A terrain-adaptive attitude stabilization control method and system for quadruped robots

By collecting multi-source vibration signals and body state data, the system identifies terrain changes and adjusts the foot force of the quadruped robot to generate joint drive commands. This solves the problem of not being able to predict terrain changes in advance in existing technologies, achieves real-time and accurate posture stability control, and improves the robot's terrain adaptability.

CN123067526APending Publication Date: 2026-11-10BEIJING XIARUI TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611163893.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-03
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing technologies cannot predict and quantify sudden changes in terrain in advance, resulting in low accuracy of posture stability control for quadruped robots when terrain changes occur.

Method used

By collecting the time spectrum diagram of multi-source vibration signals and the body state data, the system identifies terrain change events, calculates the disturbed dynamic moment vector, predicts the attitude deviation trend, generates the expected attitude correction, and adjusts the foot end force of the supporting foot through the resonance suppression reference value to generate joint drive commands and achieve attitude stability control.

Benefits of technology

It enables real-time quantification of terrain changes and advance prediction of attitude deviations, improving attitude stability and walking smoothness in complex unstructured terrain, reducing terrain-induced body resonance, and ensuring the robot's long-term walking stability.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123067526A_ABST
    Figure CN123067526A_ABST
Patent Text Reader

Abstract

This invention relates to the field of robot control technology, and discloses a terrain-adaptive attitude stabilization control method and system for a quadruped robot. The method includes: acquiring time-spectrum diagrams of multi-source vibration signals and body state data; identifying terrain change events based on the direction of sudden changes in foot force vectors; calculating the terrain disturbance vector from the data before the change; predicting attitude offset to generate a feedforward attitude correction expectation, and comparing it with the body wave sequence and frequency band energy sequence; using the feedforward attitude correction expectation and resonance suppression reference quantity to correct the expected foot force of the supporting foot; calculating the joint drive quantity based on the foot force correction quantity and position compensation quantity; generating joint drive commands and acquiring the adjusted attitude and foot force vector; comparing the adjusted attitude with the expected attitude to obtain the attitude tracking deviation and performing feedback updates. This invention solves the problem of low attitude stabilization control accuracy caused by the inability to predict and quantify terrain changes in advance.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of robot control technology, and in particular to a terrain-adaptive attitude stabilization control method and system for a quadruped robot. Background Technology

[0002] Legged robots have a superior ability to cope with complex terrain compared to wheeled robots. However, when walking on unstructured terrain (such as sand, grass, ice, gravel, etc.), the foot-ground contact state is complex and variable. Abrupt terrain changes (such as from a hard road surface to soft sand or smooth ice) can cause problems such as violent fluctuations in body posture, excessive impact on the feet, or even instability and tipping over.

[0003] In existing technologies, some solutions rely on a six-dimensional force sensor at the foot to detect changes in the foot force vector to identify abrupt terrain changes, and then estimate the equivalent perturbation torque through a perturbation observation model to generate a feedforward attitude correction expectation. These methods respond quickly to abrupt events, but they are difficult to predict the terrain material properties in advance and cannot proactively adjust the compliance parameters of foot-ground contact before the abrupt change occurs.

[0004] Another approach uses foot accelerometers and contact microphones to collect broadband vibration and acoustic signals generated by foot-to-ground contact. These signals are then used to generate a time-spectrum graph and a neural network classifier to identify the terrain material type. This method can know the terrain properties in advance and pre-adjust impedance parameters, but it lacks real-time quantitative estimation of the impact force and attitude disturbances caused by sudden terrain changes. Furthermore, its resonance suppression relies on the statistical regularity of historical gait cycles, limiting its adaptability to sudden, non-periodic terrain changes. Summary of the Invention

[0005] This invention provides a terrain-adaptive attitude stabilization control method and system for quadruped robots. Its main purpose is to solve the problem of low attitude stabilization control accuracy caused by the inability to predict and quantify sudden changes in terrain in advance.

[0006] To achieve the above objectives, the present invention provides a terrain-adaptive attitude stabilization control method for a quadruped robot, comprising: S1: Collect the time-spectrum diagram and body state data corresponding to the multi-source vibration signals of the quadruped robot in the current terrain, and identify terrain change events based on the change direction of the force vector at the foot end of the quadruped robot; S2: Based on the body state data before the terrain change event, calculate the disturbed dynamic torque vector of the quadruped robot to obtain the terrain disturbance vector of the quadruped robot; S3: Based on the terrain disturbance vector and the body state data, predict the attitude deviation trend of the quadruped robot, generate the feedforward attitude correction expectation, and based on each gait cycle of the quadruped robot, compare the body wave sequence in the body state data with the frequency band energy sequence of the time spectrum to obtain the resonance suppression reference value of the quadruped robot. S4: Using the feedforward attitude correction expectation and the resonance suppression reference, correct the expected foot end force of the supporting foot in the quadruped robot to obtain the foot end force correction amount, and calculate the target position compensation amount of the supporting foot based on the current position of the supporting foot. S5: Based on the foot force correction amount and the target position compensation amount, calculate the expected joint drive amount of each joint in the supporting foot, generate the joint drive command of the quadruped robot, and collect the adjustment posture data and adjustment foot force vector of the quadruped robot after executing the joint drive command. S6: Compare the adjusted attitude data with the attitude data corresponding to the feedforward attitude correction expectation to obtain the attitude tracking deviation, and update the feedforward attitude correction expectation and the foot force correction amount based on the attitude tracking deviation feedback.

[0007] Preferably, the step of identifying terrain change events based on the abrupt change direction of the force vector at the foot end of the quadruped robot includes: Based on the foot force vector of the quadruped robot in the body state data, the directional fluctuation range of the foot force vector is defined to obtain the directional stable range of the foot force vector; Within the stable directional range, the directional deviation of the foot force vector is detected to obtain a sudden change directional indicator; A time correlation analysis is performed on the abrupt change direction marker of each supporting foot of the quadruped robot to obtain the foot end contact perturbation event of the supporting foot. By comparing the posture changes of the quadruped robot before and after the foot contact disturbance event, the terrain change event of the quadruped robot is obtained.

[0008] Preferably, the step of calculating the disturbed dynamic moment vector of the quadruped robot based on the body state data before the terrain abrupt event, to obtain the terrain disturbance vector of the quadruped robot, includes: By using the instantaneous points marked before the occurrence of the terrain abrupt change event, the spatial distribution of the foot force vector is identified, and the spatial distribution pattern of the foot force vector is obtained; Based on the position of the torque application point in the spatial distribution pattern, the moment and rotation direction of the sudden jump in angular velocity amplitude in the body state data are identified to obtain the angular velocity jump event of the quadruped robot. By associating the point of application of torque in the spatial distribution pattern with the rotational tendency indicated by the angular velocity jump event, the direction of the disturbed action of the quadruped robot can be obtained. The stress gradient difference of the support foot is analyzed along the direction of the disturbed action, and the balance holding torque set of the support foot is calculated. By comparing the balance-maintaining torque set with the torque distribution of the supporting leg after the mutation occurs, the amplitude of the disturbed dynamic torque of the quadruped robot is obtained. By associating the direction of the disturbed action with the amplitude of the disturbed dynamic torque, the terrain disturbance vector of the quadruped robot is obtained.

[0009] Preferably, the step of predicting the attitude deviation tendency of the quadruped robot based on the terrain perturbation vector and the body state data, and generating a feedforward attitude correction expectation, includes: Based on the terrain disturbance vector, the direction of the moment of the support foot is analyzed to obtain the direction of the unbalanced moment of the support foot; Based on the direction of the unbalanced moment and the current position of the supporting foot, the ideal spatial orientation of the force vector at the foot end is derived, and the ideal force line direction of the supporting foot is obtained. By comparing the ideal force line direction with the actual direction of the foot force vector, the displacement deviation of the supporting force line of the supporting foot is obtained; Using the displacement deviation of the support force line and the body state data, the offset trend of the quadruped robot's body posture is predicted, and the posture offset prediction of the support foot is obtained. Using the predicted attitude offset as a compensation feature, the expected angle of the body attitude in the quadruped robot is adjusted in reverse to obtain the expected feedforward attitude correction of the supporting leg.

[0010] Preferably, the step of comparing the body wave sequence in the body state data and the frequency band energy sequence in the time-spectrum diagram based on each gait cycle of the quadruped robot to obtain the resonance suppression reference value of the quadruped robot includes: Based on the various gait cycles of the quadruped robot, the body wave sequence of the quadruped robot is time-aligned with the frequency band energy sequence of the time spectrum diagram to obtain a synchronization sequence pair; In the synchronization sequence pair, the overlapping positions of the fluctuation peak of the fuselage fluctuation sequence and the energy peak of the frequency band energy sequence are compared, and the energy at the overlapping positions is marked as the fluctuation-induced frequency band energy; Determine the main wave frequency range from the fuselage wave sequence; The oscillation frequency band energy is obtained by determining the frequency band energy that matches the main wave frequency range from the wave-induced frequency band energy. The energy segments in the oscillation frequency band that conform to the conventional fluctuation pattern of the fuselage fluctuation sequence are removed, and the remaining abnormal energy segments are used as the resonance suppression reference for the quadruped robot.

[0011] Preferably, the step of correcting the expected foot force of the supporting foot in the quadruped robot using the expected feedforward attitude correction and the resonance suppression reference value to obtain the foot force correction amount includes: The feedforward attitude correction expectation is decomposed along each axis of the quadruped robot's body coordinate system to obtain the quadruped robot's directional attitude recovery requirement; Based on the foot layout of the quadruped robot, the directional posture recovery requirement is mapped to the desired vector direction of the supporting foot to obtain the posture compensation amplitude of the supporting foot. Based on the phase interval of the resonance suppression reference value in each gait cycle, the dominant action space of the resonance suppression reference value in the supporting foot is identified, and the resonance sensitive phase of the supporting foot is obtained. Using the resonant sensitive phase as the priority channel for force line adjustment, the attitude compensation amplitude is deflected along the main energy transfer direction of the resonant suppression reference quantity to obtain the anti-resonance force line deflection amount of the support foot; By combining the attitude compensation amplitude and the anti-resonance force line deflection, the foot end force vector correction requirement of the supporting foot is obtained; The foot force vector correction requirement and the expected foot force of the supporting foot are superimposed to obtain the foot force correction amount of the supporting foot.

[0012] Preferably, the step of calculating the target position compensation amount of the support foot based on its current position includes: Based on the current position of the supporting foot and the spatial orientation of the foot end force correction amount, the supporting foot is displacement mapped to obtain the spatial offset range of the foot end force correction amount; Based on the desired angle in the feedforward attitude correction expectation, the current attitude of the supporting foot is calibrated to generate the attitude adjustment space range corresponding to the feedforward attitude correction expectation; Based on the energy peak time and direction of the resonance suppression reference quantity in the gait phase, the displacement path of the supporting foot is phase-marked to obtain the vibration suppression path nodes; The spatial intersection of the spatial action offset range and the attitude adjustment spatial range is taken as the effective spatial position range of the supporting foot. The remaining space area of ​​the support foot is obtained by removing the area that falls into the vibration suppression path node from the effective spatial location range; Using the remaining space area as the target, and combining it with the current position of the support foot, determine the target position compensation amount for the current position of the support foot.

[0013] Preferably, the step of calculating the desired joint actuation amount of each joint in the supporting foot based on the foot force correction amount and the target position compensation amount, and generating the joint actuation command for the quadruped robot, includes: By combining the foot force correction amount and the target position compensation amount, the desired force-position composite characteristics of the supporting foot in physical space are defined; The desired force-position composite feature is decomposed to the joint axis direction of the supporting foot to obtain the torque component and angular displacement component in each joint axis direction of the supporting foot. Work is matched between the torque component and the angular displacement component to determine the desired joint drive amount of each joint in the supporting foot; The desired joint drive quantities of each joint in the supporting leg are time-sequentially arranged into a rhythmic motion sequence to generate the joint drive commands for the quadruped robot.

[0014] Preferably, the step of comparing the adjusted attitude data with the attitude data corresponding to the expected feedforward attitude correction to obtain the attitude tracking deviation includes: Using the gait period as the segmentation period, the fuselage attitude angle trajectory of the attitude adjustment data is periodically segmented to obtain a multi-period fluctuation sequence of the attitude adjustment data. Within the same gait cycle, the envelope of the desired attitude angle trajectory of the attitude data corresponding to the desired feedforward attitude correction is extracted to obtain the ideal trajectory envelope of the desired feedforward attitude correction. By identifying the deviation time periods of the multi-period fluctuation sequence from each fluctuation angle point in the ideal trajectory envelope, the attitude deviation time periods of the adjusted attitude data are obtained. During the attitude deviation period, the discrete point implementation region contained in the multi-period fluctuation sequence is extracted to obtain the transient tracking subsequence of the attitude adjustment data; The attitude tracking deviation of the adjusted attitude data is obtained by combining the maximum offset extreme value and direction of the transient tracking subsequence with the ideal trajectory envelope.

[0015] To address the aforementioned problems, the present invention also provides a terrain-adaptive attitude stabilization control system for a quadruped robot, comprising: Signal acquisition module: Acquires the time spectrum diagram and body state data corresponding to the multi-source vibration signals of the quadruped robot in the current terrain, and identifies terrain change events based on the change direction of the force vector at the end of the foot in the quadruped robot; Disturbance calculation module: Based on the body state data before the terrain change event, calculate the disturbed dynamic torque vector of the quadruped robot to obtain the terrain disturbance vector of the quadruped robot; Feedforward prediction module: Based on the terrain disturbance vector and the body state data, predict the attitude deviation trend of the quadruped robot, generate the expected feedforward attitude correction, and based on each gait cycle of the quadruped robot, compare the body wave sequence in the body state data with the frequency band energy sequence of the time spectrum to obtain the resonance suppression reference value of the quadruped robot. Instruction generation module: Using the feedforward attitude correction expectation and the resonance suppression reference, correct the expected foot end force of the supporting foot in the quadruped robot to obtain the foot end force correction amount, and calculate the target position compensation amount of the supporting foot based on the current position of the supporting foot. Execution feedback module: Based on the foot force correction amount and the target position compensation amount, calculate the expected joint drive amount of each joint in the supporting foot, generate the joint drive command of the quadruped robot, and collect the adjustment posture data and adjustment foot force vector of the quadruped robot after executing the joint drive command; Deviation correction module: compares the adjusted attitude data with the attitude data corresponding to the feedforward attitude correction expectation to obtain the attitude tracking deviation, and updates the feedforward attitude correction expectation and the foot force correction amount based on the attitude tracking deviation feedback.

[0016] Beneficial effects This solution identifies terrain abrupt events by collecting multi-source vibration signal spectrum diagrams and body state data, combined with the direction of foot force vector abrupt change, and based on the correlation analysis of foot force spatial distribution and angular velocity jump, calculates the disturbed dynamic moment vector using pre-abrupt data, and quantifies the direction and amplitude of terrain disturbance. This enables rapid capture of ground support attribute abrupt changes and accurate determination of the direction of disturbance, achieving real-time quantification of terrain disturbance and early prediction of fuselage attitude deviation. It can generate feedforward attitude correction expectations to adjust the foot force layout in reverse, solving the problems of existing technologies that cannot predict terrain abrupt changes in advance or quantify disturbance intensity in real time, resulting in lag in attitude response and insufficient stability control accuracy. It can complete force correction before the impact of terrain abrupt changes is transmitted to the fuselage, significantly improving attitude stability and walking smoothness under complex unstructured terrain.

[0017] This solution achieves accurate calculation of the composite parameters of the support foot force and position, and orderly arrangement and output of joint drive commands by comparing the body fluctuation sequence with the time-frequency band energy sequence, filtering the energy of abnormal oscillation frequency bands to extract resonance suppression reference quantities, and combining feedforward correction to synchronously adjust the foot force and target position of the support foot. Then, by calculating the deviation between the actual posture and the desired posture, the feedforward parameters and foot force correction quantities are updated iteratively. This solves the problems of resonance suppression relying on historical statistical patterns, poor adaptability to sudden terrain changes, and disconnect between force control and position control in existing technologies. It can effectively reduce body resonance caused by terrain, continuously improve posture tracking accuracy, and ensure the long-term walking stability of the robot on varied terrain. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating a terrain-adaptive attitude stabilization control method for a quadruped robot according to an embodiment of the present invention. Figure 2 This is a functional block diagram of a terrain-adaptive attitude stabilization control system for a quadruped robot provided in an embodiment of the present invention; The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0020] This application provides a terrain-adaptive attitude stabilization control method and system for a quadruped robot. The execution entity of this method and system includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the method and system can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0021] Reference Figure 1 The diagram shown is a flowchart illustrating a terrain-adaptive attitude stabilization control method for a quadruped robot according to an embodiment of the present invention. In this embodiment, the terrain-adaptive attitude stabilization control method for a quadruped robot includes: S1: Collect the time-spectrum diagram and body state data corresponding to the multi-source vibration signals of the quadruped robot in the current terrain, and identify terrain change events based on the change direction of the force vector at the foot end of the quadruped robot; In this embodiment, identifying terrain change events based on the abrupt change direction of the force vector at the foot end of the quadruped robot includes: Based on the foot force vector of the quadruped robot in the body state data, the directional fluctuation range of the foot force vector is defined to obtain the directional stable range of the foot force vector; Within the stable directional range, the directional deviation of the foot force vector is detected to obtain a sudden change directional indicator; A time correlation analysis is performed on the abrupt change direction marker of each supporting foot of the quadruped robot to obtain the foot end contact perturbation event of the supporting foot. By comparing the posture changes of the quadruped robot before and after the foot contact disturbance event, the terrain change event of the quadruped robot is obtained.

[0022] Specifically, the multi-source vibration signal is a multi-channel time-domain vibration electrical signal that is synchronously collected by multiple sets of vibration sensing units distributed on the body, legs, and feet of the quadruped robot during its movement on the current terrain. It is the original input signal for generating the time-spectrum diagram.

[0023] The time-frequency spectrum is a two-dimensional grayscale image obtained by converting a single vibration signal into a time-frequency dimension. The horizontal axis of the image corresponds to the time dimension of signal acquisition, and the vertical axis corresponds to the frequency dimension of signal decomposition. The grayscale value of the image pixel corresponds to the energy amplitude of the vibration signal at the time-frequency intersection.

[0024] Body state data is a collection of time-series physical quantities that are synchronously collected by multiple onboard sensing units and aligned according to timestamps during the movement of a quadruped robot. It includes three core state information types: body posture angle, joint rotation angle, and foot contact force.

[0025] The foot force vector is a physical quantity collected by the foot force sensing component of the supporting foot of a quadruped robot. It represents the magnitude and three-dimensional spatial orientation of the force exerted on the foot during contact with the ground.

[0026] The directional fluctuation range is the maximum boundary range in which the spatial orientation of the foot force vector deviates normally in three-dimensional space during walking on normal, stable terrain.

[0027] The stable directional range is the range of force vector direction corresponding to the stable support phase of the foot, obtained by removing the impact offset segment at the moment of foot landing and departure from the directional fluctuation range of the force vector at the foot.

[0028] The directional deviation state is the real-time change of the spatial angle difference between the actual direction of the force vector at the foot and the center direction of the directional stability range.

[0029] The abrupt change direction marker is a recorded marker when the directional deviation of the foot force vector exceeds the boundary of the directional stability range. It contains two types of information: the time of deviation and the spatial direction of deviation.

[0030] Support legs are the leg structures that keep in contact with the ground and support the body of a quadruped robot during walking.

[0031] Temporal correlation analysis involves temporally matching the mutation direction markers of each supporting leg of a quadruped robot according to the chronological order of their occurrence to determine whether there is a sequential triggering correlation process for each leg's mutation events.

[0032] Foot contact disturbance events are foot-ground contact events confirmed by time correlation analysis to be caused by the same external ground factor, and are characterized by abnormal deviation of the foot-end force direction of one or more supporting feet.

[0033] Attitude change refers to the numerical changes and rates of change of three attitude angles of the quadruped robot body before and after a foot contact disturbance event.

[0034] A terrain change event is a terrain event identified when a foot contact disturbance event is accompanied by abnormal fluctuations in the aircraft's attitude, representing a step change in ground support properties.

[0035] Furthermore, throughout the entire process of the quadruped robot moving on the current terrain surface according to the preset gait, the onboard clock synchronization circuit outputs a unified reference clock to align the sampling clocks of all vibration sensing units, attitude sensing units, angle sensing units and force sensing units, and set the same sampling frequency and timestamp marking rules.

[0036] Vibration sensing units distributed in the fuselage frame, the middle section of each leg link, and the mounting base at each foot end sense mechanical vibration in real time and convert it into continuous analog electrical signals. After each sampling window of fixed duration, the time-domain analog electrical signal in the window is transmitted to the time-frequency conversion circuit. The hardware circuit completes the dimensional mapping from the time domain to the time-frequency domain and outputs the corresponding single-channel two-dimensional grayscale time-frequency spectrum.

[0037] The time spectrum diagrams of all acquisition channels are horizontally stitched together according to the spatial arrangement of the sensing units to form a complete multi-source vibration signal time spectrum diagram covering all vibration acquisition channels.

[0038] The force vectors at the end of each supporting foot of the quadruped robot are retrieved from the body state data. Three or more consecutive steady state cycles without posture abnormalities are selected as statistical samples. The spatial orientation distribution of the force vectors at the end of the foot in the body coordinate system is statistically analyzed.

[0039] Using the outer envelope surface pointed to by all sampling points as the boundary, the directional fluctuation range of the foot force vector is defined. Then, the impact-induced directional offset of the initial stage of foot landing and the stage before leaving the ground is removed from this range, and the directional range in which the foot is in a stable support phase is retained as the directional stability range.

[0040] Within the time interval corresponding to the stable directional range, the actual direction of the foot force vector at the current moment is compared with the center direction of the stable directional range at a fixed sampling interval. The spatial angle between the two is calculated to obtain the directional deviation state of the foot force vector. When the spatial angle exceeds the boundary threshold of the stable directional range, a sudden change direction marker is generated at the corresponding sampling moment.

[0041] Extract the mutation direction markers generated by each of the four-legged robot's supporting feet one by one, arrange and match all mutation direction markers according to the gait time axis, determine whether the mutation markers of different supporting feet conform to the sequential triggering relationship of the gait progression, and mark the foot-ground contact process corresponding to a set of mutation markers with temporal correlation as foot-end contact disturbance events.

[0042] Retrieve the body attitude data of the complete gait cycle before and after the foot contact disturbance event, and calculate the fluctuation amplitude and rate of change of the body attitude angles in the three directions of roll, pitch and yaw in the two cycles respectively. When the difference in fluctuation amplitude and the difference in rate of change between the two cycles exceed the normal fluctuation threshold of stable terrain, the foot contact disturbance event is judged as a terrain change event of the quadruped robot.

[0043] S2: Based on the body state data before the terrain change event, calculate the disturbed dynamic torque vector of the quadruped robot to obtain the terrain disturbance vector of the quadruped robot; In this embodiment, the step of calculating the disturbed dynamic moment vector of the quadruped robot based on the body state data before the terrain abrupt event, and obtaining the terrain disturbance vector of the quadruped robot, includes: By using the instantaneous points marked before the occurrence of the terrain abrupt change event, the spatial distribution of the foot force vector is identified, and the spatial distribution pattern of the foot force vector is obtained; Based on the position of the torque application point in the spatial distribution pattern, the moment and rotation direction of the sudden jump in angular velocity amplitude in the body state data are identified to obtain the angular velocity jump event of the quadruped robot. By associating the point of application of torque in the spatial distribution pattern with the rotational tendency indicated by the angular velocity jump event, the direction of the disturbed action of the quadruped robot can be obtained. The stress gradient difference of the support foot is analyzed along the direction of the disturbed action, and the balance holding torque set of the support foot is calculated. By comparing the balance-maintaining torque set with the torque distribution of the supporting leg after the mutation occurs, the amplitude of the disturbed dynamic torque of the quadruped robot is obtained. By associating the direction of the disturbed action with the amplitude of the disturbed dynamic torque, the terrain disturbance vector of the quadruped robot is obtained.

[0044] Specifically, the spatial distribution of foot force vectors is the arrangement of the spatial positions and force directions of the three-dimensional forces generated when the feet of each supporting foot of the quadruped robot come into contact with the ground in the body coordinate system.

[0045] The spatial distribution pattern of the foot force vector is the overall force layout pattern presented after integrating the spatial position, direction and amplitude of the force vectors at the foot ends of each supporting foot. It is used to characterize the force geometry of the robot in the current supporting state.

[0046] The location of the point of application of the torque is the spatial coordinate of the equivalent point of application of the torque exerted by the foot force vector on the fuselage in the fuselage coordinate system, corresponding to the contact point between the foot end and the ground.

[0047] The moment when the amplitude of angular velocity suddenly jumps is the sampling time point when the value of the fuselage angular velocity data changes significantly within the set sampling interval, corresponding to the starting moment when the fuselage begins to deflect due to terrain disturbance.

[0048] The rotational direction of a sudden jump in angular velocity amplitude is the spatial direction in which the fuselage deflects around the corresponding rotation axis when the fuselage angular velocity changes abruptly, encompassing the deflection tendencies of the pitch, roll, and yaw axes.

[0049] An angular velocity jump event is a complete disturbance process record in which the fuselage angular velocity simultaneously undergoes an abnormal jump in both amplitude and direction. It is used to characterize the rotational response of the fuselage caused by external force disturbance.

[0050] The rotational trend indicated by the angular velocity jump event is the spatial direction of the overall attitude deflection of the fuselage as reflected by the angular velocity jump event, which is used to determine the direction of the disturbance torque.

[0051] The direction of the disturbance is the spatial direction of the disturbance torque generated by the sudden change in terrain on the robot body, corresponding to the deflection direction in which the robot body becomes unstable.

[0052] The stress gradient difference of the support foot is the difference in the rate of change of the contact stress at the foot end of each support foot along the direction of the force along the direction of the disturbance. It is used to reflect the difference in the load-bearing changes of different support feet in the direction of disturbance.

[0053] The balance-maintaining torque set is the collection of counteracting torques provided by each support foot to maintain the fuselage's attitude balance, including the torque values ​​of each support foot for each spatial axis.

[0054] The moment distribution of the supporting feet after the sudden change is the spatial arrangement of the moment formed by the actual foot-end force generated by each supporting foot on the fuselage after the terrain change.

[0055] The amplitude of the disturbed dynamic moment is the magnitude of the disturbance torque exerted on the robot body by the sudden change in terrain, and is used to quantify the intensity of the terrain disturbance.

[0056] The terrain disturbance vector is a composite physical quantity that includes the direction of the disturbance and the amplitude of the disturbance torque. It is used to fully characterize the direction and intensity of the disturbance caused by abrupt changes in terrain on the robot body.

[0057] Furthermore, during the process of the quadruped robot traversing the terrain, it extracts the sampling data corresponding to the instantaneous point marked before the moment of the terrain change event is identified, extracts the three-dimensional force data of all feet in the support phase at that moment, sorts out the spatial position and direction of each foot force in the body coordinate system, and integrates them to form the spatial distribution pattern of the foot force vector.

[0058] The location of the torque application point corresponding to each foot is determined from the spatial distribution pattern. Simultaneously, the fuselage angular velocity data from the body state data is retrieved. The sampling time sequence is used to examine frame by frame for sampling moments where the angular velocity value shows a significant jump, and the corresponding rotational direction of the jump. Specifically, this involves analyzing the angular velocity sequence in the body state data and employing a dynamic threshold algorithm based on a sliding window. A threshold is set; when the change in angular velocity exceeds this threshold within a very short time (e.g., 5 milliseconds), it is identified as an angular velocity jump event. This moment is then integrated with the corresponding rotational direction and recorded as an angular velocity jump event.

[0059] The point of application of the torque in the spatial distribution pattern (i.e., the landing position of the supporting foot) is spatially geometrically correlated with the rotational tendency of the angular velocity jump event. If the angular velocity jump indicates that the fuselage deflects in the positive direction around a certain axis (e.g., the roll axis), then the direction of the disturbance is determined to be the deflection direction.

[0060] Along the determined direction of the disturbance, the rate of change of contact stress of each support foot in that direction is analyzed one by one, the stress gradient difference between each support foot is calculated, and based on the difference, the counterbalancing torque required by each support foot to maintain the balance of the fuselage is calculated. The counterbalancing torques of all support feet are summarized to form a set of balance holding torques.

[0061] Specifically, the rate of change of normal and tangential forces of the support feet located upstream and downstream of the disturbance direction is compared to quantify the non-uniformity of the support force distribution. Based on this stress gradient difference and the known robot inertia parameters, the balance holding torque required for each support foot to maintain its current attitude and position is calculated using a static or quasi-static dynamic model (e.g., the Newton-Euler recursive algorithm).

[0062] The torque values ​​of each support foot are concentrated to maintain the balance, and then compared one by one with the actual torque distribution data of each support foot after the terrain change. The torque difference between the two is calculated, and the difference is integrated to obtain the magnitude of the disturbance torque borne by the fuselage, that is, the amplitude of the disturbance torque.

[0063] The obtained direction of the disturbance and the amplitude of the disturbance moment are spatially vector-correlated and combined to form a terrain disturbance vector that contains both direction and magnitude information, thus completing the quantitative calculation of the terrain disturbance.

[0064] S3: Based on the terrain disturbance vector and the body state data, predict the attitude deviation trend of the quadruped robot, generate the feedforward attitude correction expectation, and based on each gait cycle of the quadruped robot, compare the body wave sequence in the body state data with the frequency band energy sequence of the time spectrum to obtain the resonance suppression reference value of the quadruped robot. In this embodiment, the step of predicting the attitude deviation trend of the quadruped robot based on the terrain disturbance vector and the body state data, and generating a feedforward attitude correction expectation, includes: Based on the terrain disturbance vector, the direction of the moment of the support foot is analyzed to obtain the direction of the unbalanced moment of the support foot; Based on the direction of the unbalanced moment and the current position of the supporting foot, the ideal spatial orientation of the force vector at the foot end is derived, and the ideal force line direction of the supporting foot is obtained. By comparing the ideal force line direction with the actual direction of the foot force vector, the displacement deviation of the supporting force line of the supporting foot is obtained; Using the displacement deviation of the support force line and the body state data, the offset trend of the quadruped robot's body posture is predicted, and the posture offset prediction of the support foot is obtained. Using the predicted attitude offset as a compensation feature, the expected angle of the body attitude in the quadruped robot is adjusted in reverse to obtain the expected feedforward attitude correction of the supporting leg.

[0065] Specifically, the direction of the moment of the supporting foot is the spatial direction of the moment generated by the force at the foot end of the supporting foot on the center of mass of the fuselage under the action of terrain disturbance.

[0066] The direction of the unbalanced moment is the spatial orientation of the moment that causes the fuselage to deflect after the terrain disturbance breaks the force balance of the fuselage, corresponding to the direction of rotation in which the fuselage is about to become unstable and deflect.

[0067] The current position of the support foot is the three-dimensional spatial coordinate of the point where the end of the support foot contacts the ground in the fuselage coordinate system.

[0068] Ideal spatial orientation is the optimal spatial orientation required for the foot-end force to counteract unbalanced torque and maintain fuselage attitude stability.

[0069] The ideal force line direction is the optimal orientation of the force line in three-dimensional space when the force applied at the foot is used to balance the unbalanced torque.

[0070] The actual direction of the force vector at the foot end is the spatial orientation of the ground reaction force that the foot end is actually subjected to at the current moment.

[0071] The deviation of the support force line displacement is the spatial angular difference between the actual direction of the force vector at the foot and the ideal force line direction, reflecting the degree of deviation between the current support force application direction and the optimal equilibrium direction.

[0072] The deviation trend of the quadruped robot's body attitude is the change of the body attitude angle over time under the influence of terrain disturbance, corresponding to the deflection development direction of the three axes of body pitch, roll, and yaw.

[0073] The attitude deviation prediction is calculated based on the current force deviation and fuselage status, which is the angular deviation and deflection direction that the fuselage attitude will produce in the subsequent period.

[0074] The compensation feature is a reference for adjusting the fuselage attitude setting value, corresponding to the calculated attitude offset value and direction.

[0075] The desired angle of the body posture is the preset reference posture angle value of the body pitch, roll and yaw axes when the quadruped robot moves smoothly.

[0076] The feedforward attitude correction expectation is the target angle value of the fuselage attitude obtained in advance before the attitude deviation caused by terrain disturbance is triggered, so as to offset the attitude deflection caused by terrain disturbance in advance.

[0077] The reverse adjustment is based on the predicted attitude deviation and the reference expected attitude angle of the fuselage. The predicted deflection direction and amplitude of the pitch, roll and yaw axes are separated to determine the reverse compensation direction. The corresponding compensation angle is superimposed on the reference angle to obtain the corrected expected attitude angle of the three axes. The foot force line deflection is driven in advance to generate a counteracting torque to offset the impact of terrain disturbance.

[0078] Furthermore, the torque components corresponding to the terrain disturbance vector are decomposed along the pitch, roll, and yaw axes of the fuselage coordinate system. The force application points at the feet of each support foot are matched one by one, and the direction of the torque acting on each support foot is analyzed. From this, the direction of the torque that can cause the fuselage to deviate from the equilibrium state is selected, and the direction of the unbalanced torque corresponding to each support foot is obtained.

[0079] By combining the contact point coordinates between the foot end of each support foot and the ground, and with the goal of completely offsetting the unbalanced torque and maintaining the force balance of the fuselage center of gravity, the optimal spatial orientation required for the force at the foot end is derived, and the ideal force line direction corresponding to each support foot is determined.

[0080] Specifically: Combining the direction of the unbalanced torque and the current position of the supporting feet, a force distribution strategy based on zero-space projection is adopted. The core idea is: within the robot's joint space or task space, a balancing task that is completely opposite to the unbalanced torque and used to maintain the robot's balance is predefined; then, this balancing task is mapped to the foot ends of each supporting foot, and a force vector direction that can most effectively perform the balancing task while satisfying the robot's kinematic and dynamic constraints is found.

[0081] The actual spatial direction of the foot force vector collected at the current moment is compared with the spatial angle of the derived ideal force line direction. The included angle and deviation between the two are calculated to obtain the displacement deviation of the support force line of each supporting foot.

[0082] By combining the magnitude and direction of the displacement deviation of the support force line, as well as the current attitude angle and joint rotation angle of the fuselage, the subsequent change trend and deflection amplitude of the three axial attitude angles of the fuselage under the continuous action of terrain disturbance are calculated, and the attitude offset prediction amount corresponding to each support foot is obtained.

[0083] The attitude deviation prediction is used as the basis for adjustment to perform a reverse adjustment action. First, the pitch, roll and yaw directions and yaw angle amplitudes of the three rotation axes of the fuselage are separated from the attitude deviation prediction.

[0084] The preset body reference posture angle when the quadruped robot is moving smoothly is retrieved as the initial reference value for angle adjustment.

[0085] For each fuselage rotation axis, an angle compensation amount of the corresponding amplitude is superimposed on the reference attitude angle in a direction completely opposite to the predicted deflection direction of that axis. When the fuselage is predicted to deflect to the right around the roll axis, an angle compensation amount of the same amplitude is superimposed on the left side of the roll axis. When the fuselage is predicted to tilt forward around the pitch axis, an angle compensation amount of the same amplitude is superimposed on the back side of the pitch axis.

[0086] After completing the reverse superposition of the angles of the three axes, the corrected target angles of each axis are integrated. Before the actual attitude deviation of the fuselage is caused by disturbance, the pre-corrected attitude target value is generated, and the expected feedforward attitude correction of the corresponding support foot is obtained.

[0087] This adjustment, by pre-correcting the attitude target setting value, drives the support foot to change the spatial direction of the foot end force in advance, generating a counteracting torque opposite to the predicted offset direction, thus offsetting the impact of terrain disturbances before they are transmitted to the fuselage, achieving feedforward attitude stabilization control.

[0088] In this embodiment, the step of comparing the body fluctuation sequence in the body state data and the frequency band energy sequence of the time-spectrum diagram based on the gait cycles of the quadruped robot to obtain the resonance suppression reference value of the quadruped robot includes: Based on the various gait cycles of the quadruped robot, the body wave sequence of the quadruped robot is time-aligned with the frequency band energy sequence of the time spectrum diagram to obtain a synchronization sequence pair; In the synchronization sequence pair, the overlapping positions of the fluctuation peak of the fuselage fluctuation sequence and the energy peak of the frequency band energy sequence are compared, and the energy at the overlapping positions is marked as the fluctuation-induced frequency band energy; Determine the main wave frequency range from the fuselage wave sequence; The oscillation frequency band energy is obtained by determining the frequency band energy that matches the main wave frequency range from the wave-induced frequency band energy. The energy segments in the oscillation frequency band that conform to the conventional fluctuation pattern of the fuselage fluctuation sequence are removed, and the remaining abnormal energy segments are used as the resonance suppression reference for the quadruped robot.

[0089] Specifically, the gait cycle is the length of time it takes for a quadruped robot to complete one complete walking cycle, corresponding to the entire process of a single leg sequentially completing the support phase, the swing phase, and returning to the initial support state.

[0090] The body undulation sequence is a time-series data set formed by arranging the amplitude of the body attitude angle along each axis in chronological order during the movement of a quadruped robot. It reflects the process of the body's attitude fluctuation changes as it moves forward with its gait.

[0091] A frequency band energy sequence is a time-series data set formed by arranging the energy values ​​at each moment in a fixed frequency range extracted from a time-spectrum graph in chronological order. It represents the change law of vibration energy over time within the corresponding frequency range.

[0092] Timing alignment is a process of adjusting the time correspondence of two sets of time-series data from different sources, using the same time base as a reference, so that the two sets of data points at the same time correspond one-to-one.

[0093] Synchronization sequence pairs are paired data sets consisting of a fuselage wave sequence and a frequency band energy sequence with completely consistent time bases after time alignment processing. Any data point with the same sequence number in the two sets of data corresponds to the same acquisition time.

[0094] A wave peak is the point in the fuselage wave sequence where the amplitude of the attitude wave reaches a local maximum, corresponding to the peak moment in the process of fuselage attitude fluctuation.

[0095] An energy peak is the point in a frequency band energy sequence where the vibration energy value reaches a local maximum, corresponding to the peak moment of vibration energy within that frequency band.

[0096] The overlapping position is the time point in the synchronous sequence pair where the fluctuation peak and the energy peak appear at the same moment, representing the time point when the fuselage attitude fluctuation and the corresponding frequency band vibration energy reach their peak simultaneously.

[0097] Wave-induced frequency band energy is the frequency band energy value corresponding to the overlapping position, which represents the intensity of vibration energy that occurs synchronously with the attitude fluctuation of the fuselage.

[0098] The main wave frequency range is the frequency range corresponding to the attitude wave with the highest energy proportion in the fuselage wave sequence, representing the dominant wave frequency range of fuselage attitude fluctuations.

[0099] The frequency band energy segment is the set of energy data in the wave-induced frequency band energy where the frequency falls within a specified interval, corresponding to the wave-induced vibration energy within a specific frequency range.

[0100] Oscillation frequency band energy is a frequency band energy segment that matches the main wave frequency range, representing the set of vibration energy that is consistent with the dominant wave frequency of the fuselage.

[0101] The conventional undulation pattern is a fixed correspondence between the body posture undulation and the corresponding frequency band energy when a quadruped robot is moving normally on stable terrain. It is an inherent undulation pattern brought about by the gait itself.

[0102] Abnormal energy segments are the portion of energy data in the oscillation frequency band that does not conform to the normal fluctuation pattern, corresponding to vibrational energy that exceeds the normal gait fluctuation range caused by terrain disturbance.

[0103] The resonance suppression reference value is a set of abnormal energy range data obtained through screening. It is used to characterize the intensity and frequency characteristics of abnormal vibration energy that may cause fuselage resonance, and serves as a reference for subsequent adjustment of the foot force.

[0104] Furthermore, the duration of a single gait cycle is defined as the complete cycle of ground support, leg lifting and swinging, and landing contact of a single leg of the quadruped robot. The start time of a single gait cycle is defined as the moment when the foot force sensor detects the first contact between the foot and the ground. The start and end times of all gait cycles are marked sequentially along the walking time axis.

[0105] The fuselage fluctuation time-series data continuously collected by the fuselage attitude sensing component and the energy time-series data of each frequency band obtained by the vibration sensing component after time-frequency conversion are retrieved respectively. Based on the timestamp output by the airborne unified synchronous clock, the corresponding acquisition time is marked for each set of fuselage fluctuation data points and each set of frequency band energy data points.

[0106] Using the marked start and end times of the gait cycle as the segment boundary, the two types of time series data are divided into multiple segments with the same number of gait cycles according to the same time boundary. The two types of data within the same segment are arranged one-to-one in the order of acquisition time. The fuselage wave data and frequency band energy data with the same time are paired together. All paired data are arranged in time order to form a synchronization sequence pair.

[0107] Read the paired data in the synchronization sequence group by group along the time progression direction. For the fuselage fluctuation data, compare the fluctuation amplitude of the current data point with the fluctuation amplitude of the previous and next adjacent time points. If the amplitude of the current data point is greater than the amplitude of the two adjacent data points at the same time, mark the time of the data point as the fluctuation peak position and record the corresponding amplitude.

[0108] For the frequency band energy data, the same adjacent amplitude comparison method is used to mark the data points that are simultaneously greater than the energy values ​​of the adjacent moments as the energy peak positions and record the corresponding energy values.

[0109] After marking all peak positions, match the corresponding times of all fluctuation peaks and energy peaks one by one, filter out the overlapping peak positions with completely consistent times, extract and mark the frequency band energy values ​​corresponding to these overlapping peak positions, and the marked energy values ​​are the fluctuation-induced frequency band energy.

[0110] Extract the fuselage wave sequence data within multiple consecutive complete gait cycles, divide this data segment into several consecutive frequency intervals according to a fixed frequency interval determined by the sampling frequency, and calculate the sum of the amplitudes of all wave data within each frequency interval.

[0111] Calculate the proportion of the sum of amplitudes in each frequency interval to the sum of amplitudes in all frequency intervals, select the continuous frequency interval with the highest proportion, and define the upper and lower limits of this interval as the main fluctuation frequency range. This range corresponds to the dominant fluctuation frequency range with the highest energy proportion during the fuselage attitude fluctuation process.

[0112] Using the upper and lower limits of the main fluctuation frequency range as boundaries, the corresponding frequency intervals are defined in the marked fluctuation-induced frequency band energy. All fluctuation-induced frequency band energy data with frequency values ​​falling within this interval are extracted and arranged in chronological order of acquisition time to form a continuous energy data segment, which is the oscillation frequency band energy.

[0113] When a quadruped robot walks in a stable terrain environment with the same gait phase and in the same synchrotron state, the inherent law of energy change with gait phase within the corresponding oscillation frequency band is retrieved as the conventional fluctuation mode. The upper limit of the fluctuation amplitude deviation corresponding to this mode is obtained by statistical analysis of multiple repeated walking tests and is used as the conventional fluctuation threshold.

[0114] The obtained oscillation frequency band energy is divided into several phase intervals according to the gait phase. The energy amplitude in each phase interval is compared with the energy amplitude of the corresponding phase interval in the conventional wave pattern. The energy segment whose amplitude deviation is within the conventional wave threshold is determined to be the energy segment that conforms to the conventional wave pattern, and this part of the energy is removed from the oscillation frequency band energy.

[0115] The remaining energy range whose amplitude exceeds the normal fluctuation threshold is the abnormal energy range. This abnormal energy range is used as a resonance suppression reference value for subsequent adjustment of the force and position of the support foot.

[0116] S4: Using the feedforward attitude correction expectation and the resonance suppression reference, correct the expected foot end force of the supporting foot in the quadruped robot to obtain the foot end force correction amount, and calculate the target position compensation amount of the supporting foot based on the current position of the supporting foot. In this embodiment, the step of using the expected feedforward attitude correction and the resonance suppression reference to correct the expected foot force of the supporting foot in the quadruped robot, and obtaining the foot force correction amount, includes: The feedforward attitude correction expectation is decomposed along each axis of the quadruped robot's body coordinate system to obtain the quadruped robot's directional attitude recovery requirement; Based on the foot layout of the quadruped robot, the directional posture recovery requirement is mapped to the desired vector direction of the supporting foot to obtain the posture compensation amplitude of the supporting foot. Based on the phase interval of the resonance suppression reference value in each gait cycle, the dominant action space of the resonance suppression reference value in the supporting foot is identified, and the resonance sensitive phase of the supporting foot is obtained. Using the resonant sensitive phase as the priority channel for force line adjustment, the attitude compensation amplitude is deflected along the main energy transfer direction of the resonant suppression reference quantity to obtain the anti-resonance force line deflection amount of the support foot; By combining the attitude compensation amplitude and the anti-resonance force line deflection, the foot end force vector correction requirement of the supporting foot is obtained; The foot force vector correction requirement and the expected foot force of the supporting foot are superimposed to obtain the foot force correction amount of the supporting foot.

[0117] Specifically, the body coordinate system is a three-dimensional spatial reference base fixed to the body of the quadruped robot, with the body's center of mass as the origin, and established along the three rotational axes of pitch, roll, and yaw. It is used to uniformly calibrate the spatial orientation of the body's posture and the forces acting on the feet.

[0118] The axes are the three mutually orthogonal spatial coordinate axes in the fuselage coordinate system, corresponding to the extension directions of the fuselage pitch rotation axis, roll rotation axis and yaw rotation axis, respectively.

[0119] Feedforward attitude correction is a set of target attitude angles for the fuselage that are pre-set before the terrain disturbance causes the fuselage attitude to deviate, in order to counteract the attitude deflection trend caused by the terrain disturbance in advance.

[0120] Directional attitude recovery requirement is the attitude angle recovery amount and recovery direction on each independent rotation axis obtained by splitting the feedforward attitude correction expectation along each axis of the fuselage coordinate system. It is used to characterize the attitude adjustment target on a single axis.

[0121] The foot layout refers to the spatial arrangement and relative spacing of each leg of a quadruped robot at the bottom of the robot body. It determines the lever length and direction of each supporting leg's effect on the robot's posture.

[0122] The desired vector direction is the preset spatial orientation of the force exerted on the foot end during stable walking, corresponding to the force line orientation when the supporting foot is in normal support and the body is in force balance.

[0123] The attitude compensation amplitude is the magnitude of the attitude adjustment force required by a single support foot after the directional attitude recovery requirement is mapped through the foot layout. It is used to quantify the strength of the force that the support foot needs to provide to restore the fuselage attitude.

[0124] Phase intervals are the intervals of different action stages within a single gait cycle, divided according to the time process, corresponding to different action periods such as supporting foot landing, stable support, and takeoff.

[0125] The dominant influence space is the spatial orientation range of the vibration energy corresponding to the resonance suppression reference quantity, which occupies the dominant influence in the force transmission path at the foot end of the supporting foot.

[0126] The resonance-sensitive phase is the phase interval corresponding to the energy peak of the resonance suppression reference quantity within the gait cycle. During this period, the amplification effect of vibration energy on the attitude fluctuation of the fuselage is most significant.

[0127] Force line adjustment is an operation that changes the spatial orientation of the force acting on the foot end of the support foot, and is used to adjust the direction of the torque of the reaction force at the foot end on the fuselage.

[0128] The priority channel is the time window during which force line adjustment operations are performed preferentially within the gait cycle, corresponding to the time period of the resonant sensitive phase.

[0129] The main energy transfer direction is the vibration energy corresponding to the resonance suppression reference quantity, which is the main spatial direction of the energy transfer from the foot contact point to the center of mass of the fuselage.

[0130] The anti-resonance force line deflection is the spatial deflection angle and direction of the foot force line obtained after adjusting the force line direction of the attitude compensation amplitude along the main energy transmission direction. It is used to counteract the transmission gain of resonant energy.

[0131] The foot force vector correction requirement is a complete vector description of the adjustment direction and magnitude of the supporting foot end force obtained by combining the attitude compensation amplitude and the anti-resonance force line deflection.

[0132] The expected foot force is the preset reference value of the supporting foot force when the quadruped robot is walking stably, which corresponds to the magnitude and direction of the foot reaction force required for normal support of the robot body.

[0133] The foot end force correction is the corrected target force at the foot end of the supporting foot obtained by superimposing the foot end force vector correction requirement and the baseline expected foot end force. It is used to drive the supporting foot to output a contact force that simultaneously satisfies attitude recovery and resonance suppression.

[0134] Furthermore, the expected feedforward attitude correction is decomposed along the three orthogonal axes of the quadruped robot's body coordinate system to obtain the attitude recovery angle and recovery direction in the three independent directions of pitch, roll, and yaw, thus forming the directional attitude recovery requirements corresponding to each axis.

[0135] By combining the arrangement and spacing of the legs on the bottom of the quadruped robot, the directional posture recovery requirements of the three axes are converted into the preset force direction of each supporting leg. Based on the lever arm length of each supporting leg relative to the center of mass of the robot body, the magnitude of the force output required by each supporting leg to achieve posture recovery is calculated, and the posture compensation amplitude is obtained.

[0136] Based on the energy distribution period of the resonance suppression reference value within a single gait cycle, the gait movement stage where the energy peak is located is determined, the spatial orientation in which the resonance energy has the strongest influence in the force transmission path of the supporting foot is determined, and the gait phase interval in which the supporting foot is most sensitive to the resonance energy is obtained, thus obtaining the resonance sensitive phase.

[0137] The resonant sensitive phase is taken as the priority execution period for foot force line adjustment. During this period, the spatial orientation of the force line corresponding to the attitude compensation amplitude is deflected along the main direction of the resonant energy transmission from the foot to the fuselage. The direction of the force applied to the foot is adjusted to weaken the transmission intensity of the resonant energy, thus obtaining the anti-resonance force line deflection amount corresponding to each supporting foot.

[0138] The magnitude of the force vector corresponding to the attitude compensation amplitude and the direction of the force vector corresponding to the anti-resonance force line deflection are spatially synthesized and integrated into a complete force adjustment vector that includes both adjustment magnitude and adjustment direction, thus forming the foot end force vector correction requirement for the support foot.

[0139] The foot force vector correction requirement is superimposed on the baseline expected foot force of the supporting foot, and the magnitude and direction of the original force are corrected synchronously to obtain the final target force of the supporting foot, i.e., the foot force correction amount, which is used to drive the joint to output the corresponding force.

[0140] In this embodiment, the step of calculating the target position compensation amount of the support foot based on its current position includes: Based on the current position of the supporting foot and the spatial orientation of the foot end force correction amount, the supporting foot is displacement mapped to obtain the spatial offset range of the foot end force correction amount; Based on the desired angle in the feedforward attitude correction expectation, the current attitude of the supporting foot is calibrated to generate the attitude adjustment space range corresponding to the feedforward attitude correction expectation; Based on the energy peak time and direction of the resonance suppression reference quantity in the gait phase, the displacement path of the supporting foot is phase-marked to obtain the vibration suppression path nodes; The spatial intersection of the spatial action offset range and the attitude adjustment spatial range is taken as the effective spatial position range of the supporting foot. The remaining space area of ​​the support foot is obtained by removing the area that falls into the vibration suppression path node from the effective spatial location range; Using the remaining space area as the target, and combining it with the current position of the support foot, determine the target position compensation amount for the current position of the support foot.

[0141] Specifically, the current position of the supporting foot is the three-dimensional spatial coordinate value of the point of contact between the foot end of the quadruped robot's leg and the ground in the body coordinate system when the quadruped robot is in the supporting phase.

[0142] The spatial orientation of the foot force correction is the direction in which the foot force vector, after feedforward attitude correction and resonance suppression adjustment, faces in three-dimensional space.

[0143] Displacement mapping is a mechanical geometric conversion process that transforms the vector direction of the force at the foot end into the range of spatial displacement at the foot end according to the mechanical transmission relationship of the leg linkage.

[0144] The spatial offset range of the foot force correction is the entire three-dimensional spatial region in which the foot can produce a positional offset under the constraint of the leg's mechanical structure after the foot force correction is applied.

[0145] The desired angle in the feedforward attitude correction is the target attitude angle value of the three axes of fuselage pitch, roll and yaw, which is pre-set to counteract the unbalanced moment in advance.

[0146] The current posture of the supporting foot is the overall spatial posture of the leg corresponding to the combination of the current rotation angles of the joints of the supporting foot.

[0147] The attitude adjustment space range is the set of all three-dimensional spatial positions that the supporting feet can reach when the fuselage reaches the desired attitude after feedforward correction.

[0148] The peak energy time and direction of the resonance suppression reference quantity in the gait phase are the time point when the abnormal vibration energy that causes resonance risk reaches its maximum value within the gait cycle and the spatial direction in which the vibration energy is transmitted along the fuselage.

[0149] The displacement path of the supporting foot is the continuous spatial trajectory of the foot end as it adjusts its posture during a single-step support phase.

[0150] Phase marking is a spatial positioning operation that marks the position corresponding to the resonant energy peak on the foot displacement path according to the phase division of the gait cycle.

[0151] Vibration suppression path nodes are spatial locations on the displacement path that correspond to the peak moment of resonance energy. Foot movements at these locations will exacerbate the resonance amplitude of the fuselage.

[0152] Spatial intersection is the common part of two overlapping three-dimensional spatial regions, which simultaneously satisfy the spatial constraints of both regions.

[0153] The effective spatial range is the area of ​​space that the foot can reach, which simultaneously meets the needs of foot force correction and posture adjustment.

[0154] The remaining space area is the space area where the foot position can be safely adjusted after removing the area that would cause increased resonance from the effective space location range.

[0155] The target position compensation amount is the three-dimensional spatial displacement vector required to support the foot to move from its current position to the optimal position within the remaining space region.

[0156] Furthermore, the current three-dimensional coordinates of the contact point between the foot end and the ground in the fuselage coordinate system are collected. At the same time, the orientation of the three-dimensional spatial force vector corresponding to the force correction amount at the foot end is retrieved. Based on the fixed length of each link of the supporting foot leg and the limit constraints of the rotation angle of each joint, the maximum spatial boundary of the foot end that can produce positional displacement under the force correction is deduced through kinematic geometric conversion of the series link along the direction of the force vector, and the spatial displacement range corresponding to the force correction amount at the foot end is obtained.

[0157] Extract the target angle values ​​of the three axes of fuselage pitch, roll, and yaw included in the expected feedforward attitude correction. Use these target angles as a reference to calibrate the spatial pose reference of the support leg base. In the complete process of the fuselage rotating from the current attitude to the target attitude, deduce all the three-dimensional spatial points that the support foot can reach along with the leg base. Integrate the outer envelope area of ​​all reachable points into the attitude adjustment space range corresponding to the expected feedforward attitude correction.

[0158] Extract the energy peak time point of the resonance suppression reference quantity in the phase corresponding to the current gait cycle and the direction of vibration energy transmission along the fuselage. On the complete displacement trajectory of the single-step support phase at the foot end, match the spatial point corresponding to the energy peak time point and complete the position marking to obtain the vibration suppression path node.

[0159] The spatial offset range and the attitude adjustment spatial range are uniformly mapped to the same fuselage coordinate system for regional superposition and comparison. The common part of the two spatial regions that overlap is selected as the effective spatial position range of the support foot that simultaneously meets the requirements of force correction and attitude adjustment.

[0160] The effective spatial location range is screened, and all spatial areas that coincide with the vibration suppression path nodes are eliminated. The remaining safe space that will not cause resonance amplification is retained as the remaining space area for the support foot.

[0161] Using the remaining space area as an optional boundary for foot position adjustment, and combining it with the current position coordinates of the supporting foot, a target point within the area that is closest to the current position and meets the requirements of the foot's force line orientation is selected. The magnitude and spatial direction of the displacement required for the foot to move to this point are calculated, and the target position compensation amount of the supporting foot is finally determined. S5: Based on the foot force correction amount and the target position compensation amount, calculate the expected joint drive amount of each joint in the supporting foot, generate the joint drive command of the quadruped robot, and collect the adjustment posture data and adjustment foot force vector of the quadruped robot after executing the joint drive command. In this embodiment, the step of calculating the desired joint drive amount of each joint in the supporting foot based on the foot force correction amount and the target position compensation amount, and generating the joint drive command of the quadruped robot, includes: By combining the foot force correction amount and the target position compensation amount, the desired force-position composite characteristics of the supporting foot in physical space are defined; The desired force-position composite feature is decomposed to the joint axis direction of the supporting foot to obtain the torque component and angular displacement component in each joint axis direction of the supporting foot. Work is matched between the torque component and the angular displacement component to determine the desired joint drive amount of each joint in the supporting foot; The desired joint drive quantities of each joint in the supporting leg are time-sequentially arranged into a rhythmic motion sequence to generate the joint drive commands for the quadruped robot.

[0162] Specifically, physical space is a three-dimensional physical space with the fuselage coordinate system as a reference, used to determine the actual spatial position and movable boundary range of each link, joint and foot end supporting the foot.

[0163] The desired force-position composite feature is a comprehensive spatial state that supports the foot end to simultaneously meet the target force requirements and target position requirements. It also includes vector information of the foot end force and coordinate information of the foot end position, and serves as the comprehensive input basis for subsequent joint drive parameter calculation.

[0164] The direction of the joint axis is the spatial direction extending from the rotation center axis of each rotating joint of the foot. It corresponds to the central axis of joint rotation and is the reference reference direction for calculating joint torque and angular displacement.

[0165] The torque component is the torque value obtained by decomposing the target force at the foot end along the direction of the corresponding joint axis, which characterizes the torsional strength required by the joint to output the target force at the foot end.

[0166] The angular displacement component is the rotation angle value obtained by decomposing the displacement of the foot at the target position along the direction of the corresponding joint axis. It represents the magnitude and direction of the angle that the joint needs to rotate to make the foot reach the target position.

[0167] Work matching is the process of verifying and adjusting the consistency between the direction of the torque component and the direction of rotation of the angular displacement component of the same joint, so as to ensure that the torque output of the joint is consistent with the direction of rotation and avoid energy loss and motion interference caused by the conflict between the force and the direction of displacement.

[0168] The expected joint drive quantity is the single-joint target drive parameter obtained after direction matching verification. It includes the magnitude of the torque to be output by the joint and the rotation angle to be achieved, and can be directly used to guide the output action of the joint drive component.

[0169] Timing arrangement is the process of arranging and combining the desired driving parameters of each joint according to the phase progression of the gait cycle, so that the joint movements are executed sequentially according to the preset time sequence.

[0170] A rhythmic motion sequence is a set of multi-joint linked movements divided into fixed time beats. Each beat corresponds to a set of defined joint drive parameters, which enables the leg movements to proceed in an orderly manner with uniform time steps.

[0171] The joint drive command is a set of control signals that includes the timed drive parameters of all joints of the supporting foot. It can be directly transmitted to the joint drive component to drive each joint to complete the support action in a coordinated manner according to preset requirements.

[0172] Furthermore, the force line direction indicated by the foot force correction amount is analyzed and spatially correlated with the displacement direction indicated by the target position compensation amount.

[0173] By finding the spatial intersection of the line of action of the foot end force correction and the vector of the target position compensation at the ball joint point of the supporting foot, the local force-displacement decoupling point of the supporting foot is determined, in which the foot end force correction is the desired force and the target position compensation is the desired displacement.

[0174] At the local force-displacement decoupling point, the direction of action of the foot end force correction and the displacement direction of the target position compensation together constitute the desired force-displacement composite feature of the supporting foot in the task space.

[0175] Based on the linkage structure of the supporting foot, the rotation axes of each joint from the base of the supporting foot to the ball joint point at the foot end are identified.

[0176] Based on the spatial orientation of the desired force-position composite feature at the local force-displacement decoupling point, the direction of the torque component of the foot end force correction amount along each joint axis and the direction of the angular displacement component of the target position compensation amount along each joint axis are determined.

[0177] Specifically, for the i-th joint, the corresponding foot force correction component refers to the component of the foot force correction that is perpendicular to the line connecting the joint axis and the foot ball joint point and lies in a plane perpendicular to the joint axis. This component represents the source of counteracting torque that must be provided by the joint to generate the desired force at the foot.

[0178] Meanwhile, the angular displacement caused by the target position compensation at the joint refers to the change in angle corresponding to the projection of the displacement direction of the target position compensation onto the rotation plane of the joint. This change in angle represents the spatial angle that the joint needs to rotate to reach the desired position.

[0179] Then, at each joint, the torque component generated along the joint axis by the analyzed foot force correction amount is matched with the angular displacement component caused by the target position compensation amount at that joint.

[0180] In detail, it is determined whether the directions of the torque component and the angular displacement component satisfy the same directionality. That is, when the torque generated by the foot force correction drives the foot to move in a certain direction, the angular displacement corresponding to the target position compensation also points in the same direction. Then, the expected joint drive amount of the joint is determined to be the torque component in the angular displacement direction. The magnitude of the torque component is used as the driving dominant feature, and the direction of the angular displacement component is used as the motion dominant feature.

[0181] Conversely, if the directions of the torque component and the angular displacement component are contrary, it indicates that in the current state of the joint, the foot force correction amount and the target position compensation amount are physically conflicting. It is necessary to re-adjust the posture of the supporting foot in space according to the priority of the force line direction of the foot force correction amount and the displacement direction of the target position compensation amount, so as to eliminate the state of conflicting directions until the torque component direction and the angular displacement component direction of all joints are in the same direction, thereby determining the expected joint drive amount of each joint.

[0182] Finally, the desired joint actuation values ​​determined for each joint of each supporting foot are time-sequentially arranged according to the joint numbering order and gait phase relationship of the quadruped robot. The desired joint actuation value is represented as the continuous action pulse (corresponding to the torque component of the foot force correction value) that the joint needs to output under the action of the foot force correction value, and the desired angle (corresponding to the angular displacement component of the target position compensation value) that the joint needs to achieve along with the pulse.

[0183] The continuous action pulse of each joint is combined with the desired angle to form a joint control element. The joint control elements of all joints in the supporting foot at the same time are collected to form a joint drive command frame.

[0184] Multiple joint drive command frames in a continuous time series eventually constitute a complete joint drive command. The joint drive command is sent to the corresponding joint actuator to drive the coordinated movement of each joint in the supporting foot, while simultaneously correcting the force at the foot end and compensating for the foot position.

[0185] S6: Compare the adjusted attitude data with the attitude data corresponding to the feedforward attitude correction expectation to obtain the attitude tracking deviation, and update the feedforward attitude correction expectation and the foot force correction amount based on the attitude tracking deviation feedback.

[0186] In this embodiment, the step of comparing the adjusted attitude data with the attitude data corresponding to the expected feedforward attitude correction to obtain the attitude tracking deviation includes: Using the gait period as the segmentation period, the fuselage attitude angle trajectory of the attitude adjustment data is periodically segmented to obtain a multi-period fluctuation sequence of the attitude adjustment data. Within the same gait cycle, the envelope of the desired attitude angle trajectory of the attitude data corresponding to the desired feedforward attitude correction is extracted to obtain the ideal trajectory envelope of the desired feedforward attitude correction. By identifying the deviation time periods of the multi-period fluctuation sequence from each fluctuation angle point in the ideal trajectory envelope, the attitude deviation time periods of the adjusted attitude data are obtained. During the attitude deviation period, the discrete point implementation region contained in the multi-period fluctuation sequence is extracted to obtain the transient tracking subsequence of the attitude adjustment data; The attitude tracking deviation of the adjusted attitude data is obtained by combining the maximum offset extreme value and direction of the transient tracking subsequence with the ideal trajectory envelope.

[0187] Specifically, the segmentation period is the time reference length used to segment continuous time-series data. Here, the duration of the gait period is used as the segmentation basis to divide the continuous fuselage attitude trajectory into independent segments of equal duration.

[0188] The attitude adjustment data is a set of time-series data of the attitude angles of each axis of the body collected by the onboard attitude sensing component after the quadruped robot executes the joint drive command. It reflects the actual body posture change process after the attitude adjustment action is completed.

[0189] The fuselage attitude angle trajectory is a trajectory curve formed by the continuous change of the attitude angles of the three axes of pitch, roll and yaw in the attitude data over time, which intuitively presents the fluctuating change of the fuselage attitude as the gait advances.

[0190] Periodic segmentation is a process of dividing continuous time-series trajectory data into multiple independent data segments of the same length according to a fixed time length reference, which is used to realize the correspondence and comparison of attitude data within different time periods.

[0191] The multi-period fluctuation sequence is a collection of fuselage attitude angle data segments corresponding to multiple consecutive gait cycles obtained after periodic segmentation. Each segment has the same duration and is arranged sequentially according to the walking time sequence, which is used to present the attitude fluctuation pattern during multi-walking.

[0192] The attitude data corresponding to the feedforward attitude correction expectation is a set of time-series target data containing the target attitude angles of each axis of the fuselage along the gait phase, representing the ideal attitude change process that the attitude adjustment action needs to achieve.

[0193] The desired attitude angle trajectory is a continuous target curve formed by the change of the target attitude angle of each axis in the attitude data corresponding to the expected feedforward attitude correction over time. It serves as a reference benchmark for measuring the actual attitude adjustment effect.

[0194] Envelope extraction is a process that extracts the upper and lower boundary extreme values ​​of the curve fluctuations from a continuous target trajectory curve and connects them to form a boundary profile. It is used to define the allowable fluctuation range of ideal attitude changes.

[0195] The ideal trajectory envelope is the upper and lower boundary contours of the desired attitude angular trajectory obtained by envelope extraction. The enclosed area represents the attitude fluctuation range that the feedforward attitude correction is expected to allow, and is the criterion for judging whether the actual attitude deviates.

[0196] The fluctuation angle point is the numerical position of the attitude angle at the same moment in the envelope of the multi-period fluctuation sequence and the ideal trajectory. It is used to compare the numerical difference between the actual attitude and the ideal attitude at each moment.

[0197] The deviation period is the continuous time interval in which the actual attitude angle point exceeds the boundary of the ideal trajectory envelope, corresponding to the duration of the actual attitude deviating from the ideal adjustment target.

[0198] The attitude deviation period is a continuous time interval in which the actual fuselage attitude, identified from the attitude adjustment data, exceeds the allowable range of the ideal trajectory envelope. It represents the duration of deviation during the attitude tracking process.

[0199] The discrete point implementation area is the time and angle value range covered by all discrete attitude angle sampling points contained in the multi-period fluctuation sequence within the time interval corresponding to the attitude deviation period, which is used to limit the data selection range for subsequent deviation calculation.

[0200] The transient tracking subsequence is a local data sequence composed of all discrete attitude angle sampling points extracted from the attitude deviation period of the multi-period fluctuation sequence. It reflects the transient change process of the actual attitude during the deviation period.

[0201] The maximum offset extreme value is the maximum numerical difference between the actual attitude angle and the corresponding boundary of the ideal trajectory envelope in the transient tracking subsequence, representing the maximum deviation amplitude that occurs during attitude tracking.

[0202] Deviation synthesis is a process of vector-combining the magnitude of the maximum deviation extremum with the corresponding spatial deflection direction to form a complete deviation descriptor that includes both the deviation magnitude and the deviation direction.

[0203] Attitude tracking deviation is an attitude deviation vector obtained by deviation synthesis, which includes both magnitude and spatial direction. It fully represents the degree and trend of tracking deviation between the actual adjusted attitude and the expected feedforward attitude correction.

[0204] Furthermore, using the duration of the gait cycle as the segmentation benchmark and the start time of each gait cycle as the segmentation node, the continuous trajectory of the fuselage attitude angle in the collected attitude adjustment data is segmented into segments one by one. The complete continuous attitude data is split into multiple independent data segments with the same duration corresponding one-to-one with the gait cycle, and arranged and integrated according to the walking time sequence to form a multi-cycle fluctuation sequence covering multiple continuous walking cycles.

[0205] Within the time range of a single gait cycle, all sampling points of the continuous trajectory of the target attitude angle corresponding to the expected feedforward attitude correction are traversed, and the upper and lower boundary extreme values ​​of the attitude fluctuation corresponding to each phase point are extracted. The extreme points are connected in sequence to form a continuous boundary profile, thus obtaining the ideal trajectory envelope and defining the allowable fluctuation range of the ideal attitude change.

[0206] The attitude angle sampling points at each moment in the multi-period fluctuation sequence are compared one by one with the corresponding boundary values ​​of the ideal trajectory envelope at the same moment. The continuous time intervals in which the actual attitude angle exceeds the upper and lower boundaries of the envelope are selected and marked as attitude deviation periods.

[0207] Within the marked attitude deviation time period, all actual attitude angle discrete sampling points corresponding to that time period interval are extracted, and a transient tracking subsequence containing only the deviation time period data is formed.

[0208] Finally, among all the sampling points of the transient tracking subsequence, the point with the largest difference between the actual attitude value and the boundary of the ideal trajectory envelope is located, the magnitude of the deviation and the direction of spatial deflection are determined, and the magnitude and direction are vector-combined to obtain the complete attitude tracking deviation.

[0209] First, we analyze the deviation amplitude and yaw direction of the attitude tracking error along the three orthogonal rotation axes of the fuselage: pitch, roll, and yaw.

[0210] Based on the mechanical transmission geometry between the contact points of each rotating axis of the fuselage and the foot of the support, the deviation amplitude of each axis is converted into the corresponding fuselage attitude angle compensation amount along the adjustment direction opposite to the attitude deviation direction. This compensation amount is then added to the angle value of the axis corresponding to the current feedforward attitude correction expectation, thus completing the update of the angle parameter of the feedforward attitude correction expectation.

[0211] The attitude tracking deviations of each axis are simultaneously decomposed and distributed to the three-dimensional force space of each support foot according to the lever arm length and spatial arrangement relationship of each support foot contact point relative to the fuselage center of mass. This is converted into the amplitude adjustment amount and spatial direction deflection amount of the foot end force of each support foot, and superimposed on the corresponding vector component of the current foot end force correction amount to complete the update of the force parameters of the foot end force correction amount.

[0212] After the update, the expected feedforward posture correction and the foot force correction are fed into the support foot expected foot force correction and joint drive quantity calculation stage of the next control cycle. This is used to generate the adjusted joint drive command, which drives the leg joint to perform the corrected action, thereby realizing closed-loop iterative optimization of the posture control parameters.

[0213] like Figure 2 The diagram shown is a functional block diagram of a terrain-adaptive attitude stabilization control system for a quadruped robot provided in an embodiment of the present invention.

[0214] The terrain-adaptive attitude stabilization control system 100 for a quadruped robot described in this invention can be installed in an electronic device. Depending on the functions implemented, the terrain-adaptive attitude stabilization control system 100 may include a signal acquisition module 101, a disturbance calculation module 102, a feedforward prediction module 103, an instruction generation module 104, an execution feedback module 105, and a deviation correction module 106. The module described in this invention can also be referred to as a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, and are stored in the memory of the electronic device.

[0215] In this embodiment, the functions of each module / unit are as follows: Signal acquisition module 101: Acquires the time spectrum diagram and body state data corresponding to the multi-source vibration signals of the quadruped robot in the current terrain, and identifies terrain change events based on the change direction of the force vector at the end of the foot in the quadruped robot; Disturbance calculation module 102: Based on the body state data before the terrain change event, calculate the disturbed dynamic torque vector of the quadruped robot to obtain the terrain disturbance vector of the quadruped robot; Feedforward prediction module 103: Based on the terrain disturbance vector and the body state data, predict the attitude deviation trend of the quadruped robot, generate the feedforward attitude correction expectation, and based on each gait cycle of the quadruped robot, compare the body wave sequence in the body state data and the frequency band energy sequence of the time spectrum to obtain the resonance suppression reference quantity of the quadruped robot. Instruction generation module 104: Using the feedforward attitude correction expectation and the resonance suppression reference, correct the expected foot end force of the supporting foot in the quadruped robot to obtain the foot end force correction amount, and calculate the target position compensation amount of the supporting foot based on the current position of the supporting foot. Execution feedback module 105: Based on the foot force correction amount and the target position compensation amount, calculate the expected joint drive amount of each joint in the supporting foot, generate the joint drive command of the quadruped robot, and collect the adjustment posture data and adjustment foot force vector of the quadruped robot after executing the joint drive command; Deviation correction module 106: compares the adjusted attitude data with the attitude data corresponding to the feedforward attitude correction expectation to obtain the attitude tracking deviation, and updates the feedforward attitude correction expectation and the foot force correction amount based on the attitude tracking deviation feedback.

[0216] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0217] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0218] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0219] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0220] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0221] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A terrain-adaptive attitude stabilization control method for a quadruped robot, characterized in that, The method includes: S1: Collect the time-spectrum diagram and body state data corresponding to the multi-source vibration signals of the quadruped robot in the current terrain, and identify terrain change events based on the change direction of the force vector at the foot end of the quadruped robot; S2: Based on the body state data before the terrain change event, calculate the disturbed dynamic torque vector of the quadruped robot to obtain the terrain disturbance vector of the quadruped robot; S3: Based on the terrain disturbance vector and the body state data, predict the attitude deviation trend of the quadruped robot, generate the feedforward attitude correction expectation, and based on each gait cycle of the quadruped robot, compare the body wave sequence in the body state data with the frequency band energy sequence of the time spectrum to obtain the resonance suppression reference value of the quadruped robot. S4: Using the feedforward attitude correction expectation and the resonance suppression reference, correct the expected foot end force of the supporting foot in the quadruped robot to obtain the foot end force correction amount, and calculate the target position compensation amount of the supporting foot based on the current position of the supporting foot. S5: Based on the foot force correction amount and the target position compensation amount, calculate the expected joint drive amount of each joint in the supporting foot, generate the joint drive command of the quadruped robot, and collect the adjustment posture data and adjustment foot force vector of the quadruped robot after executing the joint drive command. S6: Compare the adjusted attitude data with the attitude data corresponding to the feedforward attitude correction expectation to obtain the attitude tracking deviation, and update the feedforward attitude correction expectation and the foot force correction amount based on the attitude tracking deviation feedback.

2. The terrain-adaptive attitude stabilization control method for a quadruped robot as described in claim 1, characterized in that, The step of identifying terrain change events based on the abrupt change direction of the force vector at the foot end of the quadruped robot includes: Based on the foot force vector of the quadruped robot in the body state data, the directional fluctuation range of the foot force vector is defined to obtain the directional stable range of the foot force vector; Within the stable directional range, the directional deviation of the foot force vector is detected to obtain a sudden change directional indicator; A time correlation analysis is performed on the abrupt change direction marker of each supporting foot of the quadruped robot to obtain the foot end contact perturbation event of the supporting foot. By comparing the posture changes of the quadruped robot before and after the foot contact disturbance event, the terrain change event of the quadruped robot is obtained.

3. The terrain-adaptive attitude stabilization control method for a quadruped robot as described in claim 2, characterized in that, The step of calculating the disturbed dynamic moment vector of the quadruped robot based on the body state data before the terrain abrupt event, and obtaining the terrain disturbance vector of the quadruped robot, includes: By using the instantaneous points marked before the occurrence of the terrain abrupt change event, the spatial distribution of the foot force vector is identified, and the spatial distribution pattern of the foot force vector is obtained; Based on the position of the torque application point in the spatial distribution pattern, the moment and rotation direction of the sudden jump in angular velocity amplitude in the body state data are identified to obtain the angular velocity jump event of the quadruped robot. By associating the point of application of torque in the spatial distribution pattern with the rotational tendency indicated by the angular velocity jump event, the direction of the disturbed action of the quadruped robot can be obtained. The stress gradient difference of the support foot is analyzed along the direction of the disturbed action, and the balance holding torque set of the support foot is calculated. By comparing the balance-maintaining torque set with the torque distribution of the supporting leg after the mutation occurs, the amplitude of the disturbed dynamic torque of the quadruped robot is obtained. By associating the direction of the disturbed action with the amplitude of the disturbed dynamic torque, the terrain disturbance vector of the quadruped robot is obtained.

4. The terrain-adaptive attitude stabilization control method for a quadruped robot as described in claim 3, characterized in that, The step of predicting the attitude deviation trend of the quadruped robot based on the terrain disturbance vector and the body state data, and generating a feedforward attitude correction expectation, includes: Based on the terrain disturbance vector, the direction of the moment of the support foot is analyzed to obtain the direction of the unbalanced moment of the support foot; Based on the direction of the unbalanced moment and the current position of the supporting foot, the ideal spatial orientation of the force vector at the foot end is derived, and the ideal force line direction of the supporting foot is obtained. By comparing the ideal force line direction with the actual direction of the foot force vector, the displacement deviation of the supporting force line of the supporting foot is obtained; Using the displacement deviation of the support force line and the body state data, the offset trend of the quadruped robot's body posture is predicted, and the posture offset prediction of the support foot is obtained. Using the predicted attitude offset as a compensation feature, the expected angle of the body attitude in the quadruped robot is adjusted in reverse to obtain the expected feedforward attitude correction of the supporting leg.

5. The terrain-adaptive attitude stabilization control method for a quadruped robot as described in claim 4, characterized in that, The step of comparing the body fluctuation sequence in the body state data with the frequency band energy sequence in the time-spectrum diagram based on the gait cycles of the quadruped robot to obtain the resonance suppression reference value of the quadruped robot includes: Based on the various gait cycles of the quadruped robot, the body wave sequence of the quadruped robot is time-aligned with the frequency band energy sequence of the time spectrum diagram to obtain a synchronization sequence pair; In the synchronization sequence pair, the overlapping positions of the fluctuation peak of the fuselage fluctuation sequence and the energy peak of the frequency band energy sequence are compared, and the energy at the overlapping positions is marked as the fluctuation-induced frequency band energy; Determine the main wave frequency range from the fuselage wave sequence; The oscillation frequency band energy is obtained by determining the frequency band energy that matches the main wave frequency range from the wave-induced frequency band energy. The energy segments in the oscillation frequency band that conform to the conventional fluctuation pattern of the fuselage fluctuation sequence are removed, and the remaining abnormal energy segments are used as the resonance suppression reference for the quadruped robot.

6. The terrain-adaptive attitude stabilization control method for a quadruped robot as described in claim 5, characterized in that, The step of correcting the expected foot force of the supporting foot in the quadruped robot by using the expected feedforward attitude correction and the resonance suppression reference value to obtain the foot force correction amount includes: The feedforward attitude correction expectation is decomposed along each axis of the quadruped robot's body coordinate system to obtain the quadruped robot's directional attitude recovery requirement; Based on the foot layout of the quadruped robot, the directional posture recovery requirement is mapped to the desired vector direction of the supporting foot to obtain the posture compensation amplitude of the supporting foot. Based on the phase interval of the resonance suppression reference value in each gait cycle, the dominant action space of the resonance suppression reference value in the supporting foot is identified, and the resonance sensitive phase of the supporting foot is obtained. Using the resonant sensitive phase as the priority channel for force line adjustment, the attitude compensation amplitude is deflected along the main energy transfer direction of the resonant suppression reference quantity to obtain the anti-resonance force line deflection amount of the support foot; By combining the attitude compensation amplitude and the anti-resonance force line deflection, the foot end force vector correction requirement of the supporting foot is obtained; The foot force vector correction requirement and the expected foot force of the supporting foot are superimposed to obtain the foot force correction amount of the supporting foot.

7. The terrain-adaptive attitude stabilization control method for a quadruped robot as described in claim 6, characterized in that, The step of calculating the target position compensation amount of the support foot based on its current position includes: Based on the current position of the supporting foot and the spatial orientation of the foot end force correction amount, the supporting foot is displacement mapped to obtain the spatial offset range of the foot end force correction amount; Based on the desired angle in the feedforward attitude correction expectation, the current attitude of the supporting foot is calibrated to generate the attitude adjustment space range corresponding to the feedforward attitude correction expectation; Based on the energy peak time and direction of the resonance suppression reference quantity in the gait phase, the displacement path of the supporting foot is phase-marked to obtain the vibration suppression path nodes; The spatial intersection of the spatial action offset range and the attitude adjustment spatial range is taken as the effective spatial position range of the supporting foot. The remaining space area of ​​the support foot is obtained by removing the area that falls into the vibration suppression path node from the effective spatial location range; Using the remaining space area as the target, and combining it with the current position of the support foot, determine the target position compensation amount for the current position of the support foot.

8. The terrain-adaptive attitude stabilization control method for a quadruped robot as described in claim 7, characterized in that, The step of calculating the desired joint actuation values ​​of each joint in the supporting foot based on the foot force correction amount and the target position compensation amount, and generating the joint actuation commands for the quadruped robot, includes: By combining the foot force correction amount and the target position compensation amount, the desired force-position composite characteristics of the supporting foot in physical space are defined; The desired force-position composite feature is decomposed to the joint axis direction of the supporting foot to obtain the torque component and angular displacement component in each joint axis direction of the supporting foot. Work is matched between the torque component and the angular displacement component to determine the desired joint drive amount of each joint in the supporting foot; The desired joint drive quantities of each joint in the supporting leg are time-sequentially arranged into a rhythmic motion sequence to generate the joint drive commands for the quadruped robot.

9. The terrain-adaptive attitude stabilization control method for a quadruped robot as described in claim 8, characterized in that, The comparison of the adjusted attitude data with the attitude data corresponding to the expected feedforward attitude correction yields the attitude tracking deviation, including: Using the gait period as the segmentation period, the fuselage attitude angle trajectory of the attitude adjustment data is periodically segmented to obtain a multi-period fluctuation sequence of the attitude adjustment data. Within the same gait cycle, the envelope of the desired attitude angle trajectory of the attitude data corresponding to the desired feedforward attitude correction is extracted to obtain the ideal trajectory envelope of the desired feedforward attitude correction. By identifying the deviation time periods of the multi-period fluctuation sequence from each fluctuation angle point in the ideal trajectory envelope, the attitude deviation time periods of the adjusted attitude data are obtained. During the attitude deviation period, the discrete point implementation region contained in the multi-period fluctuation sequence is extracted to obtain the transient tracking subsequence of the attitude adjustment data; The attitude tracking deviation of the adjusted attitude data is obtained by combining the maximum offset extreme value and direction of the transient tracking subsequence with the ideal trajectory envelope.

10. A terrain-adaptive attitude stabilization control system for a quadruped robot, characterized in that, The system includes: Signal acquisition module: Acquires the time spectrum diagram and body state data corresponding to the multi-source vibration signals of the quadruped robot in the current terrain, and identifies terrain change events based on the change direction of the force vector at the end of the foot in the quadruped robot; Disturbance calculation module: Based on the body state data before the terrain change event, calculate the disturbed dynamic torque vector of the quadruped robot to obtain the terrain disturbance vector of the quadruped robot; Feedforward prediction module: Based on the terrain disturbance vector and the body state data, predict the attitude deviation trend of the quadruped robot, generate the expected feedforward attitude correction, and based on each gait cycle of the quadruped robot, compare the body wave sequence in the body state data with the frequency band energy sequence of the time spectrum to obtain the resonance suppression reference value of the quadruped robot. Instruction generation module: Using the feedforward attitude correction expectation and the resonance suppression reference, correct the expected foot end force of the supporting foot in the quadruped robot to obtain the foot end force correction amount, and calculate the target position compensation amount of the supporting foot based on the current position of the supporting foot. Execution feedback module: Based on the foot force correction amount and the target position compensation amount, calculate the expected joint drive amount of each joint in the supporting foot, generate the joint drive command of the quadruped robot, and collect the adjustment posture data and adjustment foot force vector of the quadruped robot after executing the joint drive command; Deviation correction module: compares the adjusted attitude data with the attitude data corresponding to the feedforward attitude correction expectation to obtain the attitude tracking deviation, and updates the feedforward attitude correction expectation and the foot force correction amount based on the attitude tracking deviation feedback.