Real-time motion retargeting and joint limit optimization method for biomimetic robots

Abstract

The application discloses a real-time action reorientation and joint amplitude limiting optimization method for a bionic robot, and relates to the technical field of robot motion control.The method comprises the following steps: acquiring source action data and kinematic parameters and joint limiting parameters of a target bionic robot; mapping motion primitive feature vectors of the source action into initial joint driving vectors through a motion primitive topological mapping matrix; mapping the initial joint driving vectors into joint angle vectors after amplitude limiting within the joint limiting range through an inverse tangent amplitude limiting function; calculating joint deviations caused by amplitude limiting, compensating the deviations back to redundant joints through a pseudo-inverse of a Jacobian matrix, and generating final joint target angles; dynamically adjusting a compliance coefficient according to a relative degree of deviation of the initial driving angle from a median angle; and outputting the final joint target angles as control instructions.The application realizes real-time action reorientation between different kinematic configurations, and provides continuous and derivable soft amplitude limiting without impact when the joint approaches the limiting position.

Description

Technical Field

[0001] This invention relates to the field of robot motion control technology, specifically to a method for real-time motion redirection and joint amplitude limiting optimization for biomimetic robots. Background Technology

[0002] With the rapid development of biomimetic robot technology, real-time mapping of human movements or taught movements onto biomimetic robots with different kinematic configurations, i.e., motion redirection, has become an important means of human-computer interaction and robot programming.

[0003] However, due to differences between the source and target robots in terms of the number of joint degrees of freedom, link lengths, and kinematic topology, direct mapping can lead to problems such as end-effector pose deviations or joint angle overshooting. Existing methods typically employ optimization-based inverse kinematics solutions. While these methods offer high accuracy, they suffer from the following drawbacks: the iterative optimization process is computationally intensive, making it difficult to meet real-time control requirements; when joint angles approach physical limits, traditional hard truncation or linear saturation processing can cause sudden velocity changes, resulting in robot vibration or even damage; and different joints have different safety constraints and motion accuracy requirements, and existing methods lack adaptive adjustment capabilities.

[0004] Therefore, there is an urgent need for a method that can smoothly redirect source motions to the target robot in real time and has adaptive joint limiting capabilities. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to address the shortcomings of the prior art by providing a real-time motion redirection and joint amplitude limiting optimization method for biomimetic robots.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] A method for real-time motion redirection and joint limiting optimization for biomimetic robots includes the following steps:

[0008] Step S1: Obtain source motion data and kinematic parameters and joint limit parameters of the target bionic robot;

[0009] Step S2: Map the motion element feature vectors of the source motion to the initial joint drive vectors of the target robot using the motion element topology mapping matrix;

[0010] Step S3: Map the initial joint drive vector to a limited joint angle vector within the joint limit range using an arctangent limiting function containing an adjustable compliance coefficient.

[0011] Step S4: Calculate the joint deviation caused by the amplitude limiting mapping, map the joint deviation to the end pose space through the pseudo-inverse of the Jacobian matrix and assign it back to the redundant joints to generate the compensated final joint target angle.

[0012] Step S5: Dynamically adjust the compliance coefficient according to the relative degree of deviation of the initial drive angle from the joint midpoint angle to achieve adaptive behavior of tracking and amplitude limiting protection;

[0013] Step S6: Output the compensated final joint target angle as a control command to the controller of the bionic robot.

[0014] Furthermore, step S1 specifically includes the following steps:

[0015] Step S1.1: Obtain source motion data, which is a sequence of joint angles of a human or a teachable robot, wherein the joint angle at each moment is a multi-dimensional vector;

[0016] Step S1.2: Obtain the kinematic parameters of the target bionic robot, including the total number of joints, the length of each link, and the type of each joint;

[0017] Step S1.3: Obtain the joint limit parameters of the target bionic robot, including the minimum and maximum allowable angles of each joint.

[0018] Furthermore, step S2 specifically includes the following steps:

[0019] Step S2.1: Extract motion element feature vectors from the source motion data using principal component analysis. The dimension of the motion element feature vectors is less than the dimension of the joint angles of the source motion and less than the total number of joints of the target robot.

[0020] Step S2.2: Offline calibration of the motion primitive topology mapping matrix using least squares regression from the demonstration motion data;

[0021] Step S2.3: Multiply the motion element feature vectors by the motion element topology mapping matrix to obtain the initial joint drive vectors of the target robot.

[0022] Furthermore, in step S3, the arctangent amplitude limiting function is defined for each joint, with the joint midpoint angle as the center, half the range of the joint limiting range as the amplitude, and the deviation between the initial driving angle and the joint midpoint angle as the independent variable. Through arctangent transformation, the output is strictly limited to the minimum and maximum allowable angles of the joint.

[0023] Furthermore, step S4 specifically includes the following steps:

[0024] Step S4.1: Calculate the deviation vector of the joint angles before and after the amplitude limiting;

[0025] Step S4.2: Calculate the Jacobian matrix of the target robot under the current limiting angle and the Jacobian matrix under the initial driving angle, respectively;

[0026] Step S4.3: Multiply the joint deviation vector by the pseudo-inverse of the Jacobian matrix under the initial driving angle and the Jacobian matrix under the current limiting angle in turn to obtain the compensation amount. Then, superimpose the compensation amount onto the joint angle vector after limiting to obtain the final joint target angle after compensation.

[0027] Furthermore, in step S4.3, when the redundancy of the target robot is lower than a preset threshold, a pseudo-inverse matrix with a regularized damping factor is used to replace the standard pseudo-inverse matrix. The pseudo-inverse matrix with a regularized damping factor is obtained by multiplying the product of the Jacobian matrix and its transpose, adding the damping square multiplied by the identity matrix, inverting the product, and then left-multiplying by the transpose of the Jacobian matrix.

[0028] Furthermore, step S5 specifically includes the following steps:

[0029] Step S5.1: Calculate the proximity factor for each joint. The proximity factor is the absolute value of the initial driving angle deviating from the joint midpoint angle divided by half the joint limit range, and the upper limit is set to 1.

[0030] Step S5.2: Update the compliance coefficient of each joint according to the proximity factor. The update method is to multiply the baseline compliance coefficient by the update factor, where the update factor is equal to 1 plus the adaptive gain coefficient multiplied by the power of the proximity factor.

[0031] Step S5.3: Substitute the updated compliance coefficient into the arctangent limiting function of step S3 for subsequent calculations.

[0032] Furthermore, in step S5.2, the adaptive gain coefficient is determined using an offline optimization method, so that the growth rate of the compliance coefficient matches the maximum allowable angular acceleration of the joint when the joint approaches its limit.

[0033] A storage medium storing instructions that, when read by a computer, cause the computer to execute any one of the methods described above for real-time motion redirection and joint limiting optimization for biomimetic robots.

[0034] An electronic device includes a processor and the storage medium, the processor executing instructions in the storage medium.

[0035] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0036] 1. This invention achieves linear mapping from source motion to the joint space of the target robot through a motion primitive topology mapping matrix, without the need for iterative optimization, with low computational latency, and meets the requirements of real-time control.

[0037] 2. This invention uses the arctangent function to construct a nonlinear amplitude limiting mapping. The output is strictly within the joint limit range and is continuously differentiable throughout the entire domain. When the driving angle approaches the limit boundary, the output automatically smooths and approaches the limit value and the derivative is zero, thus achieving shock-free soft amplitude limiting.

[0038] 3. This invention compensates for joint deviations caused by amplitude limiting to redundant joints by using Jacobi pseudo-inverse, thereby maximizing the pose accuracy of the end effector and resolving the contradiction between amplitude limiting and accuracy.

[0039] 4. This invention introduces a dynamic adaptive adjustment mechanism, which adjusts the compliance coefficient in real time according to the degree of the joint approaching the limit, thereby achieving a balance between high-precision tracking when far from the limit and strong amplitude limiting protection when approaching the limit. Attached Figure Description

[0040] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0041] Figure 1 This is a flowchart illustrating an embodiment of the present invention;

[0042] Figure 2 This is a schematic diagram of the arctangent limiting process according to an embodiment of the present invention;

[0043] Figure 3 This is a schematic diagram of the compliance coefficient adjustment process according to an embodiment of the present invention. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0045] like Figure 1 As shown, the real-time motion redirection and joint limiting optimization method for biomimetic robots includes the following steps:

[0046] Step S1: Obtain source motion data and kinematic parameters and joint limit parameters of the target bionic robot;

[0047] Step S2: Map the motion element feature vectors of the source motion to the initial joint drive vectors of the target robot using the motion element topology mapping matrix;

[0048] Step S3: Map the initial joint drive vector to a limited joint angle vector within the joint limit range using an arctangent limiting function containing an adjustable compliance coefficient.

[0049] Step S4: Calculate the joint deviation caused by the amplitude limiting mapping, map the joint deviation to the end pose space through the pseudo-inverse of the Jacobian matrix and assign it back to the redundant joints to generate the compensated final joint target angle.

[0050] Step S5: Dynamically adjust the compliance coefficient according to the relative degree of deviation of the initial drive angle from the joint midpoint angle to achieve adaptive behavior of tracking and amplitude limiting protection;

[0051] Step S6: Output the compensated final joint target angle as a control command to the controller of the bionic robot.

[0052] Step S1 specifically includes the following steps:

[0053] Step S1.1: Obtain source motion data, which is a sequence of joint angles of a human or a teachable robot, wherein the joint angle at each moment is a multi-dimensional vector;

[0054] Step S1.2: Obtain the kinematic parameters of the target bionic robot, including the total number of joints, the length of each link, and the type of each joint;

[0055] Step S1.3: Obtain the joint limit parameters of the target bionic robot, including the minimum and maximum allowable angles of each joint.

[0056] The source motion data and the body parameters of the target bionic robot are obtained. The source motion data is a sequence of human joint angles collected by motion capture equipment or teaching robot. Each component is arranged in the kinematic chain order of the source skeleton and includes the rotation angles of the joints of the trunk, limbs and end at the corresponding time.

[0057] Simultaneously, the kinematic parameters and joint limit parameters of the target bionic robot are acquired. The kinematic parameters include: the total number of robot joints, the length of each link, the type of each joint, whether it is a rotational or translating joint, and the robot's initial zero-position attitude in the world coordinate system. The joint limit parameters include: the minimum allowable angle and the maximum allowable angle of the joint. These two limits are determined by the mechanical structure, the output range of the actuator, and the safety redundancy, respectively.

[0058] Step S2 specifically includes the following steps:

[0059] Step S2.1: Extract motion element feature vectors from the source motion data using principal component analysis. The dimension of the motion element feature vectors is less than the dimension of the joint angles of the source motion and less than the total number of joints of the target robot.

[0060] Step S2.2: Offline calibration of the motion primitive topology mapping matrix using least squares regression from the demonstration motion data;

[0061] Step S2.3: Multiply the motion element feature vectors by the motion element topology mapping matrix to obtain the initial joint drive vectors of the target robot.

[0062] Motion primitive feature vectors are extracted from the source motion data. Principal component analysis is then used to reduce the dimensionality of the source motion dataset to obtain the motion primitive feature vectors. The specific formula is as follows:

[0063]

[0064] in, Let represent the p-dimensional motion primitive feature vector extracted at time t, where p is the number of motion primitives, i.e., the feature dimension after dimensionality reduction, determined by principal component analysis. The smallest integer that results in a cumulative contribution rate greater than or equal to 95% is selected. This represents the transpose of the projection matrix formed by the eigenvectors corresponding to the first p largest eigenvalues ​​of the source action data covariance matrix. This represents the source joint angle vector at time t in step S1;

[0065] The motion primitive topology mapping matrix is ​​calibrated offline from the demonstration motion data using least squares regression. The motion primitive topology mapping matrix satisfies:

[0066]

[0067] in, Let p represent the topological mapping matrix of motion primitives, with size n×p, where n is the total number of robot joints. Each row corresponds to one joint of the target robot, and the non-zero elements in each row represent the linear combination coefficients between the joint and each motion primitive. This represents a temporary variable in the optimization process, and... Having the same size, they represent the mapping matrices to be solved. This represents the total number of moments in the demonstration action data. This represents the actual joint angle vector of the target robot collected during the demonstration action at time t. Denotes the Frobenius norm;

[0068] Multiplying the motion element feature vectors by the motion element topology mapping matrix on the left yields the initial joint drive vectors of the target robot:

[0069]

[0070] in, Let represent the initial joint drive vector of the target robot at time t.

[0071] In step S3, the arctangent amplitude limiting function is defined for each joint, with the joint midpoint angle as the center, half the range of the joint limiting range as the amplitude, and the deviation between the initial driving angle and the joint midpoint angle as the independent variable. The output is strictly limited to the minimum and maximum allowable angles of the joint through arctangent transformation.

[0072] like Figure 2 As shown, the initial joint drive vector is mapped to a limited joint angle vector within the joint's limiting range using an arctangent limiting function that includes an adjustable compliance coefficient. For the joints of the target robot, the arctangent limiting function is defined as follows:

[0073]

[0074] in, This represents the angle value of the i-th joint at time t after amplitude limiting mapping. Let represent the median angle of the i-th joint. These represent the maximum and minimum allowable angles of the i-th joint, respectively. Represents the arctangent function. This represents the adjustable compliance coefficient of the i-th joint. This represents the i-th component of the initial joint drive vector, which is the initial drive angle of the i-th joint of the target robot.

[0075] Step S4 specifically includes the following steps:

[0076] Step S4.1: Calculate the deviation vector of the joint angles before and after the amplitude limiting. The specific formula is as follows:

[0077]

[0078] in, Let represent the joint deviation vector at time t. This represents the joint angle vector after amplitude limiting, with the i-th component being... ;

[0079] Step S4.2: Calculate the Jacobian matrix of the target robot under the current limiting angle and the Jacobian matrix under the initial driving angle, respectively;

[0080] Step S4.3: Multiply the joint deviation vector sequentially by the pseudo-inverse of the Jacobian matrix under the initial driving angle and the Jacobian matrix under the current limiting angle to obtain the compensation amount. Then, superimpose the compensation amount onto the joint angle vector after limiting to obtain the final target joint angle after compensation. The specific formula is as follows:

[0081]

[0082] in, This represents the final joint target angle vector after compensation at time t. This represents the Moore-Penrose pseudoinverse of the Jacobian matrix at the current limiting angle. This represents the Jacobian matrix at the initial driving angle.

[0083] In step S4.3, when the redundancy of the target robot is lower than a preset threshold, a pseudo-inverse matrix with a regularized damping factor is used to replace the standard pseudo-inverse matrix. The pseudo-inverse matrix with a regularized damping factor is obtained by multiplying the product of the Jacobian matrix and its transpose, adding the damping square multiplied by the identity matrix, inverting the product, and then left-multiplying by the transpose of the Jacobian matrix.

[0084] When the difference between the number of joints n and the end effector degrees of freedom of the target robot is less than or equal to 2, i.e., when the redundancy is less than or equal to 2, the Jacobian matrix is ​​prone to ill-conditioning. Therefore, a pseudo-inverse matrix with a regularized damping factor is used to replace the standard pseudo-inverse. The specific formula is as follows:

[0085]

[0086] in, This represents a pseudo-inverse matrix with a damping factor, of size n×6. Let Jacobian matrix be the value at the current time. This represents the damping coefficient, a positive decimal, typically ranging from 0.01 to 0.1. This represents a 6×6 identity matrix.

[0087] like Figure 3 As shown, step S5 specifically includes the following steps:

[0088] Step S5.1: Calculate the proximity factor for each joint. The proximity factor is the absolute value of the initial driving angle deviating from the joint midpoint angle divided by half the joint limit range, and the upper limit is set to 1.

[0089] Step S5.2: Update the compliance coefficient of each joint according to the proximity factor. The update method is to multiply the baseline compliance coefficient by the update factor, where the update factor is equal to 1 plus the adaptive gain coefficient multiplied by the power of the proximity factor.

[0090] Step S5.3: Substitute the updated compliance coefficient into the arctangent limiting function of step S3 for subsequent calculations.

[0091] In step S5.2, the adaptive gain coefficient is determined using an offline optimization method, so that the growth rate of the compliance coefficient when the joint approaches its limit matches the maximum allowable angular acceleration of the joint.

[0092] The specific formula for the proximity factor is as follows:

[0093]

[0094] in, Represents the proximity factor of the i-th joint at time t;

[0095] The compliance coefficient of each joint is updated based on the proximity factor, using the following formula:

[0096]

[0097] in, This represents the compliance coefficient of the i-th joint after the update at time t. This represents the baseline compliance coefficient of the i-th joint, obtained through offline calibration, with a value ranging from 0.5 to 8. This represents the adaptive gain coefficient, with a value ranging from 0 to 10. Represents proximity factor Power of 1 This represents the index adjustment parameter, with a value range of 1 to 4;

[0098] The specific method for determining the baseline compliance coefficient includes: for each joint, a set of test actions covering its entire range of motion are collected offline, including low-speed uniform motion, sinusoidal oscillations of different amplitudes, and rapid start-stop actions. These actions are tested with different compliance coefficients, and the tracking error and the rate of change of output when approaching the limit are recorded for each compliance coefficient. The compliance coefficient with small tracking error and smooth output transition with no obvious overshoot and jitter when approaching the limit is selected as the baseline compliance coefficient for that joint. For end joints with high motion accuracy requirements, a relatively large baseline compliance coefficient is selected to maintain linearity. For joints with high load or safety sensitivity, a relatively small baseline compliance coefficient is selected to initiate soft limiting in advance.

[0099] The updated compliance coefficient Substitute the original fixed coefficients into the arctangent limiting function in step S3 for calculation at subsequent time points.

[0100] The result calculated in step S4 As the joint angle command of the current control cycle, it is sent to the underlying servo driver of the bionic robot. The arctangent limiting function in step S3 is continuous and first-order differentiable in the entire domain. When the initial driving angle approaches the joint limit boundary, the function output value smoothly approaches the limit boundary value, and the derivative of the output with respect to the input automatically decreases to zero, so that the change in the output joint angle when it approaches the limit naturally shrinks, thus achieving impact-free soft limiting.

[0101] A storage medium storing instructions that, when read by a computer, cause the computer to execute any one of the methods described above for real-time motion redirection and joint limiting optimization for biomimetic robots.

[0102] An electronic device includes a processor and the storage medium, the processor executing instructions in the storage medium.

[0103] Any combination of one or more computer-readable media may be used. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device.

[0104] The examples described herein are merely preferred embodiments of the invention and are not intended to limit the concept and scope of the invention. Any modifications and improvements made by those skilled in the art to the technical solutions of the invention without departing from the design concept of the invention should fall within the protection scope of the invention.

Claims

1. A method for real-time motion redirection and joint limiting optimization for biomimetic robots, characterized in that, Includes the following steps: Step S1: Obtain source motion data and kinematic parameters and joint limit parameters of the target bionic robot; Step S2: Map the motion element feature vectors of the source motion to the initial joint drive vectors of the target robot using the motion element topology mapping matrix; Step S3: Map the initial joint drive vector to a limited joint angle vector within the joint limit range using an arctangent limiting function containing an adjustable compliance coefficient. Step S4: Calculate the joint deviation caused by the amplitude limiting mapping, map the joint deviation to the end pose space through the pseudo-inverse of the Jacobian matrix and assign it back to the redundant joints to generate the compensated final joint target angle. Step S5: Dynamically adjust the compliance coefficient according to the relative degree of deviation of the initial drive angle from the joint midpoint angle to achieve adaptive behavior of tracking and amplitude limiting protection; Step S6: Output the compensated final joint target angle as a control command to the controller of the bionic robot; In step S3, the arctangent amplitude limiting function is defined for each joint, with the joint midpoint angle as the center, half the range of the joint limiting range as the amplitude, and the deviation between the initial driving angle and the joint midpoint angle as the independent variable. The output is strictly limited to the minimum and maximum allowable angles of the joint through arctangent transformation. Step S4 specifically includes the following steps: Step S4.1: Calculate the deviation vector of the joint angles before and after the amplitude limiting; Step S4.2: Calculate the Jacobian matrix of the target robot under the current limiting angle and the Jacobian matrix under the initial driving angle, respectively; Step S4.3: Multiply the joint deviation vector by the pseudo-inverse of the Jacobian matrix under the initial driving angle and the Jacobian matrix under the current limiting angle in turn to obtain the compensation amount. Then, superimpose the compensation amount onto the joint angle vector after limiting to obtain the final joint target angle after compensation.

2. The method according to claim 1, characterized in that, Step S1 specifically includes the following steps: Step S1.1: Obtain source motion data, which is a sequence of joint angles of a human or a teachable robot, wherein the joint angle at each moment is a multi-dimensional vector; Step S1.2: Obtain the kinematic parameters of the target bionic robot, including the total number of joints, the length of each link, and the type of each joint; Step S1.3: Obtain the joint limit parameters of the target bionic robot, including the minimum and maximum allowable angles of each joint.

3. The method according to claim 2, characterized in that, Step S2 specifically includes the following steps: Step S2.1: Extract motion element feature vectors from the source motion data using principal component analysis. The dimension of the motion element feature vectors is less than the dimension of the joint angles of the source motion and less than the total number of joints of the target robot. Step S2.2: Offline calibration of the motion primitive topology mapping matrix using least squares regression from the demonstration motion data; Step S2.3: Multiply the motion element feature vectors by the motion element topology mapping matrix to obtain the initial joint drive vectors of the target robot.

4. The method according to claim 3, characterized in that, In step S4.3, when the redundancy of the target robot is lower than a preset threshold, a pseudo-inverse matrix with a regularized damping factor is used to replace the standard pseudo-inverse matrix. The pseudo-inverse matrix with a regularized damping factor is obtained by multiplying the product of the Jacobian matrix and its transpose, adding the damping square multiplied by the identity matrix, inverting the product, and then left-multiplying by the transpose of the Jacobian matrix.

5. The method according to claim 4, characterized in that, Step S5 specifically includes the following steps: Step S5.1: Calculate the proximity factor for each joint. The proximity factor is the absolute value of the initial driving angle deviating from the joint midpoint angle divided by half the joint limit range, and the upper limit is set to 1. Step S5.2: Update the compliance coefficient of each joint according to the proximity factor. The update method is to multiply the baseline compliance coefficient by the update factor, where the update factor is equal to 1 plus the adaptive gain coefficient multiplied by the power of the proximity factor. Step S5.3: Substitute the updated compliance coefficient into the arctangent limiting function of step S3 for subsequent calculations.

6. The method according to claim 5, characterized in that, In step S5.2, the adaptive gain coefficient is determined using an offline optimization method, so that the growth rate of the compliance coefficient when the joint approaches its limit matches the maximum allowable angular acceleration of the joint.

7. A storage medium, characterized in that, The storage medium stores instructions that, when read by a computer, cause the computer to execute the real-time motion redirection and joint amplitude limiting optimization method for biomimetic robots as described in any one of claims 1-6.

8. An electronic device, characterized in that, It includes a processor and the storage medium of claim 7, wherein the processor executes instructions in the storage medium.

Images