Output feedback robust control method, system and device for humanoid robot

Abstract

The application provides an output feedback robust control method, system and device for a humanoid robot, and belongs to the technical field of humanoid robot control. The method comprises the following steps: establishing a dynamics model of a robot limb end system, and defining state variables, control variables and output variables. Based on the state feedback optimal control principle, the dynamics model and an output feedback controller constructed by the output variables, a first robust control equation is determined; based on the mapping relationship between the state variables and the output variables in the dynamics model, the first robust control equation is converted into a second robust control equation, so as to determine a to-be-estimated parameter vector according to the second robust control equation, and the control gain matrix is obtained by online estimation based on a preset adaptive law; the second robust control equation is constructed based on the output variables; based on the control gain matrix and the output variables acquired in real time, corresponding control variables are generated to be applied to the robot limb end joint, so as to robustly control the robot limb end system.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of humanoid robot control, and in particular to an output feedback robust control method, system and device for a humanoid robot. BACKGROUND

[0002] With the development of the times and the progress of science and technology, humanoid robots have entered the prototype testing and trial application stage. With the continuous increase of research and development investment in this field, the intelligence level of humanoid robots is rising. At the same time, the core of robot limb control also needs more advanced algorithms.

[0003] The existing humanoid robot system is a complex system with multiple joints, strong coupling and high nonlinearity. The use of model-dependent control methods often requires a large number of system parameters, and the use of model-free control methods with poor interpretability has the problem of a long training process. Common control algorithms, such as PID control, adaptive control, model predictive control, etc., the PID control algorithm has low modeling requirements for the system and is widely used in industrial robot systems. However, for humanoid robots, the effectiveness of traditional control strategies represented by PID is limited, and the anti-interference ability is weak in the face of sudden external force collision, road disturbance and other scenes, and the system state recovers slowly after being disturbed.

[0004] Therefore, in view of the complex dynamic characteristics of humanoid robots, breaking through the limitations of existing methods in model dependence, anti-interference ability, control efficiency, etc. has become the key to promoting the progress of precise robust control technology. SUMMARY

[0005] The embodiments of the present application provide an output feedback robust control method, system and device for a humanoid robot, which is used to solve the technical problem of how to overcome the limitations of model dependence, anti-interference ability, control efficiency, etc. for a system with complex dynamic characteristics of humanoid robots, and to perform precise output feedback robust control.

[0006] In one aspect, the embodiments of the present application provide an output feedback robust control method for a humanoid robot, which comprises:

[0007] A dynamics model of a robot limb system is established, and state variables, control variables and output variables are defined. The dynamics model is constructed based on Lagrange's equation and at least includes an inertia matrix, a Coriolis force vector, a gravity vector and a disturbance torque. The state variables are motion state parameters of the limb joints, the control variables are control input parameters acting on the limb joints, and the output variables are measurable limb joint state parameters.

[0008] determine a first robust control equation based on a state feedback optimal control principle, the dynamic model, and an output feedback controller constructed based on the output variables; wherein the first robust control equation is added with a preset state weight matrix and a control strategy weight matrix to adjust control performance;

[0009] convert the first robust control equation into a second robust control equation based on a mapping relationship between the state variables and the output variables in the dynamic model, to determine a to-be-estimated parameter vector according to the second robust control equation, and to obtain a control gain matrix in the output feedback controller based on a preset adaptive law and online estimation;

[0010] generate a corresponding control variable based on the control gain matrix and the output variables acquired in real time, and apply the control variable to a robot limb end joint, to perform robust control on the robot limb end system.

[0011] In an implementation manner of the present application, the method is applied to output feedback robust control of a lower limb of a humanoid robot; the robot limb end system includes a plurality of lower limb joints with one or more degrees of freedom; the lower limb joints at least include a crotch joint, a hip joint, a leg joint, a knee joint, an ankle joint, and a waist joint.

[0012] establish a dynamic model of the robot limb end system, specifically including:

[0013] model kinematics equations of the lower limb of the humanoid robot based on pose vectors corresponding to the lower limb joints, to obtain kinematics equations;

[0014] establish the dynamic model based on the kinematics equations and the Lagrange equation.

[0015] In an implementation manner of the present application, the state variables are defined as:

[0016] ; wherein, , represents a position coordinate of a lower limb joint, represents a pose of the lower limb of the humanoid robot in a world coordinate system, represents a pose vector of a limb end joint, is a joint velocity vector, is a first variable in the state variables, is a second variable in the state variables;

[0017] the control variables are defined as:

[0018] ; wherein, represents a control variable, is an inertia matrix in the dynamic model, is an joint torque, is a disturbance torque, is a preset nonlinear resultant force term;

[0019] The output variable is defined as: , represents an output variable, represents an output matrix , represents a state vector of the robot limb system.

[0020] In an implementation manner of the present application, based on a state feedback optimal control principle, the dynamic model and an output feedback controller constructed by the output variable, a first robust control equation is determined, specifically comprising:

[0021] based on the state feedback optimal control principle, a state feedback controller is determined according to the dynamic model ; wherein, , is a state feedback gain matrix, represents a control strategy weight matrix, is a second coefficient matrix , is a positive definite gain matrix for solving Riccati equation;

[0022] According to the state feedback controller, a corresponding state optimal performance index function is constructed; the state optimal performance index function includes the preset state weight matrix;

[0023] The output feedback controller constructed by the output variable, the state optimal performance index function and the dynamic model are simultaneously reconstructed to obtain the first robust control equation; the output feedback controller includes the output variable and the control gain matrix.

[0024] In an implementation manner of the present application, the first robust control equation is:

[0025] ; wherein, , is a first coefficient matrix , is the control gain matrix, represents an external input torque, represents the preset state weight matrix.

[0026] In one implementation of this application, based on the mapping relationship between the state variables and the output variables in the dynamic model, the first robust control equation is transformed into a second robust control equation, specifically including:

[0027] Based on the dynamic model, the state vector of the first robust control equation is transformed to obtain: ;

[0028] Based on the mapping relationship between the state variables and the output variables The second robust governing equation is obtained: .

[0029] In one implementation of this application, determining the parameter vector to be estimated based on the second robust control equation specifically includes:

[0030] The second robust control equation is transformed by the Kronecker product to extract the estimated parameter vector based on the transformed parameter expression; the estimated parameter vector is established based on the positive definite gain matrix through vectorization operations.

[0031] The control gain matrix is ​​obtained through online estimation based on a preset adaptive law, specifically including:

[0032] Through the preset adaptive law Perform online estimation to determine the online estimation parameter vector. estimation error If the value is less than the predetermined value, determine the online estimated parameter corresponding to the parameter vector to be estimated. ;in, For online parameter estimation The time derivative; Preset learning gain; For auxiliary vectors;

[0033] Based on the determined online estimation parameters The positive definite gain matrix is ​​determined by the parameter vector to be estimated, and the control gain matrix is ​​determined based on the positive definite gain matrix and the second robust control equation.

[0034] In one implementation of this application, ,in, A predefined auxiliary regression matrix; Indicates the estimation error. ; The true parameter vector is the vector corresponding to the parameter vector to be estimated.

[0035] Secondly, embodiments of this application also provide an output feedback robust control system for a humanoid robot, the system being capable of executing the aforementioned output feedback robust control method for a humanoid robot; the system includes:

[0036] A module is established to build a dynamic model of the robot's end effector system and define state variables, control variables, and output variables. The dynamic model is based on the Lagrange equation and includes at least an inertia matrix, a Coriolis force vector, a gravity vector, and a disturbance torque. The state variables are the motion state parameters of the end effector joints, the control variables are the control input parameters acting on the end effector joints, and the output variables are the measurable state parameters of the end effector joints.

[0037] The determination module is used to determine the first robust control equation based on the state feedback optimal control principle, the dynamic model, and the output feedback controller constructed with the output variables; wherein, the first robust control equation is supplemented with a preset state weight matrix and a control strategy weight matrix to adjust the control performance;

[0038] The conversion module is used to convert the first robust control equation into a second robust control equation based on the mapping relationship between the state variables and the output variables in the dynamic model, so as to determine the parameter vector to be estimated according to the second robust control equation, and perform online estimation based on a preset adaptive law to obtain the control gain matrix; the second robust control equation is constructed based on the output variables;

[0039] The generation module is used to generate corresponding control variables based on the control gain matrix and the output variables acquired in real time, and apply them to the robot end joints to perform robust control of the robot end system.

[0040] Thirdly, embodiments of this application also provide an output feedback robust control device for a humanoid robot, the device comprising:

[0041] At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the aforementioned output feedback robust control method for a humanoid robot.

[0042] Compared with the prior art, the significant advantages of this application are as follows:

[0043] Through the above technical solutions, this application proposes a data-driven output feedback robust control scheme to achieve optimal robust control of humanoid robot systems. This application addresses the difficulty of observing the state of humanoid robots by reconstructing the robust control equations to eliminate the dependence of robot end-joint control on state variables. Secondly, it eliminates the dependence of robot control on internal states through equation transformation, enabling control solely through measurable joint state parameters. Furthermore, an adaptive law is designed to achieve real-time updates of the control gain matrix, enabling efficient and dynamically stable control even with limited data. This leads to the construction of a robust control scheme for humanoid robot end-joints integrating output feedback, robust control, and data-driven approaches. This scheme overcomes limitations in model dependence, disturbance rejection, and control efficiency, enabling precise output feedback robust control for systems with complex dynamic characteristics of humanoid robots. Attached Figure Description

[0044] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0045] Figure 1 This is a flowchart illustrating an output feedback robust control method for a humanoid robot according to an embodiment of this application.

[0046] Figure 2 This is a schematic diagram of the lower limb structure of a humanoid robot for which an output feedback robust control method for humanoid robots is applied in an embodiment of this application;

[0047] Figure 3 This is a flowchart of a robust control system for a humanoid robot according to an embodiment of this application;

[0048] Figure 4 This is a schematic diagram of the simulation curve of the stable state of the robot end effector system after using the output feedback robust control method in the embodiments of this application;

[0049] Figure 5 This is a schematic diagram of the output feedback robust control system for a humanoid robot according to an embodiment of this application;

[0050] Figure 6 This is a schematic diagram of the structure of a robust output feedback control device for a humanoid robot according to an embodiment of this application. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0052] Existing humanoid robot systems are complex systems with multiple joints, strong coupling, and high nonlinearity. Model-dependent control methods often require a large number of system parameters, while model-free control methods with poor interpretability suffer from lengthy training processes. Common control algorithms, such as PID control, adaptive control, and model predictive control, are widely used in industrial robot systems, especially PID control, which has low system modeling requirements. However, for humanoid robots, the effectiveness of traditional control strategies, represented by PID, is limited, and it has weak anti-interference capabilities in scenarios such as sudden external force collisions and road disturbances, resulting in slow system recovery after disturbances.

[0053] Based on this, embodiments of this application provide a robust output feedback control method, system, and device for humanoid robots, which solves the technical problem of how to overcome limitations in model dependence, disturbance rejection capability, and control efficiency, and achieves accurate output feedback robust control for systems with complex dynamic characteristics of humanoid robots.

[0054] The various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0055] This application provides an embodiment of a robust output feedback control method for humanoid robots, such as... Figure 1 As shown, the method may include steps S101-S104:

[0056] S101. Establish the dynamic model of the robot's end effector system and define the state variables, control variables, and output variables.

[0057] The dynamic model is constructed based on the Lagrange equations and includes at least the inertia matrix, Coriolis force vector, gravity vector, and disturbance torque. The state variables are the motion state parameters of the distal joints, the control variables are the control input parameters acting on the distal joints, and the output variables are the measurable state parameters of the distal joints.

[0058] In this embodiment, the above method is specifically applied to robust output feedback control of the lower limbs of a humanoid robot. The robot's end effector system includes multiple lower limb joints with one or more degrees of freedom. The lower limb joints include at least: hip joint, leg joint, knee joint, ankle joint, and lumbar joint. The humanoid robot's lower limbs are as follows: Figure 2As shown, the hip joint (anterior and posterior) can be understood as having two degrees of freedom, anterior and posterior, while the hip joint (valgus) refers to the degree of freedom of the hip joint to rotate outward. The robot's lower limbs in this application are configured with five joints per leg: the hip joint, hip joint, leg joint, knee joint, and ankle joint. The waist has one joint. This configuration ensures the robot's stable upright posture and walking movements.

[0059] This application establishes a dynamic model of the robot's end effector system, specifically including:

[0060] Based on the pose vectors corresponding to each lower limb joint, kinematic equations are modeled for the lower limbs of the humanoid robot, resulting in motion equations. A dynamic model is then established based on these motion equations and the Lagrange equations.

[0061] The pose vector is specifically as follows: , This can be understood as the location of the drive joints in the lower limbs. This represents the posture of the humanoid robot in the world coordinate system. Subsequently, the kinematic equations of the humanoid robot are modeled. Since the motion of the humanoid robot is also generated by the combined action of its own active force and the passive force of the environment, the following motion equations are obtained, as shown in formula (1):

[0062] (1)

[0063] in, , , The inertial matrix represents the position and orientation within the robot's end effector system. for The transpose of the matrix; and These are the acceleration vectors corresponding to the attitude and joint position, respectively; and These are the first and second biasing forces caused by centrifugal force, Coriolis force, and gravity, respectively, for posture and joint position. This corresponds to the joint drive torque of a specific joint. , These are the generalized contact forces of the external environment corresponding to the posture and joint position, respectively.

[0064] Subsequently, based on the Lagrange dynamics equations, the humanoid robot is transformed into a multi-link dynamic system. The Lagrange equations can be established from its kinetic and potential energy, as shown in formula (2):

[0065] (2)

[0066] in, Represents the system's inertia matrix. Represents the gravity matrix of the system. It is the selection matrix for robot end effector systems. It refers to the joint torque of the robot. It is a Jacobian matrix. For externally input torque, It is defined as a system parameter in a broad sense.

[0067] The general dynamic model of the humanoid robot's lower limbs can be obtained by simplifying the Lagrange dynamics equations:

[0068] (3)

[0069] in, These are the position, velocity, and acceleration vectors of the humanoid robot's limbs. The inertial matrix of a humanoid robot, The Coriolis force vector for a humanoid robot, It is the gravity vector. It is the control input torque of the robot. It is a disturbance torque.

[0070] Furthermore, for ease of calculation, the above dynamic model can also be defined as:

[0071] (4)

[0072] in, for This is a nonlinear resultant force term. In real-world scenarios, humanoid robots have a certain load capacity, and the load causes the robot's center of mass to change depending on the load. This is similar to an external force disturbance applied to the robot system; both affect the robot's center of mass, thus leading to changes in the vector matrix. and The uncertainty of the load leads to uncertainty in the model. Therefore, the following real-world conditions are used to address the problem of load variation. First, assume the existence of a positive definite matrix. and Make: Secondly, assume there exists a vector Nonnegative functions ,Right now .

[0073] Therefore, for the robot's dynamics model, state variables are defined as follows:

[0074] (5)

[0075] in, , This indicates the coordinates of the lower limb joints. This represents the pose of the humanoid robot's lower limbs in the world coordinate system. Represents the pose vector of the distal joint. For joint velocity vectors, It is the first variable in the state variables. It is the second variable in the state variables.

[0076] Define control variables as shown in formula (6):

[0077] (6)

[0078] in, Indicates control variables, The inertia matrix in the dynamic model, For joint torque, For the disturbance torque, This is a pre-defined nonlinear resultant force term.

[0079] From the state equations, we can obtain Then, using the formula of the robot system dynamics model, the state space expression is derived, as shown in formula (7):

[0080] (7)

[0081] Therefore, nonlinear variables are obtained. As shown in formula (8):

[0082] (8)

[0083] The nonlinear control model of the humanoid robot's end effector system is further obtained, as shown in formula (9):

[0084] (9)

[0085] in, The first coefficient matrix , The second coefficient matrix , Represents the output matrix , , The output variable represents the position of each lower limb joint of the humanoid robot, which is the measurable end-joint state parameter. Therefore, this application can define the output variable as: .

[0086] It should be noted that the execution subject of the above-mentioned robust control method for output feedback of humanoid robots can be a dedicated computing unit set inside the humanoid robot. For example, the dedicated computing unit can include a field programmable gate array (FPGA) chip, a graphics processing unit (GPU), a microcontroller unit (MCU), etc. The listed types are only examples, and the execution subject is not limited to these. This application does not make any specific limitations on this.

[0087] S102, based on the state feedback optimal control principle, dynamic model and output feedback controller constructed with output variables, determine the first robust control equation.

[0088] The first robust control equation includes a preset state weight matrix and a control strategy weight matrix to adjust the control performance.

[0089] In this embodiment of the application, based on the state feedback optimal control principle, dynamic model, and output feedback controller constructed with output variables, the first robust control equation is determined, specifically including:

[0090] Based on the principle of state feedback optimal control, the state feedback controller is determined according to the dynamic model. .in, , The state feedback gain matrix is... This represents the control strategy weight matrix. The second coefficient matrix , The positive definite gain matrix is ​​used to solve the Riccati equation. Based on the state feedback controller, a corresponding state-optimal performance index function is constructed. The state-optimal performance index function includes a preset state weight matrix. The output feedback controller constructed from the output variables, the state-optimal performance index function, and the dynamic model are simultaneously reconstructed to obtain the first robust control equation. The output feedback controller includes output variables and a control gain matrix.

[0091] Specifically, this application obtains the nominal system of the robot end-effector system control function based on the above formula (9), as shown in formula (10):

[0092] (10)

[0093] Based on the principle of state feedback optimal control, we obtain formula (11):

[0094] (11)

[0095] in, It is the control strategy weight matrix of the humanoid robot system. R represents an m-dimensional vector space over the real number field, and can adjust the weights of various parts in the state feedback controller.

[0096] This application solves the Riccati equation using the positive definite gain matrix P, as shown in formula (12):

[0097] (12)

[0098] The optimal performance index function is shown in formula (13):

[0099] (13)

[0100] in, The value of the performance index function represents the optimal state.

[0101] At this point, the optimal performance index function of the state feedback controller has been established.

[0102] Subsequently, for the humanoid robot system equations, the optimal performance index function for output feedback is designed as shown in formula (14):

[0103] (14)

[0104] in, To output the optimal performance index function value, and to address the fact that the state and control input of a real system may not have a 1:1 correspondence, a weight matrix is ​​introduced to regulate the system. This refers to the preset state weight matrix for the humanoid robot system. The specific value can be set by expert experience and is not limited here. Let represent an n-dimensional vector space over the real number field. The above index function consists of three parts: the first term addresses the handling of uncertainties in the system, the second term determines the convergence rate of the system state, and the third term determines the system's optimization of the control input. A preset state weight matrix Q can reasonably allocate the proportions of each state, and its weight influence rate in the performance index function is positive feedback.

[0105] In order to use the system output scalar y as the output feedback control variable, the robot output feedback controller is set as follows (15):

[0106] (15)

[0107] in, The control gain matrix to be determined is a constant matrix. The asterisk (*) is used here to distinguish it from the state feedback control variable u mentioned above. In fact, both are the same set control variable, that is, they are considered equivalent in the control process.

[0108] Subsequently, the robust control equations of the humanoid robot system are reconstructed, and the robot outputs a feedback controller. Substituting into formula (13), we get formula (16):

[0109] (16)

[0110] Differentiating the system time, we get formula (17):

[0111] (17)

[0112] Substituting the above humanoid robot nominal system into the formula, we get formula (18):

[0113] (18)

[0114] This leads to the first robust control equation for the humanoid robot, as shown in equation (19):

[0115] (19)

[0116] in, .

[0117] S103, based on the mapping relationship between state variables and output variables in the dynamic model, the first robust control equation is transformed into the second robust control equation. The parameter vector to be estimated is determined according to the second robust control equation, and online estimation is performed based on a preset adaptive law to obtain the control gain matrix in the output feedback controller. The second robust control equation is constructed based on the output variables.

[0118] In this embodiment of the application, based on the mapping relationship between state variables and output variables in the dynamic model, the first robust control equation is transformed into the second robust control equation, specifically including:

[0119] Based on the dynamic model, the state vector of the first robust control equation is transformed to obtain: Based on the mapping relationship between state variables and output variables. The second robust governing equation is obtained: .

[0120] Specifically, this application addresses the robust control equations for the constructed humanoid robot, intending to obtain the control solution using the system output variable y. Therefore, it is necessary to transform the humanoid robot system state x, as shown in formula (20):

[0121] (20)

[0122] Next, the mapping relationship between state variables and output variables is established. The second robust governing equation is obtained, as shown in equation (21):

[0123] (twenty one)

[0124] Further calculation yields formula (22):

[0125] (twenty two)

[0126] Thus, by utilizing parameter transformation to eliminate the influence of the system state variable x, precise output feedback robust control can be achieved using only the system output variable y, without requiring the state variable x. This completely eliminates the dependence on internal states, constructing a model that can be controlled solely using the measurable joint position y.

[0127] Furthermore, the humanoid robot end effector system of this application is a continuous-time system, and the obtained control function is also affected by time. Therefore, a single solution cannot obtain the solution of the system control function. Thus, the humanoid robot control function obtained in this application adopts an online solution method. During the solution process, it is necessary to determine the parameter vector to be estimated based on the second robust control equation, specifically including:

[0128] The second robust governing equation is transformed using the Kronecker product to extract the estimated parameter vector from the transformed parameter expression. The estimated parameter vector is established based on the positive definite gain matrix through vectorization operations.

[0129] In other words, this application uses the Kronecker product to obtain a parameterized expression, as shown in formula (23):

[0130] (twenty three)

[0131] To make the formula more concise, the variable is defined as formula (24):

[0132] (twenty four)

[0133] This completes the aforementioned positive definite gain matrix of variables. to the parameter vector to be estimated The online transformation is used to extract the parameter vector to be estimated. , The calculation utilizes adaptive law online estimation, by The estimated value Obtain the unknown matrix .

[0134] In one embodiment of this application, the above-mentioned online estimation based on a preset adaptive law to obtain the control gain matrix specifically includes:

[0135] By pre-setting an adaptive law Perform online estimation to determine the online estimation parameter vector. estimation error If the value is less than the predetermined value, determine the online estimated parameters corresponding to the parameter vector to be estimated. .in, For online parameter estimation The time derivative. This is the preset learning gain. This is an auxiliary vector. Based on the determined online estimation parameters... Given the parameter vector to be estimated, determine the positive definite gain matrix, and then determine the control gain matrix based on the positive definite gain matrix and the second robust control equation.

[0136] The online estimation method is as follows:

[0137] Predefined vectors With auxiliary regression matrix As in formula (25):

[0138] (25)

[0139] In this application This indicates that the value belongs to a range of (arbitrary) values ​​within the real number field, which will be explained here. Design parameters Define auxiliary vector For example, in formula (26):

[0140] (26)

[0141] in, Indicates the estimation error. , This represents the true parameter vector corresponding to the parameter vector to be estimated. It can be understood as the difference between the estimated weight matrix and the actual weight matrix. Substituting... Auxiliary vectors can be derived. .

[0142] Estimate unknown parameters The adaptive law can be designed as formula (27):

[0143] (27)

[0144] Among them, learning gain Online parameter estimation using adaptive laws. If the regression vector satisfies the continuous stimulus condition, then the estimated parameters... estimation error When the exponential convergence reaches zero, the estimated parameters are achieved. To truth value The exponential convergence is achieved. The predetermined value is 0. In actual use, the specific value is set by the user based on the scenario or expert experience, and no specific limitation is made here.

[0145] Based on this, the above can be obtained. The solution is then based on Obtain the matrix Furthermore, the control gain matrix is ​​obtained by solving the second robust control equation. .

[0146] The specific process of the above online estimation algorithm is as follows:

[0147] 1) Initial conditions: Set initial conditions for the adaptive law. and learning parameters , ;

[0148] 2) Measurement parameters: Input to the measurement system and augmentation system output (Augmentation system output) That is, the actual value of y in the above equation (22) is the change value that the robot system has generated, which can be obtained through external measurement without solving the equation), and then PW transformation is performed using equation transformation and matrix augmentation;

[0149] 3) Online solution: Calculation , , And update unknown parameters online. Obtain the control gain matrix ;

[0150] 4) Control Application: Applying output feedback control to augmented systems. .

[0151] S104 generates corresponding control variables based on the control gain matrix and the real-time acquired output variables to be applied to the robot's end joints in order to perform robust control of the robot's end system.

[0152] In this embodiment of the application, after constructing the control gain matrix using the above S101-S103, the real-time acquired output variables can be input into the output feedback controller, thereby obtaining control variables based only on the output variables, and then performing robust control of the robot's end joints through the control variables.

[0153] In the embodiments of this application, Figure 3This is a flowchart of a robust control system for a humanoid robot, which includes establishing a humanoid robot model, deriving the robot's lower limb control model (i.e., dynamic model), establishing robust control system equations (establishing the first robust control equation), constructing the optimal control performance index function (i.e., constructing the state-optimal performance index function), data-driven learning (i.e., the process of transforming the first robust control equation into the second robust control equation), and then reconstructing the control equation (obtaining the second robust control equation); finally, the controlled humanoid robot system outputs.

[0154] Based on the above scheme, the simulation curve of the stable state of the robot's end effector system obtained by applying artificial external interference to the humanoid robot system is as follows: Figure 4 As shown. By Figure 4 As shown, the curves of different colors represent the simulation curves of the robot recovering to a stable state after applying artificial external disturbances of different intensities. The robot's state parameters are initially unstable, meaning that when artificial external disturbances are applied to the humanoid robot system, the robot's lower limbs become abnormal, i.e., they are in an unstable state. Using the aforementioned output feedback robust control method to control the system, the system's state curves converge rapidly, and the system gradually sets to steady-state behavior. This highlights the ability of the proposed humanoid robot output feedback robust control method to stabilize the system, i.e., it achieves stable control of the humanoid robot system with only the system output variable y. The simulation results demonstrate the effectiveness of the optimal output feedback robust control for robot end-limb control.

[0155] Through the above technical solutions, this application proposes a data-driven output feedback robust control scheme to achieve optimal robust control of humanoid robot systems. This application addresses the difficulty of observing the state of humanoid robots by reconstructing the robust control equations to eliminate the dependence of robot end-joint control on state variables. Secondly, it eliminates the dependence of robot control on internal states through equation transformation, enabling control solely through measurable joint state parameters. Furthermore, an adaptive law is designed to achieve real-time updates of the control gain matrix, enabling efficient and dynamically stable control even with limited data. This leads to the construction of a robust control scheme for humanoid robot end-joints integrating output feedback, robust control, and data-driven approaches. This scheme overcomes limitations in model dependence, disturbance rejection, and control efficiency, enabling precise output feedback robust control for systems with complex dynamic characteristics of humanoid robots.

[0156] Furthermore, this invention applies the concept of robust output feedback optimal control to robot system control algorithms, specifically for the stabilization control of the robot's lower limbs. When a humanoid robot experiences abnormal physical disturbances, its lower limbs may become unstable. The control method of this invention can effectively control the joint angles of the humanoid robot when its posture is abnormal, preventing losses due to instability and loss of control. The application of the robust output feedback control method effectively improves the control accuracy of the humanoid robot's lower limbs and avoids the requirement for the internal state of the system in the humanoid robot control algorithm. As a control method used in precision industrial robots, applying the robust output feedback control method to the control of humanoid robot systems is an innovative work that can fully leverage its optimal control advantages and significantly improve control performance.

[0157] Figure 5 A schematic diagram of the structure of an output feedback robust control system 500 for a humanoid robot provided in this application embodiment is shown below. Figure 5 As shown, the output feedback robust control system 400 for a humanoid robot employs one of the above-described output feedback robust control methods for a humanoid robot. The output feedback robust control system 500 for a humanoid robot includes:

[0158] Module 501 is used to establish a dynamic model of the robot's end effector system and define state variables, control variables, and output variables. The dynamic model is based on the Lagrange equations and includes at least an inertia matrix, Coriolis force vector, gravity vector, and disturbance torque. State variables are the motion state parameters of the end effector joints, control variables are the control input parameters acting on the end effector joints, and output variables are the measurable end effector joint state parameters. Module 502 is used to determine the first robust control equation based on the state feedback optimal control principle, the dynamic model, and the output feedback controller constructed using the output variables. The first robust control equation includes a preset state weight matrix and a control strategy weight matrix to adjust control performance. Module 503 is used to convert the first robust control equation into a second robust control equation based on the mapping relationship between state variables and output variables in the dynamic model. The second robust control equation is used to determine the parameter vector to be estimated and to perform online estimation based on a preset adaptive law to obtain the control gain matrix. The second robust control equation is constructed based on the output variables. The generation module 504 is used to generate corresponding control variables based on the control gain matrix and the output variables acquired in real time, and apply them to the robot end joints to achieve robust control of the robot end system.

[0159] Figure 6 A schematic diagram of the structure of a robust output feedback control device for a humanoid robot provided in this application embodiment is shown below. Figure 6 As shown, the device includes:

[0160] At least one processor; and a memory communicatively connected to the at least one processor. The memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enable the at least one processor to:

[0161] A dynamic model of the robot's end effector system is established, defining state variables, control variables, and output variables. The dynamic model is based on the Lagrange equations and includes at least an inertia matrix, Coriolis force vector, gravity vector, and disturbance torque. State variables are the motion state parameters of the end effector joints, control variables are the control input parameters acting on the end effector joints, and output variables are the measurable end effector joint state parameters. Based on the state feedback optimal control principle, the dynamic model, and an output feedback controller constructed using the output variables, a first robust control equation is determined. This first robust control equation includes a preset state weight matrix and a control strategy weight matrix to adjust control performance. Based on the mapping relationship between state variables and output variables in the dynamic model, the first robust control equation is transformed into a second robust control equation. The parameter vector to be estimated is determined based on the second robust control equation, and online estimation is performed based on a preset adaptive law to obtain the control gain matrix in the output feedback controller. The second robust control equation is constructed based on the output variables. Based on the control gain matrix and the real-time acquired output variables, corresponding control variables are generated and applied to the robot's end effector joints to achieve robust control of the robot's end effector system.

[0162] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system and device embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the description of the method embodiments.

[0163] The systems, devices, and methods provided in this application are one-to-one correspondences. Therefore, the systems and devices also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the systems and devices will not be repeated here.

[0164] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0165] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A robust control method for output feedback in a humanoid robot, characterized in that, The method is applied to robust output feedback control of the lower limbs of a humanoid robot; the method includes: A dynamic model of the robot's end effector system is established, defining state variables, control variables, and output variables. The dynamic model is based on the Lagrange equations and includes at least an inertia matrix, Coriolis force vector, gravity vector, and disturbance torque. The state variables are the motion state parameters of the end effector joints, the control variables are the control input parameters acting on the end effector joints, and the output variables are the measurable end effector joint state parameters. The robot end effector system includes multiple lower limb joints with one or more degrees of freedom. The lower limb joints include at least the hip joint, thigh joint, knee joint, ankle joint, and lumbar joint. Establishing the dynamic model of the robot end effector system specifically includes: modeling the kinematic equations of the humanoid robot's lower limbs based on the pose vectors corresponding to each lower limb joint to obtain motion equations; and establishing the dynamic model based on the motion equations and the Lagrange equations. Based on the state feedback optimal control principle, the dynamic model, and the output feedback controller constructed using the output variables, a first robust control equation is determined; wherein, the first robust control equation is supplemented with a preset state weight matrix and a control strategy weight matrix to adjust the control performance; specifically, this includes: determining the state feedback controller based on the dynamic model according to the state feedback optimal control principle. ;in, , The state feedback gain matrix, This represents the control strategy weight matrix. The second coefficient matrix , For solving the Riccati equation, a positive definite gain matrix is ​​used; based on the state feedback controller, a corresponding state-optimal performance index function is constructed; the state-optimal performance index function includes the preset state weight matrix; the output feedback controller constructed from the output variables, the state-optimal performance index function, and the dynamic model are simultaneously reconstructed to obtain the first robust control equation; the output feedback controller includes the output variables and the control gain matrix; Based on the mapping relationship between the state variables and the output variables in the dynamic model, the first robust control equation is transformed into a second robust control equation. The parameter vector to be estimated is determined according to the second robust control equation, and online estimation is performed based on a preset adaptive law to obtain the control gain matrix in the output feedback controller. The second robust control equation is constructed based on the output variables. Based on the control gain matrix and the output variables acquired in real time, corresponding control variables are generated and applied to the robot's end joints to provide robust control of the robot's end system.

2. The robust output feedback control method for a humanoid robot according to claim 1, characterized in that, The state variable is defined as follows: ;in, , This indicates the coordinates of the lower limb joints. This represents the pose of the humanoid robot's lower limbs in the world coordinate system. Represents the pose vector of the distal joint. For joint velocity vectors, It is the first variable in the state variables. It is the second variable in the state variables; The control variable is defined as follows: ;in, Indicates control variables, Let be the inertia matrix in the dynamic model. For joint torque, The disturbance torque, This is to pre-determine the nonlinear resultant force term; The output variable is defined as follows: , Indicates the output variable. Represents the output matrix , This represents the state vector of the robot's end effector system.

3. The robust output feedback control method for a humanoid robot according to claim 1, characterized in that, The first robust governing equation is: ;in, , The first coefficient matrix , The control gain matrix is... Indicates the external input torque. This represents the preset state weight matrix.

4. The robust output feedback control method for a humanoid robot according to claim 3, characterized in that, Based on the mapping relationship between the state variables and the output variables in the dynamic model, the first robust control equation is transformed into a second robust control equation, specifically including: Based on the dynamic model, the state vector of the first robust control equation is transformed to obtain: ; Based on the mapping relationship between the state variables and the output variables The second robust governing equation is obtained: .

5. The robust output feedback control method for a humanoid robot according to claim 4, characterized in that, The parameter vector to be estimated is determined based on the second robust control equation, specifically including: The second robust control equation is transformed by the Kronecker product to extract the estimated parameter vector based on the transformed parameter expression; the estimated parameter vector is established based on the positive definite gain matrix through vectorization operations. The control gain matrix is ​​obtained through online estimation based on a preset adaptive law, specifically including: Through the preset adaptive law Perform online estimation to determine the online estimation parameter vector. estimation error If the value is less than the predetermined value, determine the online estimated parameter corresponding to the parameter vector to be estimated. ;in, For online parameter estimation The time derivative; Preset learning gain; For auxiliary vectors; Based on the determined online estimation parameters The positive definite gain matrix is ​​determined by the parameter vector to be estimated, and the control gain matrix is ​​determined based on the positive definite gain matrix and the second robust control equation.

6. The robust output feedback control method for a humanoid robot according to claim 5, characterized in that, ,in, A predefined auxiliary regression matrix; Indicates the estimation error. ; The true parameter vector is the vector corresponding to the parameter vector to be estimated.

7. A robust output feedback control system for a humanoid robot, characterized in that, The system is capable of executing the output feedback robust control method for a humanoid robot as described in any one of claims 1-6; the system includes: A module is established to build a dynamic model of the robot's end effector system and define state variables, control variables, and output variables. The dynamic model is based on the Lagrange equation and includes at least an inertia matrix, a Coriolis force vector, a gravity vector, and a disturbance torque. The state variables are the motion state parameters of the end effector joints, the control variables are the control input parameters acting on the end effector joints, and the output variables are the measurable state parameters of the end effector joints. The determination module is used to determine the first robust control equation based on the state feedback optimal control principle, the dynamic model, and the output feedback controller constructed with the output variables; wherein, the first robust control equation is supplemented with a preset state weight matrix and a control strategy weight matrix to adjust the control performance; The conversion module is used to convert the first robust control equation into a second robust control equation based on the mapping relationship between the state variables and the output variables in the dynamic model, so as to determine the parameter vector to be estimated according to the second robust control equation, and perform online estimation based on a preset adaptive law to obtain the control gain matrix; the second robust control equation is constructed based on the output variables; The generation module is used to generate corresponding control variables based on the control gain matrix and the output variables acquired in real time, and apply them to the robot end joints to perform robust control of the robot end system.

8. A robust control device for output feedback in a humanoid robot, characterized in that, The device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed, enables the at least one processor to perform the output feedback robust control method for a humanoid robot as described in any one of claims 1-6.

Images