Wheel-leg self-adaptive force-position coupling control method, system, medium and equipment

Abstract

The invention discloses a wheel leg self-adaptive force-position coupling control method and system, a medium and equipment, and relates to the field of robot control, and the method comprises the steps: carrying out the real-time estimation of interaction force and torque generated during the movement of a mechanical arm of a double-wheel leg-single arm robot, and determining the disturbance compensation amount generated by the movement of the mechanical arm; the method comprises the following steps: determining an actual state of a trunk attitude on the basis of a trunk dynamic model in combination with a disturbance compensation amount, introducing model predictive control MPC, determining an expected state of the trunk attitude, determining a cost function between the expected state and the actual state of the trunk attitude according to the expected state and the actual state of the trunk attitude, and determining an optimal wheel-leg active force distribution result according to the cost function; and according to a wheel-leg dynamical model used for describing the relation among the wheel-leg joint angle, speed and driving torque, the optimal wheel-leg active force distribution result is converted into leg-wheel joint torque, so that whole-body dynamic adjustment of the double-wheel-leg-single-arm robot is carried out.

Description

Technical Field

[0001] This invention relates to the field of robot control technology, and in particular to a wheel-leg adaptive force-position coupling control method, system, medium, and device. Background Technology

[0002] The dual-wheeled legged-single-arm robot combines the rapid movement of a wheeled robot on flat ground with the high adaptability of a legged robot to complex terrain, while being equipped with a robotic arm to perform a variety of tasks.

[0003] However, the movements and operations of robotic arms generate interactive forces and torques, significantly impacting the robot's dynamic stability and posing a challenge to balance control when performing tasks. Ensuring stability during robot operations has become a crucial issue in the field of robotics.

[0004] Existing bi-wheeled legged robots employ a decoupled control strategy for motion control. This involves simplifying all links except the wheels into a virtual center of mass, constructing a bi-wheeled inverted pendulum model, and achieving platform balance by controlling the wheel joint motors. Other joints are used to coordinate wheel balance or adjust the posture of the torso and legs. However, this method fails to integrate the robotic arm into a unified control system. The interactive forces and torques generated by the robotic arm during movement are not adequately modeled and compensated. Especially during dynamic tasks, inertial disturbances significantly affect system stability and can even lead to equilibrium divergence. The lack of unified planning and coordinated control of whole-body motion limits the robot's mobility and task execution efficiency in complex environments. Therefore, achieving stability in bi-wheeled legged / single-arm robots during parallel execution of dynamic movements and tasks is a pressing issue. Summary of the Invention

[0005] This invention provides a wheel-leg adaptive force-position coupling control method, system, medium, and device to solve the aforementioned problems in the prior art, namely, how to improve the stability of a dual-wheel-legged single-arm robot during motion and operation. This invention provides a wheel-leg adaptive force-position coupling control method, which includes: Real-time estimation of the interaction forces and torques generated by the robotic arm of the two-wheeled legged single-arm robot during movement, and determination of the disturbance compensation amount generated by the movement of the robotic arm; Based on a torso dynamics model describing the relationship between torso position, posture, and wheel-leg active forces, and incorporating perturbation compensation, the actual state of torso posture is determined. Then, by introducing Model Predictive Control (MPC) into the torso dynamics model, the desired state of torso posture is determined. Based on the desired and actual states of torso posture, a cost function is determined between them. Finally, based on the cost function, the optimal allocation of wheel-leg active forces is determined; wherein, the allocation result is the first term of the leg and wheel joint torque control input sequence. Based on the wheel-leg dynamics model that describes the relationship between wheel-leg joint angles, speeds, and driving torques, the optimal distribution of wheel-leg active forces is converted into wheel-leg joint torques for whole-body dynamic adjustment of the dual-wheel-leg-single-arm robot.

[0006] Optionally, the disturbance compensation amount includes a force disturbance term and a torque disturbance term, and the acquisition of the force disturbance term and the torque disturbance term specifically includes: The force disturbance term is obtained using the following formula: ; in, and These represent the ends of the robotic arm along... The acceleration components along the axis and in the direction of gravity, where g is the standard gravitational acceleration and m a For the mass of the robotic arm; The torque disturbance term is obtained using the following formula: ; in, relative position of the center of mass antisymmetric matrix, It is the acceleration vector along the torso line. This represents the set matrix of accelerations of each link in the robotic arm. and These are the torso angular velocities. With angular displacement The opposition was expressed.

[0007] Optionally, the step of determining a cost function between the desired state of the torso posture and the actual state of the torso posture, and determining the optimal allocation of the wheel-leg active force based on the cost function, specifically includes: Based on the desired state of the torso posture and the actual state of the torso posture, determine the corresponding sequence of actual state of the torso posture and the sequence of desired state of the torso posture. Based on the actual state sequence of the torso posture and the expected state sequence of the torso posture, the cost function is obtained using the following formula: ; in, ; ; in, Let cost function be , indicating from time arrive The expected state sequence of trunk attitude over the predicted time period is given, where Q represents the weighted term for the expected stiffness or impedance behavior; R represents the adjustment penalty for ground forces. and These are the maximum and minimum torque values ​​for the leg and wheel joints, respectively. X ( k () represents the actual state sequence of the torso posture during the predicted time interval from time k+1 to k+n. U ( k ) represents the input sequence for leg and wheel joint torque control from time k to k+n-1; Input sequence of leg and wheel joint torques from time k to k+n-1 U ( k The first item in ) u 0 This represents the optimal distribution of the driving force of the wheel legs.

[0008] Optionally, the conversion of the optimal distribution of the active force of the wheel and leg into the torque of the leg-wheel joint specifically includes: Based on the optimal distribution of the active force of the wheel and leg, the wheel-leg joint torque is obtained using the following formula: ; in, For the torque of the wheel-leg joint, Represents the mass matrix, The vector representing the Coriolis force and centrifugal force terms. Represents the gravity vector. For dynamic base inertial force compensation, It is the acceleration of the left and right wheels. u 0 This represents the optimal distribution of the active power of the wheel legs. To access the Jacobian matrix.

[0009] Optionally, the construction of the torso dynamics model specifically includes: Define the generalized coordinates of the torso and its derivative Describe its motion state: in It is the position of the torso. It is the posture angle of the torso. It is the linear velocity of the torso. It is the angular velocity of the torso; Generalized coordinates based on the torso and its derivative The torso dynamics model is obtained using the following formula: in, , , , For the torso The displacement along the x-axis is the forward distance of the two-wheeled legged single-arm robot. For the torso The displacement along the z-axis is the standing height of the two-wheeled legged single-arm robot; , , The torso is respectively in The roll angle, pitch angle, and yaw angle in the equation. , , The torso is respectively in The roll rate, pitch rate, and yaw rate in the equation. For the torso The velocity matrices along the x-axis and z-axis.

[0010] Optionally, the desired state of the torso attitude specifically includes: the desired position and desired velocity of the torso in the X and Z directions, the desired pitch angle, desired roll angle, desired yaw angle, desired pitch rate, desired roll rate and desired yaw rate of the torso.

[0011] This invention provides a wheel-leg adaptive force-position coupling control system, comprising: The robotic arm compensation layer is used to estimate the interactive forces and torques generated by the robotic arm of the two-wheeled legged single-arm robot in real time during movement, and to determine the amount of disturbance compensation generated by the movement of the robotic arm. The torso posture control layer, based on a torso dynamics model describing the relationship between torso position, posture, and wheel-leg active forces, and incorporating perturbation compensation, determines the actual state of the torso posture. It then determines the desired state of the torso posture by introducing Model Predictive Control (MPC) into the torso dynamics model. Based on the desired and actual states of the torso posture, a cost function is determined between them. Finally, based on the cost function, the optimal allocation of wheel-leg active forces is determined; wherein the allocation result is the first term of the leg and wheel joint torque control input sequence. The wheel-leg posture control layer is used to convert the optimal distribution of the active force of the wheel-leg into the wheel-wheel joint torque based on the wheel-leg dynamics model that describes the relationship between the wheel-leg joint angle, speed and driving torque, so as to perform whole-body dynamic adjustment of the dual-wheel-leg-single-arm robot.

[0012] The present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described wheel-leg adaptive force-position coupling control method.

[0013] The present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the above-described wheel-leg adaptive force-position coupling control method.

[0014] Compared to existing technologies, the beneficial effects of this invention are as follows: This invention provides a wheel-leg adaptive force-position coupling control method. This method effectively reduces the interference of the robotic arm's motion on the torso posture by using interactive forces and torques as feedforward compensation, thereby improving the dynamic stability of the robot when performing tasks. Simultaneously, by introducing Model Predictive Control (MPC) into the torso dynamics model that describes the relationship between torso position, posture, and wheel-leg active forces, a dynamic balance between torso posture and operational force requirements can be achieved. This maintains the overall stability of the robot in complex environments. Combined with robotic arm disturbance compensation, the optimal wheel-leg active force allocation that satisfies torso posture expectations and generalized force compensation can be obtained, improving energy utilization efficiency. Furthermore, by constructing a wheel-leg dynamics model to describe the relationship between wheel-leg joint angles, velocities, and driving torques, and converting the optimal active force allocation result into leg and wheel joint torques through wheel-leg dynamics, dynamic adjustment and stable control of the entire body of the dual-wheel-leg-single-arm robot can be achieved. This method realizes the synergistic optimization of robotic arm disturbance compensation and wheel-leg power allocation, improving the robot's stability and environmental adaptability during dynamic task execution. Attached Figure Description

[0015] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0016] Figure 1 A flowchart of a wheel-leg adaptive force-position coupling control method provided in an embodiment of the present invention; Figure 2 A three-dimensional model diagram of a two-wheeled legged single-arm robot provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the force transmission relationship of each system in the dual-wheeled legged single-arm robot provided in an embodiment of the present invention; Figure 4This is a schematic diagram of the coordinate system setup for a two-wheeled legged single-arm robot provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of the wheel-leg coordinate system setting and joint angle definition provided in an embodiment of the present invention; Figure 6 This is a schematic diagram of the robot control framework provided in an embodiment of the present invention; Figure 7 This is a schematic diagram of a robotic arm grasping experiment provided in an embodiment of the present invention; Figure 8 This is a schematic diagram of a climbing experiment provided in an embodiment of the present invention; Figure 9 This is a schematic diagram of an experiment involving a continuous staircase provided in an embodiment of the present invention; Figure 10 This is a schematic diagram of a single-sided bridge experiment provided in an embodiment of the present invention; Figure 11 This is a schematic diagram of an experiment on non-flat terrain provided in an embodiment of the present invention; Figure 12 This is a schematic diagram of a robotic arm placement experiment provided in an embodiment of the present invention; Figure 13 This is a schematic diagram of a computer device for the wheel-leg adaptive force-position coupling control method provided in an embodiment of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0018] The technical solution of the present invention and how the technical solution of the present invention solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of the present invention will now be described with reference to the accompanying drawings.

[0019] Figure 1 This is a flowchart of a wheel-leg adaptive force-position coupling control method provided in an embodiment of the present invention, as shown below. Figure 1 As shown in the figure, this embodiment illustrates a wheel-leg adaptive force-position coupling control method, including: S1: Real-time estimation of the interaction forces and torques generated by the robotic arm of the two-wheeled legged single-arm robot during movement, to determine the amount of disturbance compensation generated by the movement of the robotic arm.

[0020] For example, this invention is developed for the experimental platform—TITA-PiPER dual-wheeled legged single-arm robot, whose three-dimensional model is as follows: Figure 2 As shown, the platform integrates the TITA two-wheeled legged robot developed by DirectDrive Tech and the PiPER 6-DOF lightweight robotic arm developed by Agi leX Robotics. Figure 2 In the design, motor J1 is the rotary joint motor for the robot arm base, motor J2 is the shoulder pitch joint motor for the robot arm, motor J3 is the elbow pitch joint motor for the robot arm, motor J4 is the wrist roll joint motor for the robot arm, motor J5 is the wrist pitch joint motor for the robot arm, and motor J6 is the wrist rotation joint motor for the robot arm. The TITA robot's single-wheeled leg structure has four degrees of freedom, provided by the hip external swing joint, hip pitch joint, knee pitch joint, and wheel rotation joint. All joint motors are driven by an integrated collimation drive module, achieving a peak torque of 120 Nm per joint and possessing high dynamic response capabilities. The PiPER robot arm has six degrees of freedom, independently driven by six integrated joint motors, with a total weight of 4.2 kg and a rated load capacity of 1.5 kg. Key structural parameters of the dual-wheeled leg-single-arm robot platform are shown in Table 1.

[0021] Table 1 Robot Part Parameters Generally, in bipedal robot research, maintaining dynamic stability is the core objective. Simultaneously, the torso's pose constitutes the task space, requiring real-time tracking of the desired pose with a certain level of precision. With a robotic arm attached to the torso, the arm's movement and end-effector load cause real-time changes in the robot's center of mass. Furthermore, the interaction forces between the robotic arm and torso... ,like Figure 3 As shown, the force coupling effect acts on the wheel-leg system, weakening its trajectory tracking accuracy and impairing the dynamic stability of the overall system. This represents the interaction force between the robotic arm and the torso, i.e., the robotic arm-torso interaction force. This represents the optimal force that the left wheel leg provides to the torso. This represents the optimal force that the right wheel leg provides to the torso.

[0022] In addition, the dual-wheeled legged single-arm robot needs to maintain precise hand gripping posture control during operation, and the task places demands on the interaction force between the robotic arm and the torso, as well as the interaction force between the torso and the wheeled legs. and The application of force must be able to simultaneously meet the trunk movement task and the force requirements of the robotic arm. To achieve force-position coordinated control, it is necessary to introduce the coupling relationship between the torso position and the interaction force between the wheels and legs into the torso dynamics model.

[0023] Based on this, the torso subsystem, wheel-leg subsystem, and robotic arm subsystem ( Figure 3 The dynamic interaction shown is described using generalized forces. Within this framework, in addition to its own weight, the torso is also subjected to active generalized forces from the left and right wheel-torsock joints. and the generalized force of disturbance generated by the movement of the robotic arm From a control perspective, and For controllable input, and These are external disturbances that need to be suppressed.

[0024] For example, for compensation The impact of this needs to be analyzed on the robotic arm. Definition ,in For three-dimensional external perturbation force, This represents the three-dimensional external disturbance moment. Due to the joint motion of the robotic arm and changes in the end-effector load, the overall center of mass shifts. The position vector of the robotic arm's center of mass in the world coordinate system is defined as follows: ;

[0025] in, ; Where i (from 1 to 6) represents the link number of the robotic arm. For the mass of each link, Let this be the position of the center of mass of each link in the world coordinate system. The position of the center of mass of the torso is defined as... Then the vector of the robotic arm's center of mass relative to the torso's center of mass is: :

[0026] ; Changes in the center of mass of the robotic arm directly affect the force it exerts on the torso (the force disturbance term caused by the movement of the robotic arm). Define the acceleration of the virtual center of mass of the robotic arm as... According to Newton's second law:

[0027] ; ; in Let be the acceleration of the center of mass of link i. For the overall mass of the robotic arm, Includes trunk acceleration angular velocity of the torso and angular acceleration The components generated by coupling.

[0028] Calculated Then, the external disturbance torque (torque perturbation term) acting on the center of mass of the torso caused by the movement of the robotic arm can be further solved. As shown below: ; This torque With force These factors together constitute the main dynamic disturbance exerted by the robotic arm on the torso. To minimize the impact of this disturbance on torso posture stability, it is necessary to... and The torso dynamics model incorporates this as a feedforward compensation term to determine the actual state of the subsequent torso posture. Combined with the dynamic analysis of the wheel-leg system, a dynamic compensation strategy is implemented to ensure the robot maintains balance in complex environments and improves task execution accuracy.

[0029] For example, receiving the desired pose of the robotic arm's end effector. By calculating the spatial position and acceleration of the robotic arm's center of mass in real time, and considering the joint status, the force disturbance term caused by the robotic arm's motion is further derived. and torque disturbance term This input is introduced as a feedforward compensation into the torso attitude control layer to improve the system's robustness and adaptability to coupled dynamic effects and operational disturbances. The calculation process is shown below:

[0030] ; in, and These represent the ends of the robotic arm along... Acceleration components along the axis and in the direction of gravity, For the mass of the robotic arm, This is the standard gravitational acceleration. Defined as:

[0031] ; in, relative position of the center of mass The antisymmetric matrix is ​​of the form: , It is the acceleration vector along the torso line. The set matrix representing the accelerations of each link in the robotic arm is in the form of: , Here is the mass ratio matrix for each link. and These are the torso angular velocities. With angular displacement The opposition was expressed.

[0032] S2: Based on the torso dynamics model describing the relationship between torso position, posture, and wheel-leg active force, and combined with disturbance compensation, the actual state of torso posture is determined. By introducing model predictive control (MPC) into the torso dynamics model, the desired state of torso posture is determined. Based on the desired state and the actual state of torso posture, the cost function between the two is determined. Based on the cost function, the optimal wheel-leg active force allocation result is determined. The allocation result is the first term of the input sequence of leg and wheel joint torque control.

[0033] For example, to apply appropriate torque to the wheel-leg joints to achieve the desired torso motion and compensate for the interaction force between the robot arm and torso, it is necessary to describe the wheel-leg dynamics model and establish the transmission relationship between the generalized interaction force of the torso and the joint torque. To accurately establish the torso and wheel-leg dynamics model and describe the robot's pose motion, a model such as... Figure 4 The coordinate system shown, where It is the world coordinate system, and it is a fixed reference system. This is a torso coordinate system, with the origin fixed at the geometric center of the torso. The axis points directly in front of the torso. The axis points directly above the torso and is used to analyze torso force balance and posture control. Using a natural coordinate system, the origin and... The origins coincide, and the coordinate axes point through... The rotation is used primarily for leg dynamics calculations. For ease of control, a follower coordinate system is established. The origin of the coordinate system is located at the midpoint of the line connecting the contact points of the two wheels. The axis points in the direction of wheel movement and is parallel to the horizontal plane. The axis points in the opposite direction to gravity, and is determined using the right-hand rule. Axis orientation. See coordinate system settings. Figure 4 .

[0034] For example, because the robot is subject to nonholonomic constraints, it cannot be actively controlled. Force in the axial direction. Therefore, the dynamic model ignores the force component in this direction and assumes that the ground provides sufficiently large friction to meet the lateral passive force requirements of forward velocity and yaw motion. Given the... shaft and The torque of the shaft is mainly caused by The force generated in the axial direction is defined as the primary force exerted by the wheel leg on the torso. (i=l, r distinguishes between left and right legs). For along middle Force in the positive direction of the axis (direction of movement). For along middle Force in the positive direction of the axis (vertically upward). The main torque acting on the robot's sagittal plane is used to control the torso pitch angle, and the point of application is the torso-hip joint hinge point.

[0035] Define the generalized coordinates of the torso and its derivative Describe its motion state: ; in It is the position of the torso. It is the posture angle of the torso. It is the linear velocity of the torso. It is the angular velocity of the torso, which can be further refined into the following formula: ; ; in , , , For the torso The displacement along the x-axis is the distance the robot travels. For the torso The displacement along the z-axis is the robot's standing height; , , The torso is respectively in The roll angle, pitch angle, and yaw angle are included.

[0036] Consider the forces applied by the wheel-leg subsystem and and the disturbance force generated by the robotic arm and torque The simplified dynamic equation of the torso can be expressed as follows: ; ; in, , , i=1 represents the left leg, i=2 represents the right leg. For trunk mass, For the disturbance force of the robotic arm, exist The components along the x-axis The component in the direction of gravity, For the weight of the torso; express The trunk inertial tensor in the middle It is a vector from the torso's center of mass to the torso wheel-leg hinge point. Force-position coupling is manifested in the position... and wheeled leg power and Input item.

[0037] because The value is relatively small, formula It can be approximated by the following formula: ; ; Meanwhile, inertial tensor I It can be obtained in the following ways: ,in, yes The inertial tensor under the given conditions, From arrive The rotation matrix.

[0038] For example, based on the torso dynamics model and combined with the disturbance term output by the robotic arm compensation layer, model predictive control (MPC) can be used to achieve desired following and feedback control.

[0039] For example, this invention can be based on a torso dynamics model, combined with the disturbance term output by the robotic arm compensation layer, and employs model predictive control (MPC) to achieve desired following and feedback control. The system considers the following generalized state variables: Indicates the position and posture of the torso. Represents linear velocity and angular velocity. The control force is generated by the active force of the left and right wheel legs, defined as... , , The driving force along the direction of the robot's movement. To support vertical forces, The driving torque is used to control the pitch angle.

[0040] In addition, the robotic arm uses the disturbance forces and disturbance torques generated on the torso during operation. and This indicates that the disturbance to the torso's state should be compensated for in real time within the control model. Simultaneously, the torso's linear and angular accelerations are driven by the combined forces of the wheel-leg forces and the robotic arm disturbances, as shown in the aforementioned formula. and The established linear force coupling model of the torso can be summarized into the following state-space equations:

[0041] ; in, , , , It is a five-dimensional identity matrix. To incorporate robotic arm disturbance modeling into the control system state variables, extended state variables are introduced. The acceleration term representing the disturbance force of the robotic arm, To map the robot arm's disturbance torque to the world coordinate system, after constructing the complete extended state vector, the system's state-space variables can be uniformly represented by the following structure:

[0042] ; in, , , It is a continuous state matrix. For continuous input matrices, In The mapping matrix after introducing compensation variables has the following structure: ; The above formula The continuous-time state equations are discretized using the zeroth-order preserve method to obtain the discrete-time system model: ; ; ; in, For discrete state matrices, For discrete input matrices, Sampling time, and Let k represent the system state and input vector at time step k, respectively. The prediction time domain length is n. Based on the above discrete equations, the prediction time is... arrive The states between them are: ; in, X ( k () represents the actual state sequence of the torso posture during the predicted time interval from time k+1 to k+n; U ( k ) represents the input sequence for leg and wheel joint torque control from time k+1 to k+n-1; formula The specific forms of each matrix are as follows: ; ; ; ; in, For predicting the state transfer matrix, To predict the input transfer matrix, , To reflect the force-potential hybrid control objective under the impedance regulation strategy and to minimize the deviation between the actual state and the desired state, while considering the magnitude of the driving input, the following cost function is constructed. :

[0043] ; ; in, Indicates from time arrive The expected state sequence of trunk attitude during the prediction period. Q represents the weighted term of the expected stiffness or impedance behavior, used to control the compliance of the response to the expected pose; while R represents the adjustment penalty for ground forces, used to control energy consumption or suppression of ground-coupled disturbances. and These represent the upper and lower limits of the control input U, which, when applied to the robot, correspond to the maximum and minimum torque values ​​of the leg and wheel joints. , The optimization problem is solved in real time using QuadProg++, ultimately predicting the input sequence. The first term in the equation is taken as the optimal distribution result of the wheel-leg active force in this control cycle (optimal wheel-leg force). The specific form is as follows:

[0044] ; in, , , These are the primary driving force of the left leg to the torso in the x-axis direction, the primary driving torque in the y-axis direction, and the primary driving force in the z-axis direction, respectively. , , Do not assign the right leg the primary force to the torso in the x-axis direction, the primary torque in the y-axis direction, or the primary force in the z-axis direction.

[0045] For example, in a bi-wheeled legged single-arm robot, the movement of the robotic arm and changes in the end-effector load cause a shift in the robot's overall center of mass, exerting perturbation forces and torques on the torso. These perturbations not only affect the torso's posture stability but also weaken the trajectory tracking accuracy of the wheel-leg system, thereby impairing the overall dynamic stability of the robot. To achieve high-precision task execution and dynamic balance control, it is necessary to model and compensate for the perturbations caused by the robotic arm. By analyzing the changes in the robotic arm's center of mass and its acceleration, the perturbation forces and torques acting on the torso can be calculated and introduced as compensation terms into the torso dynamics model. This compensation mechanism lays the foundation for subsequent force-position coordinated control and is closely integrated with the modeling of the wheel-leg system, constituting a key link in the overall robot dynamics modeling.

[0046] This invention first constructs a linear time-invariant state-space model incorporating external disturbances, based on a torso dynamics model and combined with the disturbance term output from the robotic arm compensation layer. To facilitate real-time calculation by the digital controller, this model is discretized using the zero-order hold method, forming a discrete-time state-space representation. During controller design, a cost function is constructed, including state error, control input amplitude, and control input rate of change. The penalty term for the control input rate of change is used to suppress abrupt changes and ensure system stability. Subsequently, this cost function is jointly constructed with the discretized system model as a constrained optimization problem, and a high-efficiency quadratic programming solver (QuadProg++) is used for real-time solving to obtain the optimal control input sequence in the prediction time domain. In each control cycle, only the first control input of the current cycle is selected for execution, with the remaining inputs recalculated in the next cycle, achieving rolling optimization and feedback adjustment. The entire control process includes state data acquisition, disturbance term calculation, model parameter updating, optimization problem construction and solving, and control command output. Each step works collaboratively to ensure the system has good response performance and disturbance rejection capability in dynamic environments.

[0047] S3: Based on the wheel-leg dynamics model that describes the relationship between wheel-leg joint angles, speeds, and driving torques, the optimal distribution of wheel-leg active forces is converted into wheel-leg joint torques for whole-body dynamic adjustment of the dual-wheel-leg-single-arm robot.

[0048] For example, as a dynamic support mechanism for a robot, the core characteristic of a wheel-leg system lies in its adaptive capability. It can respond to external disturbances, such as human pushing or environmental collisions, through a wheel-leg dynamic response mechanism, and actively adjust the wheel-leg posture to maintain overall stability. This adaptive capability is realized based on a wheel-leg dynamic model, thereby achieving force-position coordinated balance control. A diagram of the wheel-leg structure of a bipedal wheel-arm robot is shown below. Figure 5 As shown, counterclockwise rotation is considered positive, where the wheel's deflection angle relative to the ground coordinate system is... The angle of deflection of the lower leg bar relative to the vertical direction is The deflection angle of the lower leg bar relative to the upper leg bar coordinate system is: The angle of deflection of the thigh bar relative to the vertical direction is , For the weight of the torso, These represent the x and z components of the optimal force for the wheel leg, respectively. These are the X and Z axes in the hip joint coordinate system. These are the X and Z axes in the knee joint coordinate system. These are the X and Z axes in the ankle joint coordinate system. These are the X and Z coordinate axes in the wheel joint coordinate system. These represent the rotation angles of the hip, knee, ankle, and wheel joints, respectively. Modeling is done using the wheel as the moving base, defining the leg joint angles, velocities, and driving torques as follows:

[0049] in, Indicates the angle of each joint. This represents the angular velocity of each joint. This represents the torque at each joint. , and These represent the hip joint pitch angle, the lower leg joint pitch angle, and the wheel rotation angle relative to the world coordinate system, respectively.

[0050] The dynamic interference between the wheels and legs cannot be ignored, especially when the robot's motion state changes. Therefore, it is necessary to analyze the force / torque transmission between them. Although the two wheel-leg systems act together on the torso, they have independent dynamic characteristics. The wheel joints not only provide propulsion, but their motion also drives the entire subsystem to follow the torso's movement. Using the wheel as the moving base, we model it using Newton-Euler dynamics. The initial state of the wheel is:

[0051] in, Let be the initial angular velocity vector of the wheel in the inertial frame. Let be the angular velocity of the wheel about its own axis of rotation. Let be the angular acceleration of the wheel about its own axis of rotation. Let be the linear velocity vector of the gear train at the initial moment in the inertial frame. This refers to the joint angle between the gear train and the torso. Let g be the acceleration at the center of the gear train, and g be the acceleration due to gravity.

[0052] Starting from the initial state of the wheel, the angular velocities, angular accelerations, and linear velocities of the ankle, knee, and hip joints are derived sequentially, and the motion states of the centers of mass of each link are calculated. This is based on the initial forces acting on the wheel. With torque ,in Based on the mass of the wheel, the forces and torques on each link are further derived. Finally, the forces and torques of each joint in the leg are calculated in reverse from the expected output force of the hip joint.

[0053] As a dynamic base system, changes in the acceleration and velocity of the wheels alter the robot's overall inertial distribution, and this coupling effect propagates to the leg dynamics. This can lead to leg control errors, resulting in attitude instability or decreased tracking accuracy. Definition This is the compensation term for the inertial force caused by wheel speed / acceleration on leg dynamics.

[0054] Based on the above derivation, the wheel-leg dynamics model is obtained using the following formula: ; in, Represents the mass matrix, The vector representing the Coriolis force and centrifugal force terms. Represents the gravity vector. For dynamic base inertial force compensation, It is the acceleration of the left and right wheels. The acceleration is the center acceleration of the right wheel coordinate system. The angular acceleration of the right wheel's rotation. The acceleration is the center of the left wheel coordinate system. The rotational angular acceleration of the left wheel. It is a contact Jacobian matrix.

[0055] For example, this invention models the overall motion state of a wheel-legged robot by constructing a torso dynamics model and a wheel-leg dynamics model. In torso dynamics modeling, generalized coordinates and velocities are first defined, including position, attitude, linear velocity, and angular velocity. Under the assumption of no longitudinal slippage between the wheels and the ground, nonholonomic constraints are introduced to establish the transformation relationship between the body coordinate system and the world coordinate system. Based on this, generalized variables are redefined, and a simplified linear dynamic model is constructed, considering factors such as the robot arm mass, gravity compensation, sagittal plane moment, and torso inertia tensor. The inertia tensor expression in the world coordinate system is obtained through coordinate transformation. In wheel-leg dynamics modeling, the angles, velocities, and joint torques of each joint are defined, and the coupling interference between the wheels and legs is analyzed, especially the influence of torque transmission during dynamic changes. Using the Newton-Euler method, starting from the initial state at the wheel center, the motion and force states of each joint are derived step by step, ultimately forming the standard dynamic equations, which include the mass matrix, Coriolis force and centrifugal force terms, gravity terms, inertial force compensation terms, and the contact Jacobian matrix. The construction of the torso and wheel-leg dynamics model realizes the modeling of the wheel-leg robot's dynamics system, providing a foundation for control strategy design and motion planning.

[0056] The above dynamic model provides a theoretical basis and implementation framework for suppressing dynamic disturbances caused by the movement of the robotic arm, realizing the design of feedforward compensation strategies, and constructing an overall impedance coordination control system.

[0057] For example, this invention constructs a hierarchical control architecture for the motion and control of a two-wheeled legged single-arm robot, such as... Figure 6 As shown, this architecture employs a three-level closed-loop design comprising a robotic arm dynamic compensation layer, a torso posture control layer, and a wheel-leg posture control layer, achieving robust control of the system under dynamic disturbances. The control system's operating frequency is set to 100Hz to ensure real-time performance. In terms of the control flow, the system receives the user-inputted desired torso pose and corresponding velocity, as well as the desired pose and velocity of the robotic arm's end effector, as control commands. The robotic arm compensation layer calculates the dynamic disturbance compensation term generated by the robotic arm's motion in real time. The torso posture control layer combines the torso dynamics model and the disturbance compensation output from the robotic arm compensation layer to solve for the optimal wheel-leg active force. Finally, the wheel-leg posture control layer, based on the leg dynamics model, converts the optimal wheel-leg active force into specific leg and wheel joint torques, achieving precise system control. This hierarchical control strategy, integrating robotic arm dynamic compensation, torso posture optimization control, and wheel-leg posture adjustment, provides a foundation for the stable balance and control of a dual-wheel-leg / single-arm robot system in dynamic operating environments through the synergistic effect of the multi-level closed-loop system.

[0058] For example, as a layer in a hierarchical control architecture, the robotic arm compensation layer is designed to counteract dynamic disturbances caused by the movement of the robotic arm in real time, providing a stable base input for subsequent torso and wheel-leg control.

[0059] For example, the torso posture control layer further processes the dynamic coupling between system layers to ensure that the robot maintains overall posture coordination and balance in a dynamic environment.

[0060] For example, as the final stage of hierarchical control, the wheel-leg posture control layer is responsible for mapping the optimal wheel-leg active force obtained by the torso posture control layer into the execution torque of each leg and wheel joint, and implementing it into specific actuators to achieve the goal of dynamic adjustment of the whole body. This control layer comprehensively utilizes a force-position hybrid PD impedance control strategy to cope with systematic errors caused by external disturbances and terrain changes.

[0061] Specifically, the desired angles of the leg joints are first calculated based on the robot's inverse kinematics and the desired torso state. With desired angular velocity Combined with the optimal wheel-leg force calculated from the torso posture control layer Using the above wheel-leg dynamics model, the wheel-leg joint torque can be obtained: ; Furthermore, at the joint control layer, the following impedance-type PD control law is constructed to achieve dynamic adjustment based on the target force-position behavior: ; The wheel control section also incorporates speed and acceleration feedback terms to construct an impedance response structure: ; in, This refers to the feedforward torque of the wheel-leg joint. For the angle of the leg joint. Wheel angle, wheel speed and acceleration , For the quality matrix, For the Coriolis and centrifugal force matrix, For gravity, For wheel-leg coupling compensation term, For the final control torque of the leg joint, For the final control torque of the wheel joint, This refers to the feedforward torque of the leg joint. This is the feedforward torque of the wheel joint. For the wheel speed proportional gain, The differential gain of the wheel acceleration. and For gain parameters, To achieve the desired forward speed, This represents the actual forward speed. To achieve the desired forward acceleration, This refers to the actual forward acceleration. Among them, in the above formula , , and Obtained through sensors, and adjusted and By adjusting the gain parameters, the stiffness and damping response of the joint to disturbances can be flexibly changed, giving the system a dynamic compliance capability similar to impedance.

[0062] For example, the present invention can be based on a torso dynamics model describing the relationship between torso position and attitude and the active force of the wheel-legs, combined with the perturbation force of the robotic arm. and disturbance torque The disturbance compensation amount constituted, of which Through matrix This is converted into an acceleration disturbance along the torso line. Through the inverse matrix Transformed into torso angular acceleration perturbation, both are achieved through matrix Integration is used to determine the actual state of the torso posture. By introducing Model Predictive Control (MPC) into this model, the disturbance states can be included. and The extended model is discretized using the zero-order preservation method, forming a discrete matrix. and The prediction equation is constructed to determine the desired state of the torso posture. Based on the difference between the desired state and the actual state, a cost function J(k) is constructed to minimize the state deviation and suppress the input amplitude. Based on this cost function, the optimal wheel-leg active force distribution result is solved. This result is the first term of the leg and wheel joint torque control input sequence u0, which is used as the control quantity executed in the current cycle.

[0063] In summary, this wheel-leg posture controller, based on the force-potential impedance regulation concept, forms an adaptive response mechanism for dynamic disturbance environments through feedforward guidance and feedback correction, and achieves coordinated control of the robot's whole-body posture.

[0064] To verify the effectiveness of the wheel-leg adaptive force-position coupling control method proposed in this invention, and to further evaluate the balance and motion performance of the two-wheeled-legged-single-arm robot, a simulation environment was built on Webots simulation software, and the two-wheeled-legged-single-arm robot was modeled. Simulation experiments were conducted using a control program written in C++. The experiments included multiple terrains to examine the robot's stability and adaptability in different environments. The specific experimental process was as follows: First, while maintaining its position, the robot used its robotic arm to grasp a 1kg weight from a table, and then maintained uniform motion on a flat surface to verify its walking stability. Next, the robot climbed a 2m long, 30° slope, and descended 1m down a 0.1m high staircase to reach the flat surface. Then, the robot traversed an asymmetrical single-sided bridge to examine its attitude control under non-uniform support conditions, and further crossed uneven ground to test its stability in unstructured environments. Finally, the robot reached the target area and placed the weight it was carrying on the target table. The specific content and results of each sub-experiment will be described in detail below.

[0065] (1) Robotic arm grasping experiment For example, this experiment aims to test the robot's stability during a grasping task. In the simulation environment, the robot's initial height is 0.354m, and all joints of the robotic arm are at their zero positions. A table with a height of 0.73m is built. Because the robotic arm can handle a maximum load of 1.5kg and the range of motion between the two grippers is 0-70mm, a 1kg cube with a length of 0.05m is placed on the table. The robot initially stands at a height of 0.354m and uses its robotic arm to grasp the 1kg object, as follows: Figure 7As shown, the system simultaneously relies on the force and torque compensator of the robotic arm to compensate for the impact of the load on the robot, ensuring the overall stability of the system.

[0066] Figure 7 This is a schematic diagram of a robotic arm grasping experiment, divided into two stages: the first stage is the process from the initial position to grasping, and the second stage is the process from grasping to resetting. Figure 7 As shown in (a); Figure 7 In (b), from top to bottom, the first image shows Roll, pitch, and yaw representing the three posture angles of the torso; the second image shows X as the robot's forward distance and Z as its standing height; the third image shows... , , This refers to the torque compensation terms in three directions for the robotic arm compensation. Desire is the relevant expected value.

[0067] Depend on Figure 7 As can be seen, the robot's torso attitude angles remained stable throughout the heavy object grasping experiment. Specifically, the roll angle reached its maximum value of 0.011049 rad at 20.43 s; the pitch angle reached its maximum value of 0.0027737 rad at 20.83 s; and the yaw angle reached its maximum value of 0.012481 rad at 30.97 s. Simultaneously, the robotic arm compensator could also compensate for the effects of the robotic arm on the torso in real time, mainly focusing on compensation in the robot's sagittal plane, such as... Figure 7 As shown in (b), the pitch torque compensation curves of the robotic arm in the figure show obvious arching during the process of the robotic arm picking up the heavy object and returning to the zero position. These data indicate that during the execution of the robotic arm grasping task, the robotic arm compensator can effectively compensate for interference, enabling the robot to maintain a stable torso posture and avoid the impact of posture deviation on the overall balance of the robot.

[0068] (2) 30-degree slope experiment This experiment analyzes the robot's motion and dynamic balance capabilities in a sloping environment. During the climb, the model predictive controller calculates the optimal force in real time. To ensure the torso parameters reached the desired values, the robot utilized a torque compensation mechanism in its robotic arm to stabilize its posture, enabling it to adapt to slope changes and maintain balance. The experiment was conducted with a 30° slope and a 2m length to simulate the challenges of climbing on sloping terrain. During the climb, the robot continuously adjusted the driving force and posture of its wheels and legs to maintain stable movement and effectively cope with the effects of slope inclination.

[0069] Figure 8 This is a schematic diagram of an experiment where a robot climbs a 30° slope. Figure 8 (a) shows the robot climbing the ramp from left to right. Figure 8In the second figure from top to bottom in (b), Frx and Flx represent the optimal forces provided by the right and left legs to the torso in the x-axis direction, while Frz and Flz represent the optimal forces in the z-axis direction.

[0070] like Figure 8 As shown, during the 30° slope climbing experiment, the roll angle, pitch angle, and yaw angle showed relatively small variations, with a maximum value not exceeding 0.004 rad. Simultaneously, the left and right wheel leg forces Flx and Frx increased from [60, 80] N on flat ground to [80, 180] N, and Flz and Frz increased from [180, 200] N to [200, 250] N. The robot arm's pitch torque was also compensated in real-time during the uphill process to ensure the robot's smooth ascent. This result demonstrates that the robot can adapt to slope environments and climb smoothly.

[0071] (3) Experiment on descending a continuous staircase In this experiment, the robot's coordination and posture control capabilities during the descent of stairs were tested. The height of each individual stair was set to 0.1m, and the total height of the stairs was 1m. Figure 9 It is evident that during the robot's continuous stair descent experiment, its torso attitude angle remained stable throughout, the roll angle fluctuated around the expected value with a maximum value not exceeding 0.001 rad, the yaw angle did not exceed 0.0025 rad, and the pitch angle did not exceed 0.007 rad. Simultaneously, the wheel-leg forces... The robot also makes real-time adjustments during the descent, with Frz and Flz being particularly noticeable, exhibiting a continuous convex shape. The upper and lower edges of the convex shapes represent the airborne phase of the descent. The maximum force of 250N is reached after the wheels contact the stairs. These data indicate that the robot can maintain a stable torso posture during the descent, ensuring a smooth landing and effectively coping with the impact of continuous height changes. It can adapt to changes in steps and stably complete the descent task through active power adjustment and attitude control strategies.

[0072] (4) Single-sided bridge experiment In the single-sided bridge experiment, the robot needs to cross four single-sided bridges with a slope of 15° and a bridge length of 2.5m to test its attitude control capability and stability under uneven support conditions. Figure 10 An experiment was conducted to guide a robot down a series of stairs. Figure 10 (a) shows the robot descending a series of stairs from left to right. Figure 10 In (b), Roll, Pitch, and Yaw represent the three attitude angles of the robot's torso.

[0073] During the bridge crossing, the robot's roll, pitch, and yaw angles underwent dynamic changes, but the control frame was able to adjust the main power of the wheels in real time, ensuring that the robot maintained a stable standing height. Experimental data shows that throughout the entire bridge crossing process, the robot's roll and pitch angles remained within [0, 0.002] rad, and its yaw angle remained within [0, 0.005] rad. Although the irregularity of the terrain had some impact on the robot's posture, the control frame made adjustments, enabling it to successfully complete the bridge crossing task and stably enter the next experimental scenario.

[0074] (5) Experiments on uneven terrain In non-flat terrain experiments, the robot needs to adapt to complex terrain changes to verify its stability and mobility in unstructured environments. The terrain includes undulating height differences, irregular slopes, and loose ground, simulating the uncertainties in real-world environments. Figure 11 The experimental data shown demonstrates that the robot's torso angle remained within a reasonable range throughout its movement across uneven terrain, with a maximum error not exceeding 0.008 rad. Although the irregular terrain introduced additional disturbances to the robot, model predictive control and the robotic arm effectively compensated for these external influences, enabling the robot to maintain balance and ensure stable posture.

[0075] (6) Robotic arm placement experiment After the above experiments, the robot placed the weight held by its robotic arm onto the target table. This process consisted of two stages: the first stage involved the robotic arm moving the weight from its initial position to the target position, and the second stage involved the robotic arm's joint angles returning to zero. During the placement process, the robot maintained a forward speed of 0 m / s and remained stationary.

[0076] Figure 12 The experiment on placing a heavy object with a robotic arm was divided into two stages. The first stage involved the robotic arm placing the heavy object from its initial position to the target position. The second stage involved the robotic arm's joint angles returning to zero. Figure 12 As shown in (a).

[0077] Experimental data show that the results of the weight placement experiment and the grasping experiment are consistent. Throughout the experiment, the robot's torso's attitude angle and position error remained within a reasonable range, not exceeding 0.02 rad, ensuring system stability. Simultaneously, the robotic arm compensation system adjusted the pitch moment in the sagittal plane in real time, effectively reducing the interference force and additional torque effect of the robotic arm's movement on the torso. The experiment further verified the effectiveness of the robotic arm compensation strategy, enabling the robot to better maintain a stable posture when performing tasks. This compensation mechanism not only improves the quality of task completion but also enhances the robot's adaptability in complex environments.

[0078] The above describes one or more embodiments of the wheel-leg adaptive force-position coupling control method provided in this specification. Based on the same idea, this specification also provides a corresponding wheel-leg adaptive force-position coupling control system, including: The robotic arm compensation layer is used to estimate the interactive forces and torques generated by the robotic arm of the two-wheeled legged single-arm robot in real time during movement, and to determine the amount of disturbance compensation generated by the movement of the robotic arm. The torso posture control layer, based on a torso dynamics model describing the relationship between torso position, posture, and wheel-leg active forces, and incorporating perturbation compensation, determines the actual state of the torso posture. It then determines the desired state of the torso posture by introducing Model Predictive Control (MPC) into the torso dynamics model. Based on the desired and actual states of the torso posture, a cost function is determined between them. Finally, based on the cost function, the optimal allocation of wheel-leg active forces is determined; wherein the allocation result is the first term of the leg and wheel joint torque control input sequence. The wheel-leg posture control layer is used to convert the optimal distribution of the active force of the wheel-leg into the wheel-wheel joint torque based on the wheel-leg dynamics model that describes the relationship between the wheel-leg joint angle, speed and driving torque, so as to perform whole-body dynamic adjustment of the dual-wheel-leg-single-arm robot.

[0079] Specific limitations regarding the wheel-leg adaptive force-position coupling control system can be found in the limitations of the wheel-leg adaptive force-position coupling control method described above, and will not be repeated here. Each module in the aforementioned wheel-leg adaptive force-position coupling control system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0080] The present invention also provides a computer-readable storage medium storing a computer program that can be used to execute the wheel-leg adaptive force-position coupling control method provided above.

[0081] The present invention also provides Figure 13 The schematic diagram of the computer device shown is as follows: Figure 13 As shown, at the hardware level, the computer device includes a processor, an internal bus, a network interface, memory, and non-volatile memory, and may also include other hardware required for business operations. The processor reads the corresponding computer program from the non-volatile memory into memory and then runs it to implement the wheel-leg adaptive force-position coupling control method provided in the above embodiment.

[0082] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this invention.

Claims

1. A wheel-leg adaptive force-position coupling control method, characterized in that, include: Real-time estimation of the interaction forces and torques generated by the robotic arm of the two-wheeled legged single-arm robot during movement, and determination of the disturbance compensation amount generated by the movement of the robotic arm; Based on a torso dynamics model describing the relationship between torso position, posture, and wheel-leg active forces, and incorporating perturbation compensation, the actual state of torso posture is determined. Then, by introducing Model Predictive Control (MPC) into the torso dynamics model, the desired state of torso posture is determined. Based on the desired and actual states of torso posture, a cost function is determined between them. Finally, based on the cost function, the optimal allocation of wheel-leg active forces is determined; wherein, the allocation result is the first term of the leg and wheel joint torque control input sequence. Based on the wheel-leg dynamics model that describes the relationship between wheel-leg joint angles, speeds, and driving torques, the optimal distribution of wheel-leg active forces is converted into wheel-leg joint torques for whole-body dynamic adjustment of the dual-wheel-leg-single-arm robot.

2. The wheel-leg adaptive force-position coupling control method as described in claim 1, characterized in that, The disturbance compensation amount includes a force disturbance term and a torque disturbance term. The acquisition of the force disturbance term and the torque disturbance term specifically includes: The force disturbance term is obtained using the following formula: ； in, and These represent the ends of the robotic arm along... The acceleration components along the axis and in the direction of gravity, where g is the standard gravitational acceleration and m a For the mass of the robotic arm; The torque disturbance term is obtained using the following formula: ； in, relative position of the center of mass antisymmetric matrix, It is the acceleration vector along the torso line. This represents the set matrix of accelerations of each link in the robotic arm. and These are the torso angular velocities. With angular displacement The opposition was expressed.

3. The wheel-leg adaptive force-position coupling control method as described in claim 1, characterized in that, The process involves determining a cost function between the desired and actual states of the torso posture, and then determining the optimal allocation of the wheel-leg active force based on this cost function. Specifically, this includes: Based on the desired state of the torso posture and the actual state of the torso posture, determine the corresponding sequence of actual state of the torso posture and the sequence of desired state of the torso posture. Based on the actual state sequence of the torso posture and the expected state sequence of the torso posture, the cost function is obtained using the following formula: ； in, ； ； in, Let cost function be , indicating from time arrive The expected state sequence of trunk attitude over the predicted time period is given, where Q represents the weighted term for the expected stiffness or impedance behavior; R represents the adjustment penalty for ground forces. and These are the maximum and minimum torque values ​​for the leg and wheel joints, respectively. X ( k () represents the actual state sequence of the torso posture during the predicted time interval from time k+1 to k+n. U ( k ) represents the input sequence for leg and wheel joint torque control from time k to k+n-1; Input sequence of leg and wheel joint torques from time k to k+n-1 U ( k The first item in ) u 0 This represents the optimal distribution of the driving force of the wheel legs.

4. The wheel-leg adaptive force-position coupling control method as described in claim 3, characterized in that, The process of converting the optimal distribution of the active force of the wheel and leg into the torque of the leg-wheel joint specifically includes: Based on the optimal distribution of the active force of the wheel and leg, the wheel-leg joint torque is obtained using the following formula: ； in, For the torque of the wheel-leg joint, Represents the mass matrix, The vector representing the Coriolis force and centrifugal force terms. Represents the gravity vector. For dynamic base inertial force compensation, It is the acceleration of the left and right wheels. u 0 This represents the optimal distribution of the active power of the wheel legs. To access the Jacobian matrix.

5. The wheel-leg adaptive force-position coupling control method as described in claim 1, characterized in that, The construction of the trunk dynamics model specifically includes: Define the generalized coordinates of the torso and its derivative Describe its motion state: in It is the position of the torso. It is the posture angle of the torso. It is the linear velocity of the torso. It is the angular velocity of the torso; Generalized coordinates based on the torso and its derivative The torso dynamics model is obtained using the following formula: in, , , , For the torso The displacement along the x-axis is the forward distance of the two-wheeled legged single-arm robot. For the torso The displacement along the z-axis is the standing height of the two-wheeled legged single-arm robot; , , The torso is respectively in The roll angle, pitch angle, and yaw angle in the equation. , , The torso is respectively in The roll rate, pitch rate, and yaw rate in the equation. For the torso The velocity matrices along the x-axis and z-axis.

6. The wheel-leg adaptive force-position coupling control method as described in claim 1, characterized in that, The desired state of the torso attitude specifically includes: the desired position and desired velocity of the torso in the X and Z directions, the desired pitch angle, desired roll angle, desired yaw angle, desired pitch rate, desired roll rate and desired yaw rate of the torso.

7. A wheel-leg adaptive force-position coupling control system, characterized in that, include: The robotic arm compensation layer is used to estimate the interactive forces and torques generated by the robotic arm of the two-wheeled legged single-arm robot in real time during movement, and to determine the amount of disturbance compensation generated by the movement of the robotic arm. The torso posture control layer, based on a torso dynamics model describing the relationship between torso position, posture, and wheel-leg active forces, and incorporating perturbation compensation, determines the actual state of the torso posture. It then determines the desired state of the torso posture by introducing Model Predictive Control (MPC) into the torso dynamics model. Based on the desired and actual states of the torso posture, a cost function is determined between them. Finally, based on the cost function, the optimal allocation of wheel-leg active forces is determined; wherein the allocation result is the first term of the leg and wheel joint torque control input sequence. The wheel-leg posture control layer is used to convert the optimal distribution of the active force of the wheel-leg into the wheel-wheel joint torque based on the wheel-leg dynamics model that describes the relationship between the wheel-leg joint angle, speed and driving torque, so as to perform whole-body dynamic adjustment of the dual-wheel-leg-single-arm robot.

8. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the wheel-leg adaptive force-position coupling control method according to any one of claims 1-6.

9. A computer device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the wheel leg adaptive force-position coupling control method according to any one of claims 1-6.

Images