Dynamic sole force distribution method for force-controlled quadruped robot

By modeling the dynamic sole force distribution problem of force-controlled four-legged robots into secondary optimization problems and transforming them into complementary problem models, the problem of calculation delay in the existing technology is solved, and fast and accurate force distribution is achieved, and the force is improved. The stability and task execution capabilities of four-legged robots in uncertain environments.

CN119003928APending Publication Date: 2024-11-22GUILIN UNIV OF AEROSPACE TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411124273.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-15
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

The current dynamic sole force distribution method involves complex physical models and optimization algorithms, resulting in computational delays and is not suitable for real-time applications, making it difficult to effectively adjust the sole force output of the force-controlled four-legged robot in an uncertain environment.

Method used

Model the dynamic sole force distribution problem of force-controlled four-legged robots into a quadratic optimization problem with constraints, and transform it into a complementary problem model by introducing relaxation factors and multipliers, and use complementary problem solving algorithm to simplify the calculation The process reduces the computational complexity and achieves rapid force distribution scheme.

Benefits of technology

Accurate dynamic sole force distribution solution of force-controlled quadruped robots in a short calculation time, reducing the error rate, and ensuring the stability and task execution capabilities of force-controlled quadruped robots in uncertain environments.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119003928A_ABST
    Figure CN119003928A_ABST
Patent Text Reader

Abstract

According to the dynamic plantar force distribution method for the force-controlled quadruped robot, under the supporting phase, the dynamic plantar force distribution problem of the force-controlled quadruped robot is modeled into a quadratic optimization problem model, then a relaxation factor v and multipliers u and y are introduced to enable the optimization problem model to be converted into a complementary problem model, the calculation complexity is reduced, and the calculation efficiency is improved. According to the method, the dynamic plantar force distribution of the force-controlled quadruped robot is solved under relatively simplified calculation, a solution can be obtained within relatively short calculation time, a certain error rate is reduced, and the accuracy of the dynamic plantar force distribution solution of the force-controlled quadruped robot is ensured.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of robots, and in particular to a method for distributing dynamic plantar force of a force-controlled quadruped robot. Background Art

[0002] When a force-controlled quadruped robot walks, each leg needs to dynamically distribute the force on the sole of the foot to ensure that the robot can remain stable while moving. Since the properties of the ground may change constantly, it is a challenge for the force-controlled quadruped robot to adapt to these changes. For example, when a force-controlled quadruped robot walks on uneven ground, the force output of each leg needs to be adjusted in real time to prevent tipping or slipping. The dynamic sole force distribution method is not only a core technology of the force-controlled quadruped robot, but also an important means to improve its ability to perform tasks in uncertain environments. However, current dynamic sole force distribution methods usually involve complex physical models and optimization algorithms, which lead to computational delays and are not suitable for real-time applications. Summary of the invention

[0003] The present invention aims to solve the problem that the current dynamic plantar force distribution method usually involves complex physical models and optimization algorithms, which leads to calculation delays and is not suitable for real-time applications. A dynamic plantar force distribution method for a force-controlled quadruped robot is proposed.

[0004] In order to solve the above technical problems, the present invention proposes a method for distributing dynamic plantar force of a force-controlled quadruped robot, which comprises the following steps:

[0005] A method for distributing dynamic plantar force of a force-controlled quadruped robot comprises the following steps:

[0006] Step 1: In the support phase, the dynamic plantar force distribution problem of the force-controlled quadruped robot is modeled as a constrained quadratic optimization problem model, namely:

[0007]

[0008] Step 2: By introducing a relaxation factor, a multiplier that is complementary to the relaxation factor, and a multiplier that is complementary to the normal vector of the plantar force, the optimization problem model of step 1 is transformed into a complementary problem model, namely:

[0009]

[0010] in,

[0011]

[0012]

[0013] Step 3: Use the complementary problem solving algorithm to solve the complementary problem model in step 2 to obtain the multiplier u that is complementary to the relaxation factor v and the normal vector x of the plantar force x. N ;

[0014] Step 4: Use the multiplier u obtained in step 3, which is complementary to the relaxation factor v, and the normal vector x of the plantar force x N Calculate the tangent vector x of the plantar force x T ,in:

[0015]

[0016] Step 5: The normal vector x of the plantar force x obtained in step 4 N and the tangent vector x of the plantar force x obtained in step 4 T , thus we get the solution x to the dynamic plantar force distribution problem of the force-controlled quadruped robot, where

[0017] Where: f is the objective function; x is the plantar force, x N is the normal vector of the plantar force x, x T is the tangent vector of the plantar force x; A is the mapping matrix, A1 is the first matrix block of the mapping matrix A, and A2 is the second matrix block of the mapping matrix A; S is the force following weight matrix at the center of mass of the trunk; W is the penalty matrix, W1 is the first matrix block of the penalty matrix W, and W2 is the second matrix block of the penalty matrix W; b is the force spinor at the center of mass of the trunk; T is the friction matrix, T N is the normal matrix of the friction matrix T, T T is the tangent matrix of the friction matrix T; v is the relaxation factor, y is the normal vector x to the plantar force x N M is a multiplier that is complementary to the relaxation factor v. 11 is the normal vector x of the plantar force x N The first complementary coefficient, M 12 is the first complementary coefficient of the multiplier u that is complementary to the relaxation factor v, M 21 is the normal vector x of the plantar force x N The second complementary coefficient, M 22 is the second complementary coefficient of the multiplier u that is complementary to the relaxation factor v; q1 is the normal vector x to the plantar force x N The addend of the multiplier y is complementary to each other, and q2 is the addend of the relaxation factor v.

[0018] The algorithm for solving the complementarity problem in the above step 3 is a modular matrix splitting iteration method, a projection method or a Jacobi iteration method.

[0019] Compared with the prior art, the present invention models the dynamic plantar force distribution problem of the force-controlled quadruped robot as a quadratic optimization problem model in the support phase, and then introduces a relaxation factor v and multipliers u and y to transform the optimization problem model into a complementary problem model, thereby reducing the computational complexity. The dynamic plantar force distribution of the force-controlled quadruped robot is solved under a relatively simplified calculation, and a solution can be obtained in a shorter calculation time, thereby reducing a certain error rate and ensuring the accuracy of the dynamic plantar force distribution solution of the force-controlled quadruped robot. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is the overall flow chart of the dynamic plantar force distribution method for a force-controlled quadruped robot. DETAILED DESCRIPTION

[0021] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in combination with specific examples and with reference to the accompanying drawings.

[0022] In order to better solve the problem of dynamic plantar force distribution of a force-controlled quadruped robot, the present invention proposes a method for dynamic plantar force distribution of a force-controlled quadruped robot, such as Figure 1 As shown, it includes the following steps:

[0023] Step 1: Establish an optimization problem model for the dynamic plantar force distribution problem of a force-controlled quadruped robot.

[0024] In the support phase, the dynamic plantar force distribution problem of the force-controlled quadruped robot is modeled as a constrained quadratic optimization problem model, namely:

[0025]

[0026] Where f is the objective function. x is the plantar force of a 6-dimensional column vector, where each plantar force has three dimensions and is composed of a normal force and two tangential forces. The force-controlled quadruped robot has four feet, so there are a total of 12-dimensional column vectors. Since the four feet of the force-controlled quadruped robot are symmetrical, it is sufficient to study two symmetrical feet, so it is optimized to a 6-dimensional column vector. b is the 6-dimensional force spinor at the center of mass of the trunk, the first three elements represent the force, and the last three elements represent the torque. A is a 6×6-dimensional mapping matrix, which is obtained according to the length and width of the trunk of the force-controlled quadruped robot and the height from the center of mass coordinate point to the origin of the leg coordinates. S is the force follower weight matrix at the center of mass of the trunk, which is a 6×6-dimensional diagonal matrix. W is the penalty matrix, which is a 6×6-dimensional diagonal matrix used to avoid excessive plantar force. S and W are artificially set matrices, and the values ​​in the matrix can be determined through experience or trial. T is a 6×6 dimensional friction matrix, which is obtained by linearizing the friction cone of each contact point of the quadruped robot with the ground. T is used to ensure that the tangential force of the plantar force is within the range of static friction.

[0027] Because A is not a fully rank matrix, there is no unique solution for distributing the force at the center of mass to the supporting leg. N and the tangent vector x T Two vectors: The normal vector x of x N Defined as a 2D column vector including the normal forces of the two foot-plantar forces with symmetry, the tangent vector x T is defined as a 4-dimensional column vector of the tangential forces of the two foot sole forces with symmetry. The friction matrix T corresponds to the normal vector and tangent vector of the foot sole force x and is written as: [T N T T ], where the normal matrix of T is T N is a 2×6 dimensional matrix, the tangent vector T of T T is a 4×6 dimensional matrix. The mapping matrix A corresponds to the normal vector and tangent vector of the plantar force x as: [A1 A2], where the block A1 of the A matrix is ​​a 6×2 dimensional matrix, and the block A2 of the A matrix is ​​a 6×4 dimensional matrix. The penalty matrix W corresponds to the normal vector and tangent vector of the plantar force x as: [W1 W2], where the block W1 of the W matrix is ​​a 6×2 dimensional matrix, and the block W2 of the W matrix is ​​a 6×4 dimensional matrix.

[0028] Since the quadruped robot needs support force to ensure standing in the support phase, the normal vector of the plantar force should be guaranteed to be non-negative. The friction matrix multiplied by the plantar force can obtain the friction force generated by the contact surface, and the friction force must also be guaranteed to be positive. Therefore, the constraint condition is

[0029] Step 2: Convert the established optimization problem model into a complementary problem model.

[0030] Step 2.1: Introduce the relaxation factor v to turn the inequalities in the constraints into equalities.

[0031] In order to turn the inequality in the above constraints into an equality, the relaxation factor v is introduced and substituted into the inequality so that:

[0032] T N x N +T T x T -v=0 (2)

[0033] Among them, v ≥ 0, and the dimension of v is six.

[0034] Step 2.2: Introduce multipliers u and y, and generate an augmented Lagrangian function based on the quadratic optimization problem model together with the relaxation factor v.

[0035] Introduce multipliers u and y, where y is the normal vector x to the plantar force x Nu is a multiplier that is complementary to the relaxation factor v, and u≥0, y≥0. The dimensions of u and y are 6 and are combined with the relaxation factor v to generate the augmented Lagrangian function L=f+u based on equation (1). T (T N x N +T T x T )+y T x N , we get:

[0036]

[0037] Step 2.3: Find x for the augmented Lagrangian function of equation (3): N and x T The partial derivative of x is set equal to zero. At this time, the robot is in a stable state at the equilibrium point or extreme point. N The expression and x T The expression is:

[0038]

[0039] From equations (4) and (5), we can get x T Expression and x N The expression is:

[0040]

[0041] Step 2.3: x T Substitute the expression of x into N From the expression and equation of , we can get the relationship between y and u and x. N The expression of v for u and x N expression.

[0042] Taking the partial derivative of the augmented Lagrangian function, we get the tangent vector x of the plantar force x. T Substituting the expression (6) into equation (2) and equation (4), we can get the relationship between y and u and x. N The expression of v for u and x N The expression is:

[0043] y versus u and x N The expression is:

[0044]

[0045] v for u and x N The expression is:

[0046]

[0047] Step 2.4: Based on the robot in the equilibrium state u T Complementary to v, y and x N Complementary, using y to u and x N The expression of v for u and x N The expression of is used to obtain the complementarity problem model.

[0048] When the robot is in equilibrium, u T Complementary to v, y and x N Complementary:

[0049]

[0050] According to equations (8) and (9), the complementary problem model is established:

[0051]

[0052] in,

[0053]

[0054] M 11 is the normal vector x of the plantar force x N The first complementary coefficient, M 12 is the first complementary coefficient of the multiplier u that is complementary to the relaxation factor v, M 21 is the normal vector x of the plantar force x N The second complementary coefficient, M 22 is the second complementary coefficient of the multiplier u that is complementary to the relaxation factor v; q1 is the normal vector x to the plantar force x N The addend of the multiplier y is complementary to each other, and q2 is the addend of the relaxation factor v.

[0055] Step 3: Use the complementary problem solving algorithm to solve the complementary problem model of equation (11) to obtain the multiplier u and the normal vector x of the plantar force distribution solution: N .

[0056] The complementary problem solving algorithm adopts the complementary problem solving algorithm known in the prior art, such as modular matrix splitting iteration method, projection method or Jacobi iteration method.

[0057] Step 4: Substitute u and x obtained in the complementary problem N Substitute the tangent vector x into the optimization problem. T In the expression of T .

[0058] Step 5: The normal vector x of the plantar force x obtained in step 4 N and the tangent vector x of the plantar force x obtained in step 4T , thus we get the solution to the dynamic plantar force distribution problem of the force-controlled quadruped robot, that is, the plantar force x.

[0059] It should be noted that although the embodiments of the present invention described above are illustrative, they are not intended to limit the present invention, and therefore the present invention is not limited to the above specific embodiments. Without departing from the principles of the present invention, any other embodiments obtained by those skilled in the art under the guidance of the present invention are deemed to be within the protection of the present invention.

Claims

1. A method for distributing dynamic plantar force of a force-controlled quadruped robot, characterized in that: The steps include: Step 1: In the support phase, the dynamic plantar force distribution problem of the force-controlled quadruped robot is modeled as a constrained quadratic optimization problem model, namely: Step 2: By introducing a relaxation factor, a multiplier that is complementary to the relaxation factor, and a multiplier that is complementary to the normal vector of the plantar force, the optimization problem model of step 1 is transformed into a complementary problem model, namely: in, Step 3: Use the complementary problem solving algorithm to solve the complementary problem model in step 2 to obtain the multiplier u that is complementary to the relaxation factor v and the normal vector x of the plantar force x. N ; Step 4: Use the multiplier u obtained in step 3, which is complementary to the relaxation factor v, and the normal vector x of the plantar force x N Calculate the tangent vector x of the plantar force x T ,in: Step 5: The normal vector x of the plantar force x obtained in step 4 N and the tangent vector x of the plantar force x obtained in step 4 T , thus we get the solution x to the dynamic plantar force distribution problem of the force-controlled quadruped robot, where Where: f is the objective function; x is the plantar force, x N is the normal vector of the plantar force x, x T is the tangent vector of the plantar force x; A is the mapping matrix, A1 is the first matrix block of the mapping matrix A, and A2 is the second matrix block of the mapping matrix A; S is the force following weight matrix at the center of mass of the trunk; W is the penalty matrix, W1 is the first matrix block of the penalty matrix W, and W2 is the second matrix block of the penalty matrix W; b is the force spinor at the center of mass of the trunk; T is the friction matrix, T N is the normal matrix of the friction matrix T, T T is the tangent matrix of the friction matrix T; v is the relaxation factor, y is the normal vector x to the plantar force x N M is a multiplier that is complementary to the relaxation factor v. 11 is the normal vector x of the plantar force x N The first complementary coefficient, M 12 is the first complementary coefficient of the multiplier u that is complementary to the relaxation factor v, M 21 is the normal vector x of the plantar force x N The second complementary coefficient, M 22 is the second complementary coefficient of the multiplier u that is complementary to the relaxation factor v; q1 is the normal vector x to the plantar force x N The addend of the multiplier y is complementary to each other, and q2 is the addend of the relaxation factor v.

2. The method for distributing dynamic plantar force of a force-controlled quadruped robot according to claim 1, characterized in that: The algorithm for solving the complementarity problem in step 3 is a modular matrix splitting iterative method, a projection method or a Jacobi iterative method.