A motion control method and system for a wheeled biped robot

Abstract

The application discloses a kind of wheeled biped robot motion control method and system, comprising: model predictive controller is based on the motion control instruction of input and outputs the generalized force moment F that needs to be applied to trunk;Solving the distance Δx of the trunk centroid and wheel center point in x-axis direction of robot;Based on centrifugal force compensation strategy, calculate the trunk roll angle, based on terrain adaptation strategy, calculate the wheel position under the trunk coordinate system;After inverse kinematics is solved, the expected leg joint angle, leg joint speed and swing bar pitch angle are obtained as state input;Whole body controller is based on the generalized force moment F obtained by model predictive controller and the state input obtained by inverse kinematics solving, fusion dynamics feedforward and joint feedback, finally obtain the leg joint torque and wheel joint torque for controlling robot motion.The application enhances the robustness of robot when high-speed steering, to prevent rollover.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wheeled biped robots, and particularly relates to a motion control method and system of a wheeled biped robot. BACKGROUND

[0002] The statements in this section merely provide background information related to the present application and do not necessarily constitute the prior art.

[0003] With the continuous development of science and technology, mobile robots gradually integrate into daily life and work, which leads to higher demands for the scene adaptation ability and work efficiency of robots. Combining the advantages of high energy efficiency of wheeled robots and the strong terrain adaptation ability of legged robots, in many application scenarios, wheeled-leg robots become a better choice. However, four-wheel-legged or six-wheel-legged robots have the problem of large turning radius or slow turning speed when turning, which limits their motion flexibility. Wheeled biped robots with fast movement and zero-radius turning capability can well solve this pain point.

[0004] In the existing technology, the wheel motion and leg motion of the wheeled biped robot are usually controlled separately, that is, the wheel controller realizes the balance of the robot and the leg controller realizes the terrain adaptation. This way ignores the coupling effect between the wheel and the leg, which is not conducive to the stable control of the robot. At the same time, the wheeled biped robot has the motion potential of high-speed turning, but the faster the turning speed, the greater the risk of rollover. SUMMARY

[0005] In order to solve the above problems, the present application proposes a motion control method and system of a wheeled biped robot, establishes a wheel-leg dynamics model, adopts a method of actively making the torso produce a roll angle by cooperating with the two legs, and uses the component of gravity to offset the centrifugal force to enhance the robustness of the robot when turning at high speed and prevent rollover.

[0006] In some embodiments, the following technical solutions are adopted:

[0007] A motion control method of a wheeled biped robot, comprising:

[0008] A model predictive controller is constructed based on a torso dynamics model, and a full-body controller is constructed based on a wheel-leg dynamics model;

[0009] The model predictive controller outputs a generalized force wrench F that needs to be applied to the torso based on the input motion control instruction;

[0010] The distance between the center of mass of the torso and the center point of the wheel in the x-axis direction of the robot is solved based on the WIPM dynamics ;

[0011] The torso roll angle is calculated based on a centrifugal force compensation strategy, and the wheel position in the torso coordinate system is calculated based on a terrain adaptation strategy.

[0012] Based on the obtained distance , torso roll angle and wheel position, the expected leg joint angle, leg joint speed and swing bar pitch angle are obtained through inverse kinematics solution as state input.

[0013] The generalized force screw F obtained by the model predictive controller and the state input obtained by the inverse kinematics solution are fused by the whole-body controller based on dynamics feedforward and joint feedback, and finally the leg joint torque and wheel joint torque for controlling the robot motion are obtained.

[0014] In some other embodiments, the following technical solutions are adopted:

[0015] A motion control system of a wheeled biped robot, comprising:

[0016] A controller construction module for constructing a model predictive controller based on a torso dynamics model and a whole-body controller based on a wheel-leg dynamics model;

[0017] A model predictive control module for outputting the generalized force screw F required to be applied to the torso based on the input motion control instruction by the model predictive controller; calculating the distance between the robot torso centroid and the wheel center point in the x-axis direction based on the WIPM dynamics solution ; calculating the torso roll angle based on a centrifugal force compensation strategy, and calculating the wheel position in the torso coordinate system based on a terrain adaptation strategy;

[0018] A state calculation module for obtaining the expected leg joint angle, leg joint speed and swing bar pitch angle through inverse kinematics solution as state input based on the obtained distance , torso roll angle and wheel position.

[0019] A robot control module for obtaining the leg joint torque and wheel joint torque for controlling the robot motion through the whole-body controller based on the generalized force screw F obtained by the model predictive controller and the state input obtained by the inverse kinematics solution, and fusing dynamics feedforward and joint feedback.

[0020] In some other embodiments, the following technical solutions are adopted:

[0021] A terminal device comprising a processor and a memory, the processor being configured to implement instructions; the memory being configured to store a plurality of instructions, the instructions being adapted to be loaded and executed by the processor to implement the motion control method of the wheeled biped robot described above.

[0022] In some other embodiments, the following technical solutions are adopted:

[0023] A computer readable storage medium, wherein a plurality of instructions are stored, the instructions being adapted to be loaded and executed by a processor of a terminal device to perform the motion control method of the wheeled biped robot.

[0024] Compared with the prior art, the present application has the following beneficial effects:

[0025] (1) Unlike decoupling of the wheels and the biped body, the present application emphasizes the consistency of wheel-leg motion, establishes a wheel-leg dynamics model, and reveals the force transmission relationship from the wheels to the hip joints. At the same time, a centrifugal force compensation control strategy is developed, which actively causes the torso to produce a roll angle by using the two legs, and uses the gravity component to offset the centrifugal force to enhance the robustness of the robot during high-speed turning and prevent rollover.

[0026] (2) The present application designs a hierarchical control framework based on model predictive control, which integrates the centrifugal force compensation control strategy and the terrain adaptation control strategy to enhance the high-speed turning robustness and terrain adaptation ability of the robot. Under this control framework, the wheeled biped robot can exhibit better dynamic motion ability.

[0027] Other features and advantages of the present application and additional aspects will be in part apparent and in part pointed out hereinafter. BRIEF DESCRIPTION OF DRAWINGS

[0028] Figure 1 is a schematic diagram of the motion control method of the wheeled biped robot in the embodiment of the present application;

[0029] Figure 2 is a schematic diagram of the relationship between the center of mass and the wheel axis position in the embodiment of the present application. DETAILED DESCRIPTION

[0030] It should be noted that the following detailed description is merely exemplary in nature and is intended to provide further description of the application. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0031] It should be noted that the terms used herein are merely intended to describe specific embodiments, and are not intended to limit the exemplary embodiments according to the present application. As used herein, the singular form is intended to include the plural form unless the context clearly indicates otherwise, and it should be understood that when the terms "comprise" and / or "include" are used in the specification, there is a presence of the features, steps, operations, devices, components and / or combinations thereof.

[0032] Embodiment one

[0033] In one or more embodiments, a motion control method of a wheeled biped robot is disclosed, which is combined with Figure 1 , and specifically includes the following processes:

[0034] A motion control method of a wheeled biped robot, comprising:

[0035] (1) Construct a model predictive controller based on a torso dynamics model, and construct a whole-body controller based on a wheel-leg dynamics model;

[0036] (2) The model predictive controller outputs a generalized force wrench F that needs to be applied to the torso based on an input motion control instruction;

[0037] (3) Based on the WIPM dynamics, solve the distance between the torso center of mass and the wheel center point in the x-axis direction of the robot ;

[0038] (4) Calculate the torso roll angle based on a centrifugal force compensation strategy , and calculate the wheel position in the torso coordinate system based on a terrain adaptation strategy;

[0039] (5) Based on the obtained distance , torso roll angle , and wheel position, obtain the expected leg joint angle, leg joint speed, and swing bar pitch angle through inverse kinematics, as state input; wherein the line from the wheel axis center point to the center of mass is the swing bar in the WIPM model, and the angle between the swing bar and the coordinate axis Z in the WIPM model is the swing bar pitch angle, i.e. Figure 2 ; ;

[0040] (6) The whole-body controller, based on the generalized force wrench F solved by the model predictive controller and the state input solved by the inverse kinematics, obtains the leg joint torque and wheel joint torque through dynamics and joint space joint torque feedback control, to control the motion of the robot.

[0041] As a specific embodiment, the wheeled biped robot is a typical floating base system, and in order to facilitate the construction of the controller, the dynamics models of the torso subsystem and the wheel-leg subsystem are established respectively.

[0042] 1. The construction of the torso dynamics model is as follows:

[0043] The wheeled biped robot is a typical floating base system, and in order to facilitate the construction of the controller, the dynamics models of the torso subsystem and the wheel-leg subsystem are established respectively. As determined by the mechanical structure of the wheeled biped robot, the wheel-leg subsystem applies a six-dimensional generalized force wrench to the torso, wherein . , These represent the forward thrust provided to the torso by the left and right wheel-leg subsystems, respectively. , These represent the vertical support forces provided to the torso by the left and right wheel leg subsystems, respectively. Aligned with the x-axis direction of the control coordinate system. This represents the torque used for rotation along the y-axis of the control coordinate system. The direction is consistent with the z-axis of the control coordinate system. This embodiment defines the generalized coordinates of the torso. and speed for:

[0044]

[0045]

[0046] in, It is the position of the torso. Indicates the rotation of the torso. This indicates the angular velocity of the torso. , , , , , Representing three-dimensional coordinates and rotation respectively. Representing velocity, assuming no slippage between the wheels and the ground, subject to the nonholonomic constraints of the wheel structure, the torso is in the world coordinate system. Motion in a plane involves coupling relationships as described in the formula:

[0047]

[0048] in, This represents the forward velocity of the torso in the world coordinate system. It can be observed that:

[0049]

[0050] It is the rotation matrix along the z-axis of the world coordinate system. Due to this coupling relationship, the control coordinate system is more suitable for describing the robot's motion. The generalized coordinates and velocity of the torso in the control coordinate system are defined as follows:

[0051]

[0052] in, Represents the coordinates of the torso in the z-direction. The simplified linear dynamic model of the torso is as follows:

[0053]

[0054]

[0055] wherein, , , g represents the acceleration of gravity, represents the weight of the trunk, represents the trunk inertia tensor matrix in the control coordinate system, represents the vector from the trunk center of mass to the hip joint in the control coordinate system. , , represents the angular velocity, angular acceleration, and acceleration of the trunk, respectively.

[0056] with respect to is small, so the formula can be approximated as:

[0057]

[0058] At the same time, the inertia tensor matrix in the control coordinate system can be obtained:

[0059]

[0060] wherein, is the inertia tensor matrix in the trunk coordinate system, represents the transformation matrix from the trunk to the control coordinate system.

[0061] 2, based on the trunk dynamics model to construct a model predictive controller, as follows:

[0062] Model predictive control as a model-based control method, using the model to predict the dynamic characteristics of the system in the future (prediction horizon) in each control period, and then seek the finite time open-loop optimal control input in the current control period. The simplified trunk dynamics is arranged as follows:

[0063]

[0064] , represents the position vector of the left and right hip joints pointing to the trunk.

[0065] By adding gravity to the state variable, the state space form of the dynamics equation is obtained:

[0066]

[0067] wherein, , In order to meet the solution requirements of the actual robot, the embodiment is based on the zero-order holder to discretize the linear time-varying state space model, and a discrete state space model is obtained:

[0068]

[0069] where, , , is the sampling time of the controller, and denote the state and input vectors of the system at time k, denote the state of the system at time k+1 based on the model. Assuming the prediction horizon is n, in order to obtain the optimal state input at the current time, the optimization function is defined as:

[0070]

[0071] where, denote the control input at each time within the prediction horizon, k = 0, 1, …, n-1; , and are the weight matrices, denote that the control input at adjacent times should not appear to jump, so as to avoid the instability of the robot; denote the constraint matrix, so that the robot always remains in a controllable motion state; , , is the sampling time of the controller, denote the desired state of the torso at time k, , denote the current feedback state of the torso at times k and k+1, denote the input wrench of the torso at the previous time, and denote the constraint matrix, and denote the state transition matrix in the state space equation; solve the model predictive controller, and take the optimal solution at the current time as the generalized force wrench F that needs to be applied to the torso.

[0072] 3. Maintaining dynamic balance is the basis for achieving stable motion of wheeled biped robots. In order to simplify the analysis, in combination with Figure 2 , the embodiment reduces Skater to a wheeled inverted pendulum model to analyze the relationship between the center of mass position and the motion state under the dynamics constraint, which is conducive to planning the robot motion to quickly converge to a stable attitude. The wheel and the ground can be regarded as point contact, so the contact point is the zero moment point (ZMP). WIPM and standard LIPM have similar dynamic characteristics, so their dynamic equations also adopt similar forms:

[0073]

[0074] where, , are the acceleration of the trunk center of mass in these two directions respectively, is the x-axis coordinate of the center of mass, is the x-axis coordinate of the wheel center; denotes the wheel center point, denotes the distance between the trunk center of mass and the wheel center point in the x-axis direction. Since the leg joint motors and electrical components are all integrated in the trunk, and the material supporting the legs is a light mass aluminum alloy and is hollowed out for weight reduction, the weight of the robot is mainly concentrated in the trunk, and the mass of the legs can be ignored relative to the trunk. In this embodiment, the trunk center of mass can be approximated as the center of mass in the WIPM. According to the trunk dynamics, the acceleration of the center of mass can be derived from Newton's second law:

[0075]

[0076] Substituting it into the formula gives:

[0077]

[0078] Thus, to maintain the dynamic balance of the robot, the relative position relationship between the center of mass and the wheel axis in the x-axis direction reflected by the formula needs to be satisfied.

[0079] 4. The wheel-leg dynamics model construction process is as follows:

[0080] Compared with a legged robot, a wheeled biped robot is almost always in a supporting state when it is not jumping, and needs to maintain a stable posture while providing output force to the trunk. In this embodiment, the joint angle, velocity and torque are defined as:

[0081]

[0082] where, denotes the angle value of each joint, denotes the velocity of each joint, j = leg or wheel; denotes the hip and knee joints of the legs, denotes the wheel joint, denotes the torque value of the driving joint, denotes the torque value of the driving leg joint, denotes the torque value of the driving wheel joint; i = L or R; denotes the torque value of the hip and knee joints of the legs.

[0083] During the motion, especially when the motion state of the robot changes greatly, the mutual disturbance between the wheels and the legs cannot be ignored, and the analysis of the force or torque transmitted between them is necessary. The wheel-leg subsystem can be regarded as two three-DOF manipulators connected with the trunk. Due to the characteristics of the joint structure, the wheel joint drives the subsystem to follow the motion of the trunk while providing driving force for the trunk. Therefore, taking the wheel as the moving base (base coordinate system) and the end force wrench at the hip joint as the output force, the wheel-leg dynamics equation is solved based on the Newton-Euler dynamics formula as follows:

[0084]

[0085] represents the mass matrix, represents the Coriolis force and centrifugal force term, represents the gravity term, represents the Jacobian matrix, and F represents the generalized force wrench that needs to be applied to the trunk.

[0086] 5. The centrifugal force compensation control strategy for high-speed turning is as follows:

[0087] Due to structural limitations, the wheel-leg cannot provide lateral force to the trunk to offset the influence of centrifugal force during turning. A centrifugal force compensation (CFC) control strategy is proposed, which uses the cooperation of the two legs to generate a roll angle of the trunk, and offsets the centrifugal force through the component of gravity to enhance the robustness of the robot during high-speed turning and prevent rollover. The relationship between the roll angle, the speed and the yaw rate is as follows:

[0088]

[0089] Although this method can effectively avoid the influence of centrifugal force, the robot is still limited by the zero moment point and the leg extension margin. First, the projection of the trunk centroid on the ground needs to be between and Second, the leg extension margin determines the size of the roll angle, for example, when the height of the robot is at the maximum or minimum, the trunk cannot be rolled. Therefore, the trunk roll angle is calculated as:

[0090]

[0091] where, is the trunk roll angle, represents the forward speed of the trunk in the world coordinate system, represents the yaw rate of the trunk, represents the gravity acceleration, and represent the wheel-ground contact points of the left and right wheel-legs, represents the z-direction coordinate of the trunk, represents distance and between the minimum value, represents the distance between the contact points.

[0092] 6, the terrain adaptation strategy is as follows:

[0093] Wheel-legged robot cannot estimate the terrain through the position of the landing point. In the face of undulating terrain, real-time adjustment can be made through the posture information of the torso. In the torso coordinate system, the wheel coordinates are defined as and . l represents the left leg, r represents the right leg.

[0094] According to the geometric relationship, the relationship between the roll angle and the wheel position can be derived as:

[0095]

[0096] wherein, represents the distance between the wheel-ground contact points. At the same time, the robot height needs to be kept unchanged:

[0097]

[0098] When climbing or descending, it is important to adjust the position of the body's center of gravity to adapt to different slopes. According to the analysis of the wheel-inverted pendulum model, the only is determined between the torso and the wheel. At the same time, in order to ensure the stable cooperation of the double legs, it is necessary to keep the wheel axis in the same plane, and meet the following constraint conditions:

[0099]

[0100] Combining the above formulas, the following equation can be obtained:

[0101]

[0102] This embodiment can find that the change of the torso state caused by the terrain condition can be adapted by changing the wheel position. Because is full rank, the solution is:

[0103]

[0104] 7, based on the wheel-leg dynamics model to construct the whole body controller, specifically:

[0105] Unlike legged robots, wheeled biped robots do not have a swing phase, which means that the control task only includes trunk translation and posture rotation. According to the analysis of the wheeled inverted pendulum model, each different motion state of the robot corresponds to a unique internal stable posture: CoM position, pendulum length, and tilt angle. Therefore, the translation and rotation between the trunk and the wheels are determined. Based on the formulas in the above strategy, inverse kinematics is used to solve the desired joint angles and joint velocities . However, due to the small amplitude and acceleration of the trunk motion relative to the moving base, the joint accelerations solved by inverse kinematics are ignored and solved by PD control rates

[0106]

[0107] In the dynamics equation, the related terms are mainly to compensate for the inertial force of the moving base, so only the feedback values need to be brought into the dynamics equation. Then the joint feedforward torque can be solved by the following formula:

[0108]

[0109] Due to the inevitable errors in the dynamics model of the robot, PD feedback control is introduced to the leg joint torque:

[0110]

[0111] As a kind of robot with non-minimum phase characteristics, if only the feedforward force is used to control the wheel joint of the wheeled biped robot, it is easy to deviate from the equilibrium point, and the stability of the robot during motion cannot be guaranteed. Moreover, when the robot accelerates forward or backward, the wheels need to move in the opposite direction for a distance to meet the ZMP constraint in the WIPM model. Therefore, the LQR optimal feedback control rate is introduced in this embodiment to assist the robot to quickly converge to the equilibrium point.

[0112] The state vector of LQR is selected as , wherein respectively represent the pitch angle, pitch angle velocity, average wheel hub velocity, and yaw angle velocity of the pendulum rod. At the equilibrium point, the linear state space equation of the system is obtained by approximate linearization in this embodiment: . Wherein . , respectively represent the feedback torques of the left leg and the right leg of the LQR output.

[0113]

[0114] wherein, A function representing the parameters of the wheeled inverted pendulum model. It is assumed that the control period of the controller is The state space equation is discretized in this embodiment as follows, assuming a zero-order hold:

[0115]

[0116] The optimal solution to the LQR infinite horizon problem is obtained by solving the discrete-time algebraic Riccati equation (DARE), and the optimal feedback gain matrix K is finally obtained.

[0117]

[0118] where Q and R represent weight matrices. It is worth noting that the length of the swing rod will change with the change in the robot's standing height, so the optimal feedback gain matrix K is a function of the length of the swing rod, and the control input of the system is obtained from the LQR controller:

[0119]

[0120] where the length of the swing rod L and the feedback value of the state variable are given by the state estimator, and the expected value of the state variable is where the expected values of the pitch angle and the pitch angle velocity are usually set to 0, in order to make the robot reach the dynamic equilibrium posture in the WIPM model as soon as possible and speed up the tracking of the given velocity, which can be obtained according to the ZMP constraint of the WIPM:

[0121]

[0122] Thus, the wheel joint torque can be obtained as:

[0123]

[0124] where , and are the expected leg joint acceleration, angle, and joint velocity, respectively, , are the actual leg joint angle and joint velocity, respectively, is the joint feedforward torque, represents the Jacobian matrix, and F represents the generalized force twist that needs to be applied to the torso, , , , represent the proportional coefficient and the differential coefficient in the PD control, respectively; is the torso roll angle, and represent the torque value of the driven leg joint and the feedforward torque of the leg joint, , These represent the torque value of the drive wheel joint and the feedforward torque of the wheel joint, respectively. Represents the mass matrix, Represents the Coriolis force and centrifugal force terms. Represents the gravity term. Let F represent the Jacobian matrix, and let F represent the generalized force spinor that needs to be applied to the torso. This represents the expected value of the pitch angle. This represents the optimal feedback gain matrix. , These represent the feedback value and the expected value of the state variable, respectively.

[0125] Finally, the obtained leg joint torque and wheel joint torque are processed by a state estimator (such as a Kalman filter) to obtain the robot's current motion state data, which are then fed back to the model predictive controller and the whole body controller, respectively.

[0126] Example 2

[0127] In one or more embodiments, a motion control system for a wheeled bipedal robot is disclosed, comprising:

[0128] The controller construction module is used to build a model predictive controller based on the torso dynamics model and a whole-body controller based on the wheel-leg dynamics model.

[0129] The model predictive control module is used to output the generalized force spinter F to be applied to the torso based on the input motion control commands from the model predictive controller; and to solve the distance between the robot's torso center of mass and the wheel center points in the x-axis direction based on WIPM dynamics. The torso roll angle is calculated based on a centrifugal force compensation strategy, and the wheel position in the torso coordinate system is calculated based on a terrain adaptation strategy.

[0130] The state calculation module is used to calculate the state based on the obtained distance. The torso roll angle and wheel position are used to obtain the desired leg joint angle, leg joint velocity, and swing arm pitch angle through inverse kinematics, which serve as the state inputs.

[0131] The robot control module is used to obtain the leg joint torque and wheel joint torque by integrating the generalized force spinor F obtained by the model predictive controller and the state input obtained by the inverse kinematics based on the whole body controller, and the dynamic feedforward and joint feedback.

[0132] The specific implementation methods of the above modules have been described in Example 1, and will not be detailed here.

[0133] Example 3

[0134] In one or more embodiments, a terminal device is disclosed, comprising a server, the server comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the motion control method of the wheeled biped robot in embodiment one when executing the program. For brevity, the motion control method of the wheeled biped robot in embodiment one will not be repeated here.

[0135] It should be understood that, in the embodiments, the processor can be a central processing unit (CPU), and the processor can also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), ready programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0136] The memory can include read-only memory (ROM) and random access memory (RAM) and provide instructions and data to the processor, and a portion of the memory can also include non-volatile random access memory. For example, the memory can also store device type information.

[0137] In the implementation process, each step of the above method can be completed by integrated logic circuits of hardware in the processor or instructions in the form of software.

[0138] Embodiment four

[0139] In one or more embodiments, a computer-readable storage medium is disclosed, wherein a plurality of instructions are stored, the instructions being adapted to be loaded and executed by a processor of a terminal device to implement the motion control method of the wheeled biped robot in embodiment one.

[0140] Although the specific embodiments of the application have been described in detail above with reference to the accompanying drawings, the above description is not a limitation on the scope of protection of the application, and those skilled in the art should understand that various modifications or changes made on the basis of the technical solutions of the application without creative labor are still within the scope of protection of the application.

Claims

1. A method for motion control of a wheeled biped robot, characterized by, The method comprises the following steps: a model predictive controller is constructed based on a torso dynamics model, and a whole-body controller is constructed based on a wheel-leg dynamics model; the model predictive controller outputs a generalized force wrench F required to be applied to the torso based on an input motion control instruction; Based on the WIPM dynamics to solve the distance between the center of mass of the robot's torso and the center of the wheel in the x-axis direction ; a torso roll angle is calculated based on a centrifugal force compensation strategy, and a wheel position in a torso coordinate system is calculated based on a terrain adaptation strategy; Based on the obtained distances The desired leg joint angles, leg joint velocities and swing link pitch angles are obtained by inverse kinematics solution as state inputs, based on the obtained distances, trunk roll angle and wheel position. the whole-body controller fuses dynamics feedforward and joint feedback based on the generalized force wrench F solved by the model predictive controller and state input solved by inverse kinematics, and finally obtains leg joint torque and wheel joint torque for controlling robot motion; The whole-body controller solves for leg joint accelerations, leg joint feedforward torques, wheel joint feedforward torques, and swing link pitch angles based on the resulting generalized force wrench, distance , torso roll angle, and wheel position, specifically as follows: wherein, , and are desired leg joint accelerations, angles and joint velocities, respectively, , are actual leg joint angles and joint velocities, respectively, is a joint feedforward torque, denotes a Jacobian matrix, F denotes a generalized force wrench to be applied to the torso, , , , denote a proportional and a derivative coefficient in the PD control, respectively; is a torso roll angle, and denote a torque value to drive a leg joint and a feedforward torque for the leg joint, , denote a torque value to drive a wheel joint and a feedforward torque for the wheel joint, respectively; denotes a mass matrix, denotes a Coriolis and centrifugal force term, denotes a gravitational force term; denotes a desired value of a pitch angle, denotes an optimal feedback gain matrix, , denote a state variable feedback value and a state variable desired value, respectively; , denote a forward pushing force provided by the left and right wheel-leg subsystems to the torso, respectively, , denote a vertical support force provided by the left and right wheel-leg subsystems to the torso, respectively.

2. The motion control method of the wheeled biped robot according to claim 1, wherein the model predictive controller is constructed based on a torso dynamics model, specifically: where, represents the control input at each time within the prediction horizon, k = 0, 1, …, n - 1; , and are weight matrices, represents the desired control input at the adjacent time should not appear jump phenomenon, in order to cause the robot instability; represents the constraint matrix, so that the robot always remains in a controllable motion state; , , is the sampling time of the controller, represents the desired state of the torso at time k, , represent the current feedback state of the torso at time k and k + 1, respectively, represents the input vector of the system at time k; represents the torso input wrench at the previous time, and represent the constraint matrix, and represent the state transition matrix in the state space equation; solve the model predictive controller, take the optimal solution at the current time as the generalized force wrench F that needs to be applied to the torso.

3. The motion control method of the wheeled biped robot according to claim 1, wherein the wheel-leg dynamics model, specifically: wherein denotes the angle value of a joint, j = leg or wheel; denotes the velocity of each joint, denotes the torque value driving a joint; denotes the hip and knee joints of the leg, denotes the wheel joint, denotes the torque value driving the leg joint, denotes the torque value driving the wheel joint; i = L or R; denotes the mass matrix, denotes the Coriolis and centrifugal force terms, denotes the gravity term, denotes the Jacobian matrix, F denotes the generalized force wrench to be applied to the trunk.

4. The motion control method of the wheeled biped robot according to claim 1, wherein Based on the WIPM dynamics to solve the distance between the center of mass of the robot trunk and the center of the wheel in the x-axis direction , specifically: wherein, represents a coordinate in the z direction of the torso, , respectively represent forward pushing forces provided by the left and right wheel-leg systems to the torso, , respectively represent vertical support forces provided by the left and right wheel-leg systems to the torso.

5. The motion control method of a wheeled biped robot according to claim 1, wherein, the torso roll angle is calculated based on a centrifugal force compensation strategy, specifically: wherein, is a torso roll angle, denotes a forward velocity of the torso in the world coordinate system, denotes a yaw angular velocity of the torso, denotes a gravitational acceleration, and denote the wheel-ground contact points of the left and right wheel legs, respectively, denotes a z-direction coordinate of the torso, denotes a distance and between the minimum values, denotes a distance between the contact points.

6. The motion control method of a wheeled biped robot according to claim 1, wherein, the wheel position in the torso coordinate system is calculated based on a terrain adaptation strategy, specifically: wherein, , , respectively denote the coordinates of the wheels in the trunk coordinate system; d denotes the distance between the wheel-ground contact points.

7. A motion control system of a wheeled biped robot, characterized by, The method comprises the following steps: a controller construction module is configured to construct a model predictive controller based on a torso dynamics model, and construct a whole-body controller based on a wheel-leg dynamics model; A model predictive control module is configured to output, by a model predictive controller, a generalized force wrench F to be applied to the torso based on the input motion control command; calculate a distance between a center of mass of the robot and a center of the wheel in an x-axis direction based on a WIPM dynamics solver ; calculate a torso roll angle based on a centrifugal force compensation strategy, and calculate a wheel position in a torso coordinate system based on a terrain adaptation strategy. a state computation module for computing desired leg joint angles, leg joint velocities, and swing link pitch angles as state inputs based on the obtained distances , torso roll angle, and wheel position, through inverse kinematics a robot control module is configured to fuse dynamics feedforward and joint feedback based on the generalized force wrench F solved by the model predictive controller and state input solved by inverse kinematics through the whole-body controller, and finally obtain leg joint torque and wheel joint torque for controlling robot motion. The whole-body controller solves for leg joint accelerations, leg joint feedforward torques, wheel joint feedforward torques, and swing link pitch angles based on the resulting generalized force wrench, distance , torso roll angle, and wheel position, specifically as follows: wherein , and are desired leg joint accelerations, angles and joint velocities, respectively, , are actual leg joint angles and joint velocities, respectively, is a joint feedforward torque, denotes a Jacobian matrix, F denotes a generalized force wrench to be applied to the torso, , , , denote a proportional and a derivative coefficient in the PD control, respectively; is a torso roll angle, and denote a torque value to drive a leg joint and a feedforward torque for the leg joint, , denote a torque value to drive a wheel joint and a feedforward torque for the wheel joint, respectively; denotes a mass matrix, denotes a Coriolis and centrifugal force term, denotes a gravitational force term; denotes a desired value of a pitch angle, denotes an optimal feedback gain matrix, , denote a state variable feedback value and a state variable desired value, respectively; , denote a forward pushing force provided by the left and right wheel-leg subsystems to the torso, respectively, , denote a vertical support force provided by the left and right wheel-leg subsystems to the torso, respectively.

8. A terminal device comprising a processor and a memory, the processor configured to implement instructions; the memory configured to store a plurality of instructions, the terminal device characterized by, The instructions are adapted to be loaded and executed by the processor of the terminal device to implement the motion control method of the wheeled biped robot according to any one of claims 1-6.

9. A computer-readable storage medium having stored therein a plurality of instructions, wherein the instructions, when executed by a processor, cause the processor to perform operations comprising: The instructions are adapted to be loaded and executed by the processor of the terminal device to implement the motion control method of the wheeled biped robot according to any one of claims 1-6.

Images