Adaptive gait control method and system for quadruped robot

Abstract

The present application relates to the technical field of robot gait control, and discloses a kind of quadruped robot adaptive gait control method and system, including obtaining the historical climbing data of quadruped robot and carrying out classification screening, constructs standard reference library;According to the latest climbing data of quadruped robot under various step parameters, the three-dimensional difference of standard landing point sequence is calculated, and the deviation sequence is formed, according to the deviation sequence, the deviation coefficient of mechanical foot is calculated, and the landing point compensation library is constructed;The landing point coordinates of quadruped robot and the corresponding mechanical foot are obtained, the corresponding deviation coefficient is extracted, the landing point coordinates are corrected, and the landing point reference sequence of quadruped robot climbing step is obtained.The present application improves the stability and generalization ability of robot step climbing by constructing deviation sequence and deviation coefficient;By matching deviation coefficient to correct three-dimensional coordinates one by one, the motion error of robot is offset, and the efficiency of quadruped robot step climbing is improved.

Description

Technical Field

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

[0002] As the demand for service robots continues to expand, stairs, as one of the most typical unstructured terrains in indoor and outdoor environments, present a significant challenge to the stable climbing of robots due to the differences in height and width of stairs in different scenarios.

[0003] The robot's joint drive suffers from transmission backlash and motion errors caused by load variations. The actual landing point of a single climbing action will deviate from the theoretical landing point in three-dimensional coordinates. Existing solutions for gait planning compensation strategies are mostly empirical adjustments, which cannot adaptively generate the optimal compensation scheme based on different step parameters. This results in weak generalization ability, leading to a continuous amplification of errors during multi-step climbing and reducing the climbing efficiency of quadruped robots. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides an adaptive gait control method and system for quadruped robots.

[0005] In a first aspect, the present invention provides an adaptive gait control method for a quadruped robot, the method comprising: Historical climbing data of a quadruped robot is acquired, classified, and filtered to construct a standard reference library, which includes step parameters and standard landing sequence corresponding to the mechanical legs. Based on the latest climbing data of the quadruped robot under various step parameters, the three-dimensional difference with the standard landing sequence is calculated to form a deviation sequence. Based on the deviation sequence, the deviation coefficient of the mechanical foot is calculated to construct a landing compensation library. In response to the quadruped robot performing a step climbing task, the coordinates of the landing point of the quadruped robot and the corresponding mechanical leg are obtained. The corresponding deviation coefficient is extracted from the landing point compensation library, and the coordinates of the landing point are corrected to obtain the reference sequence of landing points for the quadruped robot climbing the steps.

[0006] Furthermore, the step of acquiring and classifying the historical climbing data of the quadruped robot to construct a standard reference library includes: The running data recorded by the quadruped robot during the process of climbing the stairs is used as historical climbing data. The historical climbing data includes the stair parameters and the coordinates of the quadruped robot's landing point. The historical climbing data is classified according to the step parameters, and the landing coordinates of the quadruped robot corresponding to different step parameters are extracted. The landing point coordinates are arranged in ascending order according to the chronological sequence of the quadruped robot's climbing actions, and a corresponding mechanical leg is added to each landing point coordinate to form a standard landing point sequence. Using the step parameters as indexes and the standard landing point sequence as data, a standard reference library is constructed.

[0007] Furthermore, the step of calculating the three-dimensional difference between the standard landing point sequence and the latest climbing data of the quadruped robot under various step parameters to form a deviation sequence includes: Collect the step parameters of the current step, select the corresponding standard landing point sequence from the standard reference library, and use it as the first landing point sequence; Based on the landing point coordinates of the quadruped robot's first foot stepping on the step, calculate the three-dimensional difference between the landing point coordinates of the first foot stepping on the step in the first landing point sequence and the landing point coordinates of the first foot stepping on the step. The remaining landing point coordinates in the first landing point sequence are corrected based on the three-dimensional difference to obtain the corrected landing point sequence. Based on the landing coordinates of the mechanical foot of the quadruped robot on the step for the second time, calculate the three-dimensional difference between the landing coordinates of the mechanical foot on the step for the second time in the corrected landing sequence, and continue to correct the corrected landing sequence until all the step climbing actions are completed, and record the final corrected landing sequence. The three-dimensional difference corresponding to each step-step action, the actual landing point coordinates of the mechanical foot in that action, and the specific step level contacted by the mechanical foot are organized in the order of the actions to form a deviation sequence.

[0008] Further, the step of calculating the deviation coefficient of the mechanical foot and constructing the landing point compensation library based on the deviation sequence includes: Based on trigonometric relationships, calculate the corresponding angle of the actual landing point coordinates relative to the fuselage coordinate system in each set of data in the deviation sequence; Separate the step parameters and the corresponding step layer number of mechanical foot contact for each set of data from the deviation sequence, and calculate the relative angle of the actual landing point coordinates in the fuselage coordinate system based on trigonometric function relationships. The deviation coefficient of the mechanical foot at the relative angle is obtained by dividing the actual landing point coordinates of the mechanical foot with the corresponding landing point coordinates in the final corrected landing point sequence. Based on the deviation coefficients corresponding to each mechanical foot in each set of data, a landing point compensation library is constructed.

[0009] Furthermore, the step of calculating the deviation coefficient of the mechanical foot and constructing the landing point compensation library based on the deviation sequence further includes: By repeatedly climbing, the deviation coefficients at different relative angles were obtained; Using the mechanical foot and relative angle as indexes and the deviation coefficient as data, the landing point compensation library is updated to obtain the final landing point compensation library.

[0010] Furthermore, the step of obtaining the coordinates of the landing point of the quadruped robot and the corresponding mechanical leg, and extracting the corresponding deviation coefficient from the landing point compensation library includes: The control system of the quadruped robot obtains the coordinates of the landing point and the corresponding mechanical leg for the next step-stepping action to be performed. The first relative angle of the landing point coordinates in the fuselage coordinate system is calculated based on trigonometric function formulas. Using the first relative angle and the corresponding mechanical foot as query conditions, the corresponding deviation coefficient is extracted from the landing point compensation library.

[0011] Furthermore, the step of correcting the coordinates of the landing point to obtain the reference sequence of landing points for the quadruped robot climbing the steps includes: According to the corresponding coordinate axis, the extracted corresponding deviation coefficient is multiplied by the coordinates of the landing point to obtain the corrected coordinates of the landing point; By integrating the corrected landing point coordinates of all climbing actions according to the order of action execution, a reference sequence of landing points for the quadruped robot climbing the steps is obtained.

[0012] In a second aspect, the present invention provides an adaptive gait control system for a quadruped robot, the system comprising: The standard library construction module is used to acquire historical climbing data of the quadruped robot, classify and filter it, and build a standard reference library. The standard reference library includes step parameters and standard landing point sequences corresponding to the mechanical legs. The compensation library construction module is used to calculate the three-dimensional difference between the quadruped robot and the standard landing point sequence based on the latest climbing data of the quadruped robot under various step parameters, form a deviation sequence, and calculate the deviation coefficient of the mechanical foot based on the deviation sequence to construct the landing point compensation library. The compensation control module is used to respond to the quadruped robot performing a step climbing task, obtain the coordinates of the landing point of the quadruped robot and the corresponding mechanical leg, extract the corresponding deviation coefficient from the landing point compensation library, and correct the coordinates of the landing point to obtain the reference sequence of landing points of the quadruped robot climbing the steps.

[0013] Furthermore, the compensation library construction module is also used to collect the step parameters of the current step, select the corresponding standard landing point sequence from the standard reference library, and use it as the first landing point sequence; Based on the landing point coordinates of the quadruped robot's first foot stepping on the step, calculate the three-dimensional difference between the landing point coordinates of the first foot stepping on the step in the first landing point sequence and the landing point coordinates of the first foot stepping on the step. The remaining landing point coordinates in the first landing point sequence are corrected based on the three-dimensional difference to obtain the corrected landing point sequence. Based on the landing coordinates of the mechanical foot of the quadruped robot on the step for the second time, calculate the three-dimensional difference between the landing coordinates of the mechanical foot on the step for the second time in the corrected landing sequence, and continue to correct the corrected landing sequence until all the step climbing actions are completed, and record the final corrected landing sequence. The three-dimensional difference corresponding to each step-step action, the actual landing point coordinates of the mechanical foot in that action, and the specific step level contacted by the mechanical foot are organized in the order of the actions to form a deviation sequence.

[0014] Furthermore, the compensation library construction module is also used to calculate the corresponding angle of the actual landing point coordinates relative to the fuselage coordinate system in each group of data in the deviation sequence based on trigonometric function relationships; Separate the step parameters and the corresponding step layer number of mechanical foot contact for each set of data from the deviation sequence, and calculate the relative angle of the actual landing point coordinates in the fuselage coordinate system based on trigonometric function relationships. The deviation coefficient of the mechanical foot at the relative angle is obtained by dividing the actual landing point coordinates of the mechanical foot with the corresponding landing point coordinates in the final corrected landing point sequence. Based on the deviation coefficients corresponding to each mechanical foot in each set of data, a landing point compensation library is constructed.

[0015] This invention provides an adaptive gait control method and system for quadruped robots. By collecting historical climbing data of the quadruped robot, classifying it according to step parameters, and associating it with the coordinates of the landing points of the mechanical feet to construct a standard reference library, and combining the three-dimensional difference between the current step parameters and the first landing point, the method corrects the subsequent landing point sequence point by point and records the total deviation to form a deviation sequence. This avoids the continuous amplification of errors during multi-step climbing and improves the stability of the robot's step climbing. By separating relevant data from the deviation sequence, and calculating the deviation coefficients of the mechanical feet at different angles based on trigonometric functions, the method summarizes multiple sets of data to construct a landing point compensation library indexed by the mechanical feet and angles. This overcomes the limitations of traditional empirical adjustments, achieving adaptive compensation for steps with different parameters and enhancing the generalization ability of gait control. By obtaining the coordinates of the robot's next landing point and the corresponding mechanical foot, querying the compensation library to match the deviation coefficients, and correcting the three-dimensional coordinates one by one, the method offsets the motion errors caused by joint transmission clearance and load changes, improving the efficiency of the quadruped robot's step climbing. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the adaptive gait control method for a quadruped robot in an embodiment of the present invention; Figure 2This is a schematic diagram of the structure of the adaptive gait control system for a quadruped robot in an embodiment of the present invention.

[0017] Figure label: 10. Standard library construction module; 20. Compensation library construction module; 30. Compensation control module. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Please see Figure 1 The first embodiment of the present invention proposes an adaptive gait control method for a quadruped robot, comprising steps S10 to S30: Step S10: Obtain historical climbing data of the quadruped robot and classify and filter it to construct a standard reference library. The standard reference library includes step parameters and standard landing point sequences corresponding to the mechanical legs. Step S20: Based on the latest climbing data of the quadruped robot under various step parameters, calculate the three-dimensional difference with the standard landing sequence to form a deviation sequence, and calculate the deviation coefficient of the mechanical foot based on the deviation sequence to construct a landing compensation library. Step S30: In response to the quadruped robot performing the step climbing task, the coordinates of the landing point of the quadruped robot and the corresponding mechanical leg are obtained, the corresponding deviation coefficient is extracted from the landing point compensation library, and the coordinates of the landing point are corrected to obtain the quadruped robot climbing step landing point reference sequence.

[0020] In this embodiment, a standard reference library is first constructed based on the historical climbing data of the quadruped robot. The specific steps include: The running data recorded by the quadruped robot during the process of climbing the stairs is used as historical climbing data. The historical climbing data includes the stair parameters and the coordinates of the quadruped robot's landing point. The historical climbing data is classified according to the step parameters, and the landing coordinates of the quadruped robot corresponding to different step parameters are extracted. The landing point coordinates are arranged in ascending order according to the chronological sequence of the quadruped robot's climbing actions, and a corresponding mechanical leg is added to each landing point coordinate to form a standard landing point sequence. Using the step parameters as indexes and the standard landing point sequence as data, a standard reference library is constructed.

[0021] In this embodiment, the operational data recorded by the quadruped robot during its stair climbing process is collected as historical climbing data. This historical climbing data selects the operational data of the quadruped robot's stable climbing. The operational data includes stair parameters and the quadruped robot's landing point coordinates. The stair parameters include the number of steps, the step height, and the step width. The historical climbing data is categorized according to the stair parameters, and the landing point coordinates of each step-stepping action of the quadruped robot corresponding to different stair parameters are extracted. The landing point coordinates are then arranged in ascending order according to the chronological order of the quadruped robot's climbing actions, forming a landing point sequence. Taking the target staircase with 3 steps, a step height of 15cm, and a step width of 30cm as an example, the landing point sequence corresponding to the target staircase includes the step-stepping actions in the following order: L1, R1, L1, R1, L2, R2, L1, R1, L2, R2, L1, R1, L2, R2. That is, the robot completes the 3-step climb by alternately stepping on the staircase with its mechanical legs.

[0022] For each landing point coordinate in the landing point sequence, a corresponding mechanical foot is added. The mechanical feet include the upper left foot L1, the upper right foot R1, the lower left foot L2, and the lower right foot R2. For example, the first landing point coordinate (30cm, 10cm, 15cm) in the sequence is associated with L1, the second landing point coordinate (30cm, -10cm, 15cm) is associated with R1, the third landing point coordinate (60cm, 10cm, 30cm) is associated with L1, the fourth landing point coordinate (60cm, -10cm, 30cm) is associated with R1, the fifth landing point coordinate (30cm, 10cm, 30cm) is associated with L2, and so on to complete the association of all landing points with mechanical feet. In this embodiment, the origin of the coordinate system is defined as the midpoint of the side length of the first step. The x-axis represents the front and back coordinates, the y-axis represents the left and right coordinates, and the z-axis represents the height.

[0023] The step parameters and their corresponding landing sequence are stored in a structured format with the step parameters as the index, the corresponding landing sequence and the associated mechanical foot as the data, and a standard reference library is constructed. It can be understood that the landing sequence associated with the mechanical foot is the standard landing sequence.

[0024] Then, based on the latest state of the quadruped robot, the three-dimensional difference from the standard landing point sequence is calculated to form a deviation sequence. The specific steps include: Collect the step parameters of the current step, select the corresponding standard landing point sequence from the standard reference library, and use it as the first landing point sequence; Based on the landing point coordinates of the quadruped robot's first foot stepping on the step, calculate the three-dimensional difference between the landing point coordinates of the first foot stepping on the step in the first landing point sequence and the landing point coordinates of the first foot stepping on the step. The remaining landing point coordinates in the first landing point sequence are corrected based on the three-dimensional difference to obtain the corrected landing point sequence. Based on the landing coordinates of the mechanical foot of the quadruped robot on the step for the second time, calculate the three-dimensional difference between the landing coordinates of the mechanical foot on the step for the second time in the corrected landing sequence, and continue to correct the corrected landing sequence until all the step climbing actions are completed, and record the final corrected landing sequence. The three-dimensional difference corresponding to each step-step action, the actual landing point coordinates of the mechanical foot in that action, and the specific step level contacted by the mechanical foot are organized in the order of the actions to form a deviation sequence.

[0025] In this embodiment, the standard reference library is constructed based on the stable climbing data of the quadruped robot. Therefore, the standard landing point sequence in the standard reference library can be used as the benchmark data. Based on the quadruped robot's most recent climbing data, the three-dimensional difference between the benchmark data and the benchmark data is calculated. Specifically, the parameters of the step to be climbed are collected using a three-dimensional vision sensor installed on the quadruped robot's head. Using the first mechanical leg of the quadruped robot stepping on the step and the parameters of the step to be climbed as an index, a matching standard landing point sequence is filtered in the standard reference library. Then, the coordinates of the first mechanical leg landing on the step are collected, and the three-dimensional difference between this and the coordinates of the first mechanical leg landing on the step in the standard landing point sequence is calculated. The three-dimensional difference includes the X-axis coordinate difference, Y-axis coordinate difference, and Z-axis coordinate difference.

[0026] Then, based on the three-dimensional difference of the mechanical foot's landing point coordinates when it first steps on the step, the X-axis, Y-axis, and Z-axis coordinates of all subsequent landing points in the standard reference library for this landing point sequence are adjusted accordingly. The mechanical foot's landing point coordinates when it steps on the step again are collected, and the X-axis, Y-axis, and Z-axis coordinates of the modified second landing point position in the corresponding landing point sequence in the standard reference library are extracted. The three-dimensional difference between this and the mechanical foot's landing point coordinates in the landing point sequence in the standard reference library is calculated. Based on the three-dimensional difference of the mechanical foot's landing point coordinates when it steps on the step, the remaining landing point coordinates in the landing point sequence in the standard reference library are adjusted point by point. The above correction process is repeated until the quadruped robot completes the entire step climbing action. During the climbing process, the three-dimensional difference corresponding to each step-step action, the actual landing point coordinates of the mechanical foot for that action, and the specific step level contacted by the mechanical foot are recorded in real time and organized into a deviation sequence according to the action sequence.

[0027] For example, the 3D vision sensor on the quadruped robot's head is activated. The sensor scans the current step to be climbed, collecting parameters such as the number of steps, height, and width. After data processing, the closest target parameters for the current step are confirmed to be 3 steps, 15cm high, and 30cm wide. The robot enters the climbing preparation state and initiates the first climbing action. The first mechanical leg to contact the step is L1. The actual landing point coordinates of L1 are collected in real time by the sensor and read as (30.2cm, 10.1cm, 14.8cm). It should be noted that the z-axis may not match the target parameter height, for example, it may be less than the standard step height. This is because weathering and wear occur in reality, and the height of the step generally changes under these conditions, resulting in fluctuations in the height of each step from the original design.

[0028] From the established standard reference library, retrieve the matching landing point sequence using "3 layers, 15cm, 30cm" as the index, and extract the standard landing point coordinates of the first mechanical foot in the sequence, i.e., the standard landing point coordinates of L1, which are (30cm, 10cm, 15cm).

[0029] Based on the formula for calculating the difference between three-dimensional coordinates, the differences between the actual landing point and the standard landing point on the X-axis, Y-axis, and Z-axis are calculated respectively: the difference on the X-axis is the actual X-coordinate minus the standard X-coordinate, i.e., 30.2cm - 30cm = +0.2cm; the difference on the Y-axis is 10.1cm - 10cm = +0.1cm; and the difference on the Z-axis is 14.8cm - 15cm = -0.2cm. This completes the calculation of the three-dimensional difference of the first landing point.

[0030] Based on the calculated three-dimensional differences, the coordinates of the subsequent seven landing points in the standard reference library are adjusted axis by axis. For example, for the second landing point in the sequence, the standard coordinates of R1 are (30cm, -10cm, 15cm). After correction, the X-axis coordinate is 30cm + 0.2cm = 30.2cm, the Y-axis coordinate is -10cm + 0.1cm = -9.9cm, and the Z-axis coordinate is 15cm - 0.2cm = 14.8cm. The corrected coordinates are (30.2cm, -9.9cm, 14.8cm). For the third landing point, the standard coordinates of L1 are (60cm, 10cm, 30cm) corrected to (60.2cm, 10.1cm, 29.8cm). This process is repeated to complete the initial correction of all subsequent landing points. The principle behind this embodiment, which uses 3D difference to synchronously correct the coordinates of subsequent landing points, is that when a quadruped robot climbs a step, the 3D difference of the first landing point mainly originates from the robot's overall pose estimation error, such as the measurement deviation of the vision system's global coordinates of the step, and the initial posture deviation of the robot body. These errors have the same sign and magnitude trend in affecting the landing points of all subsequent mechanical legs. Therefore, using this difference to synchronously correct the coordinates of subsequent landing points can quickly eliminate global deviations and prevent errors from being disorderly amplified within the gait cycle.

[0031] The robot performs its second step-step action according to the corrected landing sequence, and the associated mechanical foot is R1 at this time.

[0032] The actual landing point coordinates of R1 were collected by the sensor as (30.3cm, -9.8cm, 14.7cm).

[0033] Extract the corrected coordinates (30.2cm, -9.9cm, 14.8cm) of the second landing point from the corrected landing point sequence, and recalculate the three-dimensional difference: X-axis: 30.3cm - 30.2cm = +0.1cm, Y-axis: -9.8cm - (-9.9cm) = +0.1cm, Z-axis: 14.7cm - 14.8cm = -0.1cm.

[0034] Following the above process, the remaining landing points are sequentially subjected to the "correction, execution, and difference calculation" operations until the robot completes all three steps of climbing. Throughout the climbing process, the three-dimensional difference corresponding to each step-step action, the identifier of the mechanical foot executing the action, and the specific step level contacted by the mechanical foot are recorded in real time. This data is then organized according to the order of action execution to form a complete deviation sequence. Some key data in this sequence are as follows: Action 1 corresponds to L1, 1 step, three-dimensional difference (+0.2cm, +0.1cm, -0.2cm); Action 2 corresponds to R1, 1 step, three-dimensional difference (+0.1cm, +0.1cm, -0.1cm); Action 3 corresponds to L2, 2 steps, three-dimensional difference (+0.15cm, -0.05cm, -0.15cm); Action 4 corresponds to R2, 2 steps, three-dimensional difference (+0.08cm, -0.06cm, -0.12cm).

[0035] The above steps yield the final corrected landing point sequence, the three-dimensional difference corresponding to each step-step action, the actual landing point coordinates of the mechanical foot for that action, and the specific step level that the mechanical foot contacts. Finally, these are organized into a deviation sequence according to the action sequence.

[0036] Then, based on the deviation sequence, the deviation coefficient is calculated, and the landing point compensation library is constructed. The specific steps include: Based on trigonometric relationships, calculate the corresponding angle of the actual landing point coordinates relative to the fuselage coordinate system in each set of data in the deviation sequence; Separate the step parameters and the corresponding step layer number of mechanical foot contact for each set of data from the deviation sequence, and calculate the relative angle of the actual landing point coordinates in the fuselage coordinate system based on trigonometric function relationships. The deviation coefficient of the mechanical foot at the relative angle is obtained by dividing the actual landing point coordinates of the mechanical foot with the corresponding landing point coordinates in the final corrected landing point sequence. Based on the deviation coefficients corresponding to each mechanical foot in each set of data, a landing point compensation library is constructed.

[0037] In this embodiment, the step parameters corresponding to each set of data and the corresponding step contacted by the mechanical foot are first separated from the deviation sequence. Based on trigonometric function relationships, such as the tangent function, the angle between the line connecting the coordinates to the origin and the coordinate axis is calculated. The relative angle of the landing point coordinates in each set of data with respect to the fuselage coordinate system is calculated. For example, the actual landing point coordinates of L1 in the first action are (30.2cm, 10.1cm, 14.8cm), and the angle relative to the fuselage coordinate system is calculated to be 30°. The actual landing point coordinates of L1 in the third action are (60.35cm, 10.05cm, 29.65cm), and the corresponding angle is calculated to be 18°.

[0038] For each set of data, combining the actual landing coordinates of the mechanical foot in that action with the corresponding landing coordinates in the corrected landing sequence, the deviation coefficient of the mechanical foot at the current angle is calculated by dividing the actual landing coordinates by the corresponding axis values ​​of the corrected landing coordinates. Here, the corrected landing coordinates are the landing coordinates in the final corrected landing sequence. Taking the first movement, L1, as an example, the actual landing coordinates are (30.2cm, 10.1cm, 14.8cm), and the corresponding corrected landing coordinates are (30cm, 10cm, 15cm). Dividing the actual X-axis coordinate of 30.2cm by the corrected X-axis coordinate of 30cm yields an X-axis deviation coefficient of approximately 1.0067; dividing the actual Y-axis coordinate of 10.1cm by the corrected Y-axis coordinate of 10cm yields a Y-axis deviation coefficient of 1.01; and dividing the actual Z-axis coordinate of 14.8cm by the corrected Z-axis coordinate of 15cm yields a Z-axis deviation coefficient of approximately 0.9867. Therefore, the deviation coefficients of L1 at 30° are (1.0067, 1.01, 0.9867). It should be noted that the landing coordinates of the first movement in the final corrected landing sequence are the uncorrected landing coordinates from the standard landing sequence. For ease of description, the landing coordinates in the final corrected landing sequence will be referred to as the corrected landing coordinates.

[0039] Furthermore, to improve the accuracy and generalization ability of the landing point compensation library, the robot is repeatedly executed to climb the stairs multiple times, for example, three times. The deviation coefficients generated in each task are collected, and the corresponding mechanical foot identifier and angle information are recorded for each deviation coefficient. Then, for each mechanical foot (L1, R1, L2, R2), all deviation coefficients at different angles are extracted and sorted in ascending order of angle from smallest to largest. For example, the deviation coefficients of L1 are sorted by angles of 15°, 30°, and 45°, and the deviation coefficients of R1 are sorted by angles of 12°, 28°, and 42°.

[0040] Finally, the deviation coefficients of all mechanical feet sorted in the above manner are categorized and summarized. Using "mechanical foot, angle" as a dual index and the deviation coefficient as the data value storage structure, a mechanical foot landing point compensation library is constructed. In this library, querying the deviation coefficient of L1 at 30° yields (1.0067, 1.01, 0.9867); querying the deviation coefficient of R1 at 28° yields (1.005, 0.998, 0.989). It should be noted that the deviation coefficient is correlated with the step-step sequence of landing points under this step parameter; that is, each step-step action corresponds to the deviation coefficient under this sequence. Furthermore, the landing point compensation library in this embodiment also records three-dimensional differences. These three-dimensional differences not only reconstruct the corrected landing point coordinates from the standard reference library but also retain the residual error information for each step.

[0041] After constructing the landing point compensation library through the above steps, when the quadruped robot performs a stair climbing task, the corresponding deviation coefficient is extracted from the landing point compensation library based on the coordinates of the quadruped robot's landing point and the corresponding mechanical leg, and coordinate correction is performed. The specific steps include: The control system of the quadruped robot obtains the coordinates of the landing point and the corresponding mechanical leg for the next step-stepping action to be performed. The first relative angle of the landing point coordinates in the fuselage coordinate system is calculated based on trigonometric function formulas. Using the first relative angle and the corresponding mechanical foot as query conditions, the corresponding deviation coefficient is extracted from the landing point compensation library.

[0042] In this embodiment, when the quadruped robot needs to perform a climbing task of a certain type of stairs, the robot's control system obtains the coordinates of the landing point and the corresponding mechanical foot identifier for the next step the robot will perform. For example, when the quadruped robot needs to perform a climbing task of three stairs of the same type again, the coordinates of the landing point for the first action are extracted from the standard reference library as (30cm, 10cm, 15cm), and the corresponding mechanical foot is L1.

[0043] Then, based on the X, Y, and Z axis values ​​of the landing point coordinates, the corresponding angle of the coordinates relative to the robot's body coordinate system is calculated using trigonometric function formulas. The calculated corresponding angle is 30°. The landing point compensation library is searched, and 3 steps are selected. The deviation coefficients (1.0067, 1.01, 0.9867) are matched when the mechanical foot is L1 and the angle is 30° when stepping on the step for the first time.

[0044] Then, the coordinates of the landing point are corrected according to the deviation coefficient to obtain the final landing point coordinates. The specific steps include: According to the corresponding coordinate axis, the extracted corresponding deviation coefficient is multiplied by the coordinates of the landing point to obtain the corrected coordinates of the landing point; By integrating the corrected landing point coordinates of all climbing actions according to the order of action execution, a reference sequence of landing points for the quadruped robot climbing the steps is obtained.

[0045] In this embodiment, the deviation coefficient obtained from the screening is applied to the preset landing point coordinates of the action, and the X-axis, Y-axis, and Z-axis coordinates are compensated and corrected respectively. The correction step is to multiply the coordinate of a certain axis of the landing point coordinates by the deviation coefficient of the corresponding axis. Taking the above landing point coordinates and corresponding deviation coefficients as an example, the corrected X-axis coordinate is the preset X-axis coordinate multiplied by the X-axis deviation coefficient, that is, 30cm×1.0067≈30.20cm; the corrected Y-axis coordinate is 10cm×1.01=10.10cm; and the corrected Z-axis coordinate is 15cm×0.9867≈14.80cm.

[0046] For each subsequent climbing action the robot will perform, the steps of "obtaining preset coordinates and mechanical feet, calculating corresponding angles, filtering deviation coefficients, and compensating for and correcting coordinates" are repeated. All corrected landing point coordinates are then integrated according to the action execution sequence to form a complete reference sequence for the quadruped robot's step climbing landing points. This sequence can be directly used to guide the robot in completing stable step climbing actions. It should be noted that if no corresponding step parameter is found in the standard reference library, the record with the highest step parameter similarity is selected as the most similar step, and its landing point sequence coordinates are used as the preset coordinates.

[0047] This embodiment provides an adaptive gait control method for a quadruped robot. By collecting historical climbing data of the quadruped robot, classifying it according to step parameters, and associating it with the coordinates of the mechanical foot landing points to construct a standard reference library, and combining the three-dimensional difference between the current step parameters and the first step landing point, the method corrects the subsequent landing point sequence point by point and records the entire deviation to form a deviation sequence. This avoids the continuous amplification of errors during multi-step climbing and improves the stability of the robot's step climbing. By separating relevant data from the deviation sequence, the method calculates the deviation coefficients of the mechanical feet at different angles based on trigonometric functions, and summarizes multiple sets of data to construct a landing point compensation library indexed by mechanical feet and angles. This overcomes the limitations of traditional empirical adjustments, achieving adaptive compensation for steps with different parameters and enhancing the generalization ability of gait control. By obtaining the coordinates of the robot's next landing point and the corresponding mechanical foot, querying the compensation library to match the deviation coefficients, and correcting the three-dimensional coordinates one by one, the method offsets the motion errors caused by joint transmission clearance and load changes, improving the efficiency of the quadruped robot's step climbing.

[0048] Please see Figure 2 Based on the same inventive concept, the second embodiment of this invention proposes an adaptive gait control system for a quadruped robot, comprising: The standard library construction module 10 is used to acquire historical climbing data of the quadruped robot, classify and filter it, and construct a standard reference library. The standard reference library includes step parameters and standard landing point sequences corresponding to the mechanical feet. The compensation library construction module 20 is used to calculate the three-dimensional difference between the standard landing point sequence and the latest climbing data of the quadruped robot under various step parameters, form a deviation sequence, and calculate the deviation coefficient of the mechanical foot based on the deviation sequence to construct the landing point compensation library. The compensation control module 30 is used to respond to the quadruped robot performing a step climbing task, obtain the coordinates of the landing point of the quadruped robot and the corresponding mechanical leg, extract the corresponding deviation coefficient from the landing point compensation library, and correct the coordinates of the landing point to obtain the reference sequence of landing points of the quadruped robot climbing the steps.

[0049] Furthermore, the compensation library construction module 20 is also used to collect the step parameters of the current step, select the corresponding standard landing point sequence from the standard reference library, and use it as the first landing point sequence; Based on the landing point coordinates of the quadruped robot's first foot stepping on the step, calculate the three-dimensional difference between the landing point coordinates of the first foot stepping on the step in the first landing point sequence and the landing point coordinates of the first foot stepping on the step. The remaining landing point coordinates in the first landing point sequence are corrected based on the three-dimensional difference to obtain the corrected landing point sequence. Based on the landing coordinates of the mechanical foot of the quadruped robot on the step for the second time, calculate the three-dimensional difference between the landing coordinates of the mechanical foot on the step for the second time in the corrected landing sequence, and continue to correct the corrected landing sequence until all the step climbing actions are completed, and record the final corrected landing sequence. The three-dimensional difference corresponding to each step-step action, the actual landing point coordinates of the mechanical foot in that action, and the specific step level contacted by the mechanical foot are organized in the order of the actions to form a deviation sequence.

[0050] Furthermore, the compensation library construction module 20 is also used to calculate the corresponding angle of the actual landing point coordinates relative to the fuselage coordinate system in each group of data in the deviation sequence based on trigonometric function relationships; Separate the step parameters and the corresponding step layer number of mechanical foot contact for each set of data from the deviation sequence, and calculate the relative angle of the actual landing point coordinates in the fuselage coordinate system based on trigonometric function relationships. The deviation coefficient of the mechanical foot at the relative angle is obtained by dividing the actual landing point coordinates of the mechanical foot with the corresponding landing point coordinates in the final corrected landing point sequence. Based on the deviation coefficients corresponding to each mechanical foot in each set of data, a landing point compensation library is constructed.

[0051] The technical features and effects of the quadruped robot adaptive gait control system proposed in this embodiment are the same as those of the method proposed in this embodiment, and will not be repeated here. Each module in the above-mentioned quadruped robot adaptive gait control system can be implemented entirely or partially through software, hardware, or a combination thereof. Each module can be embedded in or independent of the processor in a computer device in hardware form, or it can be stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0052] In summary, the present invention proposes an adaptive gait control method and system for a quadruped robot. The method acquires and classifies historical climbing data of the quadruped robot to construct a standard reference library, which includes step parameters and standard landing point sequences corresponding to the mechanical feet. Based on the latest climbing data of the quadruped robot under various step parameters, the method calculates the three-dimensional difference between the data and the standard landing point sequences to form a deviation sequence. Based on the deviation sequence, the method calculates the deviation coefficients of the mechanical feet to construct a landing point compensation library. In response to the quadruped robot performing a step climbing task, the method acquires the coordinates of the landing point to be climbed and the corresponding mechanical feet, extracts the corresponding deviation coefficients from the landing point compensation library, and corrects the coordinates of the landing point to be climbed to obtain a reference sequence for the landing points of the quadruped robot climbing steps. This invention collects historical climbing data of a quadruped robot, classifies it according to step parameters, and constructs a standard reference library by associating the coordinates of the landing points of the mechanical feet. Combining the current step parameters with the three-dimensional difference between the first landing point and the parameters of the step to be climbed, it corrects the subsequent landing point sequence point by point and records the entire deviation to form a deviation sequence. This avoids the continuous amplification of errors during multi-step climbing and improves the stability of the robot's step climbing. By separating relevant data from the deviation sequence and calculating the deviation coefficients of the mechanical feet at different angles based on trigonometric functions, it summarizes multiple sets of data to construct a landing point compensation library indexed by mechanical feet and angles. This overcomes the limitations of traditional empirical adjustments, achieving adaptive compensation for steps with different parameters and enhancing the generalization ability of gait control. By obtaining the coordinates of the robot's next landing point and the corresponding mechanical foot, querying the compensation library to match the deviation coefficients, and correcting the three-dimensional coordinates one by one, it offsets the motion errors caused by joint transmission clearance and load changes, improving the efficiency of the quadruped robot's step climbing.

[0053] The various embodiments in this specification are described in a progressive manner. For directly identical or similar parts of the embodiments, refer to each other. Each embodiment focuses on its differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. It should be noted that the technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0054] The embodiments described above are merely preferred embodiments of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various improvements and substitutions without departing from the technical principles of this invention, and these improvements and substitutions should also be considered within the scope of protection of this application. Therefore, the scope of protection of this patent application should be determined by the scope of the claims.

Claims

1. An adaptive gait control method for a quadruped robot, characterized in that, include: Historical climbing data of a quadruped robot is acquired, classified, and filtered to construct a standard reference library, which includes step parameters and standard landing sequence corresponding to the mechanical legs. Based on the latest climbing data of the quadruped robot under various step parameters, the three-dimensional difference with the standard landing sequence is calculated to form a deviation sequence. Based on the deviation sequence, the deviation coefficient of the mechanical foot is calculated to construct a landing compensation library. In response to the quadruped robot performing a step climbing task, the coordinates of the landing point of the quadruped robot and the corresponding mechanical leg are obtained. The corresponding deviation coefficient is extracted from the landing point compensation library, and the coordinates of the landing point are corrected to obtain the reference sequence of landing points for the quadruped robot climbing the steps.

2. The adaptive gait control method for a quadruped robot according to claim 1, characterized in that, The steps of acquiring historical climbing data of the quadruped robot, classifying and filtering it, and constructing a standard reference library include: The running data recorded by the quadruped robot during the process of climbing the stairs is used as historical climbing data. The historical climbing data includes the stair parameters and the coordinates of the quadruped robot's landing point. The historical climbing data is classified according to the step parameters, and the landing coordinates of the quadruped robot corresponding to different step parameters are extracted. The landing point coordinates are arranged in ascending order according to the chronological sequence of the quadruped robot's climbing actions, and a corresponding mechanical leg is added to each landing point coordinate to form a standard landing point sequence. Using the step parameters as indexes and the standard landing point sequence as data, a standard reference library is constructed.

3. The adaptive gait control method for a quadruped robot according to claim 1, characterized in that, The step of calculating the three-dimensional difference between the standard landing point sequence and the latest climbing data of the quadruped robot under various step parameters to form a deviation sequence includes: Collect the step parameters of the current step, select the corresponding standard landing point sequence from the standard reference library, and use it as the first landing point sequence; Based on the landing point coordinates of the quadruped robot's first foot stepping on the step, calculate the three-dimensional difference between the landing point coordinates of the first foot stepping on the step in the first landing point sequence and the landing point coordinates of the first foot stepping on the step. The remaining landing point coordinates in the first landing point sequence are corrected based on the three-dimensional difference to obtain the corrected landing point sequence. Based on the landing coordinates of the mechanical foot of the quadruped robot on the step for the second time, calculate the three-dimensional difference between the landing coordinates of the mechanical foot on the step for the second time in the corrected landing sequence, and continue to correct the corrected landing sequence until all the step climbing actions are completed, and record the final corrected landing sequence. The three-dimensional difference corresponding to each step-step action, the actual landing point coordinates of the mechanical foot in that action, and the specific step level contacted by the mechanical foot are organized in the order of the actions to form a deviation sequence.

4. The adaptive gait control method for a quadruped robot according to claim 3, characterized in that, The step of calculating the deviation coefficient of the mechanical foot and constructing the landing point compensation library based on the deviation sequence includes: Based on trigonometric relationships, calculate the corresponding angle of the actual landing point coordinates relative to the fuselage coordinate system in each set of data in the deviation sequence; Separate the step parameters and the corresponding step layer number of mechanical foot contact for each set of data from the deviation sequence, and calculate the relative angle of the actual landing point coordinates in the fuselage coordinate system based on trigonometric function relationships. The deviation coefficient of the mechanical foot at the relative angle is obtained by dividing the actual landing point coordinates of the mechanical foot with the corresponding landing point coordinates in the final corrected landing point sequence. Based on the deviation coefficients corresponding to each mechanical foot in each set of data, a landing point compensation library is constructed.

5. The adaptive gait control method for a quadruped robot according to claim 4, characterized in that, The step of calculating the deviation coefficient of the mechanical foot and constructing the landing point compensation library based on the deviation sequence further includes: By repeatedly climbing, the deviation coefficients at different relative angles were obtained; Using the mechanical foot and relative angle as indexes and the deviation coefficient as data, the landing point compensation library is updated to obtain the final landing point compensation library.

6. The adaptive gait control method for a quadruped robot according to claim 1, characterized in that, The steps of obtaining the coordinates of the landing point of the quadruped robot and the corresponding mechanical leg, and extracting the corresponding deviation coefficient from the landing point compensation library include: The control system of the quadruped robot obtains the coordinates of the landing point and the corresponding mechanical leg for the next step-stepping action to be performed. The first relative angle of the landing point coordinates in the fuselage coordinate system is calculated based on trigonometric function formulas. Using the first relative angle and the corresponding mechanical foot as query conditions, the corresponding deviation coefficient is extracted from the landing point compensation library.

7. The adaptive gait control method for a quadruped robot according to claim 6, characterized in that, The step of correcting the coordinates of the landing point to obtain the reference sequence of landing points for the quadruped robot climbing the steps includes: According to the corresponding coordinate axis, the extracted corresponding deviation coefficient is multiplied by the coordinates of the landing point to obtain the corrected coordinates of the landing point; By integrating the corrected landing point coordinates of all climbing actions according to the order of action execution, a reference sequence of landing points for the quadruped robot climbing the steps is obtained.

8. An adaptive gait control system for a quadruped robot, characterized in that, include: The standard library construction module is used to acquire historical climbing data of the quadruped robot, classify and filter it, and build a standard reference library. The standard reference library includes step parameters and standard landing point sequences corresponding to the mechanical legs. The compensation library construction module is used to calculate the three-dimensional difference between the quadruped robot and the standard landing point sequence based on the latest climbing data of the quadruped robot under various step parameters, form a deviation sequence, and calculate the deviation coefficient of the mechanical foot based on the deviation sequence to construct the landing point compensation library. The compensation control module is used to respond to the quadruped robot performing a step climbing task, obtain the coordinates of the landing point of the quadruped robot and the corresponding mechanical leg, extract the corresponding deviation coefficient from the landing point compensation library, and correct the coordinates of the landing point to obtain the reference sequence of landing points of the quadruped robot climbing the steps.

9. The adaptive gait control system for a quadruped robot according to claim 8, characterized in that, The compensation library construction module is also used to collect the step parameters of the current step, select the corresponding standard landing point sequence from the standard reference library, and use it as the first landing point sequence. Based on the landing point coordinates of the quadruped robot's first foot stepping on the step, calculate the three-dimensional difference between the landing point coordinates of the first foot stepping on the step in the first landing point sequence and the landing point coordinates of the first foot stepping on the step. The remaining landing point coordinates in the first landing point sequence are corrected based on the three-dimensional difference to obtain the corrected landing point sequence. Based on the landing coordinates of the mechanical foot of the quadruped robot on the step for the second time, calculate the three-dimensional difference between the landing coordinates of the mechanical foot on the step for the second time in the corrected landing sequence, and continue to correct the corrected landing sequence until all the step climbing actions are completed, and record the final corrected landing sequence. The three-dimensional difference corresponding to each step-step action, the actual landing point coordinates of the mechanical foot in that action, and the specific step level contacted by the mechanical foot are organized in the order of the actions to form a deviation sequence.

10. The quadruped robot adaptive gait control system according to claim 9, characterized in that, The compensation library construction module is also used to calculate the corresponding angle of the actual landing point coordinates relative to the fuselage coordinate system in each group of data in the deviation sequence based on trigonometric function relationships. Separate the step parameters and the corresponding step layer number of mechanical foot contact for each set of data from the deviation sequence, and calculate the relative angle of the actual landing point coordinates in the fuselage coordinate system based on trigonometric function relationships. The deviation coefficient of the mechanical foot at the relative angle is obtained by dividing the actual landing point coordinates of the mechanical foot with the corresponding landing point coordinates in the final corrected landing point sequence. Based on the deviation coefficients corresponding to each mechanical foot in each set of data, a landing point compensation library is constructed.

Images