Robot control method and system based on RBFNN variable parameter admittance

Abstract

The embodiment of the invention relates to the technical field of robots, and discloses a robot control method and system based on RBFNN variable parameter admittance. The control method comprises the following steps: in response to an interaction action of an operator using the rehabilitation robot, obtaining an interaction torque; in the variable admittance outer ring, inputting the interaction torque into a pre-trained RBFNN model, outputting a target inertia coefficient and a target damping coefficient of an admittance control model, and determining a position basic error according to the interaction torque based on the admittance control model; in the position inner ring, according to the position basic error and the target position, based on a sliding mode surface function, a control torque is obtained through calculation of a compensation item and a robust item; and applying a control moment to the rehabilitation robot. According to the scheme of the embodiment of the invention, the safety of rehabilitation training and the naturalness of man-machine cooperation can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of robot technology, in particular to a robot control method and system based on RBFNN variable parameter admittance. BACKGROUND

[0002] As an important branch of medical robots, rehabilitation robots have been widely used in the motor function rehabilitation training of patients with stroke, spinal cord injury, etc. Its core task is to guide or assist patients to complete specific actions under the premise of safety, so as to promote the reconstruction of neural and motor functions. In this process, the robot system and the patient form a closely coupled human-machine interaction system, and the control performance directly affects the training effect and patient experience.

[0003] Traditional robot position control strategies (such as PID control) are not capable of handling human-robot interaction tasks. Due to its high stiffness and high precision design goal, when the patient applies an unexpected force, a large interaction torque is easily generated, which not only poses a safety hazard, but also results in stiff and unnatural movement, reducing patient compliance.

[0004] To improve the interaction experience, admittance control is introduced into the field of rehabilitation robots. The basic idea is to take the interaction torque as the input, dynamically adjust the reference position of the robot through a second-order admittance control model, so that the robot exhibits a certain compliance and can produce compliant motion when subjected to external force. However, the parameters (inertia, damping, stiffness) of traditional admittance control are usually fixed. This leads to an inherent contradiction: higher inertia / damping is beneficial to movement smoothness and stability, but will reduce the response speed of the system to the interaction force, making it feel heavy; lower parameters make the system agile, but are prone to oscillation and poor anti-interference ability.

[0005] More importantly, rehabilitation training is a dynamic process, and the muscle strength level, movement pattern and interaction characteristics of different patients and different rehabilitation stages differ greatly. The fixed parameter admittance controller is difficult to maintain optimal performance in all working conditions. In addition, in the position inner loop of the robot, there are usually problems such as model parameter uncertainty, unmodeled dynamics (such as friction), and external disturbances. Simply using the admittance outer loop without designing a robust inner loop controller will cause the actual trajectory to deviate significantly from the desired trajectory, making the compliance planning of the outer loop meaningless. SUMMARY

[0006] To address the deficiencies, the embodiments of the present application disclose a robot control method and system based on RBFNN variable parameter admittance, which can improve the safety of rehabilitation training and the naturalness of human-machine collaboration.

[0007] The first aspect of the embodiments of the present application discloses a robot control method based on RBFNN variable parameter admittance, comprising: The system acquires interactive torque in response to the interactive actions of the operator using the rehabilitation robot. In the variable admittance outer loop, the interaction torque is input into the pre-trained RBFNN model, and the target inertia coefficient of the admittance control model is output. and target damping coefficient Based on the admittance control model, the position basis error is determined according to the interaction torque; In the inner position loop, based on the basic position error and the target position, the control torque is calculated using the sliding surface function and through compensation and robust terms. The control torque is applied to the rehabilitation robot.

[0008] Based on the intelligent compliance of the outer loop and the precise robustness of the inner loop, this invention constructs a hierarchical intelligent adaptive control framework, which fundamentally solves the contradiction between compliance and tracking accuracy in traditional rehabilitation robot control. This makes the robot no longer a rigid motion executor, but an intelligent partner that can understand the patient's intentions, significantly improving the safety of rehabilitation training and the naturalness of human-machine collaboration.

[0009] The admittance outer loop uses RBFNN to sense the interaction torque in real time and dynamically adjust the target inertia coefficient. and target damping coefficient This allows the robot's inertia and damping characteristics to be intelligently adjusted based on the patient's exertion. For example, when the patient intends to move actively, the RBFNN can reduce... and This makes the robot lightweight and easy to push; when stable support is needed, the parameters are increased to make the robot more stable.

[0010] The inner loop receives the adjusted instructions from the outer loop and uses advanced control algorithms that include compensation and robustness terms to ensure that the robot can resist interference such as its own model uncertainty and changes in the patient's force exertion, and accurately track this dynamically changing instruction, thus ensuring that the compliance planning of the outer loop can be faithfully executed.

[0011] As an optional implementation, in the first aspect of the present invention, the RBFNN model is pre-trained, including: by As the radial basis function of the RBFNN model, As the cost function for model training, gradient descent is used to train the RBFNN model using training samples, wherein... Let J be the radial basis function of the j-th hidden layer. For input data, and Let be the center value and standard deviation of the Gaussian function of the j-th hidden layer, respectively. Let cost function be For interactive torque, The rate of change of the interaction torque. This is the regularization parameter.

[0012] The cost function in this embodiment of the invention simultaneously penalizes the interaction force and the rate of change of the interaction force, driving the control strategy learned by the RBFNN to not only reduce the steady-state contact force to improve comfort and safety, but also suppress drastic changes in force, thereby avoiding harsh experiences such as sudden pulls or stops and ensuring smooth motion transitions.

[0013] It can also be adjusted This allows for a fine-tuning of the two objectives mentioned above. In the early stages of rehabilitation, a larger target can be set. Prioritize safety and limit the amount of interaction force as much as possible; in the later stages of recovery, the intensity can be appropriately reduced. It allows for a certain amount of interaction force for strength training while focusing more on the smoothness of movement. This gives the controller the potential to adapt to different stages of rehabilitation.

[0014] As an optional implementation, in the first aspect of the present invention, the positional basis error is determined based on the interaction torque according to the admittance control model. ,include:

[0015] in, For transfer functions, Let be the interaction torque, and s be the Laplace operator. Let be the target stiffness coefficient, and we have: .

[0016] In this embodiment of the invention, the admittance control model of second-order differential is transformed into the Laplace domain, from which the transfer function can be obtained. Based on the transfer function and interaction torque, the positional error is obtained. Then, through stability analysis, the target stiffness coefficient and target damping coefficient are integrated into a single parameter, which greatly simplifies the complexity of controller parameter tuning and ensures that the admittance control model itself is stable as a dynamic system, avoiding oscillations or instability that may be caused by arbitrary setting of outer loop parameters.

[0017] As an optional implementation, in a first aspect of the present invention, the control torque is calculated based on the positional basis error and the target position, using a sliding surface function and a compensation term and a robust term, including: The position error and velocity error are determined based on the aforementioned basic position error, actual position, and target position:

[0018]

[0019]

[0020] in, For positional error, For speed error, The nominal inertia matrix, The nominal Coriolis force and centrifugal force matrix, For the nominal gravity term, For uncertain terms, For interactive torque, To control the torque, For the positional basis error, As the reference position, For the target location, For actual location, For actual speed, For actual acceleration, Accelerate towards the target; Based on the position error and velocity error, the control torque is calculated using the sliding surface function and through compensation and robust terms.

[0021] Uncertainty It includes inertial error ΔM, Coriolis force and centrifugal force error ΔC, gravity error ΔG, and unmodeled friction force. :

[0022] By decomposing the real dynamics into a nominal model and lumped uncertainties This provides a clear approach for control design: use a nominal model for feedforward compensation and design a robust law specifically to suppress [the spread of the virus]. This unified approach enables the controller to have universal robustness against various model deviations and external disturbances.

[0023] As an optional implementation, in a first aspect of the present invention, the control torque is calculated based on the position error and velocity error, using a sliding surface function and a compensation term and a robust term, including: Adjustment quantity is constructed based on sliding surface function. :

[0024] in, Let be the sliding surface function of the position error, and:

[0025] in, For the first diagonal matrix of the design, , , , These are the three elements of the first diagonal matrix. For the design of the second diagonal matrix, , , , These are the three elements of the second diagonal matrix. The sliding mode index, , For symbolic functions, It is a diagonal matrix function; Calculate control torque :

[0026] in, To adjust the amount The sliding mode term, and: , For the design of the third diagonal matrix, , , , These are the three elements of the third diagonal matrix. The set uncertainty compensation coefficient is used to compensate for... , This is the cross-term compensation coefficient, used to compensate for cross-terms; The sliding surface function represents the velocity error.

[0027] In this embodiment of the invention, through the nominal model Perform feedforward compensation, through Interactive force compensation is performed to actively counteract measurable interactive forces and prevent them from affecting tracking accuracy, by introducing a sliding surface. and constructing adjustment amount The core of the backstepping control method is to transform a complex second-order error system into a more stable first-order system.

[0028] in, In order to offset Introduced in the definition item, In order to offset In Item, because Through with To eliminate the product term The impact, It is a sliding mode control term, whose function is to actively drive... Approaching zero enhances robustness.

[0029] and As a robust feedback term, it is based on Control and adaptive control theory design, equivalent to a nonlinear damper, when The larger the value, the greater the damping provided by the robust feedback term, thereby suppressing the effects of system oscillations and uncertainties. Used to suppress uncertainties To ensure the system meets the L2 norm gain performance, Mainly used to compensate for cross term error The final control torque will be obtained Substitution After obtaining the expression, by using inequalities such as Young's inequality, it can be finally proven that:

[0030] in It is a positive number. It is a very small positive number. According to the theory of uniformly eventual bounded stability, this inequality proves that all signals of the system are bounded, and the tracking error... Ultimately, it will be constrained to a very small neighborhood, thus ensuring the stability of the system.

[0031] As an optional implementation, in the first aspect of the present invention, the sliding surface function of the velocity error... :

[0032] in, Let be the velocity error of the i-th joint. Let be the positional error of the i-th joint, where 1 ≤ i ≤ 3. For positive integers, Let be the boundary value of the position error of the i-th joint.

[0033] By designing a piecewise function for the sliding surface function of velocity error, the core problem of control singularity in sliding mode control is cleverly solved. When the error is large, i.e. At that time, the standard terminal sliding mode item is adopted. It retains its excellent characteristic of finite-time convergence, allowing the system state to approach the equilibrium point faster than linear feedback. However, when the error is large, i.e. At that time, the gain term Freeze at a lower limit value This avoids When it approaches zero, this term tends to infinity, leading to a singularity in the control torque.

[0034] This design retains the advantage of fast terminal sliding mode convergence speed while avoiding theoretically infinite control force, enabling the controller to operate stably and reliably in actual digital systems without failing due to torque saturation or high-frequency chattering.

[0035] As an optional implementation, in a first aspect of the present invention, the interactive torque is acquired by a torque sensor mounted on a rehabilitation robot.

[0036] By collecting interactive torques through torque sensors, the system ensures that interactive intentions can be accurately and quickly captured, providing a real and reliable data foundation for subsequent intelligent algorithm processing. This is the key hardware guarantee for the entire system to move from theory to engineering application.

[0037] A second aspect of this invention discloses a robot control system based on RBFNN variable parameter admittance, comprising: The acquisition unit is used to acquire the interaction torque in response to the interactive actions of the operator using the rehabilitation robot; The determining unit is used to input the interaction torque into a pre-trained RBFNN model in the variable admittance outer loop and output the target inertia coefficient of the admittance control model. and target damping coefficient Based on the admittance control model, the position basis error is determined according to the interaction torque; The calculation unit is used to calculate the control torque in the inner position loop based on the position basis error and the target position, using the sliding surface function and through compensation and robust terms; A control unit for applying the control torque to the rehabilitation robot.

[0038] Based on the intelligent compliance of the outer loop and the precise robustness of the inner loop, this invention constructs a hierarchical intelligent adaptive control framework, which fundamentally solves the contradiction between compliance and tracking accuracy in traditional rehabilitation robot control. This makes the robot no longer a rigid motion executor, but an intelligent partner that can understand the patient's intentions, significantly improving the safety of rehabilitation training and the naturalness of human-machine collaboration.

[0039] A third aspect of the present invention discloses an electronic device, comprising: a memory storing executable program code; a processor coupled to the memory; the processor calling the executable program code stored in the memory to execute the robot control method based on RBFNN variable parameter admittance disclosed in the first aspect of the present invention.

[0040] A fourth aspect of the present invention discloses a computer-readable storage medium storing a computer program, wherein the computer program causes a computer to execute the robot control method based on RBFNN variable parameter admittance disclosed in the first aspect of the present invention. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 This is a flowchart illustrating the robot control method based on RBFNN variable parameter admittance disclosed in an embodiment of the present invention. Figure 2 This is a principle block diagram of the robot control method based on RBFNN variable parameter admittance disclosed in the embodiments of the present invention; Figure 3 This is a simulation diagram of the passive position control mode in the rehabilitation of the knee and hip joints disclosed in the embodiments of the present invention; Figure 4 This is a control effect diagram of continuous interference force applied during knee and hip joint rehabilitation, as disclosed in an embodiment of the present invention. Figure 5 This is a diagram showing the relationship between the actual position and the target position in hip joint rehabilitation as disclosed in an embodiment of the present invention; Figure 6 This is a diagram showing the relationship between the actual position and the target position in knee joint rehabilitation as disclosed in an embodiment of the present invention; Figure 7 This is a diagram showing the relationship between the actual position, target position, and control position in the variable admittance control mode during hip joint rehabilitation, as disclosed in this embodiment of the invention. Figure 8 This is a schematic diagram of the structure of a robot control system based on RBFNN variable parameter admittance provided in an embodiment of the present invention; Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0043] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0044] It should be noted that the terms first, second, third, fourth, etc., in the specification and claims of this invention are used to distinguish different objects, not to describe a specific order. The terms used in the embodiments of this invention include and have, and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products, or devices.

[0045] Based on the intelligent compliance of the outer loop and the precise robustness of the inner loop, this invention constructs a hierarchical intelligent adaptive control framework, which fundamentally solves the contradiction between compliance and tracking accuracy in traditional rehabilitation robot control. This makes the robot no longer a rigid motion executor, but an intelligent partner that can understand the patient's intentions, significantly improving the safety of rehabilitation training and the naturalness of human-machine collaboration. The following is a detailed description of this framework.

[0046] Example 1

[0047] Please see Figure 1 , Figure 1 This is a flowchart illustrating the robot control method based on RBFNN variable parameter admittance disclosed in this invention. The execution entity of the method described in this invention is an execution entity composed of software and / or hardware. This execution entity can receive relevant information via wired and / or wireless means and can send certain instructions. It may also have certain processing and storage functions. This execution entity can control multiple devices, such as remote physical servers or cloud servers and related software, or local hosts or servers and related software that perform related operations on devices located in a certain place. In some scenarios, it can also control multiple storage devices, which may be placed in the same location as the device or in different locations.

[0048] like Figure 1 As shown, the robot control method based on RBFNN-based variable parameter admittance includes the following steps: S110, In response to the interactive actions of the operator using the rehabilitation robot, acquire the interactive torque.

[0049] By collecting interactive torques through torque sensors, the system ensures that interactive intentions can be accurately and quickly captured, providing a real and reliable data foundation for subsequent intelligent algorithm processing. This is the key hardware guarantee for the entire system to move from theory to engineering application.

[0050] Taking lower limb rehabilitation robots as an example, strain-type six-dimensional force / torque sensors installed on the end effector of the lower limb rehabilitation robot can be used to collect interactive torques. The accuracy can reach 0.5%-1% FS, and the sampling frequency can be 1ms, etc.

[0051] S120. In the variable admittance outer loop, the interaction torque is input into the pre-trained RBFNN model, and the target inertia coefficient of the admittance control model is output. and target damping coefficient Based on the admittance control model, the positional basis error is determined according to the interaction torque.

[0052] Please refer to Figure 2 As shown, the admittance outer loop senses the interaction torque in real time and dynamically adjusts the target inertia coefficient through the RBFNN. and target damping coefficient This allows the robot's inertia and damping characteristics to be intelligently adjusted according to the patient's exertion, thereby achieving variable admittance control. For example, when the patient intends to move actively, the RBFNN can reduce... and This makes the robot lightweight and easy to push; when stable support is needed, the parameters are increased to make the robot more stable.

[0053] Specifically, the RBFNN model needs to be trained first. In a preferred embodiment of the present invention, using... As the radial basis function of the RBFNN model, As the cost function for model training, gradient descent is used to train the RBFNN model using training samples, wherein... Let J be the radial basis function of the j-th hidden layer. For input data, and Let be the center value and standard deviation of the Gaussian function of the j-th hidden layer, respectively. Let cost function be For interactive torque, The rate of change of the interaction torque. This is the regularization parameter.

[0054] It should be noted that the cost function in this embodiment of the invention penalizes both the interaction force and the rate of change of the interaction force. This drives the control strategy learned by the RBFNN to not only reduce the steady-state contact force to improve comfort and safety, but also to suppress drastic changes in force, thereby avoiding harsh experiences such as sudden pulls or stops and ensuring smooth motion transitions.

[0055] In addition, it can also be adjusted This allows for a fine-tuning of the two objectives mentioned above. In the early stages of rehabilitation, a larger target can be set. Prioritize safety and limit the amount of interaction force as much as possible; in the later stages of recovery, the intensity can be appropriately reduced. It allows for a certain amount of interaction force for strength training while focusing more on the smoothness of movement. This gives the controller the potential to adapt to different stages of rehabilitation.

[0056] The update rate is calculated using the gradient descent method.

[0057]

[0058] in, For the weight change value of the j-th hidden layer, The first learning rate primarily controls the sensitivity to the rate of change of force. The second learning rate primarily controls the sensitivity to the magnitude of the force. Let the base weights of the j-th hidden layer be _____. This is the updated weight value for the j-th hidden layer.

[0059] The input data to RBFNN is a vector combining the interaction torque and the rate of change of the interaction torque:

[0060] For the first The interactive torque obtained from each sampling point For the first The rate of change of the interaction torque at each sampling point, 0 ≤ p ≤ m.

[0061] After training, the data collected by the operator during use is fed into the trained RBFNN model, and the following results can be obtained:

[0062]

[0063] Among them, M d0 and B d0 The target inertia coefficients are respectively Base values ​​and target damping coefficients The base value.

[0064] Then, through stability analysis, the target stiffness coefficient and the target damping coefficient are integrated into a single parameter. This greatly simplifies the complexity of controller parameter tuning, ensures that the admittance control model itself is a stable dynamic system, and avoids oscillations or instability that may be caused by arbitrary setting of outer loop parameters.

[0065] Transforming the second-order differential admittance control model into the Laplace domain yields the transfer function:

[0066] Based on the transfer function and interaction torque, the positional basis error is obtained:

[0067] in, For transfer functions, Let be the interaction torque, and s be the Laplace operator. The target stiffness coefficient.

[0068] S130. In the inner position loop, based on the position basis error and the target position, the control torque is calculated using the sliding surface function and through compensation and robust terms.

[0069] The inner loop receives the adjusted instructions from the outer loop and uses advanced control algorithms that include compensation and robustness terms to ensure that the robot can resist interference such as its own model uncertainty and changes in the patient's force exertion, and accurately track this dynamically changing instruction, thus ensuring that the compliance planning of the outer loop can be faithfully executed.

[0070] Please refer to Figure 2 As shown, it may specifically include the following steps: The position error and velocity error are determined based on the aforementioned basic position error, actual position, and target position:

[0071]

[0072]

[0073] in, For positional error, For speed error, The nominal inertia matrix, The nominal Coriolis force and centrifugal force matrix, For the nominal gravity term, For uncertain terms, For interactive torque, To control the torque, For the positional basis error, As the reference position, For the target location, For actual location, For actual speed, For actual acceleration, Accelerate towards the target; Based on the position error and velocity error, the control torque is calculated using the sliding surface function and through compensation and robust terms.

[0074] Uncertainty It includes inertial error ΔM, Coriolis force and centrifugal force error ΔC, gravity error ΔG, and unmodeled friction force. :

[0075] By decomposing the real dynamics into a nominal model and lumped uncertainties This provides a clear approach for control design: use a nominal model for feedforward compensation and design a robust law specifically to suppress [the spread of the virus]. This unified approach enables the controller to have universal robustness against various model deviations and external disturbances.

[0076] Based on the position and velocity errors, the control torque is calculated using the sliding surface function and through compensation and robustness terms. Specifically: Adjustment quantity is constructed based on sliding surface function. :

[0077] in, Let be the sliding surface function of the position error, and:

[0078] in, For the first diagonal matrix of the design, , , , These are the three elements of the first diagonal matrix. For the design of the second diagonal matrix, , , , These are the three elements of the second diagonal matrix. The sliding mode index, , For symbolic functions, It is a diagonal matrix function; Calculate control torque :

[0079] in, To adjust the amount The sliding mode term, and: , For the design of the third diagonal matrix, , , , These are the three elements of the third diagonal matrix. The set uncertainty compensation coefficient is used to compensate for... , This is the cross-term compensation coefficient, used to compensate for cross-terms; The sliding surface function represents the velocity error.

[0080] In this embodiment of the invention, the sliding surface function of the speed error is cleverly designed as a piecewise function, thus solving the core problem of control singularity in sliding mode control:

[0081] in, Let be the velocity error of the i-th joint. Let be the positional error of the i-th joint, where 1 ≤ i ≤ 3. For positive integers, Let be the boundary value of the position error of the i-th joint.

[0082] When the error is large, that is At that time, the standard terminal sliding mode item is adopted. It retains its excellent characteristic of finite-time convergence, allowing the system state to approach the equilibrium point faster than linear feedback. However, when the error is large, i.e. At that time, the gain term Freeze at a position error boundary value (lower limit). This avoids When it approaches zero, this term tends to infinity, leading to a singularity in the control torque.

[0083] This design retains the advantage of fast terminal sliding mode convergence speed while avoiding theoretically infinite control force, enabling the controller to operate stably and reliably in actual digital systems without failing due to torque saturation or high-frequency jitter.

[0084] In this embodiment of the invention, through the nominal model Perform feedforward compensation, through Interactive force compensation is performed to actively counteract measurable interactive forces and prevent them from affecting tracking accuracy, by introducing a sliding surface. and constructing adjustment amount The core of the backstepping control method is to transform a complex second-order error system into a more stable first-order system.

[0085] In order to offset Introduced in the definition item, In order to offset In Item, because Through with To eliminate the product term The impact, It is a sliding mode control term, whose function is to actively drive... Approaching zero enhances robustness.

[0086] and As a robust feedback term, it is based on Control and adaptive control theory design, equivalent to a nonlinear damper, when The larger the value, the greater the damping provided by the robust feedback term, thereby suppressing the effects of system oscillations and uncertainties. Used to suppress uncertainties To ensure the system meets the L2 norm gain performance, Mainly used to compensate for cross term error The final control torque will be obtained Substitution After obtaining the expression, by using inequalities such as Young's inequality, it can be finally proven that:

[0087] in It is a positive number. It is a very small positive number. According to the theory of uniformly eventual bounded stability, this inequality proves that all signals of the system are bounded, and the tracking error... Ultimately, it will be constrained to a very small neighborhood, thus ensuring the stability of the system.

[0088] To verify the stability of the system, Lyapunov functions V2 were used for verification:

[0089] in The reason for introducing This is because in a mechanical system, kinetic energy It is a natural Lyapunov function term. Doing so simplifies the differentiation result. Differentiating with respect to V2, we get:

[0090] because:

[0091] but for:

[0092] Will Substitution :

[0093] Utilizing the oblique symmetry property of robot dynamics, i.e. It is obliquely symmetrical:

[0094] Substitution :

[0095] Combine the Coriolis force and centrifugal force terms:

[0096] Control torque Substitution :

[0097] Among them, the cross term is ,because If the matrix is ​​symmetric, then we have ,so: .

[0098] In the design of control torque, the following methods are used to ensure : 1. It is considered negative, that is , making ; 2. For sliding mode control items, via , making ; 3. Robust feedback items Used to suppress uncertainties Cross terms It can be done To handle, thereby ensuring .

[0099] S140. Apply the control torque to the rehabilitation robot.

[0100] By using the torque formula, the control torque is converted into a control current and applied to the rehabilitation robot, enabling the robot to cooperate with the operator to complete rehabilitation training.

[0101] in, To control the current, This is the torque coefficient.

[0102] Taking knee and hip joint rehabilitation as an example, the robot control method of the present invention is simulated, wherein, Figure 3 To simulate the effect in passive position control mode, from Figure 3 It can be seen that in the first 1-2 seconds, a disturbance force with random magnitude and direction, and an amplitude range of 0-20 N·m, is applied to the system; the trajectory is still well tracked. In the 4-5 seconds, a disturbance force of 20 N·m is applied to the hip joint. Figure 3 The third picture Figures 4-5 The green part at time s), the interference torque in the same direction as the direction of motion, the magnitude of which is 10 N·m applied to the knee joint. Figure 3 The third picture Figures 4-5 (The blue part at time s) represents the interference torque acting in the opposite direction to the motion. It can be seen that when the interference torque is consistent with the motion, the auxiliary torque decreases accordingly, and vice versa. By subtracting the auxiliary torque from the total torque to predict the interference torque, the prediction errors can be obtained as follows: hip joint RMSE = 0.0074, knee joint RMSE = 0.012.

[0103] Figure 4 The control effect under continuous disturbance force is shown, from Figure 4 It can be seen that the trajectory conforms to the disturbance torque, deviating in the same direction as the disturbance force according to the compliance of the given admittance parameters, thus avoiding rigid impact on the patient.

[0104] Figure 5 and Figure 6 The actual positions of the hip and knee joints are shown respectively. and target location Used to verify the position tracking and robustness of the position control loop. Figure 7 The actual position of the knee joint in variable admittance control mode is shown. Target location and reference position ,from Figures 5-7 It can be seen that the intelligent flexibility of the outer ring and the precise robustness of the inner ring can ensure the safety of rehabilitation training and the naturalness of human-machine collaboration.

[0105] Example 2

[0106] Please see Figure 8 , Figure 8 This is a schematic diagram of the structure of a robot control system based on RBFNN variable parameter admittance disclosed in an embodiment of the present invention. Figure 8 As shown, the robot control system based on RBFNN variable parameter admittance may include: Acquisition unit 210 is used to acquire interaction torque in response to the interactive actions of the operator using the rehabilitation robot; The determining unit 220 is used to input the interaction torque into a pre-trained RBFNN model in the variable admittance outer loop and output the target inertia coefficient of the admittance control model. and target damping coefficient Based on the admittance control model, the position basis error is determined according to the interaction torque; The calculation unit 230 is used to calculate the control torque in the inner position loop based on the position basis error and the target position, using the sliding surface function and through compensation and robust terms. Control unit 240 is used to apply the control torque to the rehabilitation robot.

[0107] Example 3

[0108] Please see Figure 9 , Figure 9 This is a schematic diagram of the structure of an electronic device disclosed in an embodiment of the present invention. The electronic device can be a computer, a server, etc. Of course, in certain cases, it can also be a mobile phone, tablet computer, monitoring terminal, or other smart device, as well as an image acquisition device with processing capabilities. Figure 9 As shown, the electronic device may include: Memory 310 storing executable program code; Processor 320 coupled to memory 310; The processor 320 calls the executable program code stored in the memory 310 to execute some or all of the steps in the robot control method based on RBFNN variable parameter admittance in Embodiment 1.

[0109] This invention discloses a computer-readable storage medium storing a computer program that causes a computer to perform some or all of the steps in the robot control method based on RBFNN variable parameter admittance in Embodiment 1.

[0110] This invention also discloses a computer program product, wherein when the computer program product is run on a computer, the computer executes some or all of the steps in the robot control method based on RBFNN variable parameter admittance in Embodiment 1.

[0111] This invention also discloses an application publishing platform, which is used to publish computer program products. When the computer program products are run on a computer, the computer executes some or all of the steps in the robot control method based on RBFNN variable parameter admittance in Embodiment 1.

[0112] In various embodiments of the present invention, it should be understood that the sequence number of each process does not necessarily imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0113] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they can be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0114] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0115] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-accessible memory. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several requests to cause a computer device (which can be a personal computer, server, or network device, specifically a processor in the computer device) to execute some or all of the steps of the methods described in the various embodiments of the present invention.

[0116] In the embodiments provided by this invention, it should be understood that B corresponding to A means that B is associated with A, and B can be determined based on A. However, it should also be understood that determining B based on A does not mean determining B solely based on A; B can also be determined based on A and / or other information.

[0117] Those skilled in the art will understand that some or all of the steps in the various methods of the embodiments described can be implemented by a program instructing related hardware. This program can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium capable of carrying or storing data.

[0118] The above provides a detailed description of the robot control method, system, electronic device, and storage medium based on RBFNN variable parameter admittance disclosed in the embodiments of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A robot control method based on RBFNN variable parameter admittance, characterized in that, include: The system acquires interactive torque in response to the interactive actions of the operator using the rehabilitation robot. In the variable admittance outer loop, the interaction torque is input into the pre-trained RBFNN model, and the target inertia coefficient of the admittance control model is output. and target damping coefficient Based on the admittance control model, the positional basis error is determined according to the interaction torque; In the inner position loop, based on the basic position error and the target position, the control torque is calculated using the sliding surface function and through compensation and robust terms. The control torque is applied to the rehabilitation robot.

2. The robot control method based on RBFNN variable parameter admittance as described in claim 1, characterized in that, Pre-trained RBFNN models include: by As the radial basis function of the RBFNN model, As the cost function for model training, gradient descent is used to train the RBFNN model using training samples, wherein... Let J be the radial basis function of the j-th hidden layer. For input data, and Let be the center value and standard deviation of the Gaussian function of the j-th hidden layer, respectively. Let cost function be For interactive torque, The rate of change of the interaction torque. This is the regularization parameter.

3. The robot control method based on RBFNN variable parameter admittance as described in claim 1, characterized in that, Based on the admittance control model, the positional foundation error is determined according to the interaction torque. ,include: in, For transfer functions, Let be the interaction torque, and s be the Laplace operator. Let be the target stiffness coefficient, and we have: .

4. The robot control method based on RBFNN variable parameter admittance as described in claim 1, characterized in that, Based on the aforementioned positional error and target position, the control torque is calculated using the sliding surface function and through compensation and robustness terms, including: The position error and velocity error are determined based on the aforementioned basic position error, actual position, and target position: in, For positional error, For speed error, The nominal inertia matrix, The nominal Coriolis force and centrifugal force matrix, For the nominal gravity term, For uncertain terms, For interactive torque, To control the torque, For the positional basis error, As the reference position, For the target location, For actual location, For actual speed, For actual acceleration, Accelerate towards the target; Based on the position error and velocity error, the control torque is calculated using the sliding surface function and through compensation and robust terms.

5. The robot control method based on RBFNN variable parameter admittance as described in claim 4, characterized in that, Based on the position error and velocity error, the control torque is calculated using the sliding surface function and through compensation and robustness terms, including: Adjustment quantity is constructed based on sliding surface function. : in, Let be the sliding surface function of the position error, and: in, For the first diagonal matrix of the design, , , , These are the three elements of the first diagonal matrix. For the design of the second diagonal matrix, , , , These are the three elements of the second diagonal matrix. The sliding mode index, , For symbolic functions, It is a diagonal matrix function; Calculate control torque : in, To adjust the amount The sliding mode term, and: , For the design of the third diagonal matrix, , , , These are the three elements of the third diagonal matrix. The set uncertainty compensation coefficient is used to compensate for... , This is the cross-term compensation coefficient, used to compensate for cross-terms; The sliding surface function represents the velocity error.

6. The robot control method based on RBFNN variable parameter admittance as described in claim 5, characterized in that, Sliding surface function of velocity error : in, Let be the velocity error of the i-th joint. Let be the positional error of the i-th joint, where 1 ≤ i ≤ 3. For positive integers, Let be the boundary value of the position error of the i-th joint.

7. The robot control method based on RBFNN variable parameter admittance as described in any one of claims 1-6, characterized in that, The interactive torque is acquired by a torque sensor installed on the rehabilitation robot.

8. A robot control system based on RBFNN variable parameter admittance, characterized in that, include: The acquisition unit is used to acquire the interaction torque in response to the interactive actions of the operator using the rehabilitation robot; The determining unit is used to input the interaction torque into a pre-trained RBFNN model in the variable admittance outer loop and output the target inertia coefficient of the admittance control model. and target damping coefficient Based on the admittance control model, the positional basis error is determined according to the interaction torque; The calculation unit is used to calculate the control torque in the inner position loop based on the position basis error and the target position, using the sliding surface function and through compensation and robust terms; A control unit for applying the control torque to the rehabilitation robot.

9. An electronic device, characterized in that, include: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the robot control method based on RBFNN variable parameter admittance as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein the computer program causes a computer to execute the robot control method based on RBFNN variable parameter admittance as described in any one of claims 1 to 7.

Images