Method and device for pre-judging distance and planning attitude before grabbing of humanoid robot

Abstract

The invention provides a distance pre-judgment and attitude planning method and device before grabbing of a humanoid robot, and the method comprises the steps: setting moment monitoring parameters of each joint of two arms based on an initial bearing moment, and carrying out the monitoring, the maximum deviation value between the actual monitoring torque value of the near-center-of-mass side joint and the preset initial bearing torque reference value serves as the torque peak value difference of the near-center-of-mass side joint; adjusting torque distribution according to the torque peak difference so as to reduce a near-center-of-mass side vertical supporting torque and increase a far-center-of-mass side compensation torque, determining target torque values of all joints, and generating a joint torque distribution scheme; and executing the joint torque distribution scheme, verifying the contact state of the near-center-of-mass side contact point, comparing the local deformation quantity of the near-center-of-mass side with a preset slippage instability limit value, and outputting a torque peak value control result containing the current torque balance state of the two arms.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a method and apparatus for predicting distance and planning posture before a humanoid robot grasps an object. Background Technology

[0002] Humanoid robot technology is gradually entering the handling scenarios requiring the manipulation of complex objects, with dual-arm collaboration in grasping and moving long, narrow objects becoming a key direction for its practical application. These tasks are crucial in various fields, including the handling of shaft parts on industrial assembly lines, the sorting of irregular pipes in logistics warehouses, assisting the elderly in moving walking sticks in elderly care settings, and assisting in the handling of slender steel bars or timber on construction sites. This is because it directly determines the reliability and safety of the robot's operation in unstructured environments with uneven mass distribution. Currently, most dual-arm collaboration strategies typically rely simply on the geometric center of the object or a pre-measured center of mass for static allocation when determining the grasping position, posture planning, and the load ratio borne by each arm. This allocation method assumes that the object has uniform overall rigidity and that the force on the two arms can be linearly proportional to the center of mass. However, in reality, many long and narrow objects often exhibit significant uneven mass distribution. Examples include a composite material tube with uneven internal filling, a metal rod with a heavy component welded to one end, a piece of wood with density gradually changing along its length, or certain irregularly shaped logistics packaging rods. These situations can cause the center of mass to deviate significantly from its geometric center. When the center of mass is significantly shifted to one side, the arm closer to the center of mass needs to bear a greater vertical supporting force, resulting in a significantly higher flexion moment at the elbow joint on that side. Meanwhile, the arm farther from the center of mass needs to significantly adjust its shoulder joint angle and even its waist posture to maintain overall moment balance. In actual handling, objects are rarely perfectly rigid, especially when the end closer to the center of mass is made of softer material, has thinner walls, or has a honeycomb structure or foam filling, resulting in significantly lower local compressive stiffness compared to the end farther from the center of mass. In such cases, once the torque applied by the arm closer to the center of mass exceeds the range that the local material stiffness can withstand, compressive deformation or even micro-creases will first occur near the gripping point. This localized deformation directly alters the originally planned normal and tangential relationship of the contact surface. The gripping point, designed for surface contact, may degenerate into point or line contact, causing a sharp decrease in friction and making it highly susceptible to slight slippage of the gripping point. Once slippage begins, it quickly disrupts the carefully calculated torque balance between the arms. The torque of the proximal arm, which was intentionally increased to compensate for the offset of the center of mass, instead becomes a catalyst for accelerating localized deformation. Increased deformation further weakens grip stability, forcing the distal arm to quickly compensate for the tilt or fall of the object. The entire system thus falls into a vicious cycle of violent torque oscillations and constant back-and-forth adjustments in the posture of the arms.Especially when the object is long and its surface stiffness exhibits a significant gradient along the axial direction, such as a long rod connected from a hard metal end to a soft rubber-coated end, or a roll of packaging material with a large difference in density between its inner and outer parts commonly used in logistics, the disturbance caused by such a small deformation near the center of mass can be rapidly transmitted along the length of the object, resulting in visible bending or twisting at the distal end. This completely disrupts the symmetrical and cooperative posture that the two arms were originally intended to achieve, leading to a situation where one side's joint torque continuously overloads and alarms, while the other side, unable to provide sufficient support, becomes uncontrollably loose and falls off – an extreme contradiction. Accurately identifying the deformation trend of low-stiffness areas due to increased force under conditions of significant center of mass offset and uneven distribution of surface stiffness along the length of the object, and accordingly coordinating the torque borne by each arm in real time to avoid chain instability caused by local deformation, thereby ensuring the symmetry of the arm posture and overall stability during the handling of long, narrow objects, becomes a key issue for achieving reliable collaborative handling by dual-arm humanoid robots. Summary of the Invention

[0003] This invention provides a method for predicting distance and planning posture before a humanoid robot grasps an object, mainly including: The initial grasping data of the object contact point is collected, the longitudinal offset distance of the object's center of mass and the elastic modulus of the surface near the center of mass are obtained, and the initial bearing torque of the arm near the center of mass is determined based on the longitudinal offset distance of the center of mass and the elastic modulus of the surface near the center of mass. Based on the initial bearing torque, the torque monitoring parameters of each joint of the two arms are set and monitored. The maximum deviation between the actual monitored torque value of the joint near the center of mass and the preset initial bearing torque benchmark value is taken as the torque peak difference of the joint near the center of mass. Adjust the torque distribution based on the peak torque difference to reduce the vertical support torque near the centroid and increase the compensation torque far from the centroid, determine the target torque value of each joint, and generate a joint torque distribution scheme.

[0004] Furthermore, determining the initial bearing moment of the arm near the center of mass based on the longitudinal offset distance of the center of mass and the elastic modulus of the surface near the center of mass includes: By combining the longitudinal offset distance of the centroid with the elastic modulus of the near-centroid side surface, the relationship between the normal deformation and contact stress at the contact point of the object is analyzed to obtain the stress distribution state of the contact area. Based on the stress distribution state, it is determined whether the contact stress near the centroid side exceeds the preset material elastic limit threshold. If it does not exceed the preset material elastic limit threshold, the initial bearing torque of the near-centroid side arm is determined according to the centroid offset ratio. If it exceeds the preset material elastic limit threshold, the torque distribution ratio of the near-centroid side is reduced and the compensating bearing torque of the far-centroid side arm is increased.

[0005] Furthermore, the step of setting and monitoring the torque monitoring parameters of each joint of both arms based on the initial bearing torque, and using the maximum deviation between the actual monitored torque value of the joint near the center of mass and the preset initial bearing torque benchmark value as the peak torque difference of the joint near the center of mass, includes: Based on the initial bearing torque value of the arm near the center of mass, the torque monitoring parameters of each joint of both arms are set and the current joint angle and corresponding drive current are obtained. The actual output torque value of each joint is calculated based on the linear relationship between motor torque and current. The actual output torque value is continuously monitored using a fixed time window. Within each time window, the real-time difference between the torque value of each joint near the centroid and the initial bearing torque reference value is calculated. The deformation development trend of the contact point is judged by the change in the difference between adjacent windows. The maximum value is extracted from the continuously monitored difference sequence as the peak torque difference.

[0006] Furthermore, after extracting the maximum value from the continuously monitored difference sequence as the peak torque difference, the process further includes: Record the contact pressure distribution data corresponding to the peak value. By comparing the peak torque difference with the preset deformation safety threshold, determine whether the contact point near the centroid is close to the critical deformation degree. If the peak torque difference exceeds the deformation safety threshold, identify the concentrated area where the pressure exceeds the preset pressure threshold from the contact pressure distribution data. Calculate the stress gradient by the ratio of the pressure difference between adjacent measuring points in the area to the distance. Evaluate the remaining bearing capacity of the contact point based on the ratio of the stress gradient to the initial elastic modulus of the material, and obtain the degree of deformation approach that reflects the contact stability near the centroid.

[0007] Furthermore, the step of adjusting the torque distribution based on the peak torque difference to reduce the vertical support torque near the centroid and increase the compensation torque far from the centroid, determining the target torque value for each joint, and generating a joint torque distribution scheme includes: Based on the peak torque difference, the value of the torque that needs to be reduced on the near-center of mass side is determined from the preset graded adjustment rules. The value of the vertical support torque reduced on the near-center of mass side and the value of the compensation torque increased on the far-center of mass side are calculated to keep the total support torque of the object constant. By combining the reduced vertical support torque value near the centroid side with the increased compensation torque value on the far centroid side, and the current joint angle configuration of the two arms, the required driving torque for the shoulder, elbow, and wrist joints is calculated backward from the required torque at the end effector through the torque transmission relationship in robot kinematics, thus obtaining the target torque value for each joint. The target torque value of each joint is compared with the preset maximum bearing torque limit of the joint. If the target torque of any joint near the center of mass exceeds its maximum bearing torque limit, the torque reduction on the near-center of mass side is reduced proportionally to the excess, and the torque on both sides is redistributed until the target torque of all joints is within the safe range.

[0008] Furthermore, the method also includes: executing the joint torque distribution scheme and verifying the contact state of the contact point near the center of mass, comparing the local deformation near the center of mass with the preset slip instability limit, and outputting the torque peak control result including the current torque balance state of the two arms.

[0009] Furthermore, the execution of the joint torque distribution scheme and verification of the contact state at the contact point near the centroid includes: The adjusted target torque value is applied by the joint motor driver, and real-time pressure distribution data of the contact point near the center of mass is collected from the pressure sensor array. The local deformation is calculated based on the relationship between the pressure change in the contact area and the elastic modulus of the material. Calculate the difference in local deformation between adjacent measuring points, compare it with the preset slip instability limit to determine the contact stability. If the maximum value in the local deformation exceeds the preset proportion of the preset slip instability limit, it is marked as having slip risk; otherwise, it is marked as a stable state.

[0010] Furthermore, the output includes the peak torque control result of the current torque balance state of the two arms, including: Based on the judgment results of stable state or slip risk, the root mean square value of the deviation is calculated as the torque balance index by combining the deviation between the current actual torque value and the target torque value of each joint of the two arms. The torque balance index, deformation value and slip risk judgment results are integrated to output the torque peak control result that includes the current torque balance state of the humanoid robot's two arms.

[0011] Furthermore, the joints include the shoulder joint, elbow joint, and wrist joint.

[0012] This invention provides a device for predicting distance and planning posture before a humanoid robot grasps an object, mainly comprising: The data acquisition and initial torque determination module is used to acquire initial gripping data at the object contact point, obtain the longitudinal offset distance of the object's center of mass and the elastic modulus of the surface near the center of mass, and determine the initial bearing torque of the arm near the center of mass based on the longitudinal offset distance of the center of mass and the elastic modulus of the surface near the center of mass. The torque monitoring and peak difference calculation module is used to set the torque monitoring parameters of each joint of the two arms based on the initial bearing torque and to monitor them. The maximum deviation between the actual monitored torque value of the joint near the center of mass and the preset initial bearing torque benchmark value is used as the peak torque difference of the joint near the center of mass. The torque distribution adjustment module is used to adjust the torque distribution according to the torque peak difference to reduce the vertical support torque on the near centroid side and increase the compensation torque on the far centroid side, determine the target torque value of each joint, and generate a joint torque distribution scheme. The scheme execution and state verification module is used to execute the joint torque distribution scheme and verify the contact state of the contact point near the center of mass. It compares the local deformation near the center of mass with the preset slip instability limit and outputs the torque peak control result including the current torque balance state of the two arms.

[0013] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: This invention discloses a method and device for predicting distance and planning posture before grasping objects using a humanoid robot. It proposes a complete solution to the problems of uneven torque and slippage instability caused by centroid offset when grasping objects with two arms. This invention collects initial data of the object's contact point, analyzes the centroid offset distance and surface elastic modulus, constructs torque distribution constraints, extracts the bearing torque of the arms near and far from the centroid, and sets joint torque monitoring parameters. By monitoring the difference between the joint torque near the centroid and the initial value in real time, the peak torque difference is calculated, and the torque distribution scheme is dynamically adjusted to reduce the vertical support torque near the centroid and increase the compensation torque on the far centroid side, ensuring contact point stability. This invention also verifies the anti-slip effect of the adjustment scheme by comparing local deformation with preset limits, and finally outputs the control results under torque balance. This invention significantly improves the stability and safety of dual-arm grasping, optimizes torque distribution efficiency, and provides reliable support for complex grasping tasks. Attached Figure Description

[0014] Figure 1 This is a flowchart of a method for predicting distance and planning posture before grasping by a humanoid robot according to the present invention.

[0015] Figure 2 This is a schematic diagram of the structure of a humanoid robot's distance prediction and posture planning device before grasping, according to the present invention. Detailed Implementation

[0016] 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.

[0017] like Figures 1-2 This embodiment of a method and apparatus for predicting distance and planning posture before a humanoid robot grasps an object may specifically include: S101. Collect initial grasping data of the object contact point, obtain the longitudinal offset distance of the object's center of mass and the elastic modulus of the surface near the center of mass, and determine the initial bearing torque of the arm near the center of mass based on the longitudinal offset distance of the center of mass and the elastic modulus of the surface near the center of mass.

[0018] The normal pressure distribution data at the object contact points is read from the pressure sensor array of the dual-arm gripper. The total force and torque at the wrist are obtained through a six-axis torque sensor. Combined with the pressure array data, the force and torque components at each contact point in three orthogonal directions are calculated. Based on static equilibrium conditions, the distance from the object's center of mass to the midpoint along the line connecting the two gripping points is calculated as the longitudinal offset distance of the center of mass. Simultaneously, the sound wave propagation velocity in the contact area near the center of mass is measured using the ultrasonic pulse-echo method. The surface elastic modulus of the contact area is determined based on the proportionality between the square of the sound velocity and the elastic modulus. The additional vertical support torque increment required by the arm near the center of mass is calculated based on the product of the longitudinal offset distance and the total weight of the object. Combined with the surface elastic modulus, the normal deformation at the contact points under the current gripping force is determined. The relationship between the deformation and contact stress is calculated using Hertzian contact theory to obtain the stress distribution state of the contact area. Based on the stress distribution state, it is determined whether the contact stress on the near-center of mass side exceeds the preset material elastic limit threshold. If it does not exceed the threshold, the bearing torque of the two arms is directly distributed according to the centroid offset ratio to determine the initial bearing torque of the near-center of mass side arm.

[0019] In one implementation, when a humanoid robot is handling a long, narrow object, it uses a sensor array on its dual-arm gripper to collect real-time mechanical information at the contact points. Based on the object's center of mass offset and the material properties of the contact area, the robot dynamically adjusts the torque distribution strategy of its arms. This approach is particularly suitable for handling long, narrow objects with uneven mass distribution, such as composite material tubes with uneven internal filling or metal rods with heavy components welded to one end.

[0020] Specifically, each contact surface of the dual-arm gripper is equipped with an 8×8 piezoresistive pressure sensor array, with a sensor spacing of 5 mm, enabling the reading of pressure distribution in the contact area at a sampling frequency of kilohertz. The voltage signal output from each sensor is converted from analog to digital to form a two-dimensional matrix reflecting the contact pressure distribution. Simultaneously, a six-axis torque sensor mounted on the gripper's wrist measures the force components in three orthogonal directions via a strain gauge bridge. , , and the corresponding torque components , , The six-axis torque sensor adopts a cross-beam structure, with strain gauges attached at specific locations on the beam. When an external force is applied, the beam undergoes elastic deformation, causing a change in the resistance of the strain gauges. This change in resistance is converted into a voltage signal via a Wheatstone bridge circuit, and then decoupled through a calibration matrix to obtain six independent force and torque components.

[0021] In one possible implementation, a system of static equilibrium equations is established based on the force and torque data measured at the two clamping points. Let the position of the left clamping point be the origin, and the distance from the right clamping point be... The total weight of the object is The horizontal distance from the center of mass to the left clamping point is Based on the conditions of vertical force balance and moment balance, the supporting force on the left side... Support force on the right side satisfy Meanwhile, taking a moment about the left clamping point yields... Solving the simultaneous equations yields the position of the centroid. The longitudinal offset distance of the centroid is and The difference.

[0022] It should be noted that the ultrasonic pulse-echo method for measuring elastic modulus is based on the propagation characteristics of sound waves in materials. An ultrasonic transducer emits longitudinal wave pulses at a megahertz frequency into the contact area; the sound waves propagate within the material and reflect back at the bottom surface. The time interval between the emitted and echo pulses is recorded using a high-speed oscilloscope. Combined with the known material thickness The speed of sound was calculated. For isotropic materials, the longitudinal wave velocity and elastic modulus... and density The relationship is Therefore, elastic modulus .

[0023] For example, Hertzian contact theory describes the nonlinear contact behavior between elastic bodies. When the fingers of a cylindrical gripper contact a long, narrow object, the contact area is elliptical, with the contact stress reaching its maximum at the center and decreasing towards the edges. For a radius of... The cylindrical fingers and the elastic modulus are When objects come into contact, under normal force Under action, maximum contact stress ,in The contact length. The normal deformation of the contact area. The relationship with contact force is ,in This is a constant related to Poisson's ratio. By measuring the actual deformation and comparing it with the theoretical calculation, it can be determined whether the material has entered the plastic deformation region. When the contact stress is detected to be close to a preset proportion of the material's yield strength, the system determines that the torque distribution strategy needs to be adjusted.

[0024] Preferably, the torque distribution strategy is adaptively adjusted based on the contact stress state. Initially, the bearing torque of both arms is distributed according to the centroid offset ratio, with the side closer to the centroid bearing a larger proportion. When the contact stress on the side closer to the centroid exceeds the safety threshold, the torque distribution coefficient on that side is gradually reduced through an iterative algorithm, while the compensation torque on the side farther from the centroid is increased, until the contact stress on both sides is within the safe range.

[0025] In one embodiment, the inverse kinematics of the robot is implemented using the Jacobian transpose method to map from end-effector torques to joint torques. Jacobian matrix It describes the linear relationship between joint velocity and end-effector velocity, and its transpose... This establishes a dual relationship between end-effector forces and joint moments. For a single arm with 7 degrees of freedom, the Jacobian matrix is ​​a 6×7 matrix, which is achieved through... The torques of each joint were calculated. ,in This represents the six-dimensional force vector at the end of the arm. The shoulder joint bears the main gravitational torque, the elbow joint adjusts the arm's configuration, and the wrist joint finely adjusts the end-effector's posture.

[0026] For example, when handling a 1.5-meter-long composite material pipe with a metal connector at one end, the system first detects that the center of gravity is shifted approximately 0.3 meters towards the metal connector. Once the contact stress on the side near the center of gravity reaches a preset proportion of the material's elastic limit, the control system automatically reduces the torque distribution ratio on that side from the initial 65% to 55%, while correspondingly increasing it from 35% to 45% on the far side, thus achieving stable and reliable dual-arm cooperative handling.

[0027] S102. Based on the initial bearing torque, set the torque monitoring parameters for each joint of the two arms and monitor them. The maximum deviation between the actual monitored torque value of the joint near the center of mass and the preset initial bearing torque benchmark value is taken as the torque peak difference of the joint near the center of mass.

[0028] Based on the initial load-bearing torque value of the arm near the centroid, the torque monitoring benchmark value of each joint is set by the Hall current sensor built into the joint motor driver. A real-time data acquisition channel with a sampling frequency of the preset frequency is configured. The current joint angle and corresponding drive current are obtained from the encoder and current sensor combination of the shoulder, elbow, and wrist joints. The actual output torque value of each joint is calculated based on the linear relationship between motor torque and current. The actual output torque value is continuously monitored using a fixed time window. In each time window, the real-time difference between the torque value of each joint near the centroid and the initial load-bearing torque benchmark value is calculated. The deformation development trend of the contact point is judged by the change in the difference between adjacent windows. If the difference continues to increase, it is marked as an intensified deformation state. If the difference remains stable, it is marked as a deformation equilibrium state. Based on the marked state of the deformation development trend, the maximum value is extracted from the continuously monitored difference sequence as the peak torque difference at the current moment. At the same time, the contact pressure distribution data corresponding to the peak value is recorded. By comparing the peak torque difference with the preset deformation safety threshold, it is determined whether the contact point near the centroid is close to the critical deformation degree. If the peak torque difference exceeds the deformation safety threshold, then the concentrated area where the pressure exceeds the preset pressure threshold is identified from the contact pressure distribution data. The stress gradient is obtained by calculating the ratio of the pressure difference between adjacent measuring points in the area to the distance. The remaining bearing capacity of the contact point is evaluated based on the ratio of the stress gradient to the initial elastic modulus of the material. The deformation approach value reflecting the contact stability near the centroid side and the corresponding peak torque difference are obtained.

[0029] In one implementation, the humanoid robot identifies the deformation trend of the contact point near the center of mass by real-time monitoring of torque changes at each joint of its two arms. This monitoring mechanism sets a monitoring benchmark based on an initial load-bearing torque, and accurately determines the degree of deformation at the contact point by continuously collecting joint torque data and analyzing the difference between the data and the benchmark value.

[0030] Specifically, the Hall current sensor built into the joint motor driver converts current to voltage based on the Hall effect principle. When the motor winding current passes through a conductor near the Hall element, the charge carriers are deflected by the Lorentz force under the action of a magnetic field applied perpendicular to the current direction, generating a potential difference across the Hall element. This potential difference is proportional to the current flowing through it, and after being processed by an amplification circuit, it outputs a voltage signal that is linearly related to the current. The motor's output torque has a definite proportional relationship with the winding current; this proportionality coefficient is the motor torque constant, which is typically determined by the motor's magnetic field strength, the number of winding turns, and the air gap length. By measuring the instantaneous current of the three-phase windings and combining it with the current electrical angle position, the actual output torque of the motor is calculated. Encoders for the shoulder, elbow, and wrist joints provide high-precision angle feedback, which is synchronously acquired with the current sensor data to form complete joint status information.

[0031] It should be noted that setting a fixed time window requires balancing monitoring sensitivity and data stability. A window width that is too small will increase noise interference, while a window that is too large will reduce the response speed to deformation. In practical applications, the window width is typically set to tens to hundreds of milliseconds. Within each window, torque data is continuously collected at a preset sampling frequency, and the average value of all sampling points within the window is calculated as the torque value at that moment.

[0032] In one possible implementation, the difference is calculated by subtracting the real-time torque value from the initial bearing torque benchmark value. The determination of the deformation development trend is based on the variation pattern of the difference across multiple adjacent time windows. If the difference in three or more consecutive windows shows a monotonically increasing trend, and the growth rate exceeds a preset threshold, it is determined to be a state of intensified deformation; if the difference fluctuates within a preset range and the rate of change is below the threshold, it is determined to be a state of deformation equilibrium. This trend determination method can filter out the influence of instantaneous disturbances and improve the reliability of deformation identification.

[0033] For example, the extraction process of peak torque difference involves dynamic scanning and extreme value detection of the difference sequence. The system maintains a buffer containing the difference values ​​of several recent time windows and updates the current maximum difference in real time using a sliding maximum value algorithm. When a new maximum value is detected, not only is this value recorded as the peak torque difference, but the contact pressure distribution data at the corresponding time is also saved simultaneously. The contact pressure distribution is obtained through a pressure sensor array, and the pressure values ​​at each sensor measuring point form a two-dimensional matrix, reflecting the stress distribution characteristics on the contact surface. The comparison between the peak torque difference and the preset deformation safety threshold adopts a graded judgment mechanism, setting multiple threshold levels corresponding to different deformation severity. When the peak difference exceeds a certain threshold level, the corresponding level of adjustment measures are triggered.

[0034] Preferably, the identification of pressure concentration areas is achieved through local extremum search of the pressure distribution matrix. Each measuring point and its neighborhood in the matrix are scanned to identify areas where the pressure value exceeds that of surrounding measuring points and is higher than a preset pressure threshold; these areas are the pressure concentration points. Further, the stress gradient is calculated using the finite difference method. For adjacent measuring points within the pressure concentration area, the ratio of their pressure difference to their spatial distance is calculated. The remaining load-bearing capacity of the material is assessed based on the relationship between the stress gradient and the initial elastic modulus. When the ratio of the stress gradient to the elastic modulus approaches the material's critical strain rate, it indicates that the contact point is close to failure. The deformation approach value is obtained by comprehensively considering the peak moment difference, stress gradient, and remaining load-bearing capacity.

[0035] For example, when handling a composite material tube with uneven internal filling density, the initial load-bearing torque near the centroid is set to 60% of the rated torque. When the peak torque difference is detected to reach 30% of the initial value, the system determines that the deformation at the contact point is close to the critical state and immediately starts the torque adjustment program to avoid grip instability caused by local overload.

[0036] S103. Adjust the torque distribution according to the peak torque difference to reduce the vertical support torque on the near centroid side and increase the compensation torque on the far centroid side, determine the target torque value of each joint, and generate a joint torque distribution scheme.

[0037] Based on the peak torque difference, the required reduction in torque near the center of mass is determined from a preset tiered adjustment rule. This rule sets corresponding torque transfer amounts according to different ranges of the peak difference, calculating the reduction in vertical support torque near the center of mass and the increase in compensation torque far from the center of mass, while maintaining the total support torque of the object unchanged. Using the reduced vertical support torque near the center of mass and the increased compensation torque far from the center of mass, combined with the current joint angle configuration of both arms, and through the torque transmission relationship in robot kinematics, the required output driving torque for each shoulder, elbow, and wrist joint is calculated backward from the required end torque, yielding the adjusted target torque values ​​for each joint. The target torque values ​​for each joint are compared with a preset maximum joint bearing torque limit. If the target torque of any joint near the center of mass exceeds its maximum bearing torque limit, the torque reduction near the center of mass is reduced proportionally, and the torque on both sides is redistributed until the target torque of all joints is within a safe range, generating a preliminary adjusted joint torque distribution scheme for both arms, including the target torque values ​​for each joint.

[0038] In one implementation, the dual-arm robot dynamically adjusts the torque distribution based on the detected peak torque difference, thereby reducing the load on the near-center of mass side and correspondingly increasing the compensation on the far-center of mass side to achieve stable support during object handling.

[0039] The tiered adjustment rules pre-define multiple peak torque difference ranges and corresponding torque transfer amounts. When the peak difference is between 0 and 30% of the preset threshold, the torque near the center of mass is reduced by 5% of its original value; between 30% and 60%, it is reduced by 10%; between 60% and 90%, it is reduced by 15%; and above 90%, it is reduced by 20%. The torque transfer amount for each range is determined based on the material properties of the object and historical handling data. The torque reduction near the center of mass is directly transferred to the far center of mass as a compensation increment to maintain a constant total supporting torque for the object.

[0040] Specifically, the torque transmission relationship is determined through the robot's forward kinematics model. Based on the current angle encoder readings of each joint, the transformation matrix from the base to the end effector is calculated. Combined with the required end torque, the driving torque that each joint should output is solved in reverse using the Jacobian transpose matrix.

[0041] Preferably, the upper limit constraint of joint torque is handled using an iterative correction method. When the target torque of a joint is detected to exceed its maximum load-bearing capacity, the torque distribution coefficient is adjusted in reverse proportionally. If the target torque of the shoulder joint exceeds the upper limit by 10%, the torque reduction on the near-center of mass side is reduced by 10%, and the torques of each joint are recalculated. This iterative process continues until all joint torques meet the safety constraints.

[0042] For example, when moving a steel pipe with its center of gravity offset by 0.4 meters from one end, the side closer to the center of gravity initially bears 70% of the supporting torque. When the peak torque difference is detected to reach 45% of the reference value, the system automatically adjusts the torque ratio on the side closer to the center of gravity to 60%, and correspondingly increases it to 40% on the side farther away, thus achieving a balanced distribution of torque.

[0043] S104. Execute the joint torque distribution scheme and verify the contact state of the contact point near the center of mass. Compare the local deformation near the center of mass with the preset slip instability limit and output the torque peak control result including the current torque balance state of the two arms.

[0044] A preliminary adjustment scheme for the torque distribution of the dual-arm joints is implemented. The adjusted target torque value is applied via the joint motor actuators. Simultaneously, real-time pressure distribution data at the contact points near the center of mass is collected from a pressure sensor array. Local deformation is calculated based on the relationship between pressure changes in the contact area and the material's elastic modulus, yielding the contact point deformation values ​​after the scheme is implemented. Based on these contact point deformation values, the maximum deformation value is extracted, and the deformation difference between adjacent measurement points is calculated. This difference is compared to a preset slip instability limit to determine contact stability. If the maximum deformation value exceeds a preset proportion of the limit, it is marked as having slip risk; otherwise, it is marked as a stable state. Based on the stable state or slip risk determination, the root mean square value of the deviation between the current actual torque value and the target torque value of each joint of the dual arms is calculated as a torque balance index. Integrating the torque balance index, deformation values, and slip risk determination results, a torque peak control result containing the current torque balance state of the humanoid robot's dual arms is output.

[0045] In one implementation, the humanoid robot executes an adjusted torque distribution scheme, verifies the effectiveness of the scheme by real-time monitoring of the contact point status, determines whether the anti-slip requirements are met, and outputs a comprehensive control result. The deformation calculation is based on the stress-strain relationship in materials mechanics. The contact pressure distribution data measured by the pressure sensor array reflects the stress state on the contact surface. Based on the elastic modulus of the material, Hooke's law is used to calculate the normal deformation at each measuring point. The specific calculation process is as follows: first, the pressure value at each sensor measuring point is obtained, divided by the contact area of ​​that point to obtain the stress value; then, the stress value is divided by the elastic modulus of the material to obtain the strain value; finally, the strain value is multiplied by the material thickness to obtain the actual deformation. The deformation difference between adjacent measuring points is calculated using the finite difference method, reflecting the spatial distribution characteristics of the deformation.

[0046] Preferably, the slip instability limit is preset based on the material surface properties and historical experimental data. This limit is typically set to 80% of the critical deformation at which the material begins to produce microslip under the current clamping force, leaving a safety margin.

[0047] Specifically, the torque balance index is obtained by calculating the root mean square (RMS) value of the deviation between the actual torque and the target torque at each joint. The smaller the RMS value of the deviation, the higher the torque control precision and the better the coordination between the two arms. Combined with deformation values ​​and slippage risk assessment, a comprehensive evaluation index is formed to reflect the stability of the current handling state.

[0048] For example, when handling a composite material tube with a center of mass offset of 0.35 meters, after implementing the adjustment scheme, the maximum deformation near the center of mass was detected to be 2.5 mm, which is lower than the slip instability limit of 3.2 mm, and the system was judged to be in a stable state. At the same time, the root mean square value of the torque deviation of each joint was 3.5% of the rated torque, indicating that the dual-arm coordinated control was good, and the output control results showed that the system was in a safe handling state.

[0049] This invention provides a device for predicting distance and planning posture before a humanoid robot grasps an object, mainly comprising: The data acquisition and initial torque determination module is used to acquire initial gripping data at the object contact point, obtain the longitudinal offset distance of the object's center of mass and the elastic modulus of the surface near the center of mass, and determine the initial bearing torque of the arm near the center of mass based on the longitudinal offset distance of the center of mass and the elastic modulus of the surface near the center of mass. The torque monitoring and peak difference calculation module is used to set the torque monitoring parameters of each joint of the two arms based on the initial bearing torque and to monitor them. The maximum deviation between the actual monitored torque value of the joint near the center of mass and the preset initial bearing torque benchmark value is used as the peak torque difference of the joint near the center of mass. The torque distribution adjustment module is used to adjust the torque distribution according to the torque peak difference to reduce the vertical support torque on the near centroid side and increase the compensation torque on the far centroid side, determine the target torque value of each joint, and generate a joint torque distribution scheme. The scheme execution and state verification module is used to execute the joint torque distribution scheme and verify the contact state of the contact point near the center of mass. It compares the local deformation near the center of mass with the preset slip instability limit and outputs the torque peak control result including the current torque balance state of the two arms.

[0050] Based on the embodiments of the present invention described above, and through the above description, those skilled in the art can make various changes and modifications without departing from the technical concept of the present invention. The technical scope of the present invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. A method for predicting distance and planning posture before grasping by a humanoid robot, characterized in that, The method includes: The initial grasping data of the object contact point is collected, the longitudinal offset distance of the object's center of mass and the elastic modulus of the surface near the center of mass are obtained, and the initial bearing torque of the arm near the center of mass is determined based on the longitudinal offset distance of the center of mass and the elastic modulus of the surface near the center of mass. Based on the initial bearing torque, the torque monitoring parameters of each joint of the two arms are set and monitored. The maximum deviation between the actual monitored torque value of the joint near the center of mass and the preset initial bearing torque benchmark value is taken as the torque peak difference of the joint near the center of mass. Adjust the torque distribution based on the peak torque difference to reduce the vertical support torque near the centroid and increase the compensation torque far from the centroid, determine the target torque value of each joint, and generate a joint torque distribution scheme.

2. The method for predicting distance and planning posture before grasping by a humanoid robot according to claim 1, characterized in that, The determination of the initial bearing moment of the arm near the center of mass based on the longitudinal offset distance of the center of mass and the elastic modulus of the surface near the center of mass includes: By combining the longitudinal offset distance of the centroid with the elastic modulus of the near-centroid side surface, the relationship between the normal deformation and contact stress at the contact point of the object is analyzed to obtain the stress distribution state of the contact area. Based on the stress distribution state, it is determined whether the contact stress near the centroid side exceeds the preset material elastic limit threshold. If it does not exceed the preset material elastic limit threshold, the initial bearing torque of the near-centroid side arm is determined according to the centroid offset ratio. If it exceeds the preset material elastic limit threshold, the torque distribution ratio of the near-centroid side is reduced and the compensating bearing torque of the far-centroid side arm is increased.

3. The method for predicting distance and planning posture before grasping by a humanoid robot according to claim 1, characterized in that, The process of setting and monitoring torque parameters for each joint of both arms based on the initial bearing torque, and using the maximum deviation between the actual monitored torque value of the joint near the center of mass and the preset initial bearing torque benchmark value as the peak torque difference of the joint near the center of mass, includes: Based on the initial bearing torque value of the arm near the center of mass, the torque monitoring parameters of each joint of both arms are set and the current joint angle and corresponding drive current are obtained. The actual output torque value of each joint is calculated based on the linear relationship between motor torque and current. The actual output torque value is continuously monitored using a fixed time window. Within each time window, the real-time difference between the torque value of each joint near the centroid and the initial bearing torque reference value is calculated. The deformation development trend of the contact point is judged by the change in the difference between adjacent windows. The maximum value is extracted from the continuously monitored difference sequence as the peak torque difference.

4. The method for predicting distance and planning posture before grasping by a humanoid robot according to claim 3, characterized in that, After extracting the maximum value from the continuously monitored difference sequence as the peak torque difference, the process further includes: Record the contact pressure distribution data corresponding to the peak value. By comparing the peak torque difference with the preset deformation safety threshold, determine whether the contact point near the centroid is close to the critical deformation degree. If the peak torque difference exceeds the deformation safety threshold, identify the concentrated area where the pressure exceeds the preset pressure threshold from the contact pressure distribution data. Calculate the stress gradient by the ratio of the pressure difference between adjacent measuring points in the area to the distance. Evaluate the remaining bearing capacity of the contact point based on the ratio of the stress gradient to the initial elastic modulus of the material, and obtain the degree of deformation approach that reflects the contact stability near the centroid.

5. The method for predicting distance and planning posture before grasping by a humanoid robot according to claim 1, characterized in that, The step of adjusting the torque distribution based on the peak torque difference to reduce the vertical support torque near the centroid and increase the compensation torque far from the centroid, determining the target torque value for each joint, and generating a joint torque distribution scheme includes: Based on the peak torque difference, the value of the torque that needs to be reduced on the near-center of mass side is determined from the preset graded adjustment rules. The value of the vertical support torque reduced on the near-center of mass side and the value of the compensation torque increased on the far-center of mass side are calculated to keep the total support torque of the object constant. By combining the reduced vertical support torque value near the centroid side with the increased compensation torque value on the far centroid side, and the current joint angle configuration of the two arms, the required driving torque for the shoulder, elbow, and wrist joints is calculated backward from the required torque at the end effector through the torque transmission relationship in robot kinematics, thus obtaining the target torque value for each joint. The target torque value of each joint is compared with the preset maximum bearing torque limit of the joint. If the target torque of any joint near the center of mass exceeds its maximum bearing torque limit, the torque reduction on the near-center of mass side is reduced proportionally to the excess, and the torque on both sides is redistributed until the target torque of all joints is within the safe range.

6. The method for predicting distance and planning posture before grasping by a humanoid robot according to claim 1, characterized in that, The method further includes: executing the joint torque distribution scheme and verifying the contact state of the contact point near the center of mass, comparing the local deformation near the center of mass with the preset slip instability limit, and outputting the torque peak control result including the current torque balance state of the two arms.

7. The method for predicting distance and planning posture before grasping by a humanoid robot according to claim 6, characterized in that, The execution of the joint torque distribution scheme and verification of the contact state of the contact point near the centroid side includes: The adjusted target torque value is applied by the joint motor driver, and real-time pressure distribution data of the contact point near the center of mass is collected from the pressure sensor array. The local deformation is calculated based on the relationship between the pressure change in the contact area and the elastic modulus of the material. Calculate the difference in local deformation between adjacent measuring points, compare it with the preset slip instability limit to determine the contact stability. If the maximum value in the local deformation exceeds the preset proportion of the preset slip instability limit, it is marked as having slip risk; otherwise, it is marked as a stable state.

8. The method for predicting distance and planning posture before grasping by a humanoid robot according to claim 7, characterized in that, The output includes the peak torque control result of the current torque balance state of the two arms, including: Based on the judgment results of stable state or slip risk, the root mean square value of the deviation is calculated as the torque balance index by combining the deviation between the current actual torque value and the target torque value of each joint of the two arms. The torque balance index, deformation value and slip risk judgment results are integrated to output the torque peak control result that includes the current torque balance state of the humanoid robot's two arms.

9. The method for predicting distance and planning posture before grasping by a humanoid robot according to claim 1, characterized in that, The joints include the shoulder joint, elbow joint, and wrist joint.

10. A device for predicting distance and planning posture before grasping by a humanoid robot, characterized in that, The device includes: The data acquisition and initial torque determination module is used to acquire initial gripping data at the object contact point, obtain the longitudinal offset distance of the object's center of mass and the elastic modulus of the surface near the center of mass, and determine the initial bearing torque of the arm near the center of mass based on the longitudinal offset distance of the center of mass and the elastic modulus of the surface near the center of mass. The torque monitoring and peak difference calculation module is used to set the torque monitoring parameters of each joint of the two arms based on the initial bearing torque and to monitor them. The maximum deviation between the actual monitored torque value of the joint near the center of mass and the preset initial bearing torque benchmark value is used as the peak torque difference of the joint near the center of mass. The torque distribution adjustment module is used to adjust the torque distribution according to the torque peak difference to reduce the vertical support torque on the near centroid side and increase the compensation torque on the far centroid side, determine the target torque value of each joint, and generate a joint torque distribution scheme.

Images