Wheel-legged robot and control method thereof

Abstract

This invention discloses a wheeled-legged robot and its control method, belonging to the field of robotics technology. The robot employs a combined wheel-leg structure, including a wheeled portion, leg portions, and hardware circuit modules. The wheeled portion utilizes a readily available bicycle frame, including a frame, front wheel, rear wheel, and steering mechanism. The leg portions are fixed to both sides of the frame, facilitating physical interaction with the ground environment and improving adaptability to complex environments. This invention combines the advantages of wheeled and legged robots, designing different controllers for different terrains, enabling the mobile robot to simultaneously travel quickly and smoothly on flat ground and dynamically overcome obstacles, thus improving the robot's mobility.

Description

Technical Field

[0001] This invention belongs to the field of robotics technology, specifically designing a novel wheeled-legged robot and its control method. This robot can be used for autonomous ground movement tasks of varying complexity. Background Technology

[0002] Ground mobile robots are one of the fastest-growing fields in scientific research. Because they can move autonomously without the assistance of external human operators, they can replace human workers in many fields, including field exploration, planetary exploration, urban patrol, emergency rescue operations, field reconnaissance, industrial automation, construction, entertainment, museum guidance, personal services, transportation, and healthcare.

[0003] Ground-based mobile robots can be divided into two main categories: legged robots (based on legs) and wheeled robots (based on wheels). While some indoor-walking legged robots excel at overcoming obstacles such as stairs or complex, uneven terrain, they typically require a significant amount of time to perform these complex movements. In contrast, wheeled robots are well-suited for flat terrain because they can move smoothly, efficiently, and quickly. However, they generally cannot handle rugged terrain, especially when encountering obstacles larger than the radius of their wheels.

[0004] Therefore, due to the aforementioned limitations of both purely wheeled and legged robots, there is an urgent need to develop a wheel-legged robot that combines the advantages of both, enabling rapid movement on flat surfaces while simultaneously navigating smoothly and overcoming obstacles in complex environments. Furthermore, research into the structure and control methods of wheel-legged hybrid robots can provide new ideas and methods for studying other unstable and nonlinear complex systems, and also offer new solutions for applications such as unmanned delivery and urban security. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a wheeled-legged robot and its control method. This invention integrates the advantages of wheeled and legged robots into a single robot through structural and control algorithm design, enabling the robot to perform different functions in various scenarios.

[0006] The specific technical solution adopted in this invention is as follows:

[0007] In a first aspect, the present invention provides a wheeled-legged robot, comprising a wheeled part, a leg part, and a hardware circuit module;

[0008] The wheeled component includes a front wheel, a rear wheel, and a steering component mounted on the frame; the front wheel serves as the robot's power wheel and is driven by a hub motor; the steering component is located above the front wheel and includes a throttle, a timing pulley, and a steering motor, which transmits torque to the throttle via the timing pulley to steer the robot.

[0009] The leg portion includes a left leg and a right leg, and each leg is equipped with a motor module consisting of a hip lateral swing joint motor, a hip lateral swing joint motor and a knee joint motor;

[0010] The hardware circuit module includes a main control board, a motor control board, an information acquisition module, and a power supply module. The power supply module provides operating voltage for the entire circuit system and the motor. The hub motor and the steering motor are both connected to the motor control board, which is connected to the main control board and communicates with it using USART. The information acquisition model communicates with the main control board using USART.

[0011] Preferably, both the left and right legs include a thigh and a calf; the upper end of the thigh is rotatably connected to a leg base, and the leg base is fixedly connected to the vehicle frame via a bracket; the upper end of the calf is rotatably connected to the thigh.

[0012] Furthermore, the hip lateral swing joint motor and the hip lateral swing joint motor are used to control the rotation mode of the connection between the thigh and the leg base, and the knee joint motor is used to control the rotation mode of the connection between the thigh and the lower leg.

[0013] Preferably, the power module includes a voltage regulator module and several model aircraft lithium batteries; the model aircraft lithium batteries can directly output 42V and 12V voltages or output 5V and 19V voltages through the voltage regulator module; wherein, the 42V voltage is used to power the hub motor and steering motor, the 12V voltage is used to power the motor module, the 5V voltage is used to power the motor control board, and the 19V voltage is used to power the main control board.

[0014] Preferably, the information acquisition module includes an inertial measurement unit, a metering wheel, and a motor encoder; the inertial measurement unit is used to collect the robot's posture and speed information and is fixed in the middle of the frame; the metering wheel is used to obtain the robot's position and is installed at the rear of the frame; the motor encoder is integrated in the motor module and is used to obtain the rotation angle and angular velocity of each motor.

[0015] Preferably, the main control board is a Jetson TX2; the motor control board is an STM32F429 development board.

[0016] Preferably, the power module, main control board, and motor control board are all placed on the rear seat of the vehicle frame.

[0017] Secondly, the present invention provides a control method for the wheeled-legged robot described in any of the first aspects, as follows:

[0018] First, a dynamic model of the wheeled robot needs to be performed, including the robot coordinates and system parameters, such as... Figure 2 As shown. Let the contact points of the front and rear wheels of the wheeled robot be C1 and C2, respectively. We choose the rear wheel's ground contact point C2 to represent the robot's planar motion, and its position in the World coordinate system is represented as r = [xy]. T The World coordinate system is fixed on the ground, with the Z-axis pointing vertically upwards. This represents the displacement of the center of gravity projection G′ to the end of the i-th leg (i = 0 for the left leg, i = 1 for the right leg). Let r represent the ground reaction force at the end of the i-th leg. i f i All are described in the World coordinate system. Let the robot's forward direction be the positive x-axis, and let the robot's tilt angle, yaw angle, and turn angle be denoted as... ψ, φ. The robot has a mass of m and is concentrated at its center of gravity G. The robot's moment of inertia about C1C2 is J. b The horizontal and vertical distances from the center of mass G to the rear wheel contact point C2 are l and l, respectively. b h G The distance between C1 and C2 is l, the fork lean angle is ε, and the front wheel wake is l. t The linear velocity of the outer edge of the rear wheel is v. The leg structure is considered as two 3-DOF robotic arms with a hip lateral swing joint, a hip lateral swing joint, and a knee joint. For ease of representation, s is used. x =sinx and c x =cosx represents the trigonometric function of angle x.

[0019] A nonholonomic constraint is given at C2. in This represents the robot's velocity along the x-axis in the World coordinate system. Let r represent the robot's velocity along the y-axis in the World coordinate system. Taking the third derivative of r, we get...

[0020]

[0021] where u=[u v u ψ ] T , robot yaw rate but

[0022]

[0023] The Lagrangian of the entire system is

[0024]

[0025] Where T represents the total kinetic energy of the system, and V represents the total potential energy of the system. Indicates the velocity of the center of mass. This represents the change in centroid height due to the change in steering angle; the projection angle is...

[0026] Using the Lagrange method, the dynamic equations of the robot without using its legs can be obtained.

[0027]

[0028] J t =mh G 2 +J b ,

[0029]

[0030] When a robot uses its legs to generate interaction forces with the ground, the ground reaction forces will affect the system's dynamics. These effects should be taken into account. yz f i Let represent the ground reaction force of the i-th leg in the yz plane of the Body coordinate system (i = 0 for the left leg, i = 1 for the right leg). Let yz r i Let represent the position vector projection of the i-th leg in the yz plane of the Body coordinate system (i = 0 for the left leg, i = 1 for the right leg). Then the corresponding ground reaction moment can be expressed as:

[0031] τ ig = yz r i × yz f i

[0032] Finally, we can obtain the torque balance equation when the leg generates interaction forces with the ground:

[0033]

[0034] The above formula can be simplified to in These are terms related to steering control and can be considered constants. It is the balancing torque of the leg-ground interaction.

[0035] Assuming the robot does not consider the power provided by the wheels and rolling friction, f aheadThe force equilibrium equation representing the external forces acting on the robot (here referring to the interaction force between the robot's legs and the ground) in its forward direction is as follows:

[0036]

[0037] Let the state variables of the system be because

[0038] Substituting these equations into the torque balance equation and the force balance equation, we obtain the robot's state-space equation as follows:

[0039]

[0040] The final dynamic equation of the system is as follows:

[0041]

[0042] S1: For flat terrain, in order to enable the robot to move quickly and maneuverably, an internal and external dynamic coordination (EIC) balance control strategy is adopted, as follows:

[0043] Let the coordinates of the rear wheel contact point C2 be... Then there is

[0044]

[0045] In the formula, Indicates the speed at the point of contact of the rear wheel (10). This represents the robot's velocity along the x-axis in the World coordinate system. Let v represent the robot's velocity along the y-axis in the World coordinate system; where the World coordinate system is fixed on the ground, the Z-axis is vertically upward, and the X-axis is the forward direction; v represents the linear velocity of the outer edge of the rear wheel (10), and c represents the velocity of the robot along the y-axis. ψ The cosine function representing the robot's yaw angle, s ψ The sine function representing the robot's yaw angle;

[0046] Continue to Taking the second derivative, and assuming the robot is moving forward at a constant speed, we can finally obtain...

[0047]

[0048] In the formula, The intermediate control variable is represented by ψ, which represents the robot's yaw angle; given the desired trajectory... Design a linear controller to track this trajectory for external dynamics.

[0049]

[0050] In the formula, xu Represents the desired x-coordinate of the robot, y-coordinate u This represents the robot's desired y-coordinate and desired robot position. Positional deviation b0, b1, b2 represent the feedback gain variables of the controller;

[0051] x (3) =x u (3) ,y (3) =y u (3) Substituting into the above formula, we can obtain the external dynamic control input.

[0052]

[0053] To consider internal dynamic balance, the above formula cannot be directly used as the system's control input; for the robot to maintain balance, the tilt angle acceleration must be zero, i.e.

[0054]

[0055] In the formula, u ψ This represents a quantity related to the yaw angle ψ;

[0056] Will Substituting into the above equation and using Newton's method, we can solve for... Again The total derivative can be obtained by taking the steps in sequence. For the obtained These represent the desired tilt angle, tilt angular velocity, and tilt angular acceleration that ensure the system's trajectory tracking.

[0057] To enable the system to track a trajectory at the desired tilt angle, achieving both robot balance and trajectory tracking, partial feedback linearization of the entire system is first required before a linear controller can be used for tracking. and its derivative;

[0058] Therefore, an internal dynamic feedback control law is designed.

[0059]

[0060] Substituting into the dynamic formula achievable

[0061]

[0062] From this point on, the tilt angle acceleration of the system can be configured arbitrarily;

[0063] In the formula, J t =mh G 2+J b ,

[0064] in, This indicates the robot's desired tilt angle acceleration. The value represents the robot's tilt angle acceleration, m represents the robot's mass, and h represents the robot's angular acceleration. G Let J represent the vertical distance from the robot's center of gravity G to the contact point C2 of the rear wheel; let C1 and C2 be the contact points of the robot's front and rear wheels, respectively. b This represents the robot's moment of inertia about C1C2; The cosine function representing the robot's tilt angle. The sine function representing the robot's tilt angle, l b This represents the horizontal distance from the robot's center of gravity G to the point of contact C2 with the rear wheel. t Indicates the front wheel wake, φ g c represents the projection angle of the throttle angle onto the ground. ε Let l represent the cosine function of the robot's fork tilt angle, and l represent the distance between C1 and C2.

[0065] In order to track and its derivative, let

[0066]

[0067] In the formula, a0 and a1 represent the trajectory tracking parameters;

[0068] As long as a0, a1 > 0, the tracking error can be guaranteed to converge to zero;

[0069] Ultimately, the internal and external dynamic coordination motion controller for driving on flat ground is...

[0070] S2: For rough terrain, a wheel-leg cooperative control strategy that integrates model prediction and internal and external dynamic coordination is adopted, as follows:

[0071] The wheel-leg coordinated control strategy integrates leg control and internal and external dynamic coordination control. It generates wheel-leg coordinated control signals based on the user-given desired trajectory and sensor information fed back by the robot, which drive the hip lateral swing joint motor, hip lateral swing joint motor, knee joint motor, steering motor and wheel hub motor respectively.

[0072] The auxiliary leg control consists of a support phase and a swing phase, with the two states scheduled by a gait generator. The support phase uses the MPC algorithm to generate the required joint motor torque as torque feedforward, and at the same time, feedback control is added to perform force-position mixing control. The swing phase uses position and velocity control.

[0073] During the support phase, the leg ends touch the ground, generating an interaction force between the legs and the ground. The function of MPC is to calculate the system input that minimizes the error of the state trajectory based on the desired state trajectory and the current sensing information, while also minimizing the system input to save energy. The MPC form is as follows:

[0074]

[0075]

[0076]

[0077] in It is τ∈[t,t+Hδ t The control sequence, δ t It is the period of model predictive control, Hδ t Represents the prediction time domain, where The error of the state vector e = xx d , x represents the state variable of the system. d Let Q and W represent the desired state variables of the system, and W be diagonal weight matrices; A represent the coefficients of the state variables in the state-space equations, B represent the coefficients of the system inputs in the state-space equations, E represent the constant term, and r represent the constant term. i This represents the displacement of the robot's center of gravity projected onto the end of the i-th leg; Indicates control input, Let i represent the ground reaction force at the end of the i-th leg, where i = 0 represents the left leg and i = 1 represents the right leg; These represent the upper and lower limits of friction force and motor torque, respectively, with the subscript j indicating the j-th element of the vector;

[0078] Matrix C is

[0079]

[0080] Where μ represents the static friction coefficient between the leg and the ground; then the above optimization problem is solved by discretization and using the qpOASES solver.

[0081] The interaction force between the legs and the ground acts on the robot like a series of balanced torques. When the i-th leg touches the ground, f i The joint torque can be calculated using the following formula.

[0082]

[0083] in, This represents the rotation matrix from the Body coordinate system to the World coordinate system; The transpose of the Jacobian matrix is ​​given; thus, the feedforward torque of the leg joint can be calculated.

[0084] Simultaneously, the supporting leg should follow a trajectory to ensure that the leg always maintains contact with the same point on the ground; let the desired joint angle, joint angular velocity, and joint angular acceleration at a certain moment be q, respectively. d , Then the feedback control quantity u fb for

[0085]

[0086] In the formula, K d K represents the differential gain. p Indicates proportional gain. q represents the joint angular velocity, and q represents the joint angle.

[0087] Therefore, the final joint motor torque

[0088]

[0089] Next, we will analyze the control of the swing phase. The key to the swing phase is determining the landing point. A heuristic formula can be used to calculate the desired leg end position, thus achieving speed regulation.

[0090]

[0091] In the formula, p des T represents the desired position of the end of the leg in the xy plane. s It is the gait period, p hip It is the position of the hip lateral joint motor in the xy plane, v G ref =[v x v y ] T It is the expected velocity of the robot's center of gravity in the xy plane, v G It corresponds to the center of gravity velocity, k p It is gain;

[0092] During the robot's movement, internal and external dynamic coordination control enables the robot to maintain balance and track its trajectory. The periodic stepping of the auxiliary legs enhances the physical interaction between the robot and the ground, greatly improving the robot's stability under the condition of limited throttle torque.

[0093] S3: For obstacle-crossing scenarios, a leg force pulse-based control strategy is adopted to enhance the robot's obstacle-crossing and anti-interference capabilities, as detailed below:

[0094] When the robot traverses obstacles, it triggers a collision detection program. When the system becomes unstable, it initiates leg force pulses. Therefore, by applying force pulses to the system, it returns to its previous equilibrium position, as detailed below:

[0095] If we denote the leg-ground interaction force as F, then the torque applied to the robot is δτ = r i ×F; Let the torque vector of a single leg joint be... but in It is the rotation matrix that transforms the Body coordinate system to the World coordinate system; therefore, the applied torque can be obtained as follows:

[0096]

[0097] The pulse balancing torque that can act on the robot is the component along the x-axis, i.e., δτ. x =δτ·e x ,e x =

[100] T For the purpose of balance, we mainly focus on δτ. x The control design, and the torque components δτ in the y-axis and z-axis directions y and δτ z The smaller the better;

[0098] Assuming the pulse torque δτ is at t τ Time passes through an extremely short period of time κ, reaching Under the action of the torque pulse, the robot's tilt angular velocity will undergo discontinuous jumps, while the tilt angle remains constant; according to the dynamic formula... have to

[0099]

[0100] Note that due to the tilt angle at The period remains unchanged, therefore According to the above formula, we can obtain

[0101]

[0102] in This represents the change in tilt angular velocity. and These represent the tilt angles before and after the torque pulse is applied;

[0103] The solution can be transformed into an optimization problem, namely, finding a suitable desired tilt angular velocity. The optimization problem aims to minimize the system's equilibrium and trajectory tracking errors, as well as the system control input; therefore, it can be described as follows:

[0104]

[0105]

[0106]

[0107] in It is an error vector, a matrix and Both are symmetric positive definite matrices, H t >0 represents the prediction time period; the objective function is expressed in H t This period involves the accumulation of robot state trajectory error and control input; this optimization problem aims to find the optimal initial tilt angular velocity so that the robot can rebalance and track its trajectory; it is solved using the Sequential Quadratic Programming (SQP) algorithm.

[0108] The value of κ should be as small as possible to ensure that the duration of leg-ground contact is short enough; during this time, assuming δτ x The result remains unchanged; it can be solved using the SQP algorithm. Then we can further obtain

[0109]

[0110] To generate δτ x It is necessary to design the leg-ground interaction force F; when the leg contacts the ground, the contact point is located at So Where F z =F·e z e z =[0 0 1] T Meanwhile, in order to eliminate δτ y and δτ z The resulting effect increased F x =F y =0, Constraints; In practice, collisions between wheels and obstacles typically prevent the robot from rotating along the y-axis and z-axis, therefore δτ y and δτ z The impact on the system is minimal and can be ignored; therefore, F can be obtained. z

[0111]

[0112] Where F z max It is the maximum vertical ground reaction force, and the position r of the leg-ground contact point. yi Adjust dynamically during the experiment;

[0113] The required joint torque can be obtained in the end.

[0114]

[0115] in i = L, R, j = 0, 1, 2.

[0116] Compared with the prior art, the present invention has the following advantages:

[0117] This invention designs a novel wheeled-legged mobile robot that combines high-speed travel on flat ground, stable travel at low speeds and on rough terrain, and strong resistance to unstructured road surface interference. It solves the problem of maintaining balance for two-wheeled robots in obstacle crossing, low-speed travel, and traversing rough terrain. Attached Figure Description

[0118] Figure 1 This is a schematic diagram of the wheeled-legged robot of the present invention, wherein (a) is a top view and (b) is a front view;

[0119] Figure 2 This is a schematic diagram of the modeling coordinates and parameters of the wheeled legged robot of the present invention, where (a) is a top view and (b) is a front view;

[0120] Figure 3 This is a schematic diagram of the hardware circuit module of the wheeled legged robot of the present invention;

[0121] Figure 4 This is a schematic diagram of the EIC control algorithm of the present invention;

[0122] Figure 5 This is a schematic diagram of the wheel-leg cooperative control algorithm of the present invention;

[0123] Figure 6 This is a schematic diagram of the leg force pulse control algorithm of the present invention;

[0124] The attached diagram is labeled as follows: 1. Frame; 2. Inertial measurement unit; 3. Steering motor; 4. Synchronous pulley; 5. Throttle; 6. Meter wheel; 7. Rear seat; 8. Bracket; 9. Hub motor; 10. Rear wheel; 11. Front wheel; 12. Leg base; 13. Hip lateral swing joint motor; 14. Hip lateral swing joint motor; 15. Knee joint motor. Detailed Implementation

[0125] The present invention will be further described and illustrated below with reference to the accompanying drawings and specific embodiments. The technical features of each embodiment of the present invention can be combined accordingly, provided that there is no mutual conflict.

[0126] like Figure 1 As shown, this invention provides a wheeled-legged robot, which mainly includes a wheeled part, a leg part, and a hardware circuit module. The structure and connection method of each component will be described in detail below.

[0127] The wheeled component comprises four parts: a frame 1, a front wheel 11, a rear wheel 10, and a steering mechanism. The front wheel 11 serves as the robot's power wheel and is driven by a hub motor 9. The steering mechanism includes a throttle 5, a timing pulley 4, and a steering motor 3. The steering motor 3 is fixed to the frame 1 and transmits torque to the throttle 5 via the timing pulley 4, enabling the robot to steer.

[0128] The leg section consists of two legs, left and right, each with three motor modules: a hip lateral swing joint motor 13, a hip lateral swing joint motor 14, and a knee joint motor 15.

[0129] Based on the above structure, a hardware circuit module is provided on the frame 1. For example... Figure 3 As shown, the hardware circuit module includes a main control board, a motor control board, an information acquisition module, and a power supply module. The power supply module provides operating voltage for the entire circuit system and the motors. Both the hub motor 9 and the steering motor 3 are connected to the motor control board, which in turn connects to the main control board and communicates via USART. The information acquisition model communicates with the main control board using USART.

[0130] In this embodiment, the information acquisition module includes an inertial measurement unit (IMU), a meter-counting wheel, and a motor encoder. The IMU 2, fixed to the middle of the frame 1, collects the robot's attitude and speed information. The meter-counting wheel 6, mounted at the rear of the frame 1, is used to acquire the robot's position. The motor encoder, integrated into the motor module, acquires the rotation angle and angular velocity of each motor. The power module includes a voltage regulator module and several model aircraft lithium batteries. The model aircraft lithium batteries can directly output 42V and 12V voltages or output 5V and 19V voltages through the voltage regulator module. The 42V voltage powers the hub motor 9 and the steering motor 3, the 12V voltage powers the motor module, the 5V voltage powers the motor control board, and the 19V voltage powers the main control board.

[0131] In this embodiment, the legs are fixed to the leg base using bolts and nuts, and the leg base is fixed to the frame bracket using the same connection method. The bracket and the frame are connected by welding. The main control board is an NVIDIA Jetson TX2; the motor control board is an STM32F429 development board. The power module, main control board, and motor control board are all located in the rear seat.

[0132] Based on the aforementioned wheeled-legged robot, this invention also designs a control method that enables the robot to simultaneously move quickly and smoothly on flat ground and dynamically overcome obstacles.

[0133] According to such Figure 2 The robot's coordinates and system parameters are shown, and a dynamic model of the robot is performed.

[0134] Let C1 and C2 be the contact points of the front and rear wheels of the wheeled robot, respectively. We choose the rear wheel's ground contact point C2 to represent the robot's planar motion, and its position in the World coordinate system is represented as r = [xy]. T The World coordinate system is fixed on the ground, with the Z-axis pointing vertically upwards. This represents the displacement of the center of gravity projection G′ to the end of the i-th leg (i = 0 for the left leg, i = 1 for the right leg). Let r represent the ground reaction force at the end of the i-th leg. i f i All are described in the World coordinate system. Let the robot's forward direction be the positive x-axis, and let the robot's tilt angle, yaw angle, and turn angle be denoted as... ψ, φ. The robot has a mass of m, concentrated at its center of mass G, and its moment of inertia about C1C2 is J. b The horizontal and vertical distances from the center of mass G to the rear wheel contact point C2 are l and l, respectively. b h G The distance between C1 and C2 is l, the fork lean angle is ε, and the front wheel wake is l. t The linear velocity of the outer edge of the rear wheel is v. The leg structure is considered as two 3-DOF robotic arms with a hip lateral swing joint, a hip lateral swing joint, and a knee joint. For ease of representation, s is used. x =sin x and c x =cosx represents the trigonometric function of angle x.

[0135] A nonholonomic constraint is given at C2. in This represents the robot's velocity along the x-axis in the World coordinate system. Let r represent the robot's velocity along the y-axis in the World coordinate system. Taking the third derivative of r, we get...

[0136]

[0137] where u=[u v u ψ ] T , robot yaw rate but

[0138]

[0139] The Lagrangian of the entire system is

[0140]

[0141] Where T represents the total kinetic energy of the system and V represents the total potential energy of the system;

[0142] Indicates the velocity of the center of mass. This represents the change in centroid height due to the change in steering angle; the projection angle is...

[0143] Using the Lagrange method, the dynamic equations of the robot without using its legs can be obtained.

[0144]

[0145] J t =mh G 2 +J b ,

[0146]

[0147] When a robot uses its legs to generate interaction forces with the ground, the ground reaction forces will affect the system's dynamics. These effects should be taken into account. yz f i Let represent the ground reaction force of the i-th leg in the yz plane of the Body coordinate system (i = 0 for the left leg, i = 1 for the right leg). Let yz r i Let represent the position vector projection of the i-th leg in the yz plane of the Body coordinate system (i = 0 for the left leg, i = 1 for the right leg). Then the corresponding ground reaction moment can be expressed as:

[0148] τ ig = yz r i × yz f i

[0149] Finally, we can obtain the torque balance equation when the leg generates interaction forces with the ground:

[0150]

[0151] The above formula can be simplified to in These are terms related to steering control and can be considered constants. It is the balancing torque of the leg-ground interaction.

[0152] Assuming the robot does not consider the power provided by the wheels and rolling friction, f ahead The force acting on the robot (here referring to the interaction force between the robot's legs and the ground) is represented by the force balance equation in its direction of movement:

[0153]

[0154] Let the state variables of the system be because

[0155] Substituting these equations into the torque balance equation and the force balance equation, we obtain the robot's state-space equation as follows:

[0156]

[0157] The final dynamic equation of the system is as follows:

[0158]

[0159] The parameters for this example are shown in the table below.

[0160]

[0161]

[0162] For flat terrain, to enable the robot to move quickly and maneuverably, an internal and external dynamic coordination (EIC) balance control strategy is adopted, such as... Figure 4 As shown. The EIC control method is as follows:

[0163] Let the coordinates of the rear wheel contact point C2 be... Then there is

[0164]

[0165] Continue to Taking the second derivative, and assuming the robot is moving forward at a constant speed, we can finally obtain...

[0166]

[0167] Given the desired trajectory Design a linear controller to track this trajectory for external dynamics.

[0168]

[0169] b0 = 3, b1 = 6, b2 = 10

[0170] in Positional deviation

[0171] x (3) =x u (3) ,y (3) =y u (3) Substituting into the above formula, we can obtain the external dynamic control input.

[0172]

[0173] To account for internal dynamic equilibrium, the above formula cannot be directly used as the system's control input. For the robot to maintain balance, the tilt angle acceleration must be zero, i.e.

[0174]

[0175] Will Substituting into the above equation and using Newton's method, we can solve for... Again The total derivative can be obtained by taking the steps in sequence. For the obtained These represent the desired tilt angle, tilt angular velocity, and tilt angular acceleration that ensure the system's trajectory tracking.

[0176] To enable the system to track a trajectory at the desired tilt angle, achieving both robot balance and trajectory tracking, partial feedback linearization of the entire system is first required before a linear controller can be used for tracking. and its derivative.

[0177] Therefore, an internal dynamic feedback control law is designed.

[0178]

[0179] Substituting into the previously derived dynamic formula achievable

[0180]

[0181] From this point on, the tilt angle acceleration of the system can be configured arbitrarily for tracking. and its derivative, let

[0182]

[0183] As long as a0, a1 > 0, the tracking error can be guaranteed to converge to zero.

[0184] Ultimately, the EIC motion controller for driving on flat ground is...

[0185] In this example, b0 = 3, b1 = 6, b2 = 10, a0 = 180, and a1 = 25.

[0186] For rough terrain, a wheel-leg coordinated control strategy that integrates model prediction and internal and external dynamic coordination is adopted, such as... Figure 5 As shown. The control method is as follows:

[0187] The wheel-leg coordinated control strategy integrates leg control and EIC control. It generates wheel-leg coordinated control signals based on the user-given desired trajectory and sensor information fed back by the robot, which drive the leg motors and steering and wheel hub motors respectively.

[0188] The auxiliary leg control consists of a support phase and a swing phase, both of which are scheduled by a gait generator. The support phase uses the MPC algorithm to generate the required joint motor torque as torque feedforward, and feedback control is added for force-position mixing control; the swing phase uses position + velocity control.

[0189] During the support phase, the leg ends touch the ground, generating interaction forces between the legs and the ground. The function of MPC (Multi-Purpose Calculation) is to calculate the system input that minimizes the error in the state trajectory based on the desired state trajectory and current sensing information, while also minimizing the system input to conserve energy. The MPC can be written as follows:

[0190]

[0191]

[0192]

[0193] in It is τ∈[t,t+Hδ t The control sequence, δ t It is the period of model predictive control (within a δ) t In this case, the differential of the system state can be approximated. (No change), Hδ t Represents the prediction time domain, where The error of the state vector e = xx d x d Let Q and W represent the desired velocity and desired tilt angle, respectively, with Q and W being diagonal weight matrices. It can be seen that... This represents the control input. Matrix C is...

[0194]

[0195] Where μ represents the coefficient of static friction between the leg and the ground. Let $\mathbf{j}$ represent the upper and lower limits of friction force and motor torque, respectively, and let $j$ represent the $j$-th element of the vector. The optimization problem is then solved by discretization and using the $qpOASES solver.

[0196] The interaction force between the legs and the ground acts on the robot as a series of balanced torques. When the i-th leg touches the ground, it generates the desired ground reaction force f. i The joint torque can be calculated using the following formula.

[0197]

[0198] in This represents the rotation matrix from the Body coordinate system to the World coordinate system. From this, the feedforward torque of the leg joint can be calculated. Simultaneously, the supporting leg should follow a trajectory to ensure that the leg remains in contact with the ground at the same point. Let the desired joint angle, joint angular velocity, and joint angular acceleration at a certain moment be q, respectively. d , Then the feedback control quantity u fb for

[0199]

[0200] Therefore, the final joint motor torque is

[0201]

[0202] Next, we will analyze the control of the swing phase. The key to the swing phase is determining the landing point. A commonly used heuristic formula can be used to calculate the desired leg end position, thus achieving speed regulation.

[0203]

[0204] Where T s It is the leg support phase time in a gait cycle, p hip It is the position of the hip lateral joint motor in the xy plane, v G ref =[v x v y ] T v is the expected velocity of the robot's center of mass in the xy plane. G It corresponds to the velocity of the center of mass, k. p It's gain.

[0205] In this example, W = diag(1e-5,1e-5,1e-5,1e-5,1e-5,1e-5,1e-5), Q = diag(200,1,1,1), μ = 0.45, H = 4, δT = 0.05s, T s =0.3s.

[0206] During robot movement, EIC control enables robot balance and trajectory tracking, while the periodic stepping of the auxiliary legs enhances the physical interaction between the robot and the ground, greatly improving the robot's stability under limited throttle torque.

[0207] For obstacle-crossing scenarios, a leg-force pulse-based control strategy is adopted to enhance the robot's obstacle-crossing and anti-interference capabilities. For example... Figure 6 As shown, the leg force pulse control method is as follows:

[0208] As the robot traverses obstacles, a collision detection program is triggered. When the system becomes unstable, leg force pulses are activated. The purpose of this control algorithm is to apply force pulses to the system, causing it to return to its previous equilibrium position. The design process of these force pulses will be analyzed below.

[0209] If we denote the leg-ground interaction force as F, then the torque applied to the robot is δτ = r i ×F. Let the torque vector of the left (right) leg joint be... but in This is the rotation matrix transforming the Body coordinate system to the World coordinate system. Therefore, the applied torque can be obtained as follows:

[0210]

[0211] The pulse balancing torque that can act on the robot is the component along the x-axis, i.e., δτ. x =δτ·e x ,e x =

[100] T For the purpose of balance, this invention focuses primarily on δτ. x The control design, and the y-axis (δτ) y ) and z-axis (δτ) z The torque component in the ) direction should be as small as possible.

[0212] Assuming the pulse torque δτ is at t τ Time passes through an extremely short period of time κ, reaching Under the action of the torque pulse, the robot's tilt angular velocity will undergo discontinuous jumps, while the tilt angle remains constant. According to the dynamic formula derived in claim 6, we obtain...

[0213]

[0214] Note that due to the tilt angle at The period remains unchanged, therefore Therefore, according to the above formula, we can obtain...

[0215]

[0216] in This represents the change in tilt angular velocity. and These represent the tilt angles before and after the torque pulse.

[0217] The solution can be transformed into an optimization problem, namely, finding a suitable desired tilt angular velocity. The goal is to minimize the system's balance and trajectory tracking errors, as well as the system control input. Therefore, the optimization problem can be described as follows:

[0218]

[0219]

[0220]

[0221] in It is an error vector, a matrix and Both are symmetric positive definite matrices, H t >0 represents the prediction time period. The objective function is expressed as follows: H t This period involves the accumulation of robot state trajectory errors and control inputs. This optimization problem aims to find the optimal initial tilt angular velocity so that the robot can rebalance and track its trajectory. This invention solves the problem using a sequential quadratic programming (SQP) algorithm.

[0222] The value of κ should be as small as possible to ensure that the duration of leg-ground contact is short enough. During this time, assume δτ x It remains unchanged. The SQP algorithm can be used to solve for... Then we can further obtain

[0223]

[0224] To generate δτ x The leg-ground interaction force F needs to be designed. When the leg contacts the ground, the contact point is located at... So Where F z =F·e z e z =[0 0 1] T At the same time, in order to eliminate δτ y and δτ z The resulting impact is that this invention increases F x =F y =0, Constraints such as wheel collisions and obstacle collisions typically prevent the robot from rotating along the y-axis and z-axis, therefore δτ y and δτ z The impact on the system is minimal and can be ignored. Therefore, F can be obtained. z

[0225]

[0226] in It is the maximum vertical ground reaction force, and the position of the leg-ground contact point. Adjust dynamically during the experiment.

[0227] The required joint torque can be obtained in the end.

[0228]

[0229] in i = L, R, j = 0, 1, 2.

[0230] In this example, H t =200ms, P=diag(1,1,1,1,10,10), Q=diag(10,10).

[0231] This invention combines the advantages of wheeled robots and legged robots, and designs different controllers for different terrains, enabling the mobile robot to move quickly and smoothly on flat ground and dynamically overcome obstacles at the same time, thereby improving the mobility of the mobile robot.

[0232] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the invention. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the invention. Therefore, all technical solutions obtained through equivalent substitution or transformation fall within the protection scope of the present invention.

Claims

1. A control method for a wheeled-legged robot, characterized in that, The wheeled-legged robot includes a wheeled part, a leg part, and a hardware circuit module; The wheeled part includes a front wheel (11), a rear wheel (10) and a steering component mounted on the frame (1); the front wheel (11) serves as the robot's power wheel and is driven by a hub motor (9); the steering component is located above the front wheel (11) and includes a throttle (5), a timing pulley (4) and a steering motor (3), which transmits torque to the throttle (5) through the timing pulley (4) to steer the robot. The leg portion includes a left leg and a right leg, and each leg is equipped with a motor module consisting of a hip lateral swing joint motor (13), a hip lateral swing joint motor (14), and a knee joint motor (15). The hardware circuit module includes a main control board, a motor control board, an information acquisition module, and a power supply module; the power supply module is used to provide working voltage for the entire circuit system and the motor; the hub motor (9) and the steering motor (3) are both connected to the motor control board, and the motor control board is connected to the main control board and communicates using USART; the information acquisition model communicates with the main control board using USART. The control method is as follows: S1: For flat terrain, in order to enable the robot to move quickly and maneuverably, an internal and external dynamic coordination balance control strategy is adopted, as follows: Let the contact point of the rear wheel (10) be... The coordinates are Then there is ； In the formula, Indicates the speed at the point of contact of the rear wheel (10). Indicates the robot's position in the World coordinate system. x Velocity in the axial direction, This represents the robot's velocity along the y-axis in the World coordinate system; where the World coordinate system is fixed on the ground, the Z-axis points vertically upwards, and the X-axis represents the forward direction. This represents the linear velocity of the outer edge of the rear wheel (10). The cosine function representing the robot's yaw angle. The sine function representing the robot's yaw angle; Continue to Taking the second derivative, and assuming the robot is moving forward at a constant speed, we can finally obtain... ； In the formula, Indicates intermediate control quantity. Indicates the robot's yaw angle; Given the desired trajectory To design a linear controller for external dynamics to track this trajectory ； In the formula, Indicates the robot's expectations x coordinate, Indicates the robot's expectations y Coordinates, desired robot position Positional deviation , The variable representing the feedback gain of the controller; Will Substituting into the above formula, we can obtain the external dynamic control input. ； In the formula, This indicates the linear velocity of the outer edge of the rear wheel (10); To consider internal dynamic balance, the above formula cannot be directly used as the system's control input; for the robot to maintain balance, the tilt angle acceleration must be zero, i.e. ； In the formula, Indicates yaw angle Relevant quantities; Will Substituting into the above equation and using Newton's method, we can solve for... And then The total derivative can be obtained by taking the steps in sequence. , For the obtained , representing the desired tilt angle, tilt angular velocity, and tilt angular acceleration that can guarantee the system's trajectory tracking; To enable the system to track a trajectory at the desired tilt angle, achieving both robot balance and trajectory tracking, partial feedback linearization of the entire system is first required before a linear controller can be used for tracking. and its derivative; Therefore, an internal dynamic feedback control law is designed. ； Substituting into the dynamic formula , can be obtained ； From this point on, the tilt angle acceleration of the system can be configured arbitrarily; In the formula, , , ;in, This indicates the robot's desired tilt angle acceleration. This represents the robot's tilt angle acceleration. Indicates the quality of the robot. Indicates from the robot's center of gravity to the contact point of the rear wheel (10) The vertical distance; let the contact points of the robot's front wheel (11) and rear wheel (10) be denoted as . and , Indicates the robot circles Moment of inertia; The cosine function representing the robot's tilt angle. The sine function representing the robot's tilt angle. Indicates from the robot's center of gravity to the contact point of the rear wheel (10) Horizontal distance, Indicates the wake of the front wheel (11), This represents the projection angle of the throttle angle onto the ground. The cosine function representing the tilt angle of the robot's front fork. express and The distance; In order to track and its derivative, let ； In the formula, Indicates trajectory tracking parameters; if only This ensures that the tracking error converges to zero; Ultimately, the internal and external dynamic coordination motion controller for driving on flat ground is... ; S2: For rough terrain, a wheel-leg cooperative control strategy that integrates model prediction and internal and external dynamic coordination is adopted, as follows: The wheel-leg coordinated control strategy integrates leg control and internal and external dynamic coordination control. It generates wheel-leg coordinated control signals based on the user-given desired state trajectory and the sensor information fed back by the robot, which drive the hip side swing joint motor (13), hip front swing joint motor (14), knee joint motor (15), steering motor (3) and wheel hub motor (9), respectively. The auxiliary leg control consists of a support phase and a swing phase, with the two states scheduled by a gait generator. The support phase uses the MPC algorithm to generate the required joint motor torque as torque feedforward, and at the same time, feedback control is added to perform force-position mixing control. The swing phase uses position and velocity control. During the support phase, the leg ends touch the ground, generating an interaction force between the legs and the ground. The function of MPC is to calculate the system input that minimizes the error of the state trajectory based on the desired state trajectory and the current sensing information, while also minimizing the system input to save energy. The MPC form is as follows: ； ； ； in yes The control sequence, It is the cycle of model predictive control. Represents the prediction time domain, where Error of the state vector , Represents the system's state variables. Represents the expected state variables of the system. and It is a diagonal weight matrix; Represents the coefficients of the state variables in the state-space equations. The coefficients representing the system inputs in the state-space equations are denoted as . Represents a constant term. This indicates that the robot's center of gravity is projected onto the first... i Displacement at the end of a leg; Indicates control input, Indicates the first i The ground reaction force at the end of each leg, of which, i =0 represents the left leg. i =1 represents the right leg; These represent the upper and lower limits of friction force and motor torque, respectively, with subscripts... The vector represents the first One element; matrix for ； in This represents the static friction coefficient between the leg and the ground; then, the optimization problem above is solved by discretization and using the qpOASES solver. When the i When one leg touches the ground, The joint torque can be calculated using the following formula. ； in, This represents the rotation matrix from the Body coordinate system to the World coordinate system; The transpose of the Jacobian matrix is ​​given; thus, the feedforward torque of the leg joint can be calculated. Simultaneously, the supporting leg should follow a trajectory to ensure that the leg remains in contact with the same point on the ground at all times; let the desired joint angle, joint angular velocity, and joint angular acceleration at a certain moment be respectively... Then the feedback control quantity for ； In the formula, Represents differential gain. Indicates proportional gain. Indicates joint angular velocity, Indicates joint angle; Therefore, the final joint motor torque ； The key to the swing phase is determining the landing point. Heuristic formulas can be used to calculate the desired leg end position, thus achieving speed regulation. ； In the formula, Indicates the desired leg end at xy Position on the plane It is the gait cycle. It is the position of the hip lateral swing joint motor (13) in the xy plane. It is the expected velocity of the robot's center of gravity in the xy plane. That is the corresponding center of gravity velocity. It is gain; During the robot's movement, internal and external dynamic coordination control enables the robot to maintain balance and track its trajectory. The periodic stepping of the auxiliary legs enhances the physical interaction between the robot and the ground, greatly improving the robot's stability under the condition of limited throttle torque. S3: For obstacle-crossing scenarios, a leg force pulse-based control strategy is adopted to enhance the robot's obstacle-crossing and anti-interference capabilities, as detailed below: When the robot traverses obstacles, it triggers a collision detection program. When the system becomes unstable, it initiates leg force pulses. Therefore, by applying force pulses to the system, it returns to its previous equilibrium position, as detailed below: Express the leg-ground interaction force as The torque applied to the robot is Let the torque vector of a single leg joint be... ,but ,in It is the rotation matrix that transforms the Body coordinate system to the World coordinate system, therefore the applied torque can be obtained as follows: ； The pulse balancing torque that can act on the robot is the component along the x-axis, i.e. For the purpose of balance, the main focus is on The control design, and the torque components in the y-axis and z-axis directions and The smaller the better; Assuming pulse torque exist A very short time passed ,achieve Therefore, under the action of the torque pulse, the robot's tilt angular velocity will undergo discontinuous jumps, while the tilt angle remains unchanged; according to the dynamic formula... ,have to ； Note that due to the tilt angle at The period remains unchanged, therefore According to the above formula, we can obtain ； in This represents the change in tilt angular velocity. and These represent the tilt angles before and after the torque pulse is applied; The solution can be transformed into an optimization problem, namely, finding a suitable desired tilt angular velocity. This allows the system's balance and trajectory tracking errors, as well as the system control input, to be minimized; therefore, the optimization problem can be described as... ； ； ； in It is an error vector, a matrix and All are symmetric positive definite matrices. It is the prediction time period; the objective function is expressed as... The accumulation of robot state trajectory error and control input during this period; this optimization problem is used to find the optimal initial tilt angular velocity so that the robot can rebalance and track the trajectory. The value should be as small as possible to ensure that the duration of leg-ground contact is short enough; during this time, assuming The result remains unchanged; it can be solved using the SQP algorithm. Then we can further obtain ； In order to generate It is necessary to design the leg-ground interaction force. When the leg touches the ground, the point of contact is located at ,So ,in , At the same time, in order to eliminate and The resulting impact increased Constraints; In practice, collisions between wheels and obstacles typically prevent the robot from rotating along the y-axis and z-axis, therefore and The impact on the system is minimal and can be ignored; therefore, we can obtain... : ； in, It is the maximum vertical ground reaction force, and the position of the leg-ground contact point. Adjust dynamically during the experiment; The required joint torque can be obtained in the end. ； in , , , .

2. The control method according to claim 1, characterized in that, The left and right legs each include a thigh and a calf; the upper end of the thigh is rotatably connected to a leg base (12), and the leg base (12) is fixedly connected to the frame (1) via a bracket (8); the upper end of the calf is rotatably connected to the thigh.

3. The control method according to claim 2, characterized in that, The hip lateral swing joint motor (13) and hip lateral swing joint motor (14) are used to control the rotation mode at the connection between the thigh and the leg base (12), and the knee joint motor (15) is used to control the rotation mode at the connection between the thigh and the lower leg.

4. The control method according to claim 1, characterized in that, The power module includes a voltage regulator module and several model aircraft lithium batteries; the model aircraft lithium batteries can directly output 42V and 12V voltages or output 5V and 19V voltages through the voltage regulator module; wherein, the 42V voltage is used to power the hub motor (9) and the steering motor (3), the 12V voltage is used to power the motor module, the 5V voltage is used to power the motor control board, and the 19V voltage is used to power the main control board.

5. The control method according to claim 1, characterized in that, The information acquisition module includes an inertial measurement unit (2), a meter wheel (6), and a motor encoder; the inertial measurement unit (2) is used to collect the robot's posture and speed information and is fixed in the middle of the frame (1); the meter wheel (6) is used to obtain the robot's position and is installed at the rear of the frame (1); the motor encoder is integrated in the motor module and is used to obtain the rotation angle and angular velocity of each motor.

6. The control method according to claim 1, characterized in that, The main control board uses a Jetson TX2; the motor control board uses an STM32F429 development board.

7. The control method according to claim 1, characterized in that, The power module, main control board and motor control board are all placed on the rear seat (7) of the frame (1).

Images