Whole body posture linkage method and device when humanoid robot reaches for objects

Abstract

The application provides a full-body posture linkage method and device when a humanoid robot reaches for an object, comprising: obtaining the arm swing angular velocity, gait phase angle and orientation deviation angle between the arm end and the target object in the walking state of the humanoid robot, determining the freeze switching judgment threshold according to the orientation deviation angle and the arm swing angular velocity; screening the preliminary candidate point set according to the comparison between the freeze switching judgment threshold and the preset threshold; updating the modified value of the freeze switching judgment threshold in the next gait cycle according to the evaluation result of the suppression result of the trunk pitch oscillation amplitude, mapping the modified value to the phase offset compensation value in the gait phase synchronization, determining the starting time of the reaching action; and fusing the reaching action starting time and the arm freezing time, evaluating the compliance degree of the preset time optimization threshold and the overall action time shortening amount, to determine the full-body posture linkage parameter.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a method and apparatus for coordinating the full-body posture of a humanoid robot when it reaches for an object. Background Technology

[0002] In the field of robotics, research on motion control for humanoid robots is of paramount importance, directly impacting their adaptability and task performance efficiency in complex environments. This is especially true when mimicking human behaviors such as reaching for objects, where coordinating the entire body's posture becomes a core challenge. Research in this area has not only advanced the naturalness of robot-human interaction but also provided broad application prospects in industrial, medical, and domestic service scenarios. However, existing methods have significant shortcomings in handling the full-body coordination control of humanoid robots reaching for objects. Many solutions focus only on optimizing a single aspect, such as simply pursuing the smoothness of arm movements, while neglecting the dynamic relationship between arm movements and gait. This neglect leads to postural imbalances or inefficiencies in actual operation due to improper timing of movement transitions, particularly when rapid responses to changes in the target object's position are required. A deeper technical challenge lies in balancing the contradiction between the timing of arm movement initiation and the dynamic rhythm of the gait. The initiation of arm movements needs to match the robot's gait rhythm, which is constantly changing, resulting in significant differences in the impact on overall stability when the arm is initiated at different times. Choosing the wrong time to initiate arm movements can cause sudden changes in shoulder strength, disrupting trunk balance and even increasing the overall movement time. For example, when a robot walks, the speed and direction of the arm swing change with the gait cycle. If the arm movement is initiated at a moment of drastic speed change, the pressure on the shoulder will increase sharply, potentially causing unnecessary trunk swaying and affecting stability. Furthermore, when walking on uneven terrain, the robot's gait cycle is shortened. If the arm movement is initiated during a high-speed phase of the cycle, the load on the shoulder joint will increase dramatically, causing the trunk to lean forward or sideways, seriously threatening overall balance. Conversely, if the arm movement is initiated during a low-speed phase of walking on a flat surface, the arm movement will be stable, but the overall response will be delayed, potentially causing the robot to miss its target. Therefore, accurately matching the arm movement initiation with the gait rhythm during robot walking to avoid sudden changes in shoulder strength and trunk swaying, while minimizing movement completion time, is a key problem that this research urgently needs to solve. Summary of the Invention

[0003] This invention provides a method for coordinating the body posture of a humanoid robot when reaching for an object, mainly including: The arm swing angular velocity, gait phase angle, and orientation deviation angle between the arm end and the target object are obtained in the walking state of the humanoid robot. The freeze switching judgment threshold is determined based on the orientation deviation angle and the arm swing angular velocity. Based on the comparison between the freeze switching determination threshold and the preset threshold, a preliminary candidate point set is selected; Obtain the shoulder joint torque mutation amount and end effector trajectory length of each candidate point in the preliminary candidate point set. Calculate the comprehensive cost value of each candidate point based on the shoulder joint torque mutation amount and end effector trajectory length of each candidate point, and search for the arm freezing time with the lowest comprehensive cost value. Obtain the torso posture data at the moment of arm freeze corresponding to the lowest comprehensive cost, evaluate the suppression result of torso pitch oscillation amplitude, analyze the timing of the robot's reach action, and then determine the whole-body posture linkage parameters that meet the preset requirements.

[0004] Furthermore, the step of acquiring the arm swing angular velocity, gait phase angle, and azimuth deviation angle between the arm end and the target object during the humanoid robot's walking state, and determining the freeze switching threshold based on the azimuth deviation angle and the arm swing angular velocity, includes: The arm angle signal is obtained by a shoulder joint encoder, and the arm angular velocity is calculated. The gait phase angle is simultaneously acquired via a hip joint encoder and mapped to a gait cycle interval; Extract the azimuth deviation angle between the end effector of the arm and the centroid of the target object; The ratio of the azimuth deviation angle to the swing arm angular velocity is used as a response sensitivity index. A quadratic cost function with the azimuth deviation angle as the variable is constructed. The minimum value is obtained by rolling time domain optimization to obtain the correction coefficient. The stabilized correction coefficient is obtained by smoothing the statistical mean and standard deviation of the correction coefficient within several gait cycle intervals. The freeze switching threshold is determined based on the stabilization correction coefficient and the response sensitivity index.

[0005] Furthermore, the step of filtering a preliminary candidate point set based on the comparison between the freeze switching determination threshold and a preset threshold includes: Obtain the freeze switching judgment threshold, compare it with the preset threshold, and calculate the extent of exceeding the limit; Based on the correspondence between the overshoot amplitude and the gait phase angle, phase moments are screened on the gait cycle time axis and marked as preliminary candidate points; Based on the gait phase angle values ​​of the preliminary candidate points, determine whether they are located in the two-leg support phase interval. If they are located in the two-leg support phase interval, they are retained; if they are located in the single-leg support phase interval, they are discarded. The retained candidate points are arranged in chronological order according to the gait phase angle values ​​to obtain the preliminary candidate point set.

[0006] Furthermore, the step of obtaining the shoulder joint torque mutation amount and end effector trajectory length of each candidate point in the preliminary candidate point set, calculating the comprehensive cost value of each candidate point based on the shoulder joint torque mutation amount and end effector trajectory length, and searching for the arm freezing moment with the lowest comprehensive cost value includes: Obtain the gait phase angle value of each candidate point in the preliminary candidate point set, extract the shoulder joint driving torque value at the corresponding phase moment, calculate the difference with the previous moment, and obtain the shoulder joint torque mutation amount. Obtain the straight-line distance between the end effector of the arm and the centroid of the target object to get the length of the bypass trajectory; The comprehensive cost is obtained by summing the products of the sudden change in shoulder joint torque and the preset torque weighting coefficient, and the products of the detour trajectory length and the preset trajectory weighting coefficient after normalization. Traverse the initial candidate point set, compare the comprehensive cost value of each candidate point, select the candidate point with the lowest cost value, and determine the swing arm freezing time.

[0007] Furthermore, the search for the swing arm freezing moment with the lowest overall cost includes: Obtain the torso posture data corresponding to the moment the swing arm is frozen, and collect the pitch angle sequence within a preset time window; The pitch angle sequence was decomposed in the frequency domain using the Fourier series fitting method to extract the oscillation amplitude of the fundamental frequency component, which was then determined as the torso pitch oscillation amplitude. The amplitude of the torso pitch oscillation is compared with a preset oscillation threshold. If it is lower than the preset oscillation threshold, the peak difference between adjacent oscillation cycles is calculated to obtain the energy decay rate.

[0008] Furthermore, the evaluation of the suppression results of torso pitch oscillation amplitude is used to analyze the timing of the robot's reaching motion, thereby determining the whole-body posture linkage parameters that meet preset requirements, including: The correction value of the freeze-switching judgment threshold in the next gait cycle is updated based on the evaluation results of the suppression of the trunk pitch oscillation amplitude. The correction value is then mapped to the phase offset compensation value in gait phase synchronization to determine the timing of the hand-reaching action. By combining the timing of the reaching motion and the moment of arm swing freeze, the overall motion time reduction and the degree to which the preset time optimization threshold are met are evaluated to determine the full-body posture linkage parameters.

[0009] This invention provides a full-body posture linkage device for a humanoid robot when reaching for an object, the device comprising: The state acquisition module is used to acquire the arm swing angular velocity, gait phase angle and azimuth deviation angle between the arm end and the target object in the humanoid robot walking state, and to determine the freeze switching judgment threshold based on the azimuth deviation angle and the arm swing angular velocity. The candidate point filtering module is used to filter a preliminary set of candidate points based on the comparison between the freeze switching determination threshold and the preset threshold. The freeze moment search module is used to obtain the shoulder joint torque mutation amount and end effector trajectory length of each candidate point in the preliminary candidate point set, calculate the comprehensive cost value of each candidate point based on the shoulder joint torque mutation amount and end effector trajectory length of each candidate point, and search for the swing arm freeze moment with the lowest comprehensive cost value. The linkage parameter determination module is used to obtain the torso posture data at the moment when the arm is frozen with the lowest comprehensive cost, evaluate the suppression result of the torso pitch oscillation amplitude, analyze the timing of the robot's reach action, and then determine the whole-body posture linkage parameters that meet the preset requirements.

[0010] Furthermore, the linkage parameter determination module includes: an oscillation suppression evaluation module, used to obtain trunk posture data at the arm freezing moment corresponding to the lowest comprehensive cost value, and to analyze the energy decay rate of trunk pitch oscillation amplitude using an oscillation evaluation mode based on Fourier series fitting, so as to evaluate the suppression result of trunk pitch oscillation amplitude. The action timing determination module is used to update the correction value of the freeze switching judgment threshold in the next gait cycle based on the evaluation result of the suppression result of the trunk pitch oscillation amplitude, and map the correction value to the phase offset compensation value in gait phase synchronization to determine the timing of the hand reaching action. The fusion module is used to fuse the timing of the reaching action and the freezing moment of the arm swing, evaluate the degree to which the overall action time reduction meets the preset time optimization threshold, and determine the full-body posture linkage parameters.

[0011] 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 coordinating the whole-body posture of a humanoid robot when reaching for an object. Addressing the complex business scenario where a humanoid robot needs to simultaneously optimize the timing of arm freezing, minimize torque mutations, constrain trajectories, and suppress torso oscillations during the reaching process, this invention collects arm angular velocity, gait phase angle, and azimuth deviation angle; calculates the deviation ratio and continuously optimizes the freezing switch threshold; combines gait phase analysis with a preliminary candidate point set; employs a cost evaluation algorithm to search for the arm freezing moment with the lowest overall cost; further, it evaluates the torso oscillation amplitude and energy attenuation rate based on Fourier series fitting to confirm the oscillation suppression effect and maps phase offset compensation values; finally, it fuses the action initiation timing and freezing moment to assess the degree of time optimization and determine the whole-body posture coordination parameters. This invention achieves synergistic optimization of minimizing torque mutations, satisfying trajectory constraints, and suppressing oscillations through multi-dimensional data fusion and optimization calculations, ensuring smooth and stable robot movements and improving overall motion efficiency and safety. Attached Figure Description

[0012] Figure 1 This is a flowchart of a method for coordinating the whole-body posture of a humanoid robot when reaching for an object, according to the present invention.

[0013] Figure 2 This is a schematic diagram of the structure of a full-body posture linkage device for a humanoid robot reaching for an object according to the present invention. Detailed Implementation

[0014] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0015] like Figure 1 This embodiment of a method for coordinating the body posture of a humanoid robot when reaching for an object may specifically include: S101. Obtain the arm swing angular velocity, gait phase angle and azimuth deviation angle between the arm end and the target object in the humanoid robot walking state, and determine the freeze switching judgment threshold based on the azimuth deviation angle and the arm swing angular velocity.

[0016] The arm swing angle signal output by the shoulder joint encoder in the current walking state of the humanoid robot is acquired. Time difference calculation is performed on the arm swing angle signal to obtain the arm swing angular velocity. Simultaneously, the gait phase angle output by the hip joint encoder is acquired. The gait phase angle is mapped to a period interval based on the temporal division of the single-leg support cycle and the double-leg support cycle. The azimuth deviation angle between the position coordinates of the arm end effector and the position coordinates of the target object's center of mass is extracted. The azimuth deviation angle is defined as the angle between the direction vector of the end effector pointing towards the target object and the current arm orientation vector. The ratio of the azimuth deviation angle and the arm swing angular velocity is calculated as a response sensitivity index. A quadratic cost function with the azimuth deviation angle as the state variable is constructed. The weight matrix of the quadratic cost function is set inversely proportional to the Euclidean distance from the end effector to the target object. A rolling temporal optimization method is used to solve for the minimum value of the cost function within the prediction temporal window to obtain the correction coefficient for the freeze-switching judgment condition. Based on the numerical distribution of the correction coefficient over multiple consecutive gait cycles, the statistical mean and standard deviation of the correction coefficient are extracted. If the standard deviation exceeds a preset fluctuation threshold, a smoothing process is applied to the correction coefficient to suppress abnormal fluctuations, resulting in a stabilized correction coefficient. The product of the stabilized correction coefficient and the response sensitivity index is mapped to a freeze-switching threshold space. The freeze-switching threshold is defined as the critical condition boundary for triggering the arm-freezing action, and the freeze-switching threshold for the current gait cycle is output.

[0017] When a humanoid robot reaches for an object while walking, the coordination between its arm swing motion and gait rhythm directly affects the stability of its overall posture. Regarding the determination of the freeze-to-freeze threshold, the following provides a specific implementation method that uses multi-source sensor data fusion and optimization algorithms to accurately determine the timing of arm freeze.

[0018] Specifically, the acquisition of the swing arm angular velocity relies on a high-precision encoder installed at the shoulder joint. The encoder outputs a discrete time sequence of the swing arm angle with a fixed sampling period. The angle values ​​at two adjacent sampling times are differentially calculated and then divided by the sampling period to obtain the swing arm angular velocity at the current moment.

[0019] For example, when a humanoid robot walks at a normal gait frequency, the arm swing angle exhibits an approximately sinusoidal variation pattern within a complete gait cycle, while the angular velocity shows a periodic fluctuation in the form of a cosine.

[0020] In one possible implementation, the gait phase angle is acquired via a hip joint encoder. The gait phase angle maps the temporal characteristics of the single-leg and double-leg support phases during walking to a complete cycle interval, ensuring a unified phase representation of the gait state at different walking speeds. The orientation deviation angle is extracted based on the spatial relationship between the end effector and the target object. It is obtained by calculating the angle between the direction vector of the end effector pointing towards the target object and the current arm orientation vector. This angle reflects the angular difference between the current arm posture and the ideal reaching direction.

[0021] It should be noted that the introduction of the response sensitivity index aims to establish a correlation between the azimuth deviation angle and the arm swing velocity. The response sensitivity index is defined as the ratio of the azimuth deviation angle to the arm swing velocity, and its physical meaning is to characterize the time scale required to eliminate a unit azimuth deviation under the current arm swing velocity condition. When the arm swing velocity is large and the azimuth deviation angle is small, the response sensitivity index value is low, indicating that the interference of initiating the reaching motion on the arm swing is small; conversely, when the arm swing velocity is small and the azimuth deviation angle is large, the response sensitivity index value is high, indicating that a longer adjustment time is required to complete the attitude switch. This index provides an important input parameter for subsequent cost function optimization.

[0022] In one embodiment, the quadratic cost function is constructed using the azimuth deviation angle as the core state variable. The cost function adopts a weighted quadratic form, wherein the weight matrix is ​​inversely proportional to the Euclidean distance from the end effector to the target object.

[0023] Specifically, when the end effector is far from the target object, the diagonal elements in the weight matrix have smaller values, allowing for larger orientation deviations without incurring excessive costs. As the end effector approaches the target object, the diagonal elements in the weight matrix increase accordingly, strengthening the penalty for orientation deviations and prompting the optimization algorithm to more actively search for control strategies to reduce deviations. This distance-adaptive weight setting mechanism allows the cost function to dynamically adjust the focus of the optimization objective according to the progress stage of the homing task, emphasizing motion stability at long distances and positioning accuracy at close distances.

[0024] Preferably, the rolling time-domain optimization method solves for the minimum value of the cost function within a prediction time-domain window. The prediction time-domain window covers a time range from the current moment to several future sampling periods. Within this window, the evolution trajectory of the azimuth deviation angle with the swing arm motion is predicted, and the correction coefficient that minimizes the integral value of the cost function is searched. The core of rolling optimization is that at each sampling moment, only the optimal correction coefficient for the current moment is executed, and then the prediction window is rolled forward by one sampling period, and the optimization calculation is re-executed based on the updated sensor data. This rolling update mechanism allows the correction coefficient to adapt to changes in the swing arm motion state and target position in real time, avoiding inaccurate judgments caused by environmental disturbances or target movement.

[0025] It is understandable that, due to the inherent speed fluctuation characteristics of arm swing motion within the gait cycle, the correction coefficients output by the rolling optimization exhibit a certain degree of fluctuation over time. By statistically analyzing the mean and standard deviation of the correction coefficients over multiple consecutive gait cycles, the stability of the correction coefficients can be assessed. If the standard deviation exceeds a preset fluctuation threshold, it indicates that the current optimization result is significantly affected by transient disturbances. In this case, a smoothing process is applied to the correction coefficient sequence to suppress high-frequency fluctuation components, resulting in more stable and reliable correction coefficient values. Based on the above processing, the freeze-switching decision threshold is determined by multiplying the stabilized correction coefficient by the response sensitivity index. The freeze-switching decision threshold is defined as the critical condition boundary for triggering the arm swing freeze action. When the real-time detected arm swing state parameters cross this boundary, a control command is immediately issued to freeze the arm swing motion, creating stable initial conditions for the subsequent reaching action. This dynamic determination of the threshold allows the freeze timing to adapt to the needs of full-body coordination control under different walking speeds and target distances.

[0026] S102. Based on the comparison between the freeze switching determination threshold and the preset threshold, a preliminary candidate point set is selected.

[0027] The freeze-switching judgment threshold is compared with a preset threshold, which is defined as the upper and lower boundaries of the phase angle that allows the arm swing freeze action to be triggered within the gait cycle. The over-limit amplitude is obtained by subtracting the upper or lower boundary of the preset threshold from the freeze-switching judgment threshold. The over-limit amplitude represents the difference between the current judgment threshold and the permissible phase range. Based on the correspondence between the over-limit amplitude and the current gait phase angle, phase moments with over-limit amplitudes within the preset tolerance threshold range are selected on the gait cycle time axis. The selected phase moments are marked as preliminary candidate points, and the corresponding gait phase angle value is recorded for each preliminary candidate point. Based on the gait phase angle values ​​of the preliminary candidate points, it is determined whether each preliminary candidate point is located within the support phase interval where both legs are on the ground simultaneously in the gait cycle. If the gait phase angle value of the preliminary candidate point is within the double-leg support phase interval, the candidate point is retained; if it is within the single-leg support phase interval, it is discarded. The retained candidate points are arranged in chronological order according to the gait phase angle values ​​to obtain the preliminary candidate point set.

[0028] In determining the initial candidate points for the arm-freezing moment, the comparison between the freeze switching judgment threshold and the preset gait phase threshold forms the basis for candidate point selection. The preset gait phase threshold includes two values: an upper boundary and a lower boundary, which correspond to the maximum and minimum phase angles allowed to trigger the arm-freezing action within the gait cycle, respectively.

[0029] Specifically, the over-limit magnitude is calculated using a difference operation. When the freeze-switching judgment threshold is greater than the upper boundary of the preset gait phase threshold, the over-limit magnitude is the difference between the two; when the freeze-switching judgment threshold is less than the lower boundary of the preset gait phase threshold, the over-limit magnitude is the difference between the lower boundary and the judgment threshold. This over-limit magnitude directly reflects the degree to which the current judgment threshold deviates from the permissible phase range.

[0030] In one possible implementation, the initial candidate point selection is based on a comparison between the out-of-limit amplitude and a preset tolerance threshold. The preset tolerance threshold is defined as the maximum allowable out-of-limit amplitude boundary. When the out-of-limit amplitude value corresponding to a certain phase moment is within the range of the preset tolerance threshold, that phase moment is marked as an initial candidate point.

[0031] For example, when a humanoid robot is walking smoothly, there are multiple phase moments within the gait cycle that meet the tolerance conditions for exceeding the limit. These moments are all included in the scope of preliminary candidate points. Each preliminary candidate point records its corresponding gait phase angle value to facilitate subsequent interval assignment judgment.

[0032] It should be noted that the two-legged support phase interval and the one-legged support phase interval in the gait cycle have different mechanical characteristics. The two-legged support phase interval refers to the period when both legs of the humanoid robot are in contact with the ground simultaneously. At this time, the robot's support base area is large, and the overall stability is high, making it suitable for performing arm-freezing movements. The one-legged support phase interval refers to the period when only one leg is in contact with the ground. At this time, the robot is in a dynamic equilibrium state, and its resistance to external disturbances is relatively weak. Based on these characteristics, the retention and elimination of preliminary candidate points are determined according to the interval type to which their gait phase angle values ​​belong. The candidate points retained after screening are arranged chronologically according to their gait phase angle values ​​from smallest to largest, forming an ordered preliminary candidate point set. Each element in this set corresponds to a moment in the two-legged support phase interval within the gait cycle that meets the tolerance condition for exceeding the limit amplitude.

[0033] S103. Obtain the shoulder joint torque mutation amount and end effector trajectory length of each candidate point in the preliminary candidate point set. Calculate the comprehensive cost value of each candidate point based on the shoulder joint torque mutation amount and end effector trajectory length of each candidate point, and search for the arm freezing time with the lowest comprehensive cost value.

[0034] The gait phase angle value corresponding to each candidate point in the preliminary candidate point set is obtained. For each candidate point, the driving torque value of the shoulder joint at that phase moment is extracted, and the driving torque value of the shoulder joint in the previous sampling period at that phase moment is also obtained. The difference between the driving torque value at the current moment and the driving torque value at the previous moment is used to obtain the change in shoulder joint torque corresponding to the candidate point. Based on the gait phase angle value corresponding to each candidate point, the position coordinates of the end effector in three-dimensional space and the position coordinates of the centroid of the target object at that phase moment are obtained. The straight-line distance between the position coordinates of the end effector and the position coordinates of the centroid of the target object is calculated to obtain the length of the bypass trajectory corresponding to the candidate point. For each candidate point in the preliminary candidate point set, the change in shoulder joint torque is multiplied by a preset torque weighting coefficient, and the length of the bypass trajectory is multiplied by a preset trajectory weighting coefficient. After maximum and minimum normalization, the products of the two are summed to obtain the comprehensive cost of the candidate point. The torque weighting coefficient and the trajectory weighting coefficient are pre-configured and their sum is a constant. The comprehensive cost value of all candidate points in the preliminary candidate point set is traversed, and the comprehensive cost value of each candidate point is compared one by one. The candidate point with the lowest comprehensive cost value is selected, and the gait phase angle value corresponding to the candidate point is determined as the swing arm freezing time.

[0035] After constructing the initial candidate point set, a comprehensive cost evaluation for each candidate point is the key step in determining the arm-freezing moment. This comprehensive cost evaluation quantifies the candidate points from two dimensions: the sudden change in shoulder joint torque and the length of the trajectories, in order to select the freezing moment that simultaneously considers mechanical stability and carrying efficiency.

[0036] Specifically, the acquisition of shoulder joint torque abrupt change is achieved based on temporal difference operations. For each candidate point in the initial candidate point set, the phase time is located according to its corresponding gait phase angle value. The driving torque value at the current moment is extracted from the torque feedback signal of the shoulder joint actuator, while simultaneously backtracking one sampling period to obtain the driving torque value at the previous moment. The driving torque value at the current moment is subtracted from the driving torque value at the previous moment, and the difference is the shoulder joint torque abrupt change corresponding to that candidate point.

[0037] It should be noted that the sudden change in shoulder joint torque reflects the degree of mechanical impact experienced by the shoulder joint at the moment the arm is frozen. During the walking process of a humanoid robot, the arm swings naturally with the rhythm of the gait, and the driving torque of the shoulder joint exhibits periodic changes. If the arm is frozen at a moment of drastic torque change, the shoulder joint actuator must abruptly change its output torque to counteract the arm's inertia. This sudden change will be transmitted through the torso to the whole body, affecting the overall posture stability.

[0038] In one possible implementation, the length of the detour trajectory is calculated using the three-dimensional Euclidean distance formula. The three-dimensional position coordinates of the end effector and the centroid of the target object at the corresponding phase of the candidate point are obtained. The coordinate differences between the two points along each coordinate axis are calculated, and the squares of these differences in each of the three directions are summed. The square root of the sum is then taken, and the resulting value is the length of the detour trajectory corresponding to that candidate point. This length represents the spatial distance that the end effector needs to traverse from its frozen position to reach the target object.

[0039] In one embodiment, the comprehensive cost value is calculated using a weighted summation method, unifying the shoulder joint torque mutation and the detour trajectory length into a single evaluation index. The torque weighting coefficient is used to adjust the proportion of the shoulder joint torque mutation in the comprehensive cost, and the trajectory weighting coefficient is used to adjust the proportion of the detour trajectory length in the comprehensive cost. Multiplying the shoulder joint torque mutation by the torque weighting coefficient yields the torque cost component, and multiplying the detour trajectory length by the trajectory weighting coefficient yields the trajectory cost component. Summing the two cost components gives the comprehensive cost value of the candidate point. Since the sum of the torque weighting coefficient and the trajectory weighting coefficient is a constant, the two weighting coefficients exhibit an inverse relationship; by adjusting the weight ratio, different balance points can be achieved between mechanical stability and material handling efficiency.

[0040] Preferably, the specific values ​​of the torque weighting coefficient and the trajectory weighting coefficient can be pre-configured according to the motion state of the humanoid robot and the urgency of the task.

[0041] For example, when the humanoid robot is walking rapidly, the gait cycle shortens, the arm swing speed increases, and the impact of sudden torque changes on posture stability becomes more significant. In this case, the torque weighting coefficient is configured to a larger value, making the overall cost assessment more focused on mechanical stability. When the target object is close to the humanoid robot or the target object is moving, the time urgency of the retrieval task increases. In this case, the trajectory weighting coefficient is configured to a larger value, making the overall cost assessment more focused on shortening the detour trajectory length to accelerate the retrieval response speed.

[0042] Understandably, the selection of the arm-freezing moment is accomplished through a traversal comparison method. The comprehensive cost value of each candidate point in the initial candidate point set is read sequentially. The comprehensive cost value of the current candidate point is compared with the recorded minimum comprehensive cost value. If the comprehensive cost value of the current candidate point is less than the recorded minimum value, the minimum value record is updated, and the current candidate point is marked as the optimal candidate point. After completing the traversal comparison of all candidate points, the gait phase angle value corresponding to the optimal candidate point is determined as the arm-freezing moment. The shoulder joint torque mutation amount and the circumduction trajectory length corresponding to this moment achieve an overall balance in the comprehensive cost evaluation, avoiding trunk swaying caused by excessive torque mutation while also considering the spatial efficiency of the end effector reaching the target object.

[0043] S104. Obtain the torso posture data at the moment of arm freeze corresponding to the lowest comprehensive cost, and use the oscillation evaluation mode based on Fourier series fitting to analyze the energy decay rate of the torso pitch oscillation amplitude in order to evaluate the suppression result of the torso pitch oscillation amplitude.

[0044] The torso posture data corresponding to the moment the arm freezes is acquired. This torso posture data is output by an inertial measurement unit (IMU) installed on the torso of the humanoid robot. Angle data of the torso in the pitch direction are collected within a preset time window starting from the moment the arm freezes, resulting in a torso pitch angle sequence. The torso pitch angle sequence is decomposed in the frequency domain using a Fourier series fitting method. This method decomposes the torso pitch angle sequence into a superposition of a fundamental frequency component and multiple harmonic components. The oscillation amplitude corresponding to the fundamental frequency component is extracted and determined as the torso pitch oscillation amplitude. The torso pitch oscillation amplitude is compared with a preset oscillation threshold. If the torso pitch oscillation amplitude is lower than the preset oscillation threshold, the peak points of two adjacent oscillation cycles are identified within the preset time window. The difference between the peak value of the later cycle and the peak value of the previous cycle is calculated. This difference is divided by the peak value of the previous cycle to obtain the energy decay rate. The oscillation suppression result is judged based on the energy decay rate. If the energy decay rate exceeds a preset decay threshold, the oscillation suppression result is confirmed as passed, and an oscillation suppression confirmation flag is output.

[0045] After determining the moment of arm swing freeze, oscillation assessment of the resulting trunk posture changes is a crucial step in verifying the rationality of the freeze timing. Quantitative analysis of the trunk pitch-to-toe oscillation characteristics determines the degree of impact of the arm swing freeze on overall stability.

[0046] Specifically, the acquisition of torso posture data relies on an inertial measurement unit (IMU) installed on the torso of the humanoid robot. The IMU includes a three-axis accelerometer and a three-axis gyroscope, and the angle data of the torso in the pitch direction is obtained by integrating the angular velocity signals from the gyroscopes. Starting from the moment the arm freezes, torso pitch angle data within a preset time window are continuously acquired at a fixed sampling frequency, forming a torso pitch angle sequence.

[0047] In one possible implementation, the Fourier series fitting method is based on the frequency domain decomposition principle of periodic signals. This method treats the torso pitch angle sequence as a periodic oscillating signal and decomposes the time-domain signal into a superposition of sinusoidal waves of different frequencies through mathematical transformation. The fundamental frequency component corresponds to the main periodic characteristics of the oscillating signal, and its frequency matches the basic rhythm of the torso pitch oscillation. The harmonic components correspond to high-frequency oscillations that are integer multiples of the fundamental frequency, reflecting the detailed features of the oscillation waveform. The specific process of Fourier series fitting includes calculating the spectrum of discrete sampling points of the torso pitch angle sequence, identifying the frequency component with the largest amplitude in the spectrum as the fundamental frequency, and extracting the amplitude coefficient corresponding to this fundamental frequency as the oscillation amplitude of the fundamental frequency component.

[0048] It should be noted that the oscillation amplitude of the fundamental frequency component directly reflects the main intensity of the torso pitch oscillation. In the Fourier series decomposition results, the fundamental frequency component usually accounts for the main share of the total oscillation energy, and its amplitude value can characterize the overall intensity of the torso pitch oscillation. The amplitude value of the fundamental frequency component is determined as the torso pitch oscillation amplitude, serving as the core indicator for subsequent oscillation evaluation. Furthermore, the torso pitch oscillation amplitude is numerically compared with a preset oscillation threshold. The preset oscillation threshold is pre-configured according to the structural parameters and stability requirements of the humanoid robot, representing the acceptable upper limit of the torso pitch oscillation.

[0049] In one embodiment, the identification of the peak value of the oscillation cycle is achieved through a time-series traversal. Within the torso pitch angle sequence covered by the preset time window, the points where the angle value changes from rising to falling are detected sequentially, and the angle value corresponding to these points is marked as the peak value of the oscillation cycle. By identifying the peak values ​​of two adjacent oscillation cycles, the specific values ​​of the peak value of the previous cycle and the peak value of the next cycle are obtained.

[0050] Preferably, the energy decay rate is calculated based on the relationship between the peak values ​​of adjacent cycles. The energy decay rate is defined as the relative attenuation of the oscillation amplitude between adjacent cycles. It is calculated by dividing the difference between the peak value of the subsequent cycle and the peak value of the preceding cycle by the peak value of the preceding cycle. Since the peak value of the subsequent cycle is smaller than the peak value of the preceding cycle during normal oscillation decay, the energy decay rate is negative; a larger absolute value indicates faster oscillation decay. This decay rate reflects the energy dissipation rate of trunk pitch oscillations during natural evolution.

[0051] Understandably, the oscillation suppression result is determined by comparing the energy decay rate with a preset decay threshold. The preset decay threshold represents the minimum required convergence speed for the oscillation. When the absolute value of the energy decay rate exceeds the preset decay threshold, it indicates that the torso pitch oscillation shows a rapid convergence trend within a preset time window, and the oscillation suppression result is confirmed as passed. The output oscillation suppression confirmation flag indicates that the selected arm freeze moment possesses good stability characteristics, and the resulting torso pitch oscillation can naturally decay to an acceptable range within a short time, without causing continuous interference to the humanoid robot's subsequent reaching movements.

[0052] S105. Based on the evaluation results of the suppression results of the torso pitch oscillation amplitude, update the correction value of the freeze switching determination threshold in the next gait cycle, map the correction value to the phase offset compensation value in gait phase synchronization, and determine the timing of the hand reaching action.

[0053] The oscillation suppression confirmation flag is obtained. If the oscillation suppression confirmation flag indicates a pass, the energy attenuation rate is subtracted from a preset attenuation threshold. The subsequent adjustment amount is obtained by multiplying this difference by a preset proportional coefficient. The subsequent adjustment amount is defined as a correction value for the freeze-switching judgment threshold within the next gait cycle; the larger the difference, the smaller the subsequent adjustment amount. Based on the ratio of the subsequent adjustment amount to the current gait cycle duration, the subsequent adjustment amount is converted into gait phase angle units to obtain a phase offset compensation value. The phase offset compensation value represents the offset angle on the gait phase axis relative to the arm-freezing time at the start of the reach action. The phase offset compensation value is added to the gait phase angle value corresponding to the arm-freezing time to obtain the reach action start timing. The reach action start timing is expressed in gait phase angle values, calibrating the phase position at which the humanoid robot triggers the reach action within the current gait cycle.

[0054] After confirming the oscillation suppression results, dynamically adjusting the freeze switching judgment conditions based on the confirmation results is a key step in achieving precise initiation of the reach action. The oscillation suppression confirmation flag serves as the trigger condition for updating the adjustment amount, and its status indicates that the selection of the current arm freeze moment has good stability characteristics.

[0055] Specifically, the subsequent adjustment amount is calculated based on the difference between the energy decay rate and the preset decay threshold. This difference reflects the extent to which the actual oscillation decay rate exceeds the minimum requirement; a larger difference indicates a better oscillation suppression effect. In this case, the freeze-switching threshold for the next state cycle does not need significant correction, so the subsequent adjustment amount is taken as a smaller value. The preset proportional coefficient serves to convert the difference scale into a threshold correction scale; this coefficient is pre-configured based on the humanoid robot's structural stiffness and walking speed range.

[0056] For example, when the energy decay rate significantly exceeds the preset decay threshold, it indicates that the selection of the current freeze time is too conservative, and the subsequent adjustment amount takes a smaller positive value, so that the judgment threshold of the next cycle is slightly relaxed; when the energy decay rate only slightly exceeds the preset decay threshold, the subsequent adjustment amount takes a larger positive value, indicating that the freeze time should be selected more cautiously in the next cycle.

[0057] In one possible implementation, the phase shift compensation value is obtained through a linear mapping from a time scale to a phase angle scale. The subsequent adjustment amount is expressed in time units, while the gait phase angle is expressed in angle units, and the conversion between the two is established based on the duration of the current gait cycle. The subsequent adjustment amount is divided by the duration of the current gait cycle, and the resulting ratio is multiplied by the phase angle range corresponding to a complete cycle to obtain the phase shift compensation value.

[0058] It should be noted that the final determination of the hand-reaching action initiation timing is achieved through phase superposition. The phase offset compensation value is added to the gait phase angle value corresponding to the arm-swinging freeze moment, and the superposition result is the hand-reaching action initiation timing. This initiation timing is expressed in the form of a gait phase angle value, which marks the phase position at which the humanoid robot triggers the hand-reaching action within the current gait cycle, ensuring that the initiation of the hand-reaching action is synchronized and coordinated with the walking gait rhythm.

[0059] S106. By integrating the timing of the reaching action and the freezing moment of the arm swing, the overall action time reduction and the degree of achievement of the preset time optimization threshold are evaluated to determine the full-body posture linkage parameters.

[0060] The timing of the reach-out action and the moment the arm swing is frozen are obtained. The phase difference between the two on the gait phase axis is calculated. The phase difference is converted into a time interval based on the current gait cycle duration. This time interval is defined as the overall action time, representing the total time taken from arm swing freezing to the start of the reach-out action. The overall action time is subtracted from a preset baseline action time to obtain the overall action time reduction. The preset baseline action time is pre-configured based on the humanoid robot's structural parameters and the target reach distance. The overall action time reduction is compared with a preset time optimization threshold. If the overall action time reduction reaches or exceeds the preset time optimization threshold, the level of compliance is determined to be qualified. Based on the qualified level of compliance, the gait phase angle value corresponding to the arm swing freezing moment, the sudden change in shoulder joint torque, and the length of the circumferential trajectory are combined and output as whole-body posture linkage parameters. These whole-body posture linkage parameters satisfy the dual conditions of minimizing sudden torque change and constraining the circumferential trajectory length.

[0061] After determining the initiation timing of the reaching motion, integrating and evaluating this initiation timing with the arm swing freeze moment is a crucial step in verifying the effectiveness of the whole-body posture coordination control. Through quantitative analysis of the overall motion time, it is determined whether the current freeze moment parameter has achieved the expected time optimization target.

[0062] Specifically, both the timing of the reach initiation and the moment of arm swing freeze are expressed in the form of gait phase angle values. The difference between the two on the gait phase axis directly reflects the phase interval between arm swing freeze and reach initiation. By dividing this phase difference by the phase angle range corresponding to a complete gait cycle and then multiplying it by the duration of the current gait cycle, the phase difference can be converted into the overall movement time expressed in time units.

[0063] In one possible implementation, the assessment of the overall motion time reduction is based on a preset baseline motion time. This preset baseline motion time is pre-configured according to the humanoid robot's structural parameters and the target reach distance, representing the standard time required to complete the arm freeze and reach initiation without any optimization. The difference between the overall motion time and the preset baseline motion time is the overall motion time reduction.

[0064] For example, when a humanoid robot walks quickly and the target object is close, the preset baseline motion time is shortened accordingly, and the overall motion time reduction requirement is increased. The preset time optimization threshold is defined as the minimum acceptable value for the overall motion time reduction. When the reduction reaches or exceeds this threshold, the level of compliance is considered acceptable.

[0065] It should be noted that the output of the whole-body posture linkage parameters integrates multiple key indicators into a unified parameter set. This parameter set includes the gait phase angle value corresponding to the arm-swing freeze moment, the sudden change in shoulder joint torque at that moment, and the corresponding path length. These three indicators characterize the freeze moment from three dimensions: temporal localization, mechanical stability, and spatial efficiency. The whole-body posture linkage parameters simultaneously satisfy the dual conditions of minimizing torque change and constraining path length, enabling the humanoid robot to maintain trunk stability while reaching for objects during walking, and to reach the target object with a shorter movement path.

[0066] like Figure 2 This invention provides a full-body posture linkage device for a humanoid robot when reaching for an object, mainly comprising: The state acquisition module is used to acquire the arm swing angular velocity, gait phase angle and azimuth deviation angle between the arm end and the target object in the humanoid robot walking state, and to determine the freeze switching judgment threshold based on the azimuth deviation angle and the arm swing angular velocity. The candidate point filtering module is used to filter a preliminary set of candidate points based on the comparison between the freeze switching determination threshold and the preset threshold. The freeze moment search module is used to obtain the shoulder joint torque mutation amount and end effector trajectory length of each candidate point in the preliminary candidate point set, calculate the comprehensive cost value of each candidate point based on the shoulder joint torque mutation amount and end effector trajectory length of each candidate point, and search for the swing arm freeze moment with the lowest comprehensive cost value. The linkage parameter determination module is used to obtain the torso posture data at the moment when the arm is frozen with the lowest comprehensive cost, evaluate the suppression result of the torso pitch oscillation amplitude, analyze the timing of the robot's reach action, and then determine the whole-body posture linkage parameters that meet the preset requirements.

[0067] Furthermore, the linkage parameter determination module includes: an oscillation suppression evaluation module, used to obtain trunk posture data at the arm freezing moment corresponding to the lowest comprehensive cost value, and to analyze the energy decay rate of trunk pitch oscillation amplitude using an oscillation evaluation mode based on Fourier series fitting, so as to evaluate the suppression result of trunk pitch oscillation amplitude. The action timing determination module is used to update the correction value of the freeze switching judgment threshold in the next gait cycle based on the evaluation result of the suppression result of the trunk pitch oscillation amplitude, and map the correction value to the phase offset compensation value in gait phase synchronization to determine the timing of the hand reaching action. The fusion module is used to fuse the timing of the reaching action and the freezing moment of the arm swing, evaluate the degree to which the overall action time reduction meets the preset time optimization threshold, and determine the full-body posture linkage parameters.

[0068] If the technical solution of this application involves personal information, the product using this technical solution has clearly informed the user of the personal information processing rules and obtained the user's voluntary consent before processing the personal information. If the technical solution of this application involves sensitive personal information, the product using this technical solution has obtained the user's separate consent before processing the sensitive personal information, and also meets the requirement of "express consent". For example, at personal information collection devices such as cameras, clear and prominent signs are set up to inform users that they have entered the scope of personal information collection and that personal information will be collected. If an individual voluntarily enters the collection scope, it is deemed that they have agreed to the collection of their personal information; or on the personal information processing device, the personal information processing rules are clearly informed through signs / information, and authorization is obtained through pop-up information or by asking the individual to upload their personal information; wherein, the personal information processing rules may include information such as the personal information processor, the purpose of personal information processing, the processing method, and the types of personal information processed.

[0069] The above description is only a preferred embodiment of the present invention. It should be noted that those skilled in the art can make several improvements and additions without departing from the principle of the present invention, and these improvements and additions should also be considered within the scope of protection of the present invention.

Claims

1. A method for coordinating the body posture of a humanoid robot when reaching for an object, characterized in that, The method includes: The arm swing angular velocity, gait phase angle, and orientation deviation angle between the arm end and the target object are obtained in the walking state of the humanoid robot. The freeze switching judgment threshold is determined based on the orientation deviation angle and the arm swing angular velocity. Based on the comparison between the freeze switching determination threshold and the preset threshold, a preliminary candidate point set is selected; Obtain the shoulder joint torque mutation amount and end effector trajectory length of each candidate point in the preliminary candidate point set. Calculate the comprehensive cost value of each candidate point based on the shoulder joint torque mutation amount and end effector trajectory length of each candidate point, and search for the arm freezing time with the lowest comprehensive cost value. Obtain the torso posture data at the moment of arm freeze corresponding to the lowest comprehensive cost, evaluate the suppression result of torso pitch oscillation amplitude, analyze the timing of robot's reach action, and then determine the whole body posture linkage parameters that meet the preset requirements. The process of acquiring the arm swing angular velocity, gait phase angle, and azimuth deviation angle between the arm end and the target object during the humanoid robot's walking state, and determining the freeze switching threshold based on the azimuth deviation angle and the arm swing angular velocity, includes: The arm angle signal is obtained by a shoulder joint encoder, and the arm angular velocity is calculated. The gait phase angle is simultaneously acquired via a hip joint encoder and mapped to a gait cycle interval; Extract the azimuth deviation angle between the end effector of the arm and the centroid of the target object; The ratio of the azimuth deviation angle to the swing arm angular velocity is used as a response sensitivity index. A quadratic cost function with the azimuth deviation angle as the variable is constructed. The minimum value is obtained by rolling time domain optimization to obtain the correction coefficient. The stabilized correction coefficient is obtained by smoothing the statistical mean and standard deviation of the correction coefficient within several gait cycle intervals. The freeze switching threshold is determined based on the stabilization correction coefficient and the response sensitivity index.

2. The method for coordinated body posture of a humanoid robot when reaching for an object, as described in claim 1, is characterized in that... The step of filtering a preliminary candidate point set based on the comparison between the freeze switching determination threshold and a preset threshold includes: Obtain the freeze switching judgment threshold, compare it with the preset threshold, and calculate the extent of exceeding the limit; Based on the correspondence between the overshoot amplitude and the gait phase angle, phase moments are screened on the gait cycle time axis and marked as preliminary candidate points; Based on the gait phase angle values ​​of the preliminary candidate points, determine whether they are located in the two-leg support phase interval. If they are located in the two-leg support phase interval, they are retained; if they are located in the single-leg support phase interval, they are discarded. The retained candidate points are arranged in chronological order according to the gait phase angle values ​​to obtain the preliminary candidate point set.

3. The method for coordinated body posture of a humanoid robot when reaching for an object according to claim 1, characterized in that, The process of obtaining the shoulder joint torque mutation amount and end effector trajectory length of each candidate point in the preliminary candidate point set, calculating the comprehensive cost value of each candidate point based on the shoulder joint torque mutation amount and end effector trajectory length, and searching for the arm freezing moment with the lowest comprehensive cost value includes: Obtain the gait phase angle value of each candidate point in the preliminary candidate point set, extract the shoulder joint driving torque value at the corresponding phase moment, calculate the difference with the previous moment, and obtain the shoulder joint torque mutation amount. Obtain the straight-line distance between the end effector of the arm and the centroid of the target object to get the length of the bypass trajectory; The comprehensive cost is obtained by summing the products of the sudden change in shoulder joint torque and the preset torque weighting coefficient, and the products of the detour trajectory length and the preset trajectory weighting coefficient after normalization. Traverse the initial candidate point set, compare the comprehensive cost value of each candidate point, select the candidate point with the lowest cost value, and determine the swing arm freezing time.

4. The method for coordinated body posture of a humanoid robot when reaching for an object, as described in claim 1, is characterized in that... The search for the swing arm freeze moment with the lowest overall cost includes: Obtain the torso posture data corresponding to the moment the swing arm is frozen, and collect the pitch angle sequence within a preset time window; The pitch angle sequence was decomposed in the frequency domain using the Fourier series fitting method to extract the oscillation amplitude of the fundamental frequency component, which was then determined as the torso pitch oscillation amplitude. The amplitude of the torso pitch oscillation is compared with a preset oscillation threshold. If it is lower than the preset oscillation threshold, the peak difference between adjacent oscillation cycles is calculated to obtain the energy decay rate.

5. A method for coordinating the whole-body posture of a humanoid robot when reaching for an object, as described in claim 1, is characterized in that... The evaluation of the suppression results of trunk pitch oscillation amplitude is used to analyze the timing of the robot's reach motion initiation, and then to determine the whole-body posture linkage parameters that meet preset requirements, including: The correction value of the freeze-switching judgment threshold in the next gait cycle is updated based on the evaluation results of the suppression of the trunk pitch oscillation amplitude. The correction value is then mapped to the phase offset compensation value in gait phase synchronization to determine the timing of the hand-reaching action. By combining the timing of the reaching motion and the moment of arm swing freeze, the overall motion time reduction and the degree to which the preset time optimization threshold are met are evaluated to determine the full-body posture linkage parameters.

6. A full-body posture linkage device for a humanoid robot reaching for an object, characterized in that, The device includes: The state acquisition module is used to acquire the arm swing angular velocity, gait phase angle and azimuth deviation angle between the arm end and the target object in the humanoid robot walking state, and to determine the freeze switching judgment threshold based on the azimuth deviation angle and the arm swing angular velocity. The candidate point filtering module is used to filter a preliminary set of candidate points based on the comparison between the freeze switching determination threshold and the preset threshold. The freeze moment search module is used to obtain the shoulder joint torque mutation amount and end effector trajectory length of each candidate point in the preliminary candidate point set, calculate the comprehensive cost value of each candidate point based on the shoulder joint torque mutation amount and end effector trajectory length of each candidate point, and search for the swing arm freeze moment with the lowest comprehensive cost value. The linkage parameter determination module is used to obtain the torso posture data at the arm freezing moment corresponding to the lowest comprehensive cost value, evaluate the suppression result of torso pitch oscillation amplitude, analyze the timing of the robot's reach action, and then determine the whole body posture linkage parameters that meet the preset requirements. The process of acquiring the arm swing angular velocity, gait phase angle, and azimuth deviation angle between the arm end and the target object during the humanoid robot's walking state, and determining the freeze switching threshold based on the azimuth deviation angle and the arm swing angular velocity, includes: The arm angle signal is obtained by a shoulder joint encoder, and the arm angular velocity is calculated. The gait phase angle is simultaneously acquired via a hip joint encoder and mapped to a gait cycle interval; Extract the azimuth deviation angle between the end effector of the arm and the centroid of the target object; The ratio of the azimuth deviation angle to the swing arm angular velocity is used as a response sensitivity index. A quadratic cost function with the azimuth deviation angle as the variable is constructed. The minimum value is obtained by rolling time domain optimization to obtain the correction coefficient. The stabilized correction coefficient is obtained by smoothing the statistical mean and standard deviation of the correction coefficient within several gait cycle intervals. The freeze switching threshold is determined based on the stabilization correction coefficient and the response sensitivity index.

7. The whole-body posture linkage device for a humanoid robot reaching for an object according to claim 6, characterized in that, The linkage parameter determination module includes: The oscillation suppression evaluation module is used to obtain the trunk posture data at the moment of arm freeze corresponding to the lowest comprehensive cost value, and to analyze the energy decay rate of the trunk pitch oscillation amplitude using an oscillation evaluation mode based on Fourier series fitting, so as to evaluate the suppression result of the trunk pitch oscillation amplitude. The action timing determination module is used to update the correction value of the freeze switching judgment threshold in the next gait cycle based on the evaluation result of the suppression result of the trunk pitch oscillation amplitude, and map the correction value to the phase offset compensation value in gait phase synchronization to determine the timing of the hand reaching action. The fusion module is used to fuse the timing of the reaching action and the freezing moment of the arm swing, evaluate the degree to which the overall action time reduction meets the preset time optimization threshold, and determine the full-body posture linkage parameters.

8. A full-body posture linkage device for a humanoid robot reaching for an object according to claim 6, characterized in that, The step of filtering a preliminary candidate point set based on the comparison between the freeze switching determination threshold and a preset threshold includes: Obtain the freeze switching judgment threshold, compare it with the preset threshold, and calculate the extent of exceeding the limit; Based on the correspondence between the overshoot amplitude and the gait phase angle, phase moments are screened on the gait cycle time axis and marked as preliminary candidate points; Based on the gait phase angle values ​​of the preliminary candidate points, determine whether they are located in the two-leg support phase interval. If they are located in the two-leg support phase interval, they are retained; if they are located in the single-leg support phase interval, they are discarded. The retained candidate points are arranged in chronological order according to the gait phase angle values ​​to obtain the preliminary candidate point set.

Images