Quadruped robot control method applied to dynamic slope topography

Through overall motion planning, coordinated motion planning and single-leg control steps, combined with virtual leg model and inverted pendulum model, the motion trajectory and posture of the four-legged robot are optimized, and the problems of large calculation load and insufficient ground contact constraints in the existing technology are solved, and the problem of large calculation load and insufficient ground contact constraints are realized. The four-legged robot walks highly stable on dynamic slopes.

CN119065367APending Publication Date: 2024-12-03NANCHANG UNIV
View PDF 0 Cites 3 Cited by

Patent Information

Application Number
CN202411127288.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-16
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

The existing four-legged robot control method has a large calculation load in dynamic slope terrain and fails to effectively consider ground contact constraints, resulting in insufficient walking stability.

Method used

Using overall motion planning, coordinated motion planning and single-leg control steps, through virtual leg model, inverted pendulum model and Kalman filtering algorithm and other technologies, the motion trajectory and posture of the four-legged robot are optimized, and the concepts of time-varying constraints and balance points are introduced to realize the Online planning and single joint control ensure optimum of foot reaction and position.

Benefits of technology

The adaptability and walking stability of the four-legged robot on dynamic slope terrain is improved, and the footing point can be selected more effectively, reducing the impact of dynamic slope on robot stability.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119065367A_ABST
    Figure CN119065367A_ABST
Patent Text Reader

Abstract

The invention discloses a quadruped robot control method applied to dynamic slope topography, and the method comprises the steps: enabling a quadruped robot to be equivalent to a virtual leg model through an overall motion planning layer, carrying out the online planning of the motion track and posture of a body according to a control instruction inputted by a user and the slope estimation fusing the data of each sensor, and carrying out the online planning of the motion track and posture of the body; the posture controller is used for controlling the posture of the body; the coordinated motion planning layer judges leg motion states, platform motion time-varying constraints are introduced into a robot dynamic model, and supporting leg joint torque and expected foot end counter-force are optimized and solved; introducing the concept of a balance point of the inverted pendulum model into planning of a swing leg, and optimizing a landing point of the robot on a dynamic slope; the single-leg control layer realizes force and position control of a single leg through kinematics and dynamics so as to execute control of a supporting phase and a swinging phase; the method adapts to a dynamic slope platform, can overcome complex road surfaces, and is good in adaptability; according to the foothold planning method, more appropriate points can be selected, and the walking stability of the quadruped robot is higher.
Need to check novelty before this filing date? Find Prior Art

Description

Technical field: The invention belongs to the technical field of quadruped robot control, and in particular relates to a quadruped robot control method applied to dynamic slope terrain. Background technology: At present, in the field of quadruped robots, some scholars have studied the control design of leg motion on dynamic rigid platforms. Henze B et al. expanded the passive factors considered in robot motion on a platform with an unstable ground environment, including the robot, controller and the bottom layer; Iqbal A et al. derived and experimentally evaluated a provable hybrid time-varying robot control system for realizing quadruped robot motion on a dynamic rigid platform; Li J et al. mapped the robot kinematic body planning from a static rigid platform to a 2-DOF swaying rigid platform while keeping the torso posture horizontal throughout the walking process.

[0003] However, there are two limitations to the effectiveness of these controllers in real-world scenarios: first, the planning algorithms are not suitable for online motion generation due to the heavy computational load imposed by their underlying full-order models. Second, the controllers do not explicitly consider the feasibility of ground contact constraints during actual walking. Summary of the invention: In order to overcome the deficiencies of the prior art, the present invention provides a control method for a quadruped robot applied to dynamic slope terrain with good adaptability and high walking stability.

[0005] An embodiment of the present invention provides a quadruped robot control method applied to dynamic slope terrain, comprising: Overall motion planning step S1: The quadruped robot is equivalent to a model with virtual legs. According to the control instructions input by the user and the slope estimation fused with the data of each sensor, the motion trajectory and posture of the main body are planned online, and the posture controller is used to control the posture of the main body; Coordinated motion planning step S2: Determine the motion state of the legs, introduce the platform motion time-varying constraints into the robot dynamics model, and optimize the support leg joint torque and the expected foot end reaction force; considering the platform motion, introduce the concept of the inverted pendulum model balance point into the planning of the swinging leg, and optimize the robot's foot landing point and foot end trajectory on the dynamic slope; Single-leg control step S3: Control of a single bionic leg is achieved through single-leg kinematics and dynamics, so that the leg reaches the desired foot-end reaction force and foot-end position to perform force control in the support phase and position control in the swing phase.

[0006] Optionally, the overall motion planning step S1 includes the following sub-steps: S11: Based on the hip, thigh and calf joint positions and body posture angles of the four bionic legs, the position of the foot end in the operating space is obtained through the homogeneous transformation method; S12: The pitch angle and roll angle of the unknown slope along the robot's forward direction are estimated by solving a set of linear equations based on the spatial positions of multiple foot endpoints. S13: using the Kalman filter algorithm to perform denoising on the preliminary estimation result of the slope, and on this basis, using the filtered slope information for attitude control when going uphill; S14: Plan the expected trajectory of the body's center of mass and the expected trajectory of the swinging legs during slope movement.

[0007] Optionally, the coordinated motion planning step S2 includes the following sub-steps: S21: Introduce the spatial linear inverted pendulum model to plan the balance point, use the platform's motion factor as compensation to maintain the robot's stable landing point, and optimize the robot's landing point on the dynamic slope; S22: When the quadruped robot moves in a Trot gait, the position and posture control of the main body is controlled by the supporting legs. The MPC model predictive control method is used to distribute the expected force of the main body to each supporting leg to obtain the expected force in the single-leg operation space. S23: Considering the time-varying parameters of the foot-platform contact point caused by the platform movement, in order to ensure the feasibility of the contact force, the time-varying complete constraint of the platform movement is added to the robot dynamics model; according to the complete time-varying robot dynamics, calculations are performed to control the force and torque of the body.

[0008] Optionally, the single-leg control step S3 includes the following sub-steps: Regarding the operation space position control S31 of the swing leg: according to the kinematic modeling method of the quadruped robot, the DH method is used to perform inverse kinematic analysis; the position of the single leg operation space is solved to the hip, thigh and calf joints of the single leg, and then the operation space position of the foot end of the swing leg is controlled by controlling the position of the single joint; for the position control of the single joint, the position sensor signal of the joint is used as the feedback signal, the actual position of the joint is used as the control target, and the calculated joint position is used as the expectation, and the PID device is used for control; Regarding the operating space force control S32 of the supporting leg: According to the quadruped robot dynamics modeling method, the Newton-Euler method is used to perform inverse dynamics modeling analysis; the force control of the single-leg operating space is solved to the hip, thigh, and calf joints of the single leg, and then the operating space force of the swing leg foot end is controlled by controlling the torque of the single joint; for the torque control of the single joint, the force sensor signal and position sensor signal of the joint are used as feedback signals to calculate the actual joint torque, the actual joint torque is used as the control target, and the solved joint torque is used as the expectation, and a PID device is used for control.

[0009] The beneficial effects of the present invention are: 1) It can not only adapt to dynamic slope terrain, but also overcome complex road surfaces, with good adaptability; 2) The foothold selection method enables a more suitable foothold to be selected, and the quadruped robot has higher walking stability; 3) According to the influence of dynamic slopes on the quadruped robot, the model predictive control (MPC) is optimized so that the reference terrain and the selected foothold match, reducing the impact of the dynamic slope terrain on the stability of the robot body. Description of the drawings: Figure 1 Schematic diagram of a quadruped robot slope motion control strategy in one embodiment of the present invention; Figure 2 This is a schematic diagram of slope information estimation in step S1 in one embodiment of the present invention; Figure 3 Schematic diagram of slope posture in step S1 in one embodiment of the present invention; Figure 4 A force analysis diagram of a single rigid body on a dynamic slope in step S2 in one embodiment of the present invention; Figure 5 3D schematic diagram of a linear inverted pendulum model on a dynamic slope in step S2 in one embodiment of the present invention. Specific implementation method: In order to enable those skilled in the art to better understand the solutions of the present invention, the following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0012] like Figure 1 As shown, a quadruped robot control method applied to dynamic slope terrain includes the following steps: Overall motion planning step S1: The quadruped robot is equivalent to a model with virtual legs. According to the control instructions input by the user and the slope estimation fused with the sensor data, the motion trajectory and posture of the main body are planned online, and the posture controller is used to control the posture of the main body.

[0013] The overall motion planning step S1 includes the following sub-steps: S11: Based on the hip, thigh, and calf joint positions and the body posture angles of the four bionic legs, the position of the quadruped robot foot end on the slope in the world coordinate system {O} is obtained by homogeneous transformation. Opi ; S12: The pitch angle and roll angle of the unknown slope along the robot's forward direction are estimated by solving a set of linear equations using multiple sets of foot endpoints to manipulate spatial positions; the slope estimation is as follows: Figure 2 As shown in the figure, the pitch angle of the slope along the forward direction of the quadruped robot is and roll angle : ;in , is the plane unit normal vector, the matrix is the rotation matrix of the yaw angle, Measured by IMU; S13: Use the Kalman filter algorithm to denoise the preliminary estimation results of the slope to obtain smooth slope estimation information and , on this basis, the filtered slope information is used for attitude control when going uphill; pass , , and the sampling interval for slope estimation To estimate the slope angular velocity: ; On this basis, the filtered slope information is used for attitude control when going uphill; finally, the uphill attitude is obtained as Figure 3 As shown in the picture on the right.

[0014] S14: Plan the expected trajectory of the body's center of mass and the expected trajectory of the swinging legs during slope movement.

[0015] Coordinated motion planning step S2: Determine the motion state of the legs and introduce the platform motion time-varying constraints into the robot dynamics model, such as Figure 4 As shown in the figure, the joint torque of the supporting leg and the expected reaction force of the foot are optimized; considering the movement of the platform, the concept of the equilibrium point of the inverted pendulum model is introduced into the planning of the swinging leg to optimize the landing point and foot trajectory of the robot on the dynamic slope.

[0016] The coordinated motion planning step S2 includes the following sub-steps: S21: Figure 5 As shown in the figure, the spatial linear inverted pendulum model is introduced to plan the equilibrium point, and the motion factor of the platform is used as the linear velocity tracking error compensation to maintain the stability of the robot's landing point , where It is the vertical acceleration of the contact point between the foot end of the supporting leg and the dynamic slope surface, which is generated by the movement of the platform; is the height of the center of mass of the robot relative to the platform when the foot end is in contact phase; The current body speed in the body coordinate system. , the velocity tracking compensation displacement based on the fuselage system can be mapped to a unified expression in the world system, and is the expected speed of the robot in the x and y directions in this system. Optimize the landing point of the robot on the dynamic slope ; S22: When the quadruped robot moves in a Trot gait, the position and posture control of the main body is controlled by the supporting legs; the MPC model predictive control method is used to distribute the expected force of the main body to each supporting leg, and the expected force in the single-leg operation space is obtained. ; The state equation is as follows: ; In the formula In order to use the traditional discretization method, The elements in must be combined into the state vector to create an augmented vector for the MPC formula. Therefore, it can be rewritten as follows: . Rewritten as an augmented vector ;but was rewritten as ; was rewritten as , in order to obtain the disturbance factor The standard single rigid body system continuous state equation is Assuming that the foot of the supporting phase does not slip on the dynamic slope, the acceleration of the foot in the world system is This is the acceleration of the contact point on the slope. When the slope is stationary or moving at a constant speed in a straight line, for .

[0017] Single-leg control step S3: Control of a single bionic leg is achieved through single-leg kinematics and dynamics, so that the leg reaches the desired foot-end reaction force and foot-end position to perform force control in the support phase and position control in the swing phase.

[0018] The single-leg control step S3 includes the following sub-steps: Regarding the operation space position control of the swing leg S31: According to the kinematic modeling method of the quadruped robot, the DH method is used to perform inverse kinematic analysis; the position of the single leg operation space is solved to the hip, thigh, and calf joints of the single leg, and then the operation space position of the swing leg foot end is controlled by controlling the position of the single joint; for the position control of the single joint, the position sensor signal of the joint is used as the feedback signal, and the actual position of the joint is used as the control target, and the calculated joint position is used. q As expected, a PID controller is used for control; Regarding the operating space force control S32 of the supporting leg: According to the quadruped robot dynamics modeling method, the Newton-Euler method is used to perform inverse dynamics modeling analysis; the force control of the single-leg operating space is solved to the hip, thigh, and calf joints of the single leg, and then the operating space force of the swing leg foot end is controlled by the torque control of the single joint; for the torque control of the single joint, the force sensor signal and position sensor signal of the joint are used as feedback signals to calculate the actual joint torque, and the actual joint torque is used as the control target, and the calculated joint torque is used to control the joint torque. τ As expected, a PID controller is used for control.

[0019] In step S3, S1 and S2 are combined to realize the control of a single bionic leg through single-leg kinematics and dynamics, so that the leg can achieve the desired foot-end reaction force and foot-end position; S23 considers the time-varying parameters of the foot-platform contact point caused by platform motion. To ensure the feasibility of contact force, the time-varying complete constraints of platform motion are added to the robot dynamics model; the force and torque of the body are controlled according to the complete time-varying robot dynamics; the planning of multi-task priority at the three levels of position, speed and acceleration is solved according to the multi-task priority control principle. At present, the expected acceleration of the joint has been obtained through kinematic whole-body control. , the expected reaction force at the foot end is solved by MPC problem , other quantities are determined by the robot kinematics, and the joint torques are calculated through the dynamic equations.

[0020] Due to the existence of virtual joints, the above quantities may not satisfy the dynamic equations, so it is necessary to The first 6 items and Introduce slack variables for optimization to find a set of variables that are as small as possible and variables To meet the conditions.

[0021] Combined with the friction cone constraint on the slope, the optimal problem is constructed and solved using the QuadProg++ library to obtain the solution that satisfies the dynamic equations of the floating body. and , substitute into the overall dynamics equation of the robot to obtain the expected value of the joint torque : , the torque of the 6-DOF virtual joint is zero, As the feedforward torque of 12 entity joints.

[0022] Based on the feedforward torque, PD control is introduced and the robot is obtained by calculation. Driving torque of each joint: , where , Desired joint positions and desired joint velocities calculated for whole-body control.

[0023] The motion host includes multiple modules, such as leg and foot odometers, swing leg planning, foothold optimization and model predictive control. The reference speed is initially input by the upper control module, such as a handle command. The present invention first introduces dynamic slope motion parameters into the foothold planning of the robot's swing leg, and compensates the foothold based on the stability condition of the inverted pendulum model. Secondly, for the support phase, a model predictive control with dynamic disturbance is designed to calculate the optimal foot-end reaction force. Finally, the hybrid control of the robot's underlying joint force and position is realized through multi-task-level whole-body control to solve the problem of affecting stability analyzed previously.

[0024] The quadruped robot walks to the dynamic slope and plans the trajectory and posture of the robot online according to the control instructions and slope estimation. The concept of the balance point of the inverted pendulum model is introduced to optimize the robot's foot landing point and foot trajectory on the dynamic slope. The control of a single bionic leg is achieved through single-leg kinematics and dynamics, so that the leg can achieve the desired foot end reaction force and foot end position.

[0025] The above specific implementation modes are used to explain the present invention rather than to limit the present invention. Any modification and change made to the present invention within the spirit of the present invention and the protection scope of the claims shall fall within the protection scope of the present invention.

Claims

1. A quadruped robot control method applied to dynamic slope terrain, characterized in that: include: Overall motion planning step S1: The quadruped robot is equivalent to a model with virtual legs. According to the control instructions input by the user and the slope estimation fused with the data of each sensor, the motion trajectory and posture of the main body are planned online, and the posture controller is used to control the posture of the main body; Coordinated motion planning step S2: Determine the motion state of the legs, introduce the platform motion time-varying constraints into the robot dynamics model, and optimize the support leg joint torque and the expected foot end reaction force; considering the platform motion, introduce the concept of the inverted pendulum model balance point into the planning of the swinging leg, and optimize the robot's foot landing point and foot end trajectory on the dynamic slope; Single-leg control step S3: Control of a single bionic leg is achieved through single-leg kinematics and dynamics, so that the leg reaches the desired foot-end reaction force and foot-end position to perform force control in the support phase and position control in the swing phase.

2. The control method according to claim 1, characterized in that: The overall motion planning step S1 includes the following sub-steps: S11: Based on the hip, thigh and calf joint positions and body posture angles of the four bionic legs, the position of the foot end in the operating space is obtained through the homogeneous transformation method; S12: The pitch angle and roll angle of the unknown slope along the robot's forward direction are estimated by solving a set of linear equations based on the spatial positions of multiple foot endpoints. S13: using the Kalman filter algorithm to perform denoising on the preliminary estimation result of the slope, and on this basis, using the filtered slope information for attitude control when going uphill; S14: Plan the expected trajectory of the body's center of mass and the expected trajectory of the swinging legs during slope movement.

3. The control method according to claim 2, characterized in that: The coordinated motion planning step S2 includes the following sub-steps: S21: Introduce the spatial linear inverted pendulum model to plan the balance point, use the platform's motion factor as compensation to maintain the robot's stable landing point, and optimize the robot's landing point on the dynamic slope; S22: When the quadruped robot moves in a Trot gait, the position and posture control of the main body is controlled by the supporting legs. The MPC model predictive control method is used to distribute the expected force of the main body to each supporting leg to obtain the expected force in the single-leg operation space. S23: Considering the time-varying parameters of the foot-platform contact point caused by the platform movement, in order to ensure the feasibility of the contact force, the time-varying complete constraint of the platform movement is added to the robot dynamics model; according to the complete time-varying robot dynamics, calculations are performed to control the force and torque of the body.

4. The control method according to claim 1, characterized in that: The single-leg control step S3 includes the following sub-steps: Regarding the operation space position control S31 of the swing leg: According to the kinematic modeling method of the quadruped robot, the DH method is used to perform inverse kinematic analysis; the position of the single leg operation space is solved to the hip, thigh and calf joints of the single leg, and then the operation space position of the swing leg foot end is controlled by controlling the position of the single joint; for the position control of the single joint, the position sensor signal of the joint is used as the feedback signal, the actual position of the joint is used as the control target, and the calculated joint position is used as the expectation, and the PID device is used for control; Regarding the operating space force control of the supporting leg S32: Based on the quadruped robot dynamics modeling method, the Newton-Euler method is used to perform inverse dynamics modeling analysis; the force control of the single-leg operating space is solved to the hip, thigh, and calf joints of the single leg, and then the operating space force of the swing leg foot end is controlled by controlling the torque of the single joint; For the torque control of a single joint, the force sensor signal and position sensor signal of the joint are used as feedback signals to calculate the actual joint torque. The actual joint torque is used as the control target, and the calculated joint torque is used as the expectation, and a PID device is used for control.

Citation Information

Cited By

  • Quadruped single-arm robot cooperative motion control method and system

    CN119820576A

  • Multi-legged robot slope marching control method and system based on double-peak Gaussian fusion

    CN120066104A

  • Quadruped robot rapid motion control method under discrete terrain

    CN120886237A