimposed on the original slip recognition threshold to obtain an updated slip recognition threshold value. The original slip recognition threshold is a conventional threshold value calibrated under low viscosity or no liquid film conditions. The updated slip recognition threshold value is higher than the original threshold, which can filter out false triggering signals generated by false adhesion components under high viscosity liquid film conditions, thereby accurately identifying real slip abrupt changes in the loading and unloading clamping process.

[0056] S105. Identify the abrupt change points in the shear force timing record of the humanoid robot's loading and unloading clamping process that cross the updated slip recognition threshold, and determine the actual abrupt change location.

[0057] The updated slip recognition threshold is obtained. During the loading and unloading clamping process of the humanoid robot, real-time shear force data is continuously collected by the force-sensitive element on the end clamping surface. The shear force intensity and direction at each sampling moment are recorded in real-time according to a fixed sampling period to form a real-time shear force time series record. Based on the real-time shear force time series record, the shear force intensity value is compared with the updated slip recognition threshold at each moment. The sampling moment when the shear force intensity value exceeds the updated slip recognition threshold is marked to obtain a set of candidate abrupt change moments. For each sampling moment in the candidate mutation moment set, the jump amplitude and directional deflection angle of shear force intensity between adjacent sampling moments are identified. The preset mutation amplitude threshold is set through the following process: input the historical shear force jump amplitude dataset, calculate the mean μ and standard deviation σ, and output the threshold Ta = μ + 2σ, for example, Ta = 0.5N in a typical experiment; the preset deflection angle threshold is set through a similar process: input the historical directional deflection angle dataset θ, calculate the mean μ and standard deviation σ, and output the threshold Td = μ + 2σ, for example, Td = 30 degrees in a typical experiment; if the jump amplitude exceeds Ta and the directional deflection angle exceeds Td, then the sampling moment is determined to be the true mutation location, and the true mutation location is determined.

[0058] After obtaining the updated slip recognition threshold, this threshold is applied to the real-time monitoring of the humanoid robot's loading and unloading clamping process. The updated slip recognition threshold has been adaptively adjusted based on the classification conclusions of adhesion effects under the current clamping conditions, and has a higher value than the conventional threshold, which can filter out false fluctuation signals caused by adhesion of high-viscosity liquid films.

[0059] Specifically, the force-sensitive element on the end clamping surface operates continuously during the loading and unloading clamping process, collecting shear force data according to a fixed sampling period to form a real-time shear force time-series record. This real-time shear force time-series record maintains the same data structure as the shear force time-series record used for Kalman filtering, containing the shear force intensity value and direction information at each sampling moment. In one embodiment, the shear force intensity value in the real-time shear force time-series record is compared with the updated slip recognition threshold time-by-time. When the shear force intensity value at a certain sampling moment exceeds the updated slip recognition threshold, that sampling moment is marked as a candidate abrupt change moment. Since the updated threshold has been raised, it can filter out most false exceedance signals caused by adhesion disturbances, and the candidate abrupt change moment set retains time points with a high probability of slippage.

[0060] In another embodiment, not all time points in the candidate abrupt change time set correspond to actual workpiece slippage events, and further screening and confirmation are still required. For each candidate time point in the set, the jump amplitude of shear force intensity and the deflection angle of its direction between that time point and adjacent sampling time points are identified. The jump amplitude reflects the degree of sudden change in shear force within a short period of time, and the deflection angle reflects the degree of sudden change in the direction of shear force. When actual workpiece slippage occurs, the friction state between the clamping surface and the workpiece undergoes an abrupt change, usually manifested as a sharp change in shear force intensity accompanied by a significant deflection of its direction; while limits exceeded by adhesion fluctuations or motion inertia usually do not possess this dual abrupt change characteristic.

[0061] Preferably, if the jump amplitude at a candidate time exceeds a preset abrupt change amplitude threshold, and the directional deflection angle exceeds a preset deflection angle threshold, then the candidate time is determined to be the actual abrupt change location. This determination condition employs a dual-threshold joint constraint, which can distinguish between actual slip abrupt changes and threshold crossing events caused by other interference signals, thereby accurately locating the moment when the workpiece begins to slip under high-viscosity liquid film conditions.

[0062] S106. Analyze the changes in shear force direction and intensity at the mutation point based on the actual mutation location, confirm whether it conforms to the characteristics of slip signal, and obtain the slip signal confirmation conclusion.

[0063] Based on the actual abrupt change location, the shear force intensity value and direction angle at that location and its adjacent sampling times are extracted. The jump variable of the shear force intensity and the deflection of the direction angle at the abrupt change point are obtained, thus acquiring the shear force change characteristics at the abrupt change point. These shear force change characteristics are then compared with preset slip signal characteristics. The slip signal characteristics include a combination of a decreasing trend in shear force intensity at the abrupt change point and a significant deflection of the direction angle. It is determined whether the shear force change characteristics at the abrupt change point conform to the slip signal characteristics. If the shear force change characteristics conform to the slip signal characteristics, the abrupt change point is determined to correspond to a real workpiece slippage event, and a slip signal confirmation conclusion of positive is output. If the shear force change characteristics do not conform to the slip signal characteristics, the abrupt change point is determined to be an interference signal, and a slip signal confirmation conclusion of negative is output, thus obtaining the slip signal confirmation conclusion.

[0064] After determining the actual location of the abrupt change, the shear force variation characteristics at that location are further extracted. These shear force variation characteristics include the jump in shear force intensity and the deflection of the direction angle. These two values ​​are obtained from the actual abrupt change location and the sampling time data of the adjacent times before and after it, reflecting the instantaneous change in the stress state of the clamping surface when the abrupt change occurs.

[0065] Specifically, the jump variable of shear force intensity is obtained by calculating the difference between the shear force intensity at the abrupt change moment and the previous sampling moment. The larger the absolute value of this difference, the more drastic the change in shear force at that moment. The deflection of the direction angle is obtained by comparing the shear force direction angle at the abrupt change moment and the previous sampling moment. The deflection reflects the degree of sudden change in the direction of shear force. In one embodiment, the preset slip signal characteristics are predetermined based on the mechanical behavior of the workpiece during slippage. When the workpiece slips on the clamping surface, the friction state between the clamping surface and the workpiece changes from static friction to dynamic friction, which is usually manifested as a decreasing trend in shear force intensity, while the shear force direction deflects towards the direction of relative workpiece movement. The slip signal characteristics include the combined conditions of a decreasing trend in shear force intensity and a significant deflection of the direction angle.

[0066] It should be noted that when matching and comparing the shear force change characteristics with the preset slip signal characteristics, it is determined whether the jump variable is negative and its absolute value exceeds the preset intensity change threshold, and simultaneously whether the deflection exceeds the preset angle change threshold. When both conditions are met, the shear force change characteristics at the abrupt change point are determined to match the slip signal characteristics.

[0067] In one embodiment, a slip signal confirmation conclusion is output based on the matching comparison result. If the shear force change characteristics match the slip signal characteristics, the slip signal confirmation conclusion is positive, indicating that an actual slip event has occurred on the workpiece corresponding to the real abrupt change position; if the shear force change characteristics do not match the slip signal characteristics, the slip signal confirmation conclusion is negative, indicating that although the abrupt change point has exceeded the slip recognition threshold, it is not a real slip signal, but an abnormal fluctuation caused by other interference factors.

[0068] S107. After the evaluation of the slip signal confirms a positive result, immediately send a trigger command to the humanoid robot control system to activate the emergency sub-task to perform the stop and then re-clamping operation.

[0069] If the slip signal confirmation is positive, a trigger command is immediately sent to the humanoid robot control system. This trigger command carries the actual abrupt change position of the current clamping condition and a slip time marker, activating a pre-configured emergency subtask. According to the execution flow of the emergency subtask, the humanoid robot stops its current handling action and maintains the workpiece position. Then, the clamping force is increased to re-clamp the workpiece, completing the stop and re-clamping operation.

[0070] When the slip signal confirmation is positive, it indicates that the humanoid robot has detected an actual slippage event in the workpiece during loading and unloading. At this point, a trigger command is immediately sent to the humanoid robot control system. This trigger command carries the actual abrupt change location and the time marker of the slippage occurrence, used to record the specific occurrence point of the slippage event. In one embodiment, the emergency subtask is an emergency response program pre-configured in the humanoid robot control system, which is activated and executed upon receiving the trigger command. The execution flow of the emergency subtask includes two consecutive actions: a stop action and a re-clamping action. The stop action refers to controlling the humanoid robot to immediately stop its current handling movement, maintaining the relative position of the end effector and the workpiece unchanged, preventing the workpiece from continuing to move in the slipped state and falling, thereby achieving automatic subdivision and scheduling of the humanoid robot's loading and unloading tasks.

[0071] It should be noted that after the stopping action is completed, a re-clamping action is immediately performed. The re-clamping action increases the clamping force applied to the workpiece by the clamping surface, re-establishing a stable clamping state, so that the friction between the workpiece and the clamping surface is restored to a level that can resist various external forces during subsequent handling, thereby ensuring the safe continuation of loading and unloading operations.

[0072] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for automatic subdivision and scheduling of loading and unloading tasks for humanoid robots, characterized in that, include: Shear force data of the contact area between the end-effector gripping surface of the humanoid robot and the workpiece is acquired and fused to obtain a shear force time-series record reflecting the force evolution during the gripping process; Based on the shear force time-series record, Kalman filtering is used to suppress noise and spurious adhesion components, separate potential adhesion disturbances, and determine the interference level. Based on the interference level, the strength of the adhesion effect is analyzed, and a classification conclusion on the adhesion effect is obtained. Based on the classification conclusions of adhesion, assess the degree to which false adhesion components obscure real slip features and update the slip recognition threshold; Based on the updated slip recognition threshold, identify the abrupt change points in the shear force time series record and determine the actual abrupt change location; Based on the actual abrupt change location, analyze the shear force change characteristics to confirm whether it conforms to the characteristics of a slip signal, and obtain a slip signal confirmation conclusion; If the sliding signal confirmation is positive, a trigger command is sent to the humanoid robot control system to activate the emergency subtask to perform the stop and re-clamping operation.

2. The automatic subdivision and scheduling method for loading and unloading tasks of humanoid robots as described in claim 1, characterized in that, The process involves acquiring shear force data in the contact area between the end effector of the humanoid robot and the workpiece, and fusing this data to obtain a shear force time-series record reflecting the force evolution during the clamping process, including: The shear force intensity value and vector direction of each measuring point in the contact area between the clamping surface and the workpiece are collected, and the measurement data are obtained according to a fixed sampling period to obtain shear force distribution data covering the entire contact area. Based on the shear force distribution data, the adhesion force reading between the clamping surface and the workpiece surface is read in the initial stage of gripping, the thickness and viscosity of the residual liquid film at the contact interface before the workpiece is unloaded are identified, and the liquid film viscosity level is divided according to the preset range. The shear force distribution data, the adhesion force readings, and the liquid film viscosity level are fused using a unified time reference to generate a shear force time series record that includes changes in shear force intensity, direction deflection, and adhesion characteristics.

3. The automatic subdivision and scheduling method for loading and unloading tasks of humanoid robots as described in claim 1, characterized in that, The process of using Kalman filtering to suppress noise and spurious adhesion components, separating potential adhesion disturbances, and determining the interference level based on the shear force time-series record includes: Based on the shear force time series record, the shear force intensity value and direction deflection data at each sampling time are extracted, a state space representation is established, and the measured values ​​are used as observation inputs to obtain the state observation sequence; Kalman filtering is used to recursively process the state observation sequence. The shear force state at the current moment is predicted based on the state estimate of the previous moment. The predicted value is compared with the actual observed value to obtain the observation residual. The predicted value is corrected and updated based on the observation residual. After iteration, a smooth shear force sequence is obtained. The difference between the smooth shear force sequence and the original record is calculated to obtain the filtered residual. The measurement noise component and the adhesion disturbance component are identified based on the changing frequency, and the potential adhesion disturbance sequence is separated. For the adhesion perturbation sequence, the average perturbation amplitude and the number of continuous sampling times are statistically analyzed. If the preset conditions are met, the interference level of the false adhesion component is determined to be high; otherwise, it is determined to be low.

4. The method for automatic subdivision and scheduling of loading and unloading tasks for humanoid robots as described in claim 1, characterized in that, The process of analyzing the strength of adhesion based on the interference level to obtain a classification conclusion on adhesion includes: The interference level is compared with a preset threshold. If the interference level exceeds the preset threshold, the current clamping condition is marked as a high viscosity liquid film state. The adhesion effect is divided into levels according to the magnitude of the exceedance. If the exceedance is greater than the preset threshold, it is classified as a strong adhesion level; otherwise, it is classified as a weak adhesion level, thus obtaining the adhesion effect classification conclusion. If the interference level does not exceed the preset threshold, the cutting fluid dragging effect is determined to be weak, and the conventional slip recognition threshold is used to obtain the adhesion classification conclusion. Based on the aforementioned classification conclusions of adhesion, the strength of adhesion under high-viscosity liquid films is distinguished, thus forming a classification result of adhesion.

5. The method for automatic subdivision and scheduling of loading and unloading tasks for humanoid robots as described in claim 1, characterized in that, The process of analyzing the strength of adhesion based on the interference level to obtain a classification conclusion on adhesion includes: When the interference level exceeds a preset threshold, the thickness and viscosity of the liquid film at the contact interface are obtained; Based on the thickness and viscosity of the liquid film, the spread range of the viscous liquid on the clamping surface is evaluated. The spread range value is calculated by weighted summation of the number of measuring points and the distribution area. At the same time, the duration of workpiece pulling and stopping is recorded. Based on the spreading range and the duration of the pulling pause, the motion characteristics of the workpiece when it leaves the clamping surface are identified. If the spreading range and the duration of the pulling pause both meet the preset range threshold and duration threshold respectively, it is determined to be a delayed detachment and classified into the strong adhesion category; otherwise, it is determined to be an instantaneous detachment and classified into the weak adhesion category, thus obtaining the adhesion effect classification conclusion.

6. The method for automatic subdivision and scheduling of loading and unloading tasks for humanoid robots as described in claim 1, characterized in that, The step of evaluating the degree to which false adhesion components obscure real slip features based on the adhesion classification conclusion, and updating the slip recognition threshold, includes: Based on the classification conclusion of adhesion, the judgment content of strong adhesion and weak adhesion is retrieved. For the performance of delayed detachment and instantaneous detachment, the degree of occlusion of false adhesion components on real slip characteristics is evaluated. The occlusion degree value is calculated based on the adhesion disturbance amplitude and the normal fluctuation amplitude. Based on the stated degree of obscuration, a preset adjustment coefficient is matched, and the increase in the original slip recognition threshold by the high viscosity liquid film is calculated, wherein the adjustment coefficient corresponding to the strong adhesion level is greater than the adjustment coefficient corresponding to the weak adhesion level. By superimposing the aforementioned lifting amplitude on the original sliding recognition threshold, an updated sliding recognition threshold is obtained.

7. The method for automatic subdivision and scheduling of loading and unloading tasks for humanoid robots as described in claim 1, characterized in that, The step of evaluating the degree to which false adhesion components obscure real slip features based on the adhesion classification conclusion, and updating the slip recognition threshold, includes: Retrieve the strong adhesion and weak adhesion determination content from the adhesion classification conclusion, and obtain the spreading range value and pulling pause time corresponding to the two determination contents of strong adhesion and weak adhesion. Based on the spread range value and the duration of slack, the degree of occlusion of the false adhesion component is evaluated to cover the real slip change point. The occlusion value is determined by comparing the adhesion disturbance amplitude with the normal fluctuation amplitude. The occlusion degree of the strong adhesion is higher than that of the weak adhesion. Based on the stated degree of obscuration, a preset lifting coefficient is matched to calculate the additional lifting amplitude of the high-viscosity liquid film on the original slip recognition threshold. The additional lifting amplitude is superimposed on the original slip recognition threshold to obtain the updated slip recognition threshold.

8. The method for automatic subdivision and scheduling of loading and unloading tasks for humanoid robots as described in claim 1, characterized in that, The step of identifying abrupt change points in the shear force time series record and determining the actual abrupt change location based on the updated slip recognition threshold includes: Based on the shear force time sequence record, the shear force intensity value is compared with the updated slip recognition threshold at each time step, and the sampling time that exceeds the threshold is marked to obtain a set of candidate mutation times; For each sampling time in the candidate mutation time set, the amplitude of the shear force intensity jump and the deflection angle are identified. If both the jump amplitude and the deflection angle exceed a preset threshold, the sampling time is determined to be the actual mutation location.

9. The method for automatic subdivision and scheduling of loading and unloading tasks for humanoid robots as described in claim 1, characterized in that, The step of analyzing the shear force change characteristics based on the actual abrupt change location, confirming whether it conforms to the characteristics of a slip signal, and obtaining a slip signal confirmation conclusion includes: Extract the shear force intensity values ​​and direction angles at the actual mutation location and the sampling times before and after it, obtain the shear force intensity jump and direction deflection at the mutation point, and obtain the shear force change characteristics; The shear force variation characteristics are matched and compared with the preset slip signal characteristics, wherein the slip signal characteristics include a combination of a decreasing trend in shear force intensity and a significant deflection of the direction angle. If the shear force change characteristics match the slip signal characteristics, it is determined to be a real workpiece slip event and a slip signal confirmation conclusion is output as positive; otherwise, it is determined to be an interference signal and a confirmation conclusion is output as negative, thus obtaining a slip signal confirmation conclusion.

10. The method for automatic subdivision and scheduling of loading and unloading tasks for humanoid robots as described in claim 1, characterized in that, If the sliding signal confirmation is positive, a trigger command is sent to the humanoid robot control system to activate the emergency subtask to perform a stop and re-clamping operation, including: If the confirmation conclusion of the slip signal is positive, a trigger command is sent to the humanoid robot control system. The trigger command carries the actual change position of the current clamping condition and the slip time mark. According to the trigger command, the pre-configured emergency sub-task is activated, controlling the humanoid robot to stop the current handling action and maintain the workpiece position; According to the emergency sub-task execution process, the clamping force is increased to re-clamp the workpiece, completing the stop and re-clamping operation, and obtaining the emergency adjustment result of the clamping process.

Images