Four-foot wheel-leg robot attitude balancing method based on force-position hybrid fuzzy control

Abstract

The invention discloses a force-position hybrid fuzzy control-based attitude balancing method for a four-foot wheel-leg robot, which comprises the following steps of: S1, controlling a robot body to keep the robot body stable; and S2, on the basis that the fuselage is kept stable, the motion state of the fuselage is dynamically adjusted according to different attitude angles and attitude angular velocities of the fuselage, so that the robot adapts to different terrains. According to the method, response can be rapidly carried out only through self-sensing of the robot, calculation consumption is small, and the method dynamically adapts to different terrains.

Description

Technical Field

[0001] The present invention relates to the field of robot control, and in particular to a posture balancing method of a quadruped wheel-legged robot based on force-position mixed fuzzy control. Background Art

[0002] A quadruped wheeled robot has both the leg structure of a quadruped robot and the wheel structure of a wheeled robot. In certain mission scenarios, such as cargo handling, when the quadruped wheeled robot is driving, in order to ensure that the body remains level, the leg structure can act as an active suspension, flexibly adjusting the position of the foot end, so that the foot end of the quadruped wheeled robot can adapt to uneven roads, while also keeping the robot body level, and maintaining a stable posture when passing through terrain such as ramps or single-sided bridges. Figure 1 shown.

[0003] To achieve such a function, different control methods are needed. Currently, the existing methods for this function are: 1. Equipped with vision or lidar, it obtains external terrain information through perception, and directly controls the leg movement through the known terrain information to adapt to the terrain; 2. Install pressure sensors or gyroscopes and other sensors on the legs to monitor the data of the legs during movement, such as plantar pressure, leg angle, etc., to adjust the leg posture in real time; 3. Establish a dynamic model for the four-legged wheeled robot, and adjust the control quantity in real time by solving the dynamic model to make the current state quantity related to the posture reach the expected position; 4. In combination with the changes in the body posture, adjust the expected position of the supporting leg or the foot end force through motor position control or torque control; 5. Reinforcement learning training.

[0004] Among them, the disadvantages of methods 1 and 2 are that they require the installation of additional sensors, which increases hardware consumption; the shortcoming of method 3 is that the establishment of the dynamic model needs to be sufficiently accurate, and at the same time, the dynamic solution requires strong main control performance and a reliable state observer for the four-legged wheeled robot with a more complex model, otherwise real-time performance and stability are difficult to guarantee; the shortcoming of method 4 is that the four-legged wheeled robot can also stand on three legs, so if the motor position control or torque control is performed alone, there will be a problem of one leg being in the air; the problem of method 5 is that the underlying strategy after reinforcement learning itself is not transparent, and its reliability is not as good as traditional control. At the same time, the deployment of reinforcement learning requires a dedicated hardware platform.

[0005] Therefore, in order to solve the above problems, a posture balancing method for a quadruped wheeled robot based on force-position hybrid fuzzy control is needed, which can respond quickly and consume little computational energy, and can dynamically adapt to different terrains by relying solely on the robot's own perception. Summary of the Invention

[0006] In view of this, the purpose of the present invention is to overcome the defects in the prior art and provide a posture balancing method for a four-legged wheeled robot based on force-position mixed fuzzy control, which can respond quickly and consume little computational energy and dynamically adapt to different terrains relying solely on the robot's own perception.

[0007] The posture balancing method of a quadruped wheeled robot based on force-position hybrid fuzzy control of the present invention comprises the following steps:

[0008] S1. Control the robot body to keep it stable;

[0009] S2. While maintaining a stable body, the robot dynamically adjusts its motion state according to its different posture angles and angular velocities, allowing it to adapt to different terrains.

[0010] Furthermore, the step S1 specifically includes:

[0011] Constructing a robot body coordinate system; wherein the robot body coordinate system has the center of the body as the origin, the robot's forward direction as the positive x-axis, the direction perpendicular to the robot body and upward as the positive z-axis, and the plane perpendicular to the x-axis and the z-axis as the y-axis;

[0012] Assume that the z-axis coordinates of the four legs are z1, z2, z3, and z4. When the robot is driving normally on a horizontal road, the z-axis coordinates of the four legs are the desired positions of the four legs. The legs of the quadruped wheeled robot are equivalent to robotic arms, and impedance control is used to keep the ends of the four legs in the desired positions.

[0013] The impedance control law of the impedance control is:

[0014]

[0015] Where F is the controlled foot end force; Kp is the spring coefficient, P d is the desired foot end position, P is the current foot end position; Kd is the damping coefficient, is the desired foot end velocity, is the current foot speed; f s is the average supporting force of the robot's four legs in a stable state.

[0016] Furthermore, the step S2 specifically includes:

[0017] From front to back and from left to right, the four legs of the robot are named leg 1 fl, leg 2 fr, leg 3 bl, and leg 4 br.

[0018] According to the different forces and expected position gains required by different legs when the pitch angle α and roll angle β change, a PD control model based on fuzzy control is constructed.

[0019] According to the actual situation of the body posture change, the PD control model is used to adjust the two quantities of force and desired position to achieve adaptive control and obtain the final foot end force.

[0020] Furthermore, a PD control model based on fuzzy control is constructed according to the following formula:

[0021]

[0022] Where ΔP i , Δf i They represent the expected position gain and the expected force gain respectively; i represents the leg number, and the values ​​of i corresponding to fl, fr, bl and br are 1 to 4 respectively; represents the force compensation gain coefficient for leg i to pitch angle, represents the force compensation gain coefficient for leg i to the roll angle, represents the position compensation gain coefficient for leg i to the pitch angle, represents the position compensation gain coefficient for the roll angle of leg i; f α 、f β 、P α and P β are all intermediate transition parameters; K α p f represents the spring constant for pitch angle and force gain; K α d f represents the damping coefficient for pitch angle and force gain; K β p f Indicates the spring coefficient for roll angle and force gain; K β d f Indicates the damping coefficient for roll angle and force gain; K α p p K represents the spring constant for the pitch angle and desired position gain; α d p K represents the damping coefficient for the pitch angle and the desired position gain; β p p K represents the spring constant for the roll angle and desired position gain; β d p represents the damping coefficient for the roll angle and the desired position gain; and They represent the angular velocity of the pitch angle and the angular velocity of the roll angle respectively.

[0023] Furthermore, the final foot end force is determined according to the following formula:

[0024]

[0025] Among them, F f i is the final actual controlled foot end force; ΔP i , Δf i They represent the expected position gain and the expected force gain respectively; i represents the leg number, and the values ​​of i corresponding to fl, fr, bl and br are 1 to 4 respectively; Kp is the spring coefficient, P d is the desired foot end position, P is the current foot end position; Kd is the damping coefficient, is the desired foot end velocity, is the current foot speed; f s is the average supporting force of the robot's four legs in a stable state.

[0026] The beneficial effects of the present invention are as follows: the present invention discloses a posture balancing method for a four-legged wheeled robot based on force-position mixed fuzzy control, which controls the foot-end force and the desired foot-end position of the robot based on the fuzzy PD control algorithm according to the posture changes of the four-legged wheeled robot body, so that the four-legged wheeled robot can dynamically adjust the body posture when driving through rough roads, such as single-sided bridges, slopes, etc., so that the body remains level. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The present invention will be further described below with reference to the accompanying drawings and embodiments:

[0028] Figure 1 This is a schematic diagram of the adaptive leg adjustment of the robot of the present invention while driving;

[0029] Figure 2 A schematic diagram of establishing the robot body coordinate system and the world coordinate system of the present invention;

[0030] Figure 3 This is a schematic diagram of the unstable state of the robot body of the present invention when the robot body is moving;

[0031] Figure 4 Schematic diagram of the changes in pitch angle and roll angle when the robot body posture changes. DETAILED DESCRIPTION

[0032] The present invention is further described below with reference to the accompanying drawings, as shown in the drawings:

[0033] This embodiment discloses a posture balancing method for a quadruped wheeled robot based on force-position hybrid fuzzy control, comprising the following steps:

[0034] S1. Control the robot body to keep it stable;

[0035] S2. While maintaining a stable body, the robot dynamically adjusts its motion state according to its different posture angles and angular velocities, allowing it to adapt to different terrains.

[0036] In this embodiment, in step S1, Figure 2 As shown, the robot body coordinate system O{b} is established and fixedly connected to the robot body. It moves with the movement of the robot body. When the four-legged wheeled robot maintains a normal driving state on a horizontal road, the z-axis parameters of the four legs in the robot body coordinate system are the same, and the body is horizontal at this time.

[0037] The robot body coordinate system O{b} takes the center of the body as its origin, the robot's forward direction as the positive x-axis, the direction perpendicular to the robot body and upward as the positive z-axis, and the plane perpendicular to the x-axis and z-axis as the y-axis. Figure 2 The world coordinate system O{s} in is used to represent the position of the robot in the global environment;

[0038] Assume that the z-axis coordinates of the four legs are z1, z2, z3, and z4. When the robot is driving normally on a horizontal road, the z-axis coordinates of the four legs are the desired positions of the four legs. The legs of the quadruped wheeled robot are equivalent to robotic arms, and impedance control is used to keep the ends of the four legs in the desired positions.

[0039] The impedance control law of the impedance control is:

[0040]

[0041] Where F is the controlled foot end force; Kp is the spring coefficient, P d is the desired foot end position, P is the current foot end position; Kd is the damping coefficient, is the desired foot end velocity, is the current foot speed; f s is the average support force of the robot's four legs in a stable state. For example, when the robot's mass is m, the average support force of the four legs is f s = mg / 4. Kp is the spring coefficient, which acts like a spring, applying a force toward the desired position when the actual position is not in the desired position. Kd is the damping coefficient, which acts like a force that brings the actual speed closer to the desired speed when the actual speed is not equal to the desired speed.

[0042] The current foot position can be read through the joint motor encoder to read the current joint position and obtained through forward kinematics solution; the current foot velocity can be read through the joint motor encoder to read the current joint velocity and mapped to the foot velocity through Jacobian matrix solution; in addition, after the desired foot torque is calculated through impedance control, the desired foot force is converted into joint torque through the Jacobian matrix, thereby performing torque control on the joint.

[0043] Under the above control method, when the robot stands on the horizontal ground, each leg has a certain flexibility. After the desired position of the foot end is given, the farther the foot end is from the desired position, the greater the foot end force generated, and the direction of the force points along the current position to the desired position.

[0044] In this embodiment, if you want to adapt to the terrain by controlling the movement of the foot end while keeping the body stable, you can start from two aspects: 1. Directly give force feedback to the foot end, and adjust the feedback force based on the force of the foot end when standing stably, so that the foot end can adapt to the road surface; 2. Directly adjust the desired position, and let the foot end track the desired position by dynamically adjusting the desired position, so that the foot end can adapt to the road surface.

[0045] Of the two adjustment methods mentioned above, 1. Direct force feedback adjustment has a fast response, but due to the flexibility of force adjustment, it is difficult to adjust the fuselage to a completely horizontal position, so there is an adjustment error; 2. Direct position feedback adjustment directly adjusts the desired position of the foot end. As long as the fuselage is not in a horizontal position, the desired position of the foot end will continue to change, so the response is slow.

[0046] Therefore, since force adjustment has flexibility and position adjustment has precision, controlling both force and position simultaneously can achieve better results.

[0047] In step S2, if Figure 3 、 Figure 4 As shown in the figure, when the robot moves on an uneven road, the posture changes of the fuselage mainly involve the pitch angle α and the roll angle β. Based on the robot body coordinate system, the positive and negative directions of the angle changes are defined according to the right-hand rule. When the pitch angle α is greater than 0, the fuselage is lower in front and higher in the back, and the front legs need to be extended and the rear legs need to be shortened, that is, the foot-end force of the front legs increases, the desired position decreases, and the foot-end force of the rear legs decreases, and the desired position increases. When the roll angle β is greater than 0, the fuselage is higher on the left and lower on the right, and the left leg needs to be shortened and the right leg extended, that is, the foot-end force of the left leg decreases, the desired position increases, and the foot-end force of the right leg increases, and the desired position decreases.

[0048] Then, in order from front to back and from left to right, the four legs of the robot are named leg 1 fl, leg 2 fr, leg 3 bl, and leg 4 br;

[0049] Because the four legs are distributed in the front, back, left, and right directions, the body attitude involves two attitude angles: pitch angle α and roll angle β. The targets that the legs can adjust are the foot-end force and the desired foot-end position. When the two different attitude angles change, in order to restore the attitude to balance, the gains of the foot-end force and the desired foot-end position of the four legs need to be positive or negative.

[0050] For example, when the pitch angle α increases, the front of the fuselage is lower and the back is higher. At this time, for the foot-end force, the front leg needs to be increased and the rear leg needs to be reduced. For the desired position of the foot-end, the front leg needs to be lengthened (expressed as a position reduction relative to the fuselage coordinate system) and the rear leg needs to be shortened (expressed as a position increase relative to the fuselage coordinate system).

[0051] According to the different forces and expected position gains required when the pitch angle α and roll angle β of different legs change, a PD control model based on fuzzy control is constructed:

[0052]

[0053] Where ΔP i , Δf i They represent the expected position gain and the expected force gain respectively; i represents the leg number, and the values ​​of i corresponding to fl, fr, bl and br are 1 to 4 respectively; represents the force compensation gain coefficient for leg i to pitch angle, represents the force compensation gain coefficient for leg i to the roll angle, represents the position compensation gain coefficient for leg i to the pitch angle, represents the position compensation gain coefficient for the roll angle of leg i;

[0054] set up That is, when i=1, the force compensation gain coefficient for leg 1 to pitch angle is 1; when i=2, the force compensation gain coefficient for leg 2 to pitch angle is 1; when i=3, the force compensation gain coefficient for leg 3 to pitch angle is -1; when i=4, the force compensation gain coefficient for leg 4 to pitch angle is -1. Similarly, you can set f α 、f β 、P α and P β These are all intermediate transition parameters.

[0055] K α p f represents the spring constant for pitch angle and force gain; K α d f represents the damping coefficient for pitch angle and force gain; K β p f Indicates the spring coefficient for roll angle and force gain; Kβ d f represents the damping coefficient for the roll angle and the force gain; K α p p represents the spring coefficient for the pitch angle and the desired position gain; K α d p represents the damping coefficient for the pitch angle and the desired position gain; K β p p represents the spring coefficient for the roll angle and the desired position gain; K β d p represents the damping coefficient for the roll angle and the desired position gain;

[0056] In the PD control, if the PD parameters are fixed, the adjustment effect is limited when the body attitude angle changes greatly or the change speed is not certain. In order to make the body attitude adjustment process more stable and adaptive to different conditions when the attitude changes, K α p f , K α d f , K β p f , K β d f , K α p p , K α d p , K β p p and K β d p These eight PD parameters can be dynamically adjusted according to the actual situation of the body attitude change. Among them, the PD parameter refers to a control gain parameter in proportional-differential control.

[0057] According to different body attitude angles and attitude angular velocities, for example, when the body pitch angle suddenly changes from 0° to 45° and suddenly changes to 5°, the adjustment effect will be different for the same PD parameter. Using the above PD control model for fuzzy control can dynamically adjust the eight PD parameters of the body force gain and the desired position gain for different body attitude changes, thereby achieving better adaptability. α and β represent the angular velocity of the pitch angle and the angular velocity of the roll angle, respectively.

[0058] According to the actual situation of the body attitude change, the PD control model is used to adjust the force and the desired position to achieve adaptive control and obtain the final foot end force. The final foot end force can be determined according to the following formula:

[0059]

[0060] where F fi is the final actual controlled foot end force; ΔP i , Δf i They represent the expected position gain and the expected force gain respectively; i represents the leg number, and the values ​​of i corresponding to fl, fr, bl and br are 1 to 4 respectively; Kp is the spring coefficient, P d is the desired foot end position, P is the current foot end position; Kd is the damping coefficient, is the desired foot speed, is the current foot speed; f s is the average supporting force of the robot's four legs in a stable state.

[0061] To better understand the PD parameters set in the present invention, the foot end force adjustment is taken as an example to illustrate the kp value for the pitch angle as follows:

[0062] When the fuselage is stable, the pitch angle α = 0. When the fuselage roll angle is too large or too small, a larger K is required. α p f To make the pitch angle converge quickly, when the pitch angle changes little, K α p f It should also be reduced accordingly to avoid system overshoot.

[0063] The range of the absolute value of the pitch angle is defined as α∈[0,45°]. If the absolute value of the pitch angle exceeds 50 degrees, it is defined as unrecoverable. Therefore, the fuzzy subset is defined as follows:

[0064] Small pitch angle (SA), α∈[0°,15°]; Medium pitch angle (MA), α∈[10°,30°];

[0065] Large elevation angle (LA), α∈[30°,45°].

[0066] For K α p f The value of should increase with the change of α, so the following fuzzy set can be set:

[0067] (SA): Smaller K α p f Value; (MA): Medium K α p f value;

[0068] (LA): Larger K α p f value.

[0069] The corresponding fuzzy output K can be calculated based on the read pitch angle α value according to the fuzzy rules α p fThe fuzzy output is converted into a specific numerical value through the defuzzification algorithm, which serves as the input of the system PD parameter.

[0070] The method of the present invention has the advantages of low computational consumption and rapid response, and is suitable for control tasks with high real-time requirements. In the force-position hybrid control mode, force control is used to achieve flexible adaptability to the environment, thereby improving the safety and stability of the control; while position control ensures the accuracy and stability of the actuator's movements, taking into account both compliance and precision. At the same time, the method has good dynamic adaptability, and can dynamically adjust the control strategy according to different posture change amplitudes and rates to ensure robustness and consistency in various motion states. In addition, the method has good terrain adaptability, and does not rely on additional sensors to perceive terrain changes. The controller's own adjustment strategy can be used to adapt to complex or uneven ground, thereby improving overall environmental adaptability and practicality.

[0071] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the purpose and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A posture balancing method for a quadruped wheeled robot based on force-position hybrid fuzzy control, characterized by: The steps include: S1. Control the robot body to keep it stable; S2. While maintaining a stable body, the robot dynamically adjusts its motion state according to its different posture angles and angular velocities, allowing it to adapt to different terrains.

2. The posture balancing method of a quadruped wheeled robot based on force-position hybrid fuzzy control according to claim 1, characterized in that: The step S1 specifically includes: Constructing a robot body coordinate system; wherein the robot body coordinate system has the center of the body as the origin, the robot's forward direction as the positive x-axis, the direction perpendicular to the robot body and upward as the positive z-axis, and the plane perpendicular to the x-axis and the z-axis as the y-axis; Assume that the z-axis coordinates of the four legs are z1, z2, z3, and z4. When the robot is driving normally on a horizontal road, the z-axis coordinates of the four legs are the desired positions of the four legs. The legs of the quadruped wheeled robot are equivalent to robotic arms, and impedance control is used to keep the ends of the four legs in the desired positions. The impedance control law of the impedance control is: Where F is the controlled foot end force; Kp is the spring coefficient, P d is the desired foot end position, P is the current foot end position; Kd is the damping coefficient, is the desired foot end velocity, is the current foot speed; f s is the average supporting force of the robot's four legs in a stable state.

3. The posture balancing method of a quadruped wheeled robot based on force-position hybrid fuzzy control according to claim 1, characterized in that: The step S2 specifically includes: From front to back and from left to right, the four legs of the robot are named leg 1 fl, leg 2 fr, leg 3 bl, and leg 4 br. According to the different forces and expected position gains required by different legs when the pitch angle α and roll angle β change, a PD control model based on fuzzy control is constructed. According to the actual situation of the body posture change, the PD control model is used to adjust the two quantities of force and desired position to achieve adaptive control and obtain the final foot end force.

4. The posture balancing method of a quadruped wheeled robot based on force-position hybrid fuzzy control according to claim 3, characterized in that: The PD control model based on fuzzy control is constructed according to the following formula: Where ΔP i , Δf i They represent the expected position gain and the expected force gain respectively; i represents the leg number, and the values ​​of i corresponding to fl, fr, bl and br are 1 to 4 respectively; represents the force compensation gain coefficient for leg i to pitch angle, represents the force compensation gain coefficient for leg i to the roll angle, represents the position compensation gain coefficient for leg i to the pitch angle, represents the position compensation gain coefficient for the roll angle of leg i; f α 、f β 、P α and P β are all intermediate transition parameters; K α p f represents the spring constant for pitch angle and force gain; K α d f represents the damping coefficient for pitch angle and force gain; K β p f Indicates the spring coefficient for roll angle and force gain; K β d f Indicates the damping coefficient for roll angle and force gain; K α p p K represents the spring constant for the pitch angle and desired position gain; α d p K represents the damping coefficient for the pitch angle and the desired position gain; β p p K represents the spring constant for the roll angle and desired position gain; β d p represents the damping coefficient for the roll angle and the desired position gain; and They represent the angular velocity of the pitch angle and the angular velocity of the roll angle respectively.

5. The posture balancing method of a quadruped wheeled robot based on force-position hybrid fuzzy control according to claim 3, characterized in that: The final foot-end force is determined according to the following formula: Among them, F f i is the final actual controlled foot end force; ΔP i , Δf i They represent the expected position gain and the expected force gain respectively; i represents the leg number, and the values ​​of i corresponding to fl, fr, bl and br are 1 to 4 respectively; Kp is the spring coefficient, P d is the desired foot end position, P is the current foot end position; Kd is the damping coefficient, is the desired foot end velocity, is the current foot speed; f s is the average supporting force of the robot's four legs in a stable state.

Images