A dynamic energy saving gripping method and system for humanoid robot handling processes

Abstract

The application provides a kind of humanoid robot carrying process dynamic energy-saving clamping method and system, it is related to robot control field, the method comprises: in stationary state, by controlling the end effector of robot to target goods gradually decreasing clamping force determines static slip critical force;When robot moves, according to the real-time kinematics data of leg joint, the touchdown shock time window that robot generates foot touchdown impact is predicted;And according to the moving speed of robot and step height, the vertical impact acceleration of foot touchdown instant and the inertia clamping force increment required to maintain the stability of goods are calculated;Control end effector to superimpose inertia clamping force increment on target clamping force in touchdown shock time window, and at the end of touchdown shock time window, control the clamping force of end effector and fall back to target clamping force.The method can guarantee that robot is in response to walking high-frequency impact stability, while reducing the redundant power consumption generated by maintaining constant large clamping force throughout.

Description

Technical Field

[0001] This application relates to the field of robot control, and in particular to a dynamic energy-saving clamping method and system for humanoid robot handling processes. Background Technology

[0002] With the rapid development of robotics technology, humanoid robots, due to their ability to adapt to unstructured environments, have shown great application potential in fields such as logistics, household services, and disaster relief. Unlike traditional wheeled or tracked mobile robots, humanoid robots primarily rely on alternating bipedal walking for movement. This biomimetic locomotion characteristic allows them to overcome obstacles such as steps and thresholds, enabling them to perform handling tasks in complex application scenarios. During handling operations, the robot's end effector (such as a dexterous hand or gripper) needs to apply a clamping force to the goods, and the magnitude of this clamping force directly determines the stability of the goods during handling.

[0003] In related technologies, robot gripping control technology mainly adopts two strategies: one is constant force control with a high safety factor, which presets a fixed large gripping force sufficient to resist the maximum potential disturbance force; the other is closed-loop control based on force / visual feedback, which monitors the object's state in real time through sensors and dynamically increases the gripping force when a sliding tendency is detected. In normal smooth movement or static grasping scenarios, both methods can maintain a certain success rate to some extent. For industrial applications, it is usually preferable to increase the gripping stiffness of the robotic arm or set a larger static friction safety threshold to ensure that the goods will not fall off even when slight disturbances occur.

[0004] However, the walking of humanoid robots is essentially a continuous, controlled collision process. The foot contact with the ground instantly generates a high-frequency impact force transmitted upwards along the robot's body. The feedback-based closed-loop control of related technologies is limited by the physical delays in sensing, transmission, and execution mechanisms, often making it difficult to establish the necessary anti-interference gripping force in time at the millisecond-level impact moment. This causes the goods to loosen or fall off before the motor can respond. To compensate for the safety hazards caused by this lag, related technologies generally employ a constant force control strategy, forcing the motor to maintain an extremely high gripping force capable of resisting the peak impact at the moment of ground contact throughout the entire handling cycle, including numerous smooth oscillating phases. This method of defending against transient impact risks with continuous high redundancy output results in the end effector operating in an overload state far exceeding actual load requirements for most of the time, causing significant ineffective energy loss and shortening the robot's operating endurance. Summary of the Invention

[0005] This application provides a dynamic energy-saving gripping method and system for humanoid robot handling processes, which addresses the problem that related technologies apply excessive gripping force to goods continuously throughout the handling process to ensure handling safety, resulting in significant energy waste in the robot during goods handling.

[0006] In a first aspect, this application provides a dynamic energy-saving gripping method for a humanoid robot during handling, the method comprising: When the robot is stationary, the end effector of the robot is controlled to apply a gradually decreasing clamping force to the target cargo, and the static slip critical force when the clamping force is applied to the target cargo is determined. The static slip critical force is the clamping force corresponding to the micro-slip signal of the contact surface between the end effector and the target cargo reaches a preset threshold. When the robot moves the target cargo by using the target clamping force, the real-time kinematic data of the robot's leg joints is decomposed into gait phase to predict the ground impact window when the robot's feet make contact with the ground during the movement. The target clamping force is calculated based on the static slip critical force and the preset safety margin. The vertical impact acceleration at the moment the foot touches the ground is calculated based on the robot's moving speed and stride height, as well as the weight of the target cargo, and the increment of inertial clamping force required to maintain the stability of the cargo is calculated based on the vertical impact acceleration. When the current system time enters the preset feedforward time period before the ground impact time window, the end effector is controlled to instantaneously superimpose the inertial clamping force increment on the target clamping force, and at the end of the ground impact time window, the clamping force of the end effector is controlled to fall back to the target clamping force.

[0007] By employing the above technical solution, the system utilizes gait phase prediction to determine the timing of foot impact, superimposing the dynamically calculated inertial clamping force increment only within a short time window of the impact, while maintaining a lower reference force based on the static critical force during other stable periods. This method overcomes the limitations of sensor feedback lag, enabling the robot to maintain stability against high-frequency impacts during walking while reducing redundant power consumption caused by maintaining a constant large clamping force throughout the entire process.

[0008] In some embodiments, the step of the robot moving the target cargo by means of a target clamping force specifically includes: Calculate the composite centroid coordinates corresponding to the robot carrying the target cargo, wherein the composite centroid coordinates are the centroid coordinates corresponding to the robot and the target cargo being regarded as a rigidly connected whole; Calculate the target backward tilt angle corresponding to the robot's torso posture when the projection point of the composite centroid coordinates on the horizontal plane overlaps with the projection area of ​​the robot's current supporting leg on the horizontal plane. The robot's hip joint is driven to tilt in the opposite direction to the target cargo until the robot's torso posture reaches the target tilt angle.

[0009] By adopting the above technical solution, the system treats the robot and the cargo as a rigid whole and calculates the position of the composite center of mass. Based on the relationship between the projection of the composite center of mass on the horizontal plane and the support area, it actively adjusts the torso's backward tilt angle. This method can reduce the forward tilting and overturning torque caused by carrying heavy objects to a certain extent by adjusting the robot's own posture, prompting the overall center of gravity of the system to return to the effective balance domain of the supporting legs. This optimizes the dynamic stability when walking with a load and reduces the risk of tipping over due to the forward shift of the center of gravity.

[0010] In some embodiments, the step of calculating the target backward tilt angle corresponding to the robot's torso posture when the projection point of the composite centroid coordinates on the horizontal plane overlaps with the projection area of ​​the robot's current supporting leg on the horizontal plane specifically includes: The projection point of the ankle joint center of the robot's current supporting leg onto the horizontal plane is determined as the zero-torque stagnation point; Calculate the horizontal deviation distance of the projection point of the composite centroid coordinates onto the horizontal plane relative to the zero-moment stagnation point; The angle change required to eliminate the horizontal deviation distance is calculated based on the preset geometric mapping relationship between the trunk center of gravity height and the hip joint pitch angle. The angle change is superimposed on the current robot's torso posture angle to obtain the target tilt angle that allows the projection point of the composite centroid coordinates to converge to the zero torque stagnation point.

[0011] By adopting the above technical solution, the system, based on the zero-torque stationary point theory, quantitatively calculates the horizontal deviation between the composite centroid projection and the ankle joint stationary point, and uses geometric mapping relationships to solve for the posture correction required to eliminate the deviation. This process achieves quantitative control of the torso tilt amplitude, ensuring that the robot's overall center of pressure always converges to the foot support area. By offsetting the effect of off-center loading through posture compensation, the robot's balance maintenance ability during handling is further improved.

[0012] In some embodiments, the step of determining the static slip threshold force of the target cargo when the end effector of the controlled robot applies a gradually decreasing clamping force to the target cargo includes: The end effector is controlled to clamp the target cargo with a preset initial safety clamping force, so that the target cargo remains in a stationary suspended state; According to the preset force unloading gradient, the output torque of the drive motor of the end effector is controlled to decrease linearly. During the torque reduction process, the tangential micro-vibration data of the contact surface collected by the tactile sensor set on the contact surface of the end effector is determined as the micro-slip signal. When the micro-motion slip signal is detected to exceed the preset signal threshold, the clamping force is immediately stopped from decreasing and the output torque of the drive motor at the current moment is determined as the static slip critical force.

[0013] By adopting the above technical solution, the system can automatically detect the true physical boundary (static slip critical force) required to maintain the current cargo's stillness by linearly unloading the clamping force and combining it with high-frequency tactile monitoring in a static state. This mechanism can adaptively obtain the optimal reference clamping force for a specific work object without prior knowledge of the cargo's friction coefficient or weight.

[0014] In some embodiments, after the step of immediately stopping the reduction of the clamping force and determining the current output torque of the drive motor as the static slip critical force when the detected micro-slip signal exceeds a preset signal threshold, the method further includes: The tactile sensor acquires the pressure distribution data on the contact surface of the end effector at the current moment. Calculate the eccentricity distance of the pressure center point of the pressure distribution data relative to the geometric center of the end effector; Based on the eccentric distance, the off-center load compensation coefficient corresponding to the preset safety margin is matched from the preset database.

[0015] By adopting the above technical solution, the system analyzes the pressure distribution through tactile sensors to identify the degree of eccentricity of the cargo relative to the end effector, and adjusts the safety margin accordingly by matching the corresponding compensation coefficient. This allows the force control strategy to fully consider the additional torque effect generated by non-ideal gripping positions (such as gripping edges), specifically enhancing the gripping robustness under eccentric loading conditions, and reducing the probability of the cargo rotating or accidentally falling off during movement due to uneven force distribution.

[0016] In some embodiments, the step of calculating the increment of inertial clamping force required to maintain the stability of the cargo based on the vertical impact acceleration specifically includes: The equivalent coefficient of friction between the end effector and the target cargo is determined based on the gravity index of the target cargo and the static slip critical force. The vertical inertial force generated by the target cargo at the moment of contact with the ground is calculated based on the vertical impact acceleration and the gravity index. The anti-disturbance clamping force required to suppress the slippage of the target cargo is calculated based on the vertical inertial force and the equivalent friction coefficient. The difference between the anti-disturbance clamping force and the target clamping force is determined as the inertial clamping force increment.

[0017] By adopting the above technical solution, the system uses the equivalent friction coefficient of the contact surface and the vertical impact acceleration at the moment of contact with the ground to deduce the anti-disturbance clamping force required to suppress the slippage of the goods. This ensures that the superimposed clamping force is sufficient to resist the vertical inertial force at the moment of contact with the ground, and will not damage fragile goods or cause unnecessary energy waste due to blindly applying excessive force, thus achieving precise force control.

[0018] In some embodiments, after the step of controlling the clamping force of the end effector to fall back to the target clamping force, the method further includes: The robot monitors the actual moment its foot touches the ground by using force sensors mounted on its feet. Calculate the timing deviation between the actual moment and the time center point of the ground impact time window; The timing deviation is used as a correction factor to shift and correct the timing parameters of the ground impact window predicted for the next walking cycle, so that the ground impact window of the next cycle remains synchronized with the robot's actual gait phase.

[0019] By employing the above technical solution, the system compares the actual ground contact time with the predicted time window and then corrects the time window parameters for the next walking cycle in a closed loop. This adaptive correction mechanism can reduce prediction drift caused by uneven ground or accumulated gait errors, ensuring that the application of high clamping force is always precisely synchronized with the actual physical impact time, preventing the clamping force from not being established in time when the impact arrives due to timing misalignment.

[0020] Secondly, this application provides a robot control system, the system comprising: one or more processors and a memory; The memory is coupled to the one or more processors. The memory is used to store computer program code, which includes computer instructions. The one or more processors call the computer instructions so that the system can implement the dynamic energy-saving clamping method for humanoid robot handling provided in the above embodiments, which will not be described in detail here.

[0021] Thirdly, this application provides a computer-readable storage medium including instructions that, when executed on a robot control system, enable the system to implement a dynamic energy-saving gripping method for humanoid robot handling provided in the above embodiments, which will not be elaborated here.

[0022] Fourthly, this application provides a computer program product, including a computer program that, when running on a robot control system, enables the system to implement a dynamic energy-saving gripping method for humanoid robot handling provided in the above embodiments, which will not be elaborated here.

[0023] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: 1. By decomposing and predicting the gait phase in real time, the system can more accurately pinpoint the microsecond-level impact window of the foot striking the ground. It only superimposes the dynamically calculated inertial gripping force increment at the moment the impact arrives and occurs, while maintaining a lower baseline gripping force during other stable motion phases. This method overcomes the limitations of sensor feedback lag, enabling the robot to significantly reduce ineffective high-load gripping time while maintaining stability against high-frequency impacts during walking, thus lowering the robot's energy consumption.

[0024] 2. During the static clamping phase, by using a linear unloading drive torque and a high-sensitivity tactile sensor to monitor tangential micro-vibrations, the system can automatically explore and lock the "static slip critical force" that keeps the cargo stationary, using it as the physical benchmark for subsequent control. Combined with compensation calculations for the eccentricity of the pressure distribution on the contact surface, this mechanism can autonomously generate optimal initial force control parameters for unknown cargoes with different friction coefficients, weights, and non-ideal gripping postures, improving the reliability of gripping diverse objects in unstructured scenarios.

[0025] 3. The system treats the cargo and robot as a rigid whole, calculating in real time the deviation between the projection of the composite center of mass onto the horizontal plane and the zero-moment stagnation point on the foot, and then inversely solving for the torso tilt angle required to eliminate the deviation. By actively driving the hip joint to adjust the torso posture, the system can use the robot's own gravitational torque to counteract the overturning torque of the cargo, forcing the overall center of pressure to always converge within the effective support domain, thus ensuring balance during the loaded movement from an overall dynamic perspective. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating a dynamic energy-saving clamping method for a humanoid robot during handling, as described in an embodiment of this application. Figure 2 This is another flowchart illustrating a dynamic energy-saving clamping method for handling humanoid robots in an embodiment of this application; Figure 3 This is a schematic diagram of the physical device structure of a robot control system in an embodiment of this application. Detailed Implementation

[0027] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification and appended claims of this application, the singular expressions “a,” “an,” “the,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to any or all possible combinations including one or more of the listed items.

[0028] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0029] In unstructured scenarios such as logistics sorting, household services, and disaster relief, humanoid robots need to move on two legs and grip goods with end effectors. These scenarios often involve uneven ground, diverse goods (different weights / shapes / materials), and the need to cross obstacles, which places extremely high demands on gripping stability and energy consumption.

[0030] The relevant technologies mainly employ two clamping strategies: one is constant high clamping force control, which presets a fixing force sufficient to withstand extreme impacts and maintains high load output throughout the process; the other is feedback-type closed-loop control, which relies on sensors to monitor the slippage trend and then dynamically increases the force. However, the former leads to high energy consumption throughout the process, shortens the robot's endurance, and may also damage fragile items; the latter, due to physical delays in sensor perception, signal transmission, and motor response, is generally designed to handle millisecond-level instantaneous impacts when the feet touch the ground, which can easily cause the goods to fall off, and both are difficult to adapt to non-ideal clamping postures (such as off-center loading).

[0031] To address this, this application obtains the minimum reference force (static slip critical force) for maintaining cargo stability through static detection, and determines the low-energy target clamping force by combining it with off-center load compensation. During movement, the timing of the foot contact with the ground and the required additional clamping force are predicted, and the incremental force is only briefly added before and after the impact. After the impact, the reference force is applied to the fall back, while time window deviations are corrected to ensure synchronization. For ease of understanding, the method provided in this implementation is described in process below. Please refer to... Figure 1 This is a flowchart illustrating a dynamic energy-saving clamping method for handling humanoid robots in an embodiment of this application.

[0032] S101. Control the end effector to clamp the target cargo with a preset initial safety clamping force, so that the target cargo remains in a stationary suspended state.

[0033] The initial safety clamping force refers to the system-preset initial clamping force that ensures the target cargo will not slip or fall off when it is stationary and hovering. Its value is preset based on the rated load of the robot's end effector, the weight range of common cargoes, and the friction coefficient range. The end effector is the component of the robot used to directly clamp the target cargo (such as a dexterous hand). The stationary hovering state refers to the static state of the target cargo after it is clamped, detached from the support surface, without significant displacement or shaking.

[0034] This step is performed at the initial stage when the robot begins its handling task. At this point, the robot has already used its end effector to grip the target cargo and is stationary, but has not yet started moving.

[0035] Specifically, the system sets an initial safe clamping force based on pre-stored parameters (such as the maximum clamping capacity of the end effector and past clamping experience data for similar goods). This force must meet the core requirement of "sufficient to lift the target goods off the ground and keep them stationary," avoiding situations where the initial clamping force is too small, causing the goods to slip directly, or too large, resulting in wasted initial energy. Subsequently, the system controls the drive motor of the end effector to output torque according to this initial safe clamping force, clamping the target goods until they are smoothly lifted and in a stationary, unsupported suspended state.

[0036] In some embodiments, the system can first identify the size and approximate weight range of the target cargo using a vision sensor, and match the initial safe clamping force for the corresponding weight range from a preset clamping force database; control the gripper of the end effector to open to an opening size suitable for the cargo size, slowly approach and contact the cargo surface; drive the motor to output torque according to the matched initial safe clamping force, clamp the cargo, and slowly lift the end effector until the cargo is detached from the support surface, confirming with a force sensor that the cargo has no tendency to slip, and maintain a suspended state; optionally, the system can also first control the end effector to contact the cargo with a small initial force, collect pressure data of the contact surface through a tactile sensor to determine whether the cargo is stably clamped; if it is not stably clamped (e.g., the pressure data fluctuates greatly), the clamping force is gradually increased until the cargo is smoothly lifted and suspended, and the clamping force at this time is the initial safe clamping force.

[0037] S102. During the linear decrease of the output torque of the drive motor of the end effector controlled by the preset force unloading gradient, the tangential micro-vibration data of the contact surface collected by the tactile sensor set on the contact surface of the end effector is determined as the micro-slip signal.

[0038] Among them, the force unloading gradient refers to the rate at which the output torque of the end effector drive motor decreases linearly. Its value is preset according to the sampling frequency of the tactile sensor and the stability requirements of the cargo to ensure that the decreasing process is smooth and can accurately capture the slip signal; the tangential micro-vibration data refers to the small vibration signal generated in the horizontal direction between the target cargo and the contact surface of the end effector. This signal is a precursor to the cargo slipping; the micro-slip signal is a key signal used to determine whether the target cargo is about to slip, and it is directly determined by the tangential micro-vibration data collected by the tactile sensor.

[0039] Specifically, after the cargo is stably suspended, the system does not suddenly reduce the clamping force. Instead, it controls the output torque of the drive motor to decrease uniformly and linearly according to the preset force unloading gradient, avoiding sudden torque changes that could cause the cargo to slip unexpectedly. At the same time, the tactile sensor installed on the contact surface of the end effector continuously and frequently collects vibration data of the contact surface. Since the cargo will first generate a small vibration along the tangential direction of the contact surface before the clamping force decreases to a critical value, the system directly defines the collected tangential micro-vibration data as a micro-slip signal.

[0040] In some embodiments, the system can read a preset force unloading gradient (e.g., decreasing by 5N per second) from the memory, and drive the motor to gradually reduce the output torque according to the gradient. After each small decrease (e.g., 0.1N), it pauses for 10 milliseconds to stabilize the state. At the same time, the system controls the tactile sensor to collect tangential vibration data of the contact surface at a preset sampling frequency (e.g., 1000Hz), filters the collected data (to remove environmental interference noise), and obtains pure tangential micro-vibration data, which is then identified as micro-slip signal.

[0041] S103. When the micro-motion slip signal is detected to exceed the preset signal threshold, immediately stop reducing the clamping force and determine the current drive motor output torque as the static slip critical force.

[0042] Among them, the preset signal threshold refers to the critical value of the micro-motion slip signal that the system has set in advance to determine that the cargo is about to slip. This value is determined based on a large amount of experimental data, which ensures that the slip precursor can be captured in time and avoids misjudgment. The static slip critical force refers to the clamping force corresponding to the micro-motion slip signal of the contact surface between the end effector and the target cargo reaches the preset threshold. It is the minimum clamping force required to keep the cargo stationary.

[0043] Specifically, during the force unloading process, the system compares the collected micro-slip signals with preset signal thresholds in real time. As the clamping force gradually decreases, the static friction between the cargo and the end effector gradually approaches its maximum value. When the clamping force decreases to a certain value, the static friction is insufficient to keep the cargo stationary, and the cargo begins to produce slight tangential vibrations. At this point, the micro-slip signal suddenly increases and exceeds the preset signal threshold. After detecting this trigger condition, the system immediately sends a stop unloading command to the drive motor to prevent the clamping force from continuing to decrease and causing the cargo to slip. At the same time, it records the output torque of the drive motor at the current moment. The clamping force corresponding to this torque is the static slip critical force, which is the minimum force value benchmark for keeping the cargo stationary.

[0044] S104. Obtain the pressure distribution data on the contact surface of the end effector at the current moment, and the eccentricity distance of the pressure center point of the pressure distribution data relative to the geometric center of the end effector.

[0045] Among them, pressure distribution data refers to the specific values ​​and distribution of pressure borne by each area of ​​the contact surface between the target cargo and the end effector, which can reflect the uniformity of force on the cargo on the clamping surface; pressure center point refers to the equivalent point of action of the pressure distribution on the contact surface, that is, the point of action of the resultant force of all pressures; end effector geometric center refers to the physical center position of the end effector clamping surface, which is a preset reference center point; eccentricity distance refers to the straight-line distance between the pressure center point and the geometric center, which is used to measure the degree of deviation of the cargo clamping position.

[0046] Specifically, at the instant the static slip critical force is determined, the system does not immediately enter the movement phase. Instead, it uses tactile sensors (or dedicated pressure distribution sensors) on the end effector contact surface to simultaneously collect pressure distribution data from multiple sampling points on the clamping surface. Subsequently, the system calculates the coordinates of the pressure center point using weighted averaging and other algorithms, extracts the preset coordinates of the end effector's geometric center, and calculates the straight-line distance between the two coordinates. This distance is the eccentricity distance. Through this process, the system can identify whether the cargo is clamped at the geometric center position. If the eccentricity distance is large, it indicates that the cargo is unevenly loaded, requiring compensation within the subsequent safety margin.

[0047] Optionally, the system controls an array of tactile sensors on the contact surface of the end effector to collect the pressure values ​​of each sensor unit, forming a pressure distribution matrix, i.e., pressure distribution data; based on the pressure value and coordinate position of each unit in the matrix, the system uses the centroid calculation method, with the pressure value of each unit as the weight, to calculate the coordinates of the pressure center point; and calls the preset coordinate parameters of the geometric center of the end effector to calculate the eccentric distance using the distance formula between two points.

[0048] S105. Match the off-center load compensation coefficient corresponding to the preset safety margin from the preset database based on the eccentricity distance.

[0049] Among them, the preset database refers to the data set that the system pre-stores, which contains the correspondence between eccentric distance and off-center load compensation coefficient. This data set is established based on a large amount of experimental data under different off-center load conditions. The preset safety margin refers to the proportion or value of the clamping force added on the basis of the static sliding critical force to ensure stable clamping of goods. The off-center load compensation coefficient is a coefficient determined according to the eccentric distance and used to correct the preset safety margin. The larger the eccentric distance, the larger the compensation coefficient is usually to offset the unstable effects caused by off-center load.

[0050] Specifically, after obtaining the eccentricity distance, the system queries a preset database. This database stores eccentricity compensation coefficients corresponding to different eccentricity distance ranges. For example, the compensation coefficient is 1.0 for eccentricity distances of 0-5mm, and 1.2 for eccentricity distances of 5-10mm. Based on the calculated eccentricity distance, the system determines the range to which it belongs and then matches the corresponding eccentricity compensation coefficient. This coefficient corrects the preset safety margin because when goods are eccentrically loaded, the basic preset safety margin alone may not guarantee stability. The compensation coefficient increases the safety margin to offset the torque effect caused by the eccentricity, ensuring that the subsequent target clamping force can adapt to non-ideal clamping conditions.

[0051] Not limited to fixed table lookup matching methods, in order to adapt to completely unknown and irregular cargo, the system uses a built-in off-center load moment penalty function to smoothly and continuously calculate the off-center load compensation coefficient. .

[0052] Specifically, when the eccentricity distance between the pressure center point and the geometric center is calculated... Then, set the maximum effective gripping half-width of the gripper to... At this time, the eccentric load compensation coefficient is calculated as follows: ,in, The system's preset gripper torque amplification constant; k is the off-center load sensitivity index (usually k=2). Calculate... Then, the preset safety margin is multiplied by this coefficient to inject an additional safety threshold against rotational torque on the basis of static critical equilibrium. When there is no off-center load ( When the deviation is greater than 1, the compensation coefficient is 1, which does not cause additional power consumption; the greater the deviation, the greater the nonlinear supplementary force, so as to dynamically solve the risk of grasping and falling off under non-ideal posture.

[0053] S106. When the robot moves by carrying the target cargo with the target clamping force, perform gait phase decomposition on the real-time kinematic data of the robot's leg joints to predict the ground impact window when the robot's feet touch the ground during the movement.

[0054] Among them, kinematic data refers to the real-time motion parameters of the robot's leg joints (such as hip joints, knee joints, and ankle joints), such as angles, angular velocities, and angular accelerations, which are used to reflect the motion state of the joints; gait phase decomposition refers to the process of dividing the complete cycle of the robot's bipedal walking into different stages (such as support phase, swing phase, ground contact phase, and ground lift phase), with each stage corresponding to specific joint motion characteristics; the ground contact impact window refers to the time interval between when the robot's foot is about to touch the ground, at the moment of ground contact, and for a short period of time after ground contact. High-frequency impact forces will be generated within this interval, and its start and end times are determined based on gait prediction.

[0055] Specifically, during robot movement, the system continuously collects real-time kinematic data of each joint in the legs. This data is analyzed using a pre-defined gait phase recognition algorithm, precisely decomposing the walking cycle into multiple consecutive gait phases. Since foot impact mainly occurs during the transition from the end of the swing phase to the beginning of the stance phase, the system predicts the specific time interval from when the foot will touch the ground to the end of the impact in each walking cycle—the ground impact window—based on the decomposed gait phase timing and the current joint movement speed and angle change trends.

[0056] Optionally, the system pre-stores standard gait phase templates (including the range of joint motion parameters for each phase) at different walking speeds; collects leg joint angle and angular velocity data in real time, compares and matches them with the standard templates to determine the current gait phase; and calculates the start and end times of the ground impact window based on the time interval between the current phase and the ground contact phase, combined with the duration of the historical walking cycle. Optionally, the system can also use machine learning models (such as LSTM neural networks) to train gait phase recognition and time prediction models with historical kinematic data, walking speed, stride length, etc. as input features. During movement, the real-time collected joint kinematic data is input into the model, and the current gait phase and the corresponding ground impact time window are directly output. It is understood that other methods can also be used to achieve this, such as using the pre-ground impact signal of the foot force sensor to assist in the judgment, or combining acceleration sensor data to optimize the time window prediction accuracy. This is not limited here.

[0057] It should be noted that before performing this step of movement, the system has pre-calculated the target clamping force for smooth handling based on the static sliding critical force obtained in step S103 and a preset safety margin matched from a preset database. This target clamping force can serve as the reference low-energy torque required for the end effector to output during the robot's start-up and smooth movement phases. Subsequently, the system controls the robot to cyclically execute the predicted actions of the gait phase and impact window during the movement of the target goods using the target clamping force.

[0058] S107. Calculate the vertical impact acceleration at the moment the foot touches the ground based on the robot's moving speed, stride height, and the weight of the target cargo.

[0059] Among them, the movement speed refers to the real-time movement rate of the robot in the horizontal direction, which is directly obtained by the robot's motion control system or calculated by the odometer; the stride height refers to the maximum vertical height of the foot from the ground when the robot's swing leg is lifted, which is derived from the leg joint kinematic data or measured by the visual sensor; the vertical impact acceleration refers to the instantaneous acceleration generated in the vertical direction at the moment the foot touches the ground, which is the core mechanical factor that causes the risk of cargo slippage.

[0060] Specifically, the system acquires the robot's current real-time movement speed and stride height. These two parameters directly determine the impact intensity when the foot touches the ground—the faster the movement speed and the greater the stride height, the greater the kinetic energy of the foot colliding with the ground, and the greater the vertical impact acceleration. Subsequently, the system inputs a preset dynamic calculation model, which is based on the principles of collision mechanics and comprehensively considers factors such as ground stiffness and the cushioning characteristics of the robot's legs, to derive the vertical impact acceleration at the moment the foot touches the ground through movement speed and stride height.

[0061] In some embodiments, the vertical impact acceleration can be calculated in a variety of ways: Optionally, the system reads the real-time movement speed from the robot motion controller, calculates the stride height by combining the leg joint angle sensor data with the forward kinematics algorithm; calls the preset empirical formula model (which is fitted based on a large amount of impact experimental data at different speeds and strides), substitutes the movement speed and stride height into the formula, and directly calculates the vertical impact acceleration. Optionally, the system can also construct a multibody dynamics simulation model, inputting the robot's body mass, leg structure parameters, current movement speed and stride height, as well as preset ground stiffness parameters; by simulating the collision process of the foot touching the ground, it outputs the vertical impact acceleration value at the moment of contact; it is understood that other methods can also be used, such as collecting historical impact data through foot accelerometers, establishing a mapping database of speed, stride and acceleration, and querying and matching in real time, which is not limited here.

[0062] Furthermore, in actual dynamic calculations, the vertical impact acceleration at the moment the foot touches the ground is not only related to the movement speed and stride, but the system also compensates for the absorption of impact kinetic energy by the leg structure stiffness through a preset spring-loaded inverted pendulum (SLIP) model. This allows for a more accurate output of the effective vertical impact acceleration acting on the target cargo, ensuring accurate calculation of the required inertial clamping force increment and avoiding ineffective power consumption waste caused by overestimating the acceleration.

[0063] In one specific embodiment, the system can use a momentum-impulse conversion algorithm based on the equivalent spring damping model of the leg for calculation, and the specific calculation formula is as follows: , where a is the vertical impact acceleration; This refers to the robot's horizontal movement speed; stride height; g is the acceleration due to gravity; For the robot's static mass; The characteristic weight of the target cargo; This is the basic buffer time constant for the robot's legs; The equivalent stiffness damping coefficient is determined by the configuration of the robot's mechanical legs.

[0064] S108. Determine the equivalent friction coefficient between the end effector and the contact surface of the target cargo based on the gravity index and static slip critical force of the target cargo.

[0065] Among them, the gravity index refers to the relevant parameters of the gravity acting on the target cargo, which is usually directly taken as the weight value of the cargo (which can be indirectly measured by the force sensor when the robot is gripping it); the equivalent friction coefficient refers to the equivalent friction parameter between the end effector and the contact surface of the target cargo, which takes into account factors such as material friction characteristics and surface condition, and is used to quantify the anti-slip ability of the contact surface.

[0066] Specifically, the system knows that the static slip critical force is the minimum clamping force required to keep the cargo stationary. According to the principle of static friction balance, the static slip critical force is directly related to the cargo's weight and the friction coefficient of the contact surface. Therefore, based on the target cargo's weight index and the determined static slip critical force, the system calculates the equivalent friction coefficient through reverse calculation. This coefficient comprehensively reflects the actual anti-slip capability of the current clamping contact surface without prior knowledge of the material friction characteristics of the cargo or end effector, and can adaptively adapt to different cargo clamping scenarios.

[0067] In a specific implementation, the target cargo gravity index G, the total number of clamping surfaces N of the two end effectors (e.g., N=2 for a two-finger gripper), and the static sliding critical force are taken as the factors. Calculate the equivalent friction coefficient of the contact surface. .

[0068] Optionally, the system collects the weight of the target cargo (i.e., gravity index related data) through the force sensor of the end effector; calls the preset static friction balance formula, takes the static sliding critical force as the maximum static friction force, and substitutes it into the formula to solve for the equivalent friction coefficient.

[0069] S109. Calculate the vertical inertial force generated by the target cargo at the moment of contact with the ground based on the vertical impact acceleration and gravity index.

[0070] Vertical inertial force refers to the inertial force generated by the vertical impact acceleration of the target cargo at the moment of impact. Its direction is the same as the direction of impact acceleration, and it is an additional force that causes the cargo to have a tendency to slide.

[0071] Specifically, according to Newton's second law, the inertial force of an object is directly related to its acceleration and mass. The system knows the target cargo's weight (from which its mass can be derived) and the vertical impact acceleration at the moment of contact with the ground. The vertical inertial force experienced by the target cargo at the moment of contact is calculated by multiplying these two values. This inertial force is superimposed on the cargo's weight, leading to an increase in the normal force at the contact surface between the cargo and the end effector. This necessitates a larger clamping force to suppress slippage; therefore, this value is a key input for calculating the anti-disturbance clamping force.

[0072] In some embodiments, the system divides the weight (gravity index) of the target cargo by the gravitational acceleration to obtain the mass of the cargo; obtains the calculated vertical impact acceleration, multiplies the mass by the acceleration to obtain the magnitude of the vertical inertial force; determines the direction of the vertical inertial force (consistent with the impact direction) based on the direction of the impact acceleration; optionally, the system considers the cushioning and vibration reduction characteristics of the robot's legs and corrects the vertical impact acceleration (e.g., by multiplying by a cushioning coefficient) to obtain the effective impact acceleration actually acting on the cargo; calculates the effective vertical inertial force in conjunction with the cargo mass; simultaneously, it monitors the pressure changes on the contact surface in real time using tactile sensors to verify and fine-tune the calculated inertial force.

[0073] Optionally, based on Newton's second law, and using the vertical impact acceleration *a* calculated in step S107 and the target cargo mass *m* = *G* / *g*, calculate the additional vertical inertial force generated at the moment of impact. .

[0074] S110. Calculate the anti-disturbance clamping force required to suppress the slippage of the target cargo based on the vertical inertial force and the equivalent friction coefficient, and determine the difference between the anti-disturbance clamping force and the target clamping force as the inertial clamping force increment.

[0075] Among them, the anti-disturbance clamping force refers to the minimum clamping force that can completely counteract the slippage tendency of the target cargo and ensure the stability of the cargo when the foot touches the ground; the inertial clamping force increment refers to the additional clamping force added to cope with the impact, that is, the difference between the anti-disturbance clamping force and the target clamping force; the vertical inertial force is the additional force generated by the vertical impact acceleration of the cargo at the moment of ground contact, which will increase the force on the contact surface between the cargo and the end effector; the equivalent friction coefficient is a parameter reflecting the anti-slip capability of the contact surface, which is derived from the cargo gravity index and the static slip critical force.

[0076] Upon impact, vertical inertial forces cause the cargo to tend to move downwards, leading to a horizontal slippage risk at the contact surface. This slippage risk is significantly greater than during the steady-state phase. To prevent cargo slippage, the clamping force applied by the end effector must generate sufficient static friction. The magnitude of this static friction depends on the clamping force and the equivalent coefficient of friction—a higher equivalent coefficient of friction results in stronger anti-slip capability at the contact surface, requiring a smaller clamping force; conversely, a greater vertical inertial force leads to a more pronounced slippage tendency, requiring a larger static friction force and consequently a larger clamping force. Therefore, the system calculates the minimum clamping force, or anti-disturbance clamping force, by correlating the vertical inertial force and the equivalent coefficient of friction. The target clamping force is a low-energy baseline force during the steady-state phase, capable only of balancing the slippage risk caused by cargo gravity and insufficient to handle impacts. Therefore, subtracting the target clamping force from the anti-disturbance clamping force yields the incremental inertial clamping force, which is the additional clamping force required during an impact.

[0077] In some embodiments, the system can obtain the specific value of the vertical inertial force, combine it with the determined equivalent friction coefficient and the cargo weight index obtained in the preceding steps, and calculate according to the logic of "total anti-disturbance clamping force = (vertical inertial force + cargo weight index) ÷ equivalent friction coefficient" to ensure that the static friction force generated by the clamping force is sufficient to offset the total sliding driving force including gravity and vertical inertial force, i.e., the optimal anti-disturbance clamping force. Subsequently, the target clamping force determined for the steady phase is retrieved from the memory; the target clamping force is subtracted from the total anti-disturbance clamping force calculated above to obtain the inertial clamping force increment used only to resist the impact, while ensuring that the total force after superposition does not exceed the maximum clamping force limit of the end effector. Optionally, the system can also classify risk levels (such as low, medium, and high) based on the magnitude of the vertical inertial force, and set corresponding safety factors (such as 1.1, 1.2, and 1.3) for different levels; multiply the vertical inertial force by the safety factor, and then divide by the equivalent friction coefficient to obtain a more reliable anti-disturbance clamping force; subtract the target clamping force from this anti-disturbance clamping force to obtain the inertial clamping force increment with safety redundancy; it is understood that other methods can also be used to achieve this, such as combining clamping force data under historical impact conditions for correction, or fine-tuning the calculation results according to the surface roughness of the cargo, which is not limited here.

[0078] S111. When the current system time enters the preset feedforward time period before the ground impact time window, the end effector is controlled to instantaneously superimpose the inertial clamping force increment on the target clamping force, and when the ground impact time window ends, the clamping force of the end effector is controlled to fall back to the target clamping force.

[0079] Among them, the preset feedforward time period refers to a preset advance time before the ground impact time window begins, which is used to offset the response delay of the sensor and motor, and ensure that the clamping force has reached the anti-interference requirement when the impact occurs. Its duration is preset based on the experimental data of system response delay (usually 10-50 milliseconds); instantaneous superposition refers to superimposing the inertial clamping force increment onto the target clamping force at an extremely fast rate (millisecond level) to avoid the force value changing too slowly and causing the anti-interference to be untimely.

[0080] Specifically, the system compares the current system time with the start time of the impact window in real time. When it detects that the current time has entered a preset feedforward period (e.g., 30 milliseconds before the start of the window), it immediately sends a force increment superposition command to the drive motor of the end effector. The motor responds quickly, instantaneously superimposing the inertial clamping force increment on the target clamping force, so that the clamping force quickly reaches the disturbance-resistant clamping force level, ensuring that the goods can be stably clamped within milliseconds of the impact. When the system determines that the current time has exceeded the end time of the impact window (the impact has disappeared), it sends a fallback command, controlling the motor to quickly reduce the clamping force to the target clamping force, restoring the low-energy operation state and avoiding energy waste caused by continuous increment superposition.

[0081] Optionally, the system includes a time determination module in the processor to synchronize the system time with the ground impact time window parameters in real time. When it detects that "current time = ground impact time window start time - preset feedforward time period", it triggers a motor control signal and quickly adjusts the motor output torque through PWM (pulse width modulation) technology to achieve instantaneous superposition of the inertial clamping force increment. When it detects that the current time exceeds the ground impact time window end time, it sends a reverse adjustment signal to control the motor torque to fall back to the value corresponding to the target clamping force.

[0082] S112. Calculate the timing deviation between the actual moment when the robot's foot touches the ground and the time center point of the impact window.

[0083] Among them, the actual moment refers to the instant when the robot's foot truly contacts the ground, which is determined by the pressure change signal triggered by the foot force sensor; the time center point of the ground impact time window refers to the middle time point of the preset ground impact time window, which serves as the reference point for time sequence comparison; the time sequence deviation refers to the time difference between the actual moment and the time center point (which can be positive or negative), used to reflect the degree of deviation between the predicted time window and the actual working condition.

[0084] Specifically, the system monitors the contact pressure between the robot's foot and the ground in real time using force sensors mounted on its feet. When the pressure value suddenly jumps from 0 (or a minimum value) to a preset contact threshold, the system time at this moment is recorded, which is the actual moment the foot contacts the ground. Subsequently, the system extracts the start and end times of the previously predicted ground impact window, calculates the midpoint between the two (time center point), and calculates the difference between the actual moment and this center point to obtain the timing deviation—if the actual moment is later than the center point, the deviation is positive; if it is earlier than the center point, the deviation is negative.

[0085] Optionally, the system sets the sampling frequency of the foot force sensor to 1000Hz to ensure that millisecond-level pressure changes are captured; when the pressure values ​​of three consecutive sampling points all exceed the preset contact threshold, it is confirmed that the foot has touched the ground, and the system time of the third sampling point is recorded as the actual time; the average start and end time of the ground impact window (time center point) is calculated; the timing deviation is obtained by subtracting the time center point from the actual time.

[0086] S113. The timing deviation is used as a correction factor to shift and correct the timing parameters of the ground impact window predicted for the next walking cycle, so that the ground impact window of the next cycle is synchronized with the robot's actual gait phase.

[0087] Translation correction refers to shifting the predicted ground impact time window of the next walking cycle forward or backward by a certain amount of time to match the actual ground contact time; walking cycle refers to the complete time for the robot to complete one alternating step (from the ground contact of one foot to the ground contact of the same foot again); actual gait phase refers to the true timing state of each stage during the robot's walking process.

[0088] Specifically, after predicting the ground contact impact window for the next gait cycle, the system does not directly use the original time window parameters. Instead, it uses the timing deviation calculated in the previous round as a correction factor to shift the start and end times of the original time window: if the timing deviation is positive (the actual time is later than the center point), it means the predicted time window is too early, and the next round's time window needs to be shifted backward by the time corresponding to the deviation; if the deviation is negative (the actual time is earlier than the center point), it means the predicted time window is too late, and the next round's time window needs to be shifted forward. Through this shifting correction, the ground contact impact window for the next gait cycle will move closer to the actual ground contact time, ensuring that the time center point of the time window coincides with the actual ground contact time as much as possible, thus achieving synchronization between the time window and the actual gait phase.

[0089] In some embodiments, the system adds a timing deviation to the start and end times of the predicted ground impact window for the next travel cycle to obtain the corrected start and end times of the new time window; for example, the original time window is [100ms, 150ms], the timing deviation is 10ms, and the corrected time window is [110ms, 160ms]; after correction, it checks whether the new time window exceeds the time range of the next travel cycle. If it does, boundary constraints are applied (not exceeding the start and end times of the cycle).

[0090] In the above embodiments, the system uses gait phase to predict the timing of foot contact with the ground and generates impact. It only superimposes the dynamically calculated inertial clamping force increment within a short time window of the impact, while maintaining a lower reference force based on a static critical force during other stable periods. This method overcomes the limitations of sensor feedback lag, enabling the robot to maintain stability against high-frequency impacts during walking while reducing redundant power consumption caused by maintaining a constant large clamping force throughout the entire process.

[0091] When a robot transports target goods, the weight and gripping position of the goods may alter the overall center of mass distribution of the robot and the goods. If the projection of the combined center of mass onto the horizontal plane exceeds the effective support area of ​​the current supporting leg, a tilting moment will be generated, causing the robot to become unbalanced and tip over. This application can also quantify and calculate the overall center of mass offset and lock the balance reference point during robot movement, thereby deriving and adjusting the torso tilt angle. The robot's own posture compensation can then counteract the off-center load effect caused by the goods, ensuring that the overall pressure center always converges within the effective domain of the supporting leg during loaded movement, thus guaranteeing the stability of the transport balance from a dynamic perspective. Please refer to [link / reference] for details. Figure 2 This is another flowchart illustrating a dynamic energy-saving clamping method for humanoid robot handling in an embodiment of this application.

[0092] It should be noted that, in this embodiment, the robot's movement by using the target gripping force to move the target cargo does not merely refer to the periodic stepping motion of the legs, but rather includes a complex and coordinated process of movement execution and dynamic balance maintenance. To counteract the forward bias torque caused by the cargo gripping in front, the system, while controlling the leg joints to generate forward movement speed, needs to include trunk posture compensation logic as a parallel subtask within the movement step. Specifically, this coordinated balancing movement sub-step includes the following: S201. Calculate the composite centroid coordinates corresponding to the robot's handling of the target cargo.

[0093] This step is performed when the robot has stably gripped the target cargo with the target clamping force through the end effector and is in a stationary state about to start moving, or when the calculation is updated in real time during each step cycle during the movement to ensure that the center of mass data is synchronized with the current motion state.

[0094] The system treats the robot and the target cargo as a rigid, immovable whole, rather than calculating their centers of mass separately. The core logic of the calculation is to combine the individual center of mass positions and mass percentages of both, and obtain the overall equivalent center of mass coordinates through weighted fusion. Specifically, the system acquires the robot's own center of mass parameters (including the coordinates and corresponding masses of the torso, limbs, and other components; this data is the basic parameters preset at the robot's factory). Then, it measures the weight (i.e., the cargo mass) of the target cargo using the force sensor of the end effector, and uses visual or tactile sensors to determine the cargo's installation position relative to the robot's end effector, deriving the cargo's center of mass coordinates. Finally, based on the calculation logic of the rigid body composite center of mass, the system integrates the center of mass coordinates and mass data of both to solve for the overall composite center of mass in three dimensions (including position information in the x, y, and z axes).

[0095] S202. Determine the projection point of the ankle joint center of the robot's current supporting leg onto the horizontal plane as the zero-torque stationary point.

[0096] It should be noted that the "overlap of projection point and projection area" mentioned in the embodiments of this application refers to the projection point of the composite centroid on the horizontal plane falling within the effective support polygon area formed by the projection of the robot's current supporting leg on the horizontal plane. Specifically, the system can determine which leg is the current supporting leg based on the pressure data from the foot force sensor. For example, the leg with a pressure value greater than a preset support threshold (i.e., the minimum pressure value sufficient to support the total weight of the robot and the cargo) is identified as the current supporting leg. Subsequently, the three-dimensional coordinates of the ankle joint of the supporting leg are obtained (the angle data of the hip, knee, and ankle joints are collected by the leg joint angle sensor, combined with the length parameters of the robot's leg links, and derived through the forward kinematics algorithm). Finally, the z-axis (vertical direction) coordinate of the ankle joint is zeroed out, and only the x-axis and y-axis (in the horizontal plane) coordinates are retained. This coordinate point is the zero-torque stationary point, and its core function is to serve as the reference origin for judging the centroid offset. When the projection of the composite centroid coincides with this point, the overturning moment at the ankle joint of the supporting leg is zero, and the robot is in the most stable equilibrium state.

[0097] In some embodiments, the system collects pressure data from the force sensors at the ends of the two legs in real time, compares the values ​​of the two, and determines the leg with the larger pressure value that continuously exceeds the support threshold as the current supporting leg; calls the linkage length parameters of the supporting leg (preset lengths of the thigh, calf, and foot), and calculates the three-dimensional coordinates of the ankle joint center by combining the real-time angle of the ankle joint collected by the joint angle sensor; projects the three-dimensional coordinates onto the horizontal plane through the coordinate transformation module (i.e., sets the z-axis coordinate to 0), outputs the coordinate point in the xy plane, and determines it as the zero torque stationary point.

[0098] S203. Calculate the horizontal deviation distance of the projection point of the composite centroid coordinates onto the horizontal plane relative to the zero-moment stationary point.

[0099] Among them, the horizontal deviation distance refers to the straight-line distance between the projection point of the composite centroid coordinates on the horizontal plane and the zero-moment stationary point. It is used to quantify the degree of deviation of the overall centroid relative to the equilibrium reference point. The larger the value, the more serious the centroid deviation and the higher the risk of imbalance.

[0100] Specifically, the system extracts the x-axis and y-axis coordinates from the three-dimensional coordinates of the composite centroid calculated in S201 to obtain the coordinates of the projection point of the composite centroid on the horizontal plane (ignoring the vertical data of the z-axis, focusing only on the positional relationship on the horizontal plane). Then, it retrieves the coordinates of the zero-moment stationary point determined in S202 (also in the xy-plane coordinates). Finally, based on the distance formula between two points in the plane, it calculates the straight-line distance between the two coordinate points; this distance is the horizontal deviation distance, and its result directly reflects the current offset of the centroid relative to the equilibrium reference point.

[0101] Optionally, the system can map the coordinates of the synthesized centroid horizontal projection point and the zero-torque stationary point to the robot's own coordinate system (a local coordinate system with the robot's torso center as the origin); unify the coordinates of the two points to the same dimension (e.g., millimeter-level units) through coordinate transformation; and use an incremental calculation method to first calculate the offset in the x-axis direction (Δx=|x1-x0|) and the offset in the y-axis direction (Δy=|y1-y0|); then, through vector synthesis, treat Δx and Δy as the two legs of a right triangle, calculate the length of the hypotenuse as the horizontal deviation distance, and simultaneously record the offset direction (positive / negative x-axis direction, positive / negative y-axis direction). It is understandable that a "real-time sampling + moving average" method can also be used to improve calculation accuracy, for example, continuously collecting 10 sets of synthesized centroid projection coordinates and zero-torque stationary point coordinates, calculating the deviation distance for each set, and then taking the average value; this is not limited here.

[0102] S204. Calculate the angle change required to eliminate the horizontal deviation distance based on the preset geometric mapping relationship between the trunk center of gravity height and the hip joint pitch angle.

[0103] Among them, the height of the torso center of mass refers to the vertical distance from the robot's own torso center of mass to the horizontal plane. This data is related to the robot's current torso posture and hip joint angle, and can be measured in real time by the torso posture sensor; the hip joint pitch angle refers to the angle of rotation of the hip joint around the forward and backward direction (parallel to the ground and perpendicular to the walking direction), with the positive direction being the torso tilting forward and the negative direction being the torso tilting backward; the geometric mapping relationship refers to the data set preset by the system that reflects the corresponding law between the height of the torso center of mass and the hip joint pitch angle. It is obtained by fitting a large amount of experimental data, and the core is to quantify "the amount of offset of the horizontal position of the torso center of mass when the hip joint rotates by a certain angle".

[0104] Specifically, the system collects the current torso center of gravity height using attitude sensors (such as IMU inertial measurement units) mounted on the torso, ensuring that this height data matches the robot's current standing and cargo-holding states. Then, it retrieves a pre-set geometric mapping database, which stores the horizontal offset of the torso center of gravity corresponding to a 1° change in hip joint pitch angle at different torso center of gravity heights (e.g., at a torso center of gravity height of 80cm, a 1° hip tilt can cause the torso center of gravity to shift horizontally backward by 2cm). Finally, based on the magnitude of the horizontal deviation, the required change in hip joint pitch angle is derived in reverse—that is, calculating "how much horizontal movement of the torso center of gravity is needed to offset the current horizontal deviation." This angle value is the angle change required to eliminate the deviation, and since the deviation needs to be offset by torso tilting backward, the angle change is a reverse (tilting backward) angle value.

[0105] Optionally, the system acquires three-dimensional posture data of the torso through an IMU sensor, extracts height information in the z-axis direction, and calculates the current torso center of mass height H; retrieves the "pitch angle - horizontal offset" mapping curve corresponding to H from a preset database (e.g., when H=80cm, the mapping curve is y=2x, where y is the horizontal offset in cm and x is the tilt angle); based on the horizontal deviation distance D (e.g., D=4cm), substitutes it into the mapping curve to solve for x=D / 2=2°, that is, the angle change is 2° of tilt; performs boundary verification on the solution to ensure that the angle change does not exceed the maximum tilt angle limit of the hip joint (e.g., maximum tilt 10°).

[0106] In some specific implementation scenarios, an inverted pendulum model can also be used to approximate the solution: The system uses the robot's supporting ankle joint as the pivot point for rotation. Let the current height of the torso's center of mass be L, and the current torso posture angle be... The calculated horizontal deviation distance of the composite centroid that needs to be eliminated is: To eliminate this deviation, based on the spatial trigonometric geometric relationships of the inverted pendulum: ,in, The horizontal deviation distance of the trunk's center of mass. The required angular change in the robot's torso is determined by the amount of compensation needed, and then the required angular change to eliminate this horizontal deviation is derived using inverse trigonometric functions. .

[0107] S205. The angle change is superimposed on the current robot's torso posture angle to obtain the target tilt angle that allows the projection point of the composite centroid coordinates to converge to the zero torque stagnation point.

[0108] Among them, the current torso posture angle refers to the initial tilt angle of the robot's torso relative to the vertical direction before adjustment. It can be collected in real time by the posture sensor. The positive direction is forward tilt and the negative direction is backward tilt (e.g., if the current posture is forward tilt of 1°, the angle value is +1°). The target backward tilt angle refers to the final posture angle that the robot's torso needs to achieve after the angle is superimposed. Its core function is to make the composite centroid projection converge to the zero torque stationary point to ensure balance.

[0109] Specifically, the system collects the current torso posture angle through attitude sensors, determining its initial tilt direction (forward or backward) and specific angle value. Then, it determines the direction of the angle change superposition—since the angle change is to achieve a backward tilt of the torso to offset the forward tilt deviation of the center of mass, the angle change is superimposed in the "backward tilt direction" regardless of whether the current torso is tilting forward or backward (i.e., when the current angle is forward, a negative angle is superimposed; when the current angle is backward, a negative angle is superimposed to increase the backward tilt amplitude). Finally, the current torso posture angle and the angle change are algebraically superimposed to obtain the target backward tilt angle. For example: if the current torso is tilted forward by 1° (angle value + 1°), and the angle change is a backward tilt of 2° (angle value - 2°), the superimposed target backward tilt angle is +1° + (-2°) = -1° (i.e., a backward tilt of 1°); if the current torso is tilted backward by 0.5° (angle value -0.5°), the superimposed target backward tilt angle is -0.5° + (-2°) = -2.5° (i.e., a backward tilt of 2.5°).

[0110] S206. Drive the robot's hip joint to tilt in the opposite direction of the target cargo until the robot's torso posture reaches the target tilt angle.

[0111] Specifically, the system determines the tilt direction of the hip joint based on the gripping position of the target cargo: since the cargo is gripped by an end effector (such as a hand), it is usually located in front of or to the side of the robot, and the torso tilt needs to be in the opposite direction to the cargo (for example: if the cargo is directly in front of the robot, the torso tilts backward; if the cargo is to the left front of the robot, the torso tilts backward and to the right, the core being to shift the composite center of mass towards the supporting leg). Subsequently, a posture adjustment command is sent to the hip joint's drive motor, including the target tilt angle and drive speed (set according to the gait cycle to avoid excessively fast movements that could cause impact). Finally, during the drive process, the current torso posture angle is collected in real time by a torso posture sensor and compared with the target tilt angle to form a closed-loop control: if the current angle does not reach the target value, the motor continues to rotate; if the difference between the current angle and the target tilt angle is less than a preset error threshold (e.g., ±0.1°), the motor drive stops, completing the posture adjustment.

[0112] In some embodiments, hip joint actuation and posture calibration can be achieved in multiple ways: Optionally, the system uses a vision sensor to locate the gripping position of the target cargo (e.g., 50cm in front of the robot) and determines that the hip joint tilt direction is directly backward; sends a PWM control command to the hip joint drive motor, sets the motor rotation speed to 5° / s, and the target angle to -2.5° (2.5° backward tilt); initiates a posture feedback closed loop, and collects the current torso posture angle every 10ms through an IMU sensor; when the current angle is collected as -2.45° (0.05° difference from the target angle, less than the error threshold of 0.1°), a stop command is sent to the motor; maintains the current posture for 100ms, and after confirming that the posture is stable, completes this drive adjustment.

[0113] Optionally, the system can also decompose the target tilt angle into angular displacement commands of the hip joint (e.g., a 3° tilt corresponds to 150 pulses of motor rotation). Considering the inertia of the cargo, a segmented drive strategy is adopted: the first segment drives the motor to rotate 2° at a low speed of 3° / s. After collecting posture data to confirm that there is no impact, the second segment drives the remaining 1° at a speed of 6° / s. During the drive, the pressure distribution of the supporting leg is monitored by the foot force sensor. If the uniformity of the pressure distribution exceeds the preset threshold (e.g., the pressure difference between different areas is less than 5%), the drive continues. If the pressure distribution is abnormal (e.g., excessive pressure on one side), the drive is paused and the drive direction is finely adjusted. When the current posture angle reaches the target tilt angle (e.g., -3°) and stabilizes for 3 sampling cycles, the drive stops, and the posture adjustment is completed.

[0114] It is understandable that a "gait coordination drive" approach can also be adopted, which combines the robot's walking gait and completes the torso tilt adjustment in stages during the period when the supporting leg is under stable force, so as to avoid conflict with the gait action. This is not limited here.

[0115] It should be noted that the above steps of eliminating the composite centroid deviation by tilting the hip joint, in addition to maintaining walking balance, change the robot's longitudinal rigid force line. This increases the non-vertical dissipation component of the high-frequency impact force from the feet when it is transmitted to the torso and end effector, thereby effectively reducing the absolute vertical impact acceleration transmitted to the target cargo. This works in conjunction with the aforementioned dynamic energy-saving clamping force algorithm to further reduce the peak demand for inertial clamping force increment.

[0116] For ease of understanding, such as Figure 3 A schematic diagram of an exemplary architecture for implementing an embodiment of the present invention is shown. It should be noted that... Figure 3 The schematic diagram shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0117] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by instructions (computer programs), or by instructions (computer programs) controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor. The robot control system of this embodiment includes a storage medium and a processor. The storage medium stores multiple instructions that can be loaded by the processor to execute any step of the method provided in this embodiment.

[0118] Specifically, the storage medium and the processor are electrically connected directly or indirectly to enable data transmission or interaction. For example, these components can be electrically connected to each other via one or more signal lines. The storage medium stores computer-executable instructions that implement data access control methods, including at least one software functional module that can be stored in the storage medium in the form of software or firmware. The processor executes various functional applications and data processing by running the software program and module stored in the storage medium. The storage medium can be, but is not limited to, Random Access Memory (RAM), Read-Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), etc. The storage medium stores the program, and the processor executes the program after receiving the execution instructions.

[0119] Furthermore, the software programs and modules within the aforementioned storage medium may also include an operating system, which may include various software components and / or drivers for managing system tasks (e.g., memory management, storage device control, power management, etc.) and can communicate with various hardware or software components to provide an operating environment for other software components. The processor may be an integrated circuit chip with signal processing capabilities. The aforementioned processor may be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc., which can implement or execute the methods, steps, and logic flowcharts disclosed in this embodiment. The general-purpose processor may be a microprocessor or any conventional processor.

[0120] Since the instructions stored in the storage medium can execute the steps in any of the methods provided in the embodiments of the present invention, the beneficial effects of any of the methods provided in the embodiments of the present invention can be achieved, as detailed in the preceding embodiments, and will not be repeated here.

[0121] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A dynamic energy-saving gripping method for humanoid robot handling processes, characterized in that, The method includes: When the robot is stationary, the end effector of the robot is controlled to apply a gradually decreasing clamping force to the target cargo, and the static slip critical force when the clamping force is applied to the target cargo is determined. The static slip critical force is the clamping force corresponding to the micro-slip signal of the contact surface between the end effector and the target cargo reaches a preset threshold. When the robot moves the target cargo by using the target clamping force, the real-time kinematic data of the robot's leg joints is decomposed into gait phase to predict the ground impact window when the robot's feet make contact with the ground during the movement. The target clamping force is calculated based on the static slip critical force and the preset safety margin. The vertical impact acceleration at the moment the foot touches the ground is calculated based on the robot's moving speed and stride height, as well as the weight of the target cargo, and the increment of inertial clamping force required to maintain the stability of the cargo is calculated based on the vertical impact acceleration. When the current system time enters the preset feedforward time period before the ground impact time window, the end effector is controlled to instantaneously superimpose the inertial clamping force increment on the target clamping force, and at the end of the ground impact time window, the clamping force of the end effector is controlled to fall back to the target clamping force.

2. The method according to claim 1, characterized in that, The step of the robot moving the target cargo by using the target clamping force specifically includes: Calculate the composite centroid coordinates corresponding to the robot carrying the target cargo, wherein the composite centroid coordinates are the centroid coordinates corresponding to the robot and the target cargo being regarded as a rigidly connected whole; Calculate the target backward tilt angle corresponding to the robot's torso posture when the projection point of the composite centroid coordinates on the horizontal plane overlaps with the projection area of ​​the robot's current supporting leg on the horizontal plane. The robot's hip joint is driven to tilt in the opposite direction to the target cargo until the robot's torso posture reaches the target tilt angle.

3. The method according to claim 2, characterized in that, The step of calculating the target backward tilt angle corresponding to the robot's torso posture when the projection point of the composite centroid coordinates on the horizontal plane overlaps with the projection area of ​​the robot's current supporting leg on the horizontal plane specifically includes: The projection point of the ankle joint center of the robot's current supporting leg onto the horizontal plane is determined as the zero-torque stagnation point; Calculate the horizontal deviation distance of the projection point of the composite centroid coordinates onto the horizontal plane relative to the zero-moment stagnation point; The angle change required to eliminate the horizontal deviation distance is calculated based on the preset geometric mapping relationship between the trunk center of gravity height and the hip joint pitch angle. The angle change is superimposed on the current robot's torso posture angle to obtain the target tilt angle that allows the projection point of the composite centroid coordinates to converge to the zero torque stagnation point.

4. The method according to claim 1, characterized in that, The step of determining the static slip critical force of the target cargo when the end effector of the controlled robot applies a gradually decreasing clamping force to the target cargo includes: The end effector is controlled to clamp the target cargo with a preset initial safety clamping force, so that the target cargo remains in a stationary suspended state; According to the preset force unloading gradient, the output torque of the drive motor of the end effector is controlled to decrease linearly. During the torque reduction process, the tangential micro-vibration data of the contact surface collected by the tactile sensor set on the contact surface of the end effector is determined as the micro-slip signal. When the micro-motion slip signal is detected to exceed the preset signal threshold, the clamping force is immediately stopped from decreasing and the output torque of the drive motor at the current moment is determined as the static slip critical force.

5. The method according to claim 4, characterized in that, After the step of immediately stopping the reduction of clamping force and determining the current output torque of the drive motor as the static slip critical force when the detected micro-slip signal exceeds the preset signal threshold, the method further includes: The tactile sensor acquires the pressure distribution data on the contact surface of the end effector at the current moment. Calculate the eccentricity distance of the pressure center point of the pressure distribution data relative to the geometric center of the end effector; Based on the eccentric distance, the off-center load compensation coefficient corresponding to the preset safety margin is matched from the preset database.

6. The method according to claim 4, characterized in that, The step of calculating the increment of inertial clamping force required to maintain cargo stability based on the vertical impact acceleration specifically includes: The equivalent coefficient of friction between the end effector and the target cargo is determined based on the gravity index of the target cargo and the static slip critical force. The vertical inertial force generated by the target cargo at the moment of contact with the ground is calculated based on the vertical impact acceleration and the gravity index. The anti-disturbance clamping force required to suppress the slippage of the target cargo is calculated based on the vertical inertial force and the equivalent friction coefficient. The difference between the anti-disturbance clamping force and the target clamping force is determined as the inertial clamping force increment.

7. The method according to claim 1, characterized in that, After the step of controlling the clamping force of the end effector to return to the target clamping force, the method further includes: The robot monitors the actual moment its foot touches the ground by using force sensors mounted on its feet. Calculate the timing deviation between the actual moment and the time center point of the ground impact time window; The timing deviation is used as a correction factor to shift and correct the timing parameters of the ground impact window predicted for the next walking cycle, so that the ground impact window of the next cycle remains synchronized with the robot's actual gait phase.

8. A robot control system, characterized in that, The system includes: one or more processors and memory; The memory is coupled to the one or more processors, the memory being used to store computer program code, the computer program code including computer instructions, the one or more processors invoking the computer instructions to cause the system to perform the method as described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that, When the instructions are executed on the robot control system, the system performs the method as described in any one of claims 1-7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is run on the robot control system, it causes the system to perform the method as described in any one of claims 1-7.

Images