Industrial scene humanoid robot multi-modal collaborative decision method and device

Abstract

The application is suitable for the technical field of industrial robot control, and provides an industrial scene humanoid robot multi-modal collaborative decision method and device, wherein the industrial scene humanoid robot multi-modal collaborative decision method comprises the following steps: according to the workpiece weight, the estimated friction coefficient, the current actual clamping pressure, the comprehensive damage risk index and the real-time deformation rate of the workpiece obtained in real time, the acceleration limiting value of the manipulator is calculated through a preset acceleration limiting model. The industrial scene humanoid robot multi-modal collaborative decision method provided by the application realizes the collaborative calculation of the manipulator pressure optimization value and the acceleration limiting value in the fourth step and the fifth step. The pressure optimization model considers the rigidity of the workpiece and the comprehensive damage risk while increasing the clamping pressure to cope with the slip risk, so that the pressure is prevented from being too large.

Description

Technical Field

[0001] This invention belongs to the field of industrial robot control technology, and in particular relates to a multimodal collaborative decision-making method and device for humanoid robots in industrial scenarios. Background Technology

[0002] Humanoid robots, with their human-like form and movement capabilities, have shown broad application prospects in industrial settings, and are particularly suitable for complex tasks such as precision assembly, material handling, and equipment operation.

[0003] Traditional robot control methods typically rely on single-modal feedback or simple threshold judgment mechanisms, making them ill-suited to dynamically changing industrial environments. For example, in workpiece grasping scenarios, operating solely based on preset fixed pressure values ​​may lead to insufficient friction coefficients and slippage risks when the workpiece surface hardness is high or roughness is low; conversely, excessive clamping pressure may cause overload damage when the workpiece material is soft or heavy. Existing multimodal fusion decision-making methods, while attempting to integrate multiple information sources, suffer from significant drawbacks: parameter processing is often oversimplified, employing linear weighting or single exponential compression, failing to preserve the nonlinear relationships of key physical characteristics; the lack of independent assessment mechanisms for slippage and overload risks prevents differentiated control strategies for different risk types; the pressure control module and acceleration control module are designed independently, neglecting their close physical coupling, potentially resulting in control commands exceeding the robot's actual execution capabilities; furthermore, these methods fail to construct a complete decision chain from workpiece attribute perception to risk index calculation to actuator parameter optimization, leading to poor system adaptability to sudden changes in workpiece attributes or environmental disturbances. These problems severely restrict the reliability and operational efficiency of humanoid robots in industrial settings.

[0004] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0005] The purpose of this invention is to provide a multimodal collaborative decision-making method and apparatus for humanoid robots in industrial scenarios, aiming to solve the above-mentioned problems.

[0006] This invention is implemented as follows: a multimodal collaborative decision-making method for humanoid robots in industrial scenarios, comprising the following steps: S1: Based on the obtained workpiece surface hardness and surface roughness, output the estimated friction coefficient through the preset first mapping model; based on the obtained workpiece surface hardness and workpiece weight, output the contact stiffness coefficient through the preset second mapping model. S2: Based on the obtained gripping surface area, grasping stability, and pressure uniformity of the robotic arm, calculate the robotic arm working state index using a pre-set robotic arm working state evaluation model. S3: Based on the obtained workpiece surface hardness, the estimated friction coefficient, the preset movement path curvature, and the equipment foundation vibration intensity, determine the slippage risk factor; based on the obtained workpiece surface hardness, the contact stiffness coefficient, and the robot arm working state index, determine the overload risk factor; based on the slippage risk factor and the overload risk factor, output the comprehensive damage risk index through the preset comprehensive damage risk assessment model. S4: Based on the estimated friction coefficient, the contact stiffness coefficient, the workpiece weight, the comprehensive damage risk index, and the real-time acceleration of the robot arm, the optimized pressure value of the robot arm is calculated through a preset pressure optimization model. S5: Based on the workpiece weight, the estimated friction coefficient, the current actual clamping pressure, the comprehensive damage risk index, and the real-time deformation rate of the workpiece, calculate the acceleration limit value of the robot arm through a preset acceleration limit model.

[0007] In a further technical solution, in step S1, the estimated friction coefficient is obtained in the following way: Based on the reference friction coefficient, multiply by a first attenuation term and a second attenuation term in sequence; The first attenuation term is (1 minus the product of the hardness influence coefficient and the normalized value of the workpiece surface hardness). The second attenuation term is a negative exponential function with the natural constant e as the base and (the product of the negative roughness attenuation coefficient and the normalized value of the workpiece surface roughness) as the exponent, which yields the estimated friction coefficient. The contact stiffness coefficient is obtained in the following way: Based on the reference contact stiffness, multiply by the γ power of the normalized value of the workpiece surface hardness, and then multiply by an amplification term determined by the normalized value of the workpiece weight. This amplification term is (1 plus the product of the weight influence coefficient and the normalized value of the workpiece weight) to obtain the contact stiffness coefficient. Among them, the normalized value of workpiece surface hardness is obtained by dividing the measured hardness by the upper limit of the hardness range, the normalized value of workpiece surface roughness is obtained by dividing the measured roughness by the upper limit of the roughness range, and the normalized value of workpiece weight is obtained by dividing the measured weight by the upper limit of the weight range.

[0008] A further technical solution is to obtain the robotic arm's working status index through the following methods: The basic state value is obtained by multiplying the normalized value of the clamping surface area, the normalized value of the gripping stability, and the normalized value of the pressure uniformity, and then taking the cube root. The attenuation term is obtained by multiplying (1 minus the product of the normalized value of gripping stability and the normalized value of pressure uniformity) with the preset state attenuation coefficient. Multiply the base state value by (1 minus the decay term) to obtain the robot's working state index; Among them, the normalized value of the clamping surface area is obtained by dividing the measured clamping surface area by the maximum clamping surface area, the normalized value of the gripping stability is obtained by dividing the measured stability value by the preset maximum stability value, and the normalized value of the pressure uniformity is obtained by dividing the measured pressure uniformity by the preset maximum uniformity value.

[0009] In a further technical solution, in step S3, the slip risk factor is obtained by multiplying the normalized value of the movement path curvature and the normalized value of the equipment foundation vibration intensity, and then dividing by [the product of the estimated friction coefficient and (1 plus the normalized value of the workpiece surface hardness)] to obtain the slip risk factor; The overload risk factor is obtained by calculating the reciprocal of the product of the contact stiffness coefficient, the robot working state index, and (1 plus the normalized value of the workpiece surface hardness). The normalized value of the movement path curvature is obtained by dividing the measured movement path curvature by the set maximum value of the movement path curvature, and the normalized value of the equipment foundation vibration intensity is obtained by dividing the measured equipment foundation vibration intensity by the set maximum value of the foundation vibration intensity.

[0010] In a further technical solution, step S3, the comprehensive damage risk index is obtained through the following method: The slip risk factor and the overload risk factor are summed, and then the negative exponent of the sum is calculated. Finally, the negative exponent is subtracted from 1 to obtain the comprehensive damage risk index.

[0011] In a further technical solution, in step S4, the optimized value of the robotic arm pressure is obtained through the following method: Based on the minimum anti-slip pressure, the difference between the maximum anti-damage pressure and the minimum anti-slip pressure is multiplied by (1 minus the comprehensive damage risk index) to obtain the optimized pressure value for the robotic arm; The minimum anti-slip pressure is calculated as follows: the sum of gravitational acceleration and (dynamic load coefficient multiplied by the real-time acceleration of the robot arm), multiplied by the actual weight of the workpiece, and then divided by (the sum of the estimated friction coefficient and a very small positive number to prevent division by zero). The maximum damage prevention pressure is the product of the contact stiffness coefficient and the maximum allowable deformation rate of the workpiece.

[0012] In a further technical solution, in step S5, the acceleration limit value of the robotic arm is obtained in the following way: First, calculate the theoretical maximum acceleration that can be achieved under the current clamping force. This theoretical maximum acceleration is the product of the estimated friction coefficient and the current actual clamping pressure, and then divided by (the sum of the actual weight of the workpiece and a very small positive number to prevent division by zero). Then, the theoretical maximum acceleration is multiplied by (1 minus the comprehensive damage risk index) and (1 minus the normalized value of the real-time deformation rate of the workpiece) in sequence to obtain the acceleration limit value of the robot arm; The normalized value of the real-time deformation rate of the workpiece is obtained by dividing the real-time deformation rate of the workpiece by the set maximum deformation rate limit.

[0013] A multimodal collaborative decision-making device for humanoid robots in industrial scenarios, applied to the aforementioned multimodal collaborative decision-making method for humanoid robots in industrial scenarios, includes: The data acquisition unit is used to acquire workpiece surface hardness, workpiece surface roughness, workpiece weight, robot gripping surface area, gripping stability, pressure uniformity, preset movement path curvature, equipment foundation vibration intensity, robot real-time acceleration, current actual gripping pressure, and workpiece real-time deformation rate. The parameter calculation unit, connected to the data acquisition unit, is used for: Based on the workpiece surface hardness and surface roughness, the estimated coefficient of friction is calculated. The contact stiffness coefficient is calculated based on the workpiece surface hardness and workpiece weight. Based on the gripping surface area, grasping stability, and pressure uniformity of the robotic arm, the working state index of the robotic arm is calculated. The risk assessment unit, connected to both the data acquisition unit and the parameter calculation unit, is used for: Based on the workpiece surface hardness, the estimated friction coefficient, the preset movement path curvature, and the vibration intensity of the equipment foundation, the slip risk factor is calculated. The overload risk factor is calculated based on the workpiece surface hardness, the contact stiffness coefficient, and the robot arm working state index. Based on the slip risk factor and the overload risk factor, the comprehensive damage risk index is calculated; The decision optimization unit, connected to the data acquisition unit, the parameter calculation unit, and the risk assessment unit respectively, is used for: Based on the estimated friction coefficient, the contact stiffness coefficient, the workpiece weight, the comprehensive damage risk index, and the real-time acceleration of the robot arm, the optimized value of the robot arm pressure is calculated. Based on the workpiece weight, the estimated friction coefficient, the current actual clamping pressure, the comprehensive damage risk index, and the real-time deformation rate of the workpiece, the acceleration limit value of the robot arm is calculated. The execution control unit, connected to the decision optimization unit, is used to generate control commands based on the optimized pressure value of the manipulator and the limited acceleration value of the manipulator, and send them to the manipulator actuator.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention provides a multimodal collaborative decision-making method for humanoid robots in industrial scenarios. Through first and second steps, it achieves a refined and multi-dimensional assessment of the physical properties of the workpiece and the working state of the robot arm. For example, by estimating the friction coefficient, contact stiffness coefficient, and robot arm working state index, it provides a more accurate physical basis for subsequent decisions. This effectively avoids the information loss caused by "overly simplistic parameter processing" in traditional methods.

[0015] This invention provides a multimodal collaborative decision-making method for humanoid robots in industrial scenarios. In the third step, this application clearly distinguishes and quantifies slip risk factors and overload risk factors, and outputs a comprehensive damage risk index based on these. In the example of the ceramic workpiece mentioned above, due to its smooth surface, the slip risk factor is assessed as moderately high, while the overload risk factor is low. This mechanism of separate assessment contrasts sharply with the background art's finding that "the lack of separate assessment of slip risk and overload risk makes it difficult to adopt targeted control strategies for different risk types." This application can specifically identify the main risks, providing a clear direction for subsequent pressure and acceleration adjustments.

[0016] This invention provides a multimodal collaborative decision-making method for humanoid robots in industrial scenarios. In steps four and five, this application achieves the collaborative calculation of optimized pressure values ​​and acceleration limits for the robotic arm. The pressure optimization model increases clamping pressure to address slippage risks while considering workpiece stiffness and overall damage risks to avoid excessive pressure. Simultaneously, the acceleration limit model dynamically adjusts the upper limit of acceleration based on current pressure, friction coefficient, and real-time deformation rate. This mechanism of coupling and collaboratively optimizing pressure and acceleration effectively solves the problem in the prior art where "pressure control and acceleration control are independent, failing to consider the physical coupling between them, leading to physically unfeasible control commands." In this way, the robot can generate physically feasible, safe, and efficient control commands. Attached Figure Description

[0017] Figure 1 A schematic diagram illustrating the steps of a multimodal collaborative decision-making method for humanoid robots in industrial scenarios; Figure 2 This is a schematic diagram of a multimodal collaborative decision-making device for humanoid robots in industrial settings. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0019] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0020] like Figure 1 As shown, an embodiment of the present invention provides a multimodal collaborative decision-making method for humanoid robots in industrial scenarios, comprising the following steps: S1: Based on the obtained workpiece surface hardness and surface roughness, output the estimated friction coefficient through the preset first mapping model; based on the obtained workpiece surface hardness and workpiece weight, output the contact stiffness coefficient through the preset second mapping model. S2: Based on the obtained gripping surface area, grasping stability, and pressure uniformity of the robotic arm, calculate the robotic arm working state index using a pre-set robotic arm working state evaluation model. S3: Based on the obtained workpiece surface hardness, the estimated friction coefficient, the preset movement path curvature, and the equipment foundation vibration intensity, determine the slippage risk factor; based on the obtained workpiece surface hardness, the contact stiffness coefficient, and the robot arm working state index, determine the overload risk factor; based on the slippage risk factor and the overload risk factor, output the comprehensive damage risk index through the preset comprehensive damage risk assessment model. S4: Based on the estimated friction coefficient, the contact stiffness coefficient, the workpiece weight, the comprehensive damage risk index, and the real-time acceleration of the robot arm, the optimized pressure value of the robot arm is calculated through a preset pressure optimization model. S5: Based on the workpiece weight, the estimated friction coefficient, the current actual clamping pressure, the comprehensive damage risk index, and the real-time deformation rate of the workpiece, calculate the acceleration limit value of the robot arm through a preset acceleration limit model.

[0021] In this embodiment, the multimodal collaborative decision-making method for humanoid robots in industrial scenarios refers to a decision-making process applied to humanoid robots in industrial production environments. This process integrates and analyzes data from different sensors (i.e., multimodal) to achieve intelligent and coordinated optimization of robot operation behavior. The method aims to improve the robot's operational accuracy, safety, and adaptability in complex tasks.

[0022] The first and second mapping models refer to pre-established mathematical or statistical models used to convert one or more input parameters (such as workpiece surface hardness, roughness, and weight) into specific output parameters (such as estimated friction coefficient and contact stiffness coefficient). These models are trained and calibrated using experimental data or physical principles.

[0023] A robotic arm working status evaluation model is a mathematical model used to comprehensively analyze parameters such as the gripping surface area, grasping stability, and pressure uniformity of a robotic arm to quantify its current grasping state. A robotic arm working status index is a quantitative indicator calculated using the robotic arm working status evaluation model, reflecting the overall performance and reliability of the workpiece being grasped by the robotic arm.

[0024] A comprehensive damage risk assessment model is a mathematical model that integrates slip risk factors and overload risk factors to output a comprehensive reflection of the total risk of damage to a workpiece during operation. The comprehensive damage risk index is an indicator calculated using the comprehensive damage risk assessment model, quantifying the overall damage risk faced by the workpiece during operation.

[0025] A pressure optimization model is a mathematical model used to calculate the optimal clamping pressure based on various input parameters (such as friction coefficient, stiffness coefficient, risk index, and real-time acceleration). The optimal pressure value for a robotic arm refers to the ideal clamping pressure that the robotic arm should apply, calculated using the pressure optimization model, while ensuring workpiece safety and operational efficiency.

[0026] An acceleration limit model is a mathematical model used to calculate the maximum permissible acceleration of a robot arm based on various input parameters (such as workpiece weight, friction coefficient, current pressure, risk index, and real-time deformation rate). The robot arm acceleration limit value refers to the maximum allowable motion acceleration of the robot arm under current operating conditions, calculated by the acceleration limit model, to avoid workpiece damage or slippage.

[0027] This application proposes a multimodal collaborative decision-making method for humanoid robots in industrial scenarios. Its main feature is that it achieves precise and safe operation of workpieces through a series of collaborative steps.

[0028] In a preferred embodiment of the present invention, the calculation method for the estimated friction coefficient in step S1 is as follows: ; The contact stiffness coefficient is calculated as follows: ; in This is the hardness influence coefficient. This is the roughness attenuation coefficient. , , As the reference friction coefficient, This is the normalized value of the workpiece surface hardness, which is obtained by dividing the measured hardness by the upper limit of the hardness range. This is the normalized value of the workpiece surface roughness, which is obtained by dividing the measured roughness by the upper limit of the roughness measurement range. To estimate the coefficient of friction; in This is the hardness influence coefficient. The weight influence coefficient. , , As the reference contact stiffness, This is the normalized value of the workpiece weight, which is obtained by dividing the measured weight by the upper limit of the weight range. The contact stiffness coefficient has the same properties as... The same actual physical unit.

[0029] In this embodiment, the method for calculating the estimated friction coefficient aims to quantify the influence of workpiece surface hardness and surface roughness on the friction coefficient through a mathematical model. This calculation method can be constructed based on empirical formulas, physical models, or machine learning models, and its purpose is to provide a relatively accurate estimate of the friction coefficient to support subsequent slip risk assessment and pressure optimization.

[0030] The calculation method for the contact stiffness coefficient aims to quantify the influence of workpiece surface hardness and workpiece weight on contact stiffness through a mathematical model. This calculation method can be based on the principles of materials mechanics, finite element analysis, or fitting of experimental data. Its purpose is to provide a relatively accurate estimate of contact stiffness to support subsequent overload risk assessment and pressure optimization. In addition to the above formula, the contact stiffness coefficient can also be obtained through indentation testing, ultrasonic testing, or theoretical calculations based on workpiece geometry and material properties.

[0031] Hardness Influence Coefficient and roughness attenuation coefficient It is used to adjust the normalized value of the surface hardness of a workpiece. and the normalized value of workpiece surface roughness For the estimated coefficient of friction Parameters that influence the degree of influence. These coefficients are usually determined through experimental calibration, data fitting, or expert experience to ensure that the model accurately reflects the frictional characteristics of different materials and surface conditions.

[0032] Reference friction coefficient This refers to the coefficient of friction between the workpiece and the gripping surface of the robot arm under standard or ideal conditions. This value can be preset based on factors such as material type and surface treatment process, serving as the starting point for calculating the coefficient of friction.

[0033] Normalized value of workpiece surface hardness This method converts the measured surface hardness of the workpiece into a dimensionless numerical value for standardized processing within the model. It is obtained by dividing the measured hardness by the upper limit of the hardness measurement range, ensuring compatibility with different hardness measurement units and ranges.

[0034] Normalized surface roughness of workpiece It converts the measured surface roughness of a workpiece into a dimensionless numerical value. It is obtained by dividing the measured roughness by the upper limit of the roughness measurement range, so that data from different roughness measurement methods and standards can be used uniformly.

[0035] Hardness Influence Coefficient and weight influence coefficient It is used to adjust the normalized value of the surface hardness of a workpiece. and normalized value of workpiece weight For contact stiffness coefficient Parameters that influence the degree of influence. These coefficients are usually determined through experimental calibration, data fitting, or expert experience to ensure that the model accurately reflects the stiffness characteristics under different material and weight conditions.

[0036] Reference contact stiffness This refers to the contact stiffness between the workpiece and the gripping surface of the robot arm under standard or ideal conditions. This value can be preset according to factors such as material type and contact geometry, serving as the starting point for contact stiffness calculation.

[0037] Normalized value of workpiece weight It converts the measured workpiece weight into a dimensionless value. This value is obtained by dividing the measured weight by the upper limit of the weight range, ensuring compatibility across different weight ranges.

[0038] This application provides specific mathematical formulas to calculate the estimated friction coefficient and contact stiffness coefficient, solving the problem of undefined pre-defined mapping models. This ensures that the calculation of the friction coefficient and contact stiffness coefficient has a clear physical basis and mathematical operability, thereby improving the accuracy of subsequent risk assessment and the reliability of decision-making. In the calculation of the estimated friction coefficient, friction changes are modeled using a multiplicative exponential form based on the baseline friction coefficient combined with the normalized values ​​of hardness and roughness. The hardness influence coefficient and roughness attenuation coefficient control the nonlinear effects of hardness and roughness on friction, enabling the model to more accurately reflect the physical characteristics of friction reduction due to increased hardness and friction attenuation due to increased roughness, avoiding the bias that may be caused by simple linear models. At the same time, the normalized values ​​are standardized by dividing the measured values ​​by the upper limit of the range, ensuring that the input data is within a reasonable range and facilitating stable model operation. In the calculation of the contact stiffness coefficient, stiffness variation is modeled using an exponential function and linear terms, based on the baseline contact stiffness combined with normalized hardness and weight values. The hardness influence coefficient controls the exponential enhancement of stiffness by hardness, while the weight influence coefficient controls the linear contribution of weight to stiffness. This allows the model to more accurately capture the physical laws governing the increase in stiffness due to increased hardness and weight. The normalized values ​​also ensure data consistency, and the constraint of coefficients greater than zero allows the model to adapt to different industrial scenarios through parameter adjustments, enhancing overall adaptability. This modeling approach based on physical laws makes the calculation results of the estimated friction coefficient and contact stiffness coefficient more reliable and accurate, laying the foundation for the accurate assessment of subsequent slip risk factors, overload risk factors, and comprehensive damage risk indices. This, in turn, improves the decision-making effectiveness of the robot's pressure optimization values ​​and acceleration limits.

[0039] In a preferred embodiment of the present invention, the working state index of the robotic arm is calculated as follows: ; in The state decay coefficient, , This is a normalized value for the clamping surface area, which is obtained by dividing the measured clamping surface area by the maximum clamping surface area. To capture the stability normalization value, it is obtained by dividing the measured stability value by the preset maximum stability value; This is the normalized value for pressure uniformity, which is obtained by dividing the measured pressure uniformity by the preset maximum uniformity value. This is the working status index of the robotic arm.

[0040] In this embodiment, the solution effectively solves the problem of inaccurate work status assessment by introducing a comprehensive method for calculating the robot's working status index, ensuring the accuracy of subsequent risk decisions. Specifically, this calculation method first performs geometric averaging based on the normalized values ​​of the gripping surface area, gripping stability, and pressure uniformity. This balances the contributions of each factor and avoids evaluation bias caused by a single parameter dominating the assessment. At the same time, by combining the state decay coefficient and the decay adjustment of the product term, the nonlinear effects in the actual working state are simulated. When stability and uniformity are insufficient, the index value is automatically reduced, thus more realistically reflecting the dynamic performance of the robot. Each technical feature specifically enhances the assessment accuracy: the normalized value of the clamping surface area is standardized by the ratio of the measured value to the maximum value, ensuring comparability under different size scenarios; the normalized value of gripping stability is based on the preset maximum value, quantifying the impact of stability on the working state; the normalized value of pressure uniformity is also processed proportionally to capture key details of pressure distribution; the state decay coefficient, as a positive coefficient, dynamically adjusts the degree of exponential decay to adapt to fluctuations in the industrial environment; the final output working state index integrates all factors, providing reliable input for overload risk factors and strengthening the robustness of the entire decision chain.

[0041] Among them, the state decay coefficient It is a positive parameter used to adjust the working status index of the robot arm. For crawling stability and pressure uniformity The sensitivity of the product term reflects the degree of attenuation of the overall working condition index when gripping stability and pressure uniformity deviate from the ideal state. This coefficient can be obtained through experimental calibration, for example, by conducting numerous gripping tests under different workpieces and gripping conditions, recording the actual working performance of the robot, and combining this with expert experience, using optimization algorithms (such as least squares method, genetic algorithm, etc.) to fit the coefficient that best reflects the actual situation. The value. Furthermore, this coefficient can be preset according to specific industrial application scenarios and risk tolerance. For example, for precision assembly tasks requiring high precision and high safety, a larger value can be set. The value is set to more strictly penalize deficiencies in stability and uniformity; while for more fault-tolerant, rough handling tasks, a smaller value can be set. value.

[0042] Normalized value of clamping surface area This refers to the ratio of the actual contact area of ​​the workpiece held by the robotic arm to the maximum gripping area of ​​the robotic arm, quantifying the degree of contact between the robotic arm and the workpiece. This value can be obtained by measuring the actual gripping area in real time using a contact sensor array integrated inside the robotic arm, and then dividing it by the maximum gripping surface area determined during the robotic arm's design. Alternatively, the theoretical gripping area can be calculated and normalized by recognizing the workpiece's geometry and the robotic arm's grasping posture using a vision system, combined with a pre-set robotic arm model.

[0043] Capture stability normalization value This refers to the ratio of the stability of the gripping state of a robotic arm when grasping a workpiece to a preset maximum stability level. It reflects the robotic arm's ability to resist external disturbances and maintain its gripping posture. This value can be calculated and normalized by real-time monitoring of gripping torque fluctuations using force / torque sensors on the robotic arm's end effector, combined with the geometric center and mass distribution of the workpiece.

[0044] Normalized value of pressure uniformity This refers to the ratio of the uniformity of pressure distribution on the contact surface when a robotic arm grips a workpiece to a preset maximum uniformity. A uniform pressure distribution helps prevent damage to the workpiece due to excessive localized stress. This value can be obtained by deploying multiple pressure sensors on the gripping surface of the robotic arm to collect pressure data at each point in real time, then calculating the standard deviation or variance of these pressures, and comparing it with the preset maximum uniformity to obtain a normalized value. Alternatively, a high-resolution pressure distribution map can be obtained through a tactile sensor array, and then image processing techniques can be used to analyze the concentration or dispersion of the pressure distribution and convert it into a normalized value.

[0045] Robotic Arm Working Status Index This is a comprehensive indicator used to quantify the overall performance of a robotic arm when performing grasping tasks. It combines the aforementioned parameters to provide a unified evaluation standard. This index can be directly calculated using the formula described above and used as input for subsequent risk assessment models. It can also be visualized, for example, displayed on the robot control interface as a dashboard or progress bar, providing operators with intuitive feedback on the robotic arm's working status.

[0046] Traditional methods for assessing the working condition of robotic arms often employ simple linear weighting or single-parameter judgments, failing to fully capture the nonlinear coupling relationship between the three key factors: gripping surface area, grasping stability, and pressure uniformity. This leads to discrepancies between the assessment results and the actual working condition, consequently affecting subsequent risk assessment and decision-making. The solution proposed in this application uses a nonlinear mathematical model to normalize the gripping surface area. Capture stability normalized values and pressure uniformity normalized value This approach organically integrates various methods to provide a more accurate and robust assessment of the robot's operational status. Specifically, this calculation method first involves... Cube root calculations were performed to achieve a geometric mean for the three key parameters. This approach effectively balances the contributions of each parameter, preventing extreme values ​​of a single parameter from having an excessive impact on the overall evaluation result, thus ensuring the comprehensiveness and objectivity of the evaluation. Based on this, a decay term was introduced into the calculation method. ,in This is the state decay coefficient. The ingenious design of this decay term lies in its ability to simulate the stability issues encountered during grasping in real-world industrial scenarios. and pressure uniformity When insufficient, the robot's working status index This phenomenon will be subject to non-linear penalty. When When the value is close to 1, the decay term is close to 1 and has little impact on the exponent; while when... When the value is reduced, the decay term decreases significantly, making Rapid decline. This nonlinear decay mechanism allows the evaluation results to more accurately reflect the true performance decline of the robot under complex or suboptimal grasping conditions, thus providing a more reliable input for subsequent overload risk factor calculations. Through the above calculation method, the scheme of this application can overcome the limitations of traditional evaluation methods in handling multi-factor coupling, providing a more refined and accurate robot working state index. This index, as a key input for overload risk factor calculations in subsequent steps, can significantly improve the accuracy and reliability of the multimodal collaborative decision-making method for humanoid robots in the entire industrial scenario. For example, in step S3, the overload risk factor... The calculation directly depends on A more accurate Enabling The assessment is more realistic, thus avoiding workpiece damage or robot malfunction due to inaccurate assessments. This precise assessment mechanism enables the robot to better adapt to the challenges posed by workpiece diversity and environmental complexity, ensuring efficient and stable operation while guaranteeing workpiece safety.

[0047] In a preferred embodiment of the present invention, the slip risk factor and overload risk factor are calculated in step S3 as follows: , ; in This is the normalized value of the movement path curvature, which is obtained by dividing the measured movement path curvature by the set maximum value of the movement path curvature. This is the normalized value of the equipment foundation vibration intensity, which is obtained by dividing the measured equipment foundation vibration intensity by the set maximum value of the foundation vibration intensity. To estimate the coefficient of friction, This represents the normalized value of the workpiece surface hardness. As a slip risk factor, For contact stiffness coefficient, This is an index representing the working condition of the robotic arm. This is an overload risk factor.

[0048] In this embodiment, the slip risk factor This is an indicator used to quantify the likelihood and severity of relative slippage of a workpiece during gripping and movement by a robotic arm. Its core function is to assess the balance between external disturbances (such as the curvature of the movement path and equipment vibration) and the workpiece's own frictional characteristics. This factor can serve as an important basis for the decision-making system to adjust its gripping strategy; for example, when the slippage risk factor is high, the system may need to increase gripping pressure or reduce movement speed. It can be implemented by acquiring relevant parameters in real time through sensors and inputting them into a pre-set physical model for calculation, or by training a machine learning model on historical data to predict the slippage risk under current conditions.

[0049] Overload risk factor This is an indicator used to assess the likelihood and severity of workpiece deformation or damage due to excessive force during robotic gripping. It primarily focuses on the stiffness characteristics of the interface between the robotic arm and the workpiece, as well as the robotic arm's gripping posture. A high overload risk factor indicates a potential risk of workpiece damage, and the decision-making system should consider reducing gripping pressure or adjusting the gripping attitude. This factor can be calculated by real-time monitoring of the workpiece-robotic arm contact state, combined with principles of materials mechanics, or obtained through pre-calibration and modeling.

[0050] Normalized value of movement path curvature This reflects the curvature of the preset path when the robot moves the workpiece. A greater path curvature means the robot needs to overcome greater inertial or centrifugal forces, increasing the risk of workpiece slippage. Normalizing this value eliminates the influence of different path lengths and dimensions, making it comparable across different tasks. This value can be calculated from the path data output by the robot's motion planning module, for example, by calculating the second derivative of the path function and dividing by the set maximum curvature value. Alternatively, it can be calculated by fitting the curve to discrete points on the path and then calculating the curvature of the fitted curve.

[0051] Normalized value of equipment foundation vibration intensity This indicates the vibration level of the workbench or environment where the robot is located in an industrial setting. External vibrations directly affect the stable gripping of workpieces by the robotic arm, increasing the possibility of workpiece slippage. Normalizing this vibration helps to conduct a consistent risk assessment under different vibration environments. This value can be obtained by collecting data in real time from accelerometers or vibration sensors mounted on the equipment base, and dividing the peak or root mean square value by the set maximum vibration intensity limit. Alternatively, it can be obtained by performing spectral analysis on the vibration signal, extracting the energy within a specific frequency range, and then normalizing it.

[0052] Estimated coefficient of friction This characterizes the magnitude of the frictional force between the gripping surface of the robotic arm and the workpiece surface. The coefficient of friction is a key factor in resisting slippage; the higher the value, the less likely the workpiece is to slip. In the calculation of the slippage risk factor, the estimated coefficient of friction is used as a denominator, reflecting its role in suppressing slippage risk. This coefficient can be obtained based on the workpiece's surface hardness, roughness, and other physical properties, and can be predicted using a pre-established mapping model.

[0053] Normalized value of workpiece surface hardness Hardness reflects the workpiece material's ability to resist localized plastic deformation. It not only affects the workpiece's wear resistance but also indirectly influences its frictional characteristics and compressive strength when in contact with a robotic arm. In the calculation of the slip risk factor, hardness, by influencing the coefficient of friction and as part of the denominator, contributes to slip risk. In the calculation of the overload risk factor, hardness, as part of the denominator, directly reflects the workpiece's ability to resist overload damage. This value can be obtained by measuring the workpiece's hardness with a hardness tester and then normalizing it.

[0054] Contact stiffness coefficient This describes the elastic deformation characteristics of the interface between the robot arm and the workpiece, that is, the ease with which the interface deforms under a given pressure. Greater stiffness results in less deformation of the workpiece under stress and stronger resistance to overload damage. In the calculation of the overload risk factor, the contact stiffness coefficient, as a denominator, reflects its role in suppressing overload risk. This coefficient can be obtained based on physical properties such as the workpiece's surface hardness and weight, and can be predicted using a pre-established mapping model.

[0055] Robotic Arm Working Status Index The performance of the robotic arm in grasping tasks was comprehensively evaluated, including factors such as gripping surface area, grasping stability, and pressure uniformity. This index reflects the reliability and effectiveness of the robotic arm in gripping the workpiece. In the calculation of the overload risk factor, the robotic arm's working condition index, as a denominator, reflects its role in suppressing overload risk; that is, the better the robotic arm's working condition, the lower the overload risk. This index can be obtained based on real-time sensor data of the robotic arm and calculated using a pre-established evaluation model.

[0056] The solution in this application introduces a slippage risk factor. and overload risk factors This precise calculation method solves the problem of inaccurate risk assessment in traditional methods. The scheme first uses the normalized value of the moving path curvature... Normalized value of vibration intensity of equipment foundation To quantify the impact of the external environment on workpiece stability, these two factors directly reflect the intensity of external disturbances that may cause workpiece slippage. Simultaneously, they are combined with the estimated friction coefficient. and the normalized value of workpiece surface hardness The physical properties of these workpieces themselves are taken into consideration, among which This represents the workpiece's inherent ability to resist sliding, while The slip risk is then adjusted by influencing frictional characteristics and overall stability. An external disturbance factor is constructed by placing it in the numerator and the workpiece's intrinsic resistance to slip in the denominator. The calculation model allows this factor to physically reflect the probability of a workpiece slipping due to insufficient friction under given external conditions. On the other hand, the overload risk factor... The calculations focus on assessing the potential damage to the workpiece during clamping. This model utilizes the contact stiffness coefficient. To characterize the deformation resistance of the interface between the workpiece and the robot arm, and the robot arm's working condition index. This reflects the stability and uniformity of the robotic gripper's hold. These two factors together determine the workpiece's sensitivity to deformation or damage under gripping force. Similarly, the normalized value of the workpiece surface hardness... This was also introduced, further strengthening the modulating effect of workpiece material properties on overload risk. By placing these inherent damage resistance capabilities and the reliability of the robotic gripper in the denominator, an overload risk factor was constructed. The computational model physically quantifies the likelihood of workpiece damage due to excessive force during clamping. Through these two independent but interconnected risk factor calculation models, this application enables a refined and physical assessment of slippage and overload risks. This differentiated assessment avoids the risk confusion caused by simple weighting in traditional methods, allowing subsequent decision optimizations (such as pressure optimization and acceleration limiting) to be targeted. For example, when the slippage risk is high, the decision system can prioritize increasing friction or reducing external disturbances; when the overload risk is high, it can prioritize reducing clamping force or optimizing clamping posture. This physical model-based risk quantification method enables humanoid robots to more accurately understand the current operating state in complex industrial scenarios, providing a solid foundation for achieving precise and compliant workpiece operation, thereby significantly improving the reliability and safety of decision-making.

[0057] In a preferred embodiment of the present invention, the comprehensive damage risk index is calculated in step S3 as follows: ; As a slip risk factor, As an overload risk factor, This is a comprehensive damage risk index.

[0058] In this embodiment, the calculation method aims to integrate multiple independent risk factors into a unified, quantifiable index to comprehensively reflect the overall potential damage risk faced by humanoid robots operating workpieces in industrial scenarios. Its purpose is to provide a standardized, easily understood, and processable risk metric, offering a reliable basis for subsequent decision optimization (such as pressure and acceleration adjustments). One possible implementation is to use a nonlinear function, such as an exponential function, to handle the cumulative effect of risk factors, avoiding distortions that may result from linear superposition. Another possible implementation is to use a method based on fuzzy logic or neural networks, training a model to learn the complex relationships between different risk factors, thereby outputting a comprehensive risk index.

[0059] This formula is the specific mathematical expression used in this application to calculate the comprehensive damage risk index. Its core lies in utilizing the nonlinear characteristics of the exponential function to incorporate the slip risk factor... and overload risk factors To achieve effective integration. When and When it increases, its sum Increase, exponential term It will approach 0, thus making it The exponential term approaches 1. Conversely, when the risk factor is small, the exponential term approaches 1, making... Approaching 0. This design ensures an overall damage risk index. The value always ranges between 0 and 1, achieving risk normalization and facilitating risk comparison and management across different scenarios and workpieces. This formula can be implemented directly in the robot controller's calculation module based on standard mathematical library functions. Alternatively, it can be implemented using a pre-established lookup table to represent different ranges of... Values ​​mapped to corresponding Values ​​are adjusted to improve real-time computing efficiency.

[0060] Slippage risk factor It is an indicator that measures the potential risk of relative slippage of a workpiece during the gripping process of a robotic arm. Its function is to quantify the possibility of workpiece detachment or positional displacement due to factors such as insufficient friction, external disturbance, or improper operation. It can be obtained in several ways. For example, it can be calculated using a preset physical model or empirical formula based on physical parameters such as workpiece surface characteristics (e.g., coefficient of friction), robot movement speed, path curvature, and external vibration intensity. Another approach is to use a machine learning model, trained on historical operating data, to predict the slippage risk under the current operating conditions.

[0061] Overload risk factor It is an indicator that measures the potential risk of workpiece damage or deformation due to excessive force during the gripping process of a robotic arm. Its function is to quantify the possibility of workpiece damage caused by factors such as excessive gripping force, workpiece fragility, or insufficient rigidity of the robotic arm. This information can be obtained in several ways. For example, it can be calculated using mechanical analysis models or finite element simulation results based on parameters such as the mechanical properties of the workpiece material (e.g., contact stiffness coefficient), the gripping state of the robot (e.g., working state index), and the geometry of the workpiece. Another approach is to dynamically assess overload risk by combining real-time force sensor data and the workpiece's damage threshold.

[0062] The solution proposed in this application introduces a comprehensive damage risk assessment model based on an exponential function, which organically integrates slip risk factors and overload risk factors to generate a unified comprehensive damage risk index. Specifically, in step S3, the slip risk factor is first determined based on the workpiece surface hardness, estimated friction coefficient, preset movement path curvature, and equipment foundation vibration intensity. Simultaneously, based on the workpiece surface hardness, contact stiffness coefficient, and robot arm working state index, overload risk factors are determined. These two risk factors quantify the potential hazards that may be encountered during operation from the two dimensions of workpiece stability and integrity, respectively. Subsequently, this application will... and As input, these values ​​are substituted into a pre-set comprehensive damage risk assessment model, which uses the formula... To calculate the comprehensive damage risk index The ingenious design of this exponential function lies in its ability to non-linearly reflect the cumulative effect of risk. When any risk factor or the sum of both is small, the comprehensive risk index rises gradually; while as the risk factors increase, the comprehensive risk index accelerates and gradually approaches 1, but never exceeds 1. This characteristic allows the comprehensive damage risk index to more realistically simulate the accumulation and saturation of risk in the actual physical world, avoiding the overestimation or underestimation of risk that may result from simple linear superposition. Simultaneously, the formula naturally normalizes the comprehensive risk index to the range of 0 to 1, making the risk assessment results highly comparable and operable, facilitating direct use by subsequent decision-making modules. In this way, the proposed solution integrates the originally dispersed slippage risk and overload risk into a comprehensive, physically meaningful index, solving the problems of incomplete and inaccurate risk assessment in traditional methods. This integration not only considers the stability of the workpiece during grasping and movement but also takes into account the workpiece's inherent fragility, providing a more reliable risk basis for humanoid robots to perform delicate operations in complex industrial scenarios.

[0063] In a preferred embodiment of the present invention, the calculation method for the optimized value of the robotic arm pressure in step S4 is as follows: , ; ; in This is the dynamic load factor. , It is the acceleration due to gravity. It is a very small positive number that is not zero. To provide real-time acceleration for the robotic arm, To estimate the coefficient of friction, This is the actual weight of the workpiece. Minimum anti-slip pressure; The maximum allowable deformation rate of the workpiece. For contact stiffness coefficient, To maximize damage prevention pressure; To form a comprehensive damage risk index, Optimized pressure values ​​for the robotic arm.

[0064] In this embodiment, the actual weight of the workpiece It refers to the product of the actual mass of the workpiece and the gravitational acceleration, obtained through force sensors or preset parameters when the robot grasps the workpiece. It is the basic physical quantity for calculating the required clamping force.

[0065] gravitational acceleration It is a constant used to convert mass into gravity.

[0066] Dynamic load factor It is a dimensionless parameter used to quantify the effect of the additional inertial force generated by the robot during acceleration or deceleration on the clamping pressure. Its value is usually determined experimentally or preset according to the robot's motion characteristics.

[0067] Real-time acceleration of robotic arm It refers to the instantaneous acceleration of the robot arm during the process of grasping and moving the workpiece. It is usually obtained in real time through the inertial measurement unit installed on the robot arm or the encoder data of the robot body.

[0068] Estimated coefficient of friction It is output based on information such as workpiece surface hardness and surface roughness through a preset first mapping model. It reflects the frictional characteristics between the workpiece surface and the gripper surface of the robot, and is a key parameter to prevent workpiece slippage. (A very small positive number that is not zero) It is introduced into the denominator to avoid the estimation of the friction coefficient. This ensures the numerical stability of the calculation process by preventing division by zero errors that may occur when the result is close to or equal to zero.

[0069] Maximum allowable deformation rate of workpiece It is a preset threshold value that represents the maximum degree of deformation a workpiece can withstand under pressure without permanent damage. Its value depends on the material properties, structural strength, and process requirements of the workpiece.

[0070] Contact stiffness coefficient It is output based on information such as workpiece surface hardness and workpiece weight, through a preset second mapping model. Comprehensive Damage Risk Index It is output by a pre-set comprehensive damage risk assessment model based on slip risk factors and overload risk factors.

[0071] The solution in this application precisely calculates the minimum anti-slip pressure. and maximum damage prevention pressure In conjunction with the comprehensive damage risk index Dynamic adjustments are made to optimize the pressure of the robotic arm. The determination of minimum anti-slip pressure. The calculation takes into account the actual weight of the workpiece. Gravitational acceleration and the real-time acceleration of the robotic arm The resulting dynamic inertial force, and through the dynamic load coefficient Make adjustments to ensure that the clamping force is sufficient to overcome gravity and inertia during the robot's movement, preventing workpiece slippage. Estimate the coefficient of friction. As a denominator term, it directly affects the required clamping force; the greater the friction, the smaller the required clamping force. Secondly, the maximum damage prevention pressure... The calculation is based on the contact stiffness coefficient of the workpiece. and the maximum allowable deformation rate of the workpiece This is designed to ensure that the clamping pressure does not exceed the workpiece's tolerance limit, avoiding deformation or damage to the workpiece. Contact stiffness coefficient. It reflects the inherent properties of the workpiece material and allows for the maximum deformation rate. This sets an upper limit for safe operation. Finally, the optimized pressure value for the robotic arm. The calculations cleverly combine and It also introduced a comprehensive damage risk index. As a weighting factor. When the comprehensive damage risk index A higher level indicates a greater operational risk. Will tend to Prioritize slip resistance and safety; when the comprehensive damage risk index A lower level indicates lower operational risk. It can be appropriately increased to get closer. This dynamic adjustment mechanism allows the robotic arm to adaptively adjust its gripping pressure based on real-time operating conditions and risk assessments, achieving an optimal balance between anti-slip and damage prevention. The aforementioned estimated coefficient of friction... Contact stiffness coefficient and comprehensive injury risk index All of these are based on a comprehensive analysis and evaluation of multimodal information such as workpiece surface hardness, workpiece surface roughness, workpiece weight, robotic gripping surface area, gripping stability, pressure uniformity, preset movement path curvature, and equipment foundation vibration intensity. This allows pressure optimization decision-making to make full use of multi-source information, forming a complete closed loop from perception to decision-making, thereby significantly improving the accuracy and robustness of decision-making.

[0072] In a preferred embodiment of the present invention, in step S5, the theoretical maximum acceleration based on the current clamping force is first calculated, and the specific calculation method is as follows: ; The method for calculating the acceleration limit of the robotic arm by substituting the theoretical maximum acceleration into the pressure optimization model is as follows: ; in To estimate the coefficient of friction, The current actual clamping pressure is expressed in actual pressure units. This is the actual weight of the workpiece. It is a very small positive number that is not zero. This is the theoretical maximum acceleration; To form a comprehensive damage risk index, The normalized value of the real-time deformation rate of the workpiece is obtained by dividing the real-time deformation rate of the workpiece by the set maximum deformation rate limit. This is a limit value for the acceleration of the robotic arm.

[0073] In this embodiment, the solution addresses the problem of inaccurate acceleration control by specifying the calculation method for acceleration limits, ensuring that acceleration optimization is based on physical constraints and real-time feedback, and avoiding the risk of slippage or damage. First, calculating the theoretical maximum acceleration based on the current clamping force aims to determine, under current physical conditions, the maximum acceleration that the robot arm can apply without slippage of the workpiece. Its purpose is to provide a physical upper limit for subsequent acceleration constraints, ensuring the physical feasibility of robot movement. This can be achieved through real-time calculation by a computing module built into the robot controller, or by an external high-performance computing unit (such as an edge computing device or cloud server) receiving real-time data, performing calculations, and transmitting the results back.

[0074] Calculation method of theoretical maximum acceleration Used for precise quantification of theoretical maximum acceleration Among them, the estimated coefficient of friction This reflects the frictional characteristics between the workpiece surface and the robotic arm, and the current actual clamping pressure. It is the real-time force applied to the workpiece by the robot arm, and the actual weight of the workpiece. It is an inherent property of the workpiece, and a very small positive number that is not zero. This method avoids cases where the denominator is zero, ensuring computational stability. It guarantees that the theoretical maximum acceleration is determined based on actual physical parameters, thus providing a reliable physical constraint for subsequent acceleration limits. Substituting the theoretical maximum acceleration into the pressure optimization model to calculate the robot's acceleration limit describes how, after determining the physical upper limit, risk assessment and real-time deformation information are further combined to calculate the practically usable robot acceleration limit. Its function is to further consider operational safety and workpiece protection based on physical feasibility. This can be achieved through a decision-making module within the robot's control system, which determines the theoretical maximum acceleration. As input, it is combined with other real-time data for calculation; or through a pre-trained machine learning model, various parameters are used as feature inputs to output acceleration limit values.

[0075] How to calculate the acceleration limit of a robotic arm Used to determine the final acceleration limit value for the robotic arm. It will achieve the theoretical maximum acceleration. With comprehensive injury risk index and the normalized value of the real-time deformation rate of the workpiece Combined. Among them, the comprehensive damage risk index This reflects the combined risks of workpiece slippage and overload; the normalized value of the workpiece's real-time deformation rate. This quantifies the degree of deformation of the workpiece under the current operation. Through this multiplicative relationship, when the risk or deformation increases, the acceleration limit will decrease accordingly, thereby achieving flexible protection of the workpiece.

[0076] This application's solution further improves the multimodal collaborative decision-making method for humanoid robots in industrial scenarios by introducing a refined constraint on the robot's acceleration. Throughout the decision-making process, the system calculates the estimated friction coefficient based on acquired information such as workpiece surface hardness and roughness, using a pre-set first and second mapping model. The contact stiffness coefficient is also considered. Simultaneously, the robot's working condition index is evaluated by combining the gripping surface area, grasping stability, and pressure uniformity. Based on this, the workpiece surface hardness and estimated friction coefficient are comprehensively considered. The system presets the curvature of the movement path, the vibration intensity of the equipment foundation, the contact stiffness coefficient, and the working state index of the robot arm. Using a pre-set comprehensive damage risk assessment model, it outputs a comprehensive damage risk index. These preliminary parameter calculations and risk assessments provide comprehensive foundational data for subsequent pressure and acceleration optimization. Based on this, to ensure that the robot arm avoids both slippage and damage when moving the workpiece, this scheme, in step S5, first calculates the current actual clamping pressure... Actual weight of the workpiece and the estimated coefficient of friction Calculate the theoretical maximum acceleration that the workpiece can withstand under the current clamping force. This calculation step ensures that the acceleration setting does not exceed the limit of physical friction, thus effectively preventing slippage of the workpiece due to excessive acceleration during movement. Subsequently, the theoretical maximum acceleration under this physical constraint is... With the real-time acquired comprehensive damage risk index and the normalized value of the real-time deformation rate of the workpiece By combining these methods and employing specific calculation techniques, the final acceleration limit value for the robotic arm is obtained. Comprehensive Damage Risk Index The introduction of this feature ensures that the acceleration limit considers not only the risk of slippage but also the risk of overload damage. Meanwhile, the normalized value of the workpiece's real-time deformation rate... This consideration further ensures that when the workpiece has already undergone a certain degree of deformation, the robot can reduce its acceleration in time to prevent further deformation and damage to the workpiece. Through this layered, multi-factor coupled calculation method, this solution achieves dynamic and precise limitation of the robot's acceleration, enabling the robot to better balance efficiency and safety when performing grasping and moving tasks, especially when dealing with fragile or high-value workpieces, providing more compliant and reliable operation.

[0077] like Figure 2 As shown, this application proposes a multimodal collaborative decision-making device for humanoid robots in industrial scenarios. The device includes a data acquisition unit, a parameter calculation unit, a risk assessment unit, a decision optimization unit, and an execution control unit.

[0078] The data acquisition unit is used to acquire data on workpiece surface hardness, surface roughness, weight, gripper area, grasping stability, pressure uniformity, preset movement path curvature, equipment foundation vibration intensity, real-time robot arm acceleration, current actual gripping pressure, and real-time workpiece deformation rate. This data acquisition unit is the perception layer of the entire decision-making system, responsible for acquiring raw data from the industrial environment and the robot's own state. It can be composed of various sensors, such as: a Leeb hardness tester sensor for measuring workpiece surface hardness, a roughness meter sensor for measuring workpiece surface roughness, a six-dimensional force sensor integrated into the robot arm for measuring workpiece weight, a high-resolution vision camera system for measuring the gripper area and grasping stability, a flexible thin-film pressure sensor array for measuring pressure uniformity, a robot path planning module for acquiring the preset movement path curvature, a three-axis accelerometer for measuring equipment foundation vibration intensity, an encoder and inertial measurement unit for real-time monitoring of robot arm movement, and a laser displacement sensor for measuring the real-time workpiece deformation rate. These sensors are connected to a central data processing unit via an industrial Ethernet network.

[0079] The parameter calculation unit is connected to the data acquisition unit and is used to calculate the estimated friction coefficient based on the workpiece surface hardness and roughness; the contact stiffness coefficient based on the workpiece surface hardness and weight; and the robot arm's working state index based on the gripping surface area, grasping stability, and pressure uniformity. This unit is the core of data preprocessing and feature extraction, transforming raw sensor data into physically meaningful parameters. For example, the parameter calculation unit can be a high-performance industrial PC running a real-time operating system and deploying software modules for performing the calculations of the estimated friction coefficient, contact stiffness coefficient, and robot arm working state index in the above methods.

[0080] The risk assessment unit is connected to both the data acquisition unit and the parameter calculation unit. It calculates the slip risk factor based on the workpiece surface hardness, the estimated friction coefficient, the preset movement path curvature, and the vibration intensity of the equipment foundation; it calculates the overload risk factor based on the workpiece surface hardness, the contact stiffness coefficient, and the robot's working state index; and it calculates the comprehensive damage risk index based on the slip risk factor and the overload risk factor. This unit is responsible for quantitatively assessing the risks that may occur during robot operation. It can be a standalone calculation module or integrated into the parameter calculation unit. By executing the risk assessment model defined in the above method, it comprehensively considers workpiece characteristics, environmental factors, and the robot's state to calculate the slip risk factor and the overload risk factor, and further synthesizes the comprehensive damage risk index, providing crucial information for subsequent decision optimization.

[0081] The decision optimization unit is connected to the data acquisition unit, parameter calculation unit, and risk assessment unit, respectively. It calculates the optimal pressure value for the robotic arm based on the estimated friction coefficient, contact stiffness coefficient, workpiece weight, comprehensive damage risk index, and real-time acceleration of the robotic arm; and calculates the robotic arm acceleration limit value based on the workpiece weight, estimated friction coefficient, current actual clamping pressure, comprehensive damage risk index, and real-time deformation rate of the workpiece. This unit is the core of the entire decision system, responsible for generating optimal control commands based on the risk assessment results and real-time status. It can be a high-performance industrial controller or a dedicated computing platform, running the pressure optimization model and acceleration limit model defined in the above method to calculate in real-time the optimal clamping pressure value and motion acceleration limit value of the robotic arm that prevents slippage and avoids overload.

[0082] The execution control unit is connected to the decision optimization unit and is used to generate control commands and send them to the robot's actuators based on the optimized pressure value and the limited acceleration value of the robot. This unit is the executor of the decision results, transforming the abstract optimization values ​​into physical actions that the robot can understand and execute. It can be the robot's motion controller, receiving commands from the decision optimization unit and converting them into servo motor drive signals, thereby precisely controlling the gripping force and movement trajectory of the robot.

[0083] This application's solution modularizes and tightly connects functions such as data acquisition, parameter calculation, risk assessment, and decision optimization, forming a closed-loop system from perception to decision-making to execution. The data acquisition unit acquires multimodal information in real-time and comprehensively, providing rich and accurate input for subsequent decision-making. The parameter calculation unit transforms raw data into physically meaningful parameters, laying the foundation for risk assessment. The risk assessment unit comprehensively and accurately quantifies slip and overload risks, enabling the decision-making process to fully consider potential hazards. Based on the comprehensive risk assessment, the decision optimization unit calculates physically feasible and optimal pressure and acceleration control parameters, effectively solving the problems of independent pressure and acceleration control and insufficient handling of physical coupling relationships in traditional methods. Finally, the execution control unit accurately transmits the optimized instructions to the robotic arm, ensuring the accuracy and safety of robot operations. This integrated device design enables the efficient, real-time, and accurate implementation of the aforementioned multimodal collaborative decision-making method for humanoid robots in industrial scenarios, overcoming potential processing delays and inaccurate control issues during implementation.

[0084] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multimodal collaborative decision-making method for humanoid robots in industrial scenarios, characterized in that, Includes the following steps: S1: Based on the obtained workpiece surface hardness and surface roughness, output the estimated friction coefficient through the preset first mapping model; based on the obtained workpiece surface hardness and workpiece weight, output the contact stiffness coefficient through the preset second mapping model. S2: Based on the obtained gripping surface area, grasping stability, and pressure uniformity of the robotic arm, calculate the robotic arm working state index using a pre-set robotic arm working state evaluation model. S3: Determine the slip risk factor based on the obtained workpiece surface hardness, the estimated friction coefficient, the preset movement path curvature, and the vibration intensity of the equipment foundation; Based on the obtained workpiece surface hardness, the contact stiffness coefficient, and the robot arm working state index, the overload risk factor is determined. Based on the slip risk factor and the overload risk factor, a comprehensive damage risk index is output through a pre-set comprehensive damage risk assessment model; S4: Based on the estimated friction coefficient, the contact stiffness coefficient, the workpiece weight, the comprehensive damage risk index, and the real-time acceleration of the robot arm, the optimized pressure value of the robot arm is calculated through a preset pressure optimization model. S5: Based on the workpiece weight, the estimated friction coefficient, the current actual clamping pressure, the comprehensive damage risk index, and the real-time deformation rate of the workpiece, calculate the acceleration limit value of the robot arm through a preset acceleration limit model.

2. The multimodal collaborative decision-making method for humanoid robots in industrial scenarios according to claim 1, characterized in that, In step S1, the estimated coefficient of friction is obtained in the following way: Based on the reference friction coefficient, multiply by a first attenuation term and a second attenuation term in sequence; The first attenuation term is (1 minus the product of the hardness influence coefficient and the normalized value of the workpiece surface hardness). The second attenuation term is a negative exponential function with the natural constant e as the base and (the product of the negative roughness attenuation coefficient and the normalized value of the workpiece surface roughness) as the exponent, which yields the estimated friction coefficient. The contact stiffness coefficient is obtained in the following way: Based on the reference contact stiffness, multiply by the γ power of the normalized value of the workpiece surface hardness, and then multiply by an amplification term determined by the normalized value of the workpiece weight. This amplification term is (1 plus the product of the weight influence coefficient and the normalized value of the workpiece weight) to obtain the contact stiffness coefficient. Among them, the normalized value of workpiece surface hardness is obtained by dividing the measured hardness by the upper limit of the hardness range, the normalized value of workpiece surface roughness is obtained by dividing the measured roughness by the upper limit of the roughness range, and the normalized value of workpiece weight is obtained by dividing the measured weight by the upper limit of the weight range.

3. The multimodal collaborative decision-making method for humanoid robots in industrial scenarios according to claim 1, characterized in that, The robotic arm's working status index is obtained through the following methods: The basic state value is obtained by multiplying the normalized value of the clamping surface area, the normalized value of the gripping stability, and the normalized value of the pressure uniformity, and then taking the cube root. The attenuation term is obtained by multiplying (1 minus the product of the normalized value of gripping stability and the normalized value of pressure uniformity) with the preset state attenuation coefficient. Multiply the base state value by (1 minus the decay term) to obtain the robot's working state index; Among them, the normalized value of the clamping surface area is obtained by dividing the measured clamping surface area by the maximum clamping surface area, the normalized value of the gripping stability is obtained by dividing the measured stability value by the preset maximum stability value, and the normalized value of the pressure uniformity is obtained by dividing the measured pressure uniformity by the preset maximum uniformity value.

4. The multimodal collaborative decision-making method for humanoid robots in industrial scenarios according to claim 1, characterized in that, In step S3, the slip risk factor is obtained by multiplying the normalized value of the movement path curvature and the normalized value of the equipment foundation vibration intensity, and then dividing by [the product of the estimated friction coefficient and (1 plus the normalized value of the workpiece surface hardness)] to obtain the slip risk factor. The overload risk factor is obtained by calculating the reciprocal of the product of the contact stiffness coefficient, the robot working state index, and (1 plus the normalized value of the workpiece surface hardness). The normalized value of the movement path curvature is obtained by dividing the measured movement path curvature by the set maximum value of the movement path curvature, and the normalized value of the equipment foundation vibration intensity is obtained by dividing the measured equipment foundation vibration intensity by the set maximum value of the foundation vibration intensity.

5. The multimodal collaborative decision-making method for humanoid robots in industrial scenarios according to claim 4, characterized in that, In step S3, the comprehensive damage risk index is obtained in the following way: The slip risk factor and the overload risk factor are summed, and then the negative exponent of the sum is calculated. Finally, the negative exponent is subtracted from 1 to obtain the comprehensive damage risk index.

6. The multimodal collaborative decision-making method for humanoid robots in industrial scenarios according to claim 1, characterized in that, In step S4, the optimized value of the robotic arm pressure is obtained in the following way: Based on the minimum anti-slip pressure, the difference between the maximum anti-damage pressure and the minimum anti-slip pressure is multiplied by (1 minus the comprehensive damage risk index) to obtain the optimized pressure value for the robotic arm; The minimum anti-slip pressure is calculated as follows: the sum of gravitational acceleration and (dynamic load coefficient multiplied by the real-time acceleration of the robot arm), multiplied by the actual weight of the workpiece, and then divided by (the sum of the estimated friction coefficient and a very small positive number to prevent division by zero). The maximum damage prevention pressure is the product of the contact stiffness coefficient and the maximum allowable deformation rate of the workpiece.

7. The multimodal collaborative decision-making method for humanoid robots in industrial scenarios according to claim 1, characterized in that, In step S5, the acceleration limit value of the robotic arm is obtained in the following way: First, calculate the theoretical maximum acceleration that can be achieved under the current clamping force. This theoretical maximum acceleration is the product of the estimated friction coefficient and the current actual clamping pressure, and then divided by (the sum of the actual weight of the workpiece and a very small positive number to prevent division by zero). Then, the theoretical maximum acceleration is multiplied by (1 minus the comprehensive damage risk index) and (1 minus the normalized value of the real-time deformation rate of the workpiece) in sequence to obtain the acceleration limit value of the robot arm; The normalized value of the real-time deformation rate of the workpiece is obtained by dividing the real-time deformation rate of the workpiece by the set maximum deformation rate limit.

8. A multimodal collaborative decision-making device for humanoid robots in industrial scenarios, applied to the multimodal collaborative decision-making method for humanoid robots in industrial scenarios as described in any one of claims 1-7, characterized in that, include: The data acquisition unit is used to acquire workpiece surface hardness, workpiece surface roughness, workpiece weight, robot gripping surface area, gripping stability, pressure uniformity, preset movement path curvature, equipment foundation vibration intensity, robot real-time acceleration, current actual gripping pressure, and workpiece real-time deformation rate. The parameter calculation unit, connected to the data acquisition unit, is used for: Based on the workpiece surface hardness and surface roughness, the estimated coefficient of friction is calculated. The contact stiffness coefficient is calculated based on the workpiece surface hardness and workpiece weight. Based on the gripping surface area, grasping stability, and pressure uniformity of the robotic arm, the working state index of the robotic arm is calculated. The risk assessment unit, connected to both the data acquisition unit and the parameter calculation unit, is used for: Based on the workpiece surface hardness, the estimated friction coefficient, the preset movement path curvature, and the vibration intensity of the equipment foundation, the slip risk factor is calculated. The overload risk factor is calculated based on the workpiece surface hardness, the contact stiffness coefficient, and the robot arm working state index. Based on the slip risk factor and the overload risk factor, the comprehensive damage risk index is calculated; The decision optimization unit, connected to the data acquisition unit, the parameter calculation unit, and the risk assessment unit respectively, is used for: Based on the estimated friction coefficient, the contact stiffness coefficient, the workpiece weight, the comprehensive damage risk index, and the real-time acceleration of the robot arm, the optimized value of the robot arm pressure is calculated. Based on the workpiece weight, the estimated friction coefficient, the current actual clamping pressure, the comprehensive damage risk index, and the real-time deformation rate of the workpiece, the acceleration limit value of the robot arm is calculated. The execution control unit, connected to the decision optimization unit, is used to generate control commands based on the optimized pressure value of the manipulator and the limited acceleration value of the manipulator, and send them to the manipulator actuator.

Images