A method for fixed-time preset performance control of a robotic arm with non-affine failure

By using a hybrid kernel width learning system and fixed-time sliding mode control, the robustness and fixed-time convergence problems of non-affine fault robotic arms were solved, achieving high-precision trajectory tracking and meeting the technical needs of multiple fields.

CN123103630APending Publication Date: 2026-11-17BEIJING INFORMATION SCI & TECH UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611228776.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-13
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing fault-tolerant control methods are rarely based on non-affine systems, and they also fail to simultaneously address the issues of fixed-time convergence and robustness to non-affine faults.

Method used

A hybrid kernel width learning system (BLS) is used to estimate the total uncertainty of the system. Combined with fixed-time sliding mode control and a preset performance function, the position and velocity information of the robotic arm is obtained through sensors, the tracking error is calculated, and a fixed-time sliding surface is designed. The width learning system is used to approximate the model uncertainty and non-affine faults, and an adaptive update law and update control law are designed to generate the control torque of the robotic arm.

Benefits of technology

It achieves accurate and rapid trajectory tracking of the robotic arm under non-affine fault conditions, reduces the computational burden, and ensures that the tracking error converges to the preset performance boundary within a fixed time, meeting the technical needs of intelligent manufacturing equipment industries such as industrial robots, special operation robots, service and consumer robots, additive manufacturing equipment manufacturing, precision assembly, human-machine collaboration, and aerospace.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123103630A_ABST
    Figure CN123103630A_ABST
Patent Text Reader

Abstract

This invention discloses a fixed-time preset performance control method for a robotic arm with non-affine faults, relating to the field of robot control technology. The method includes: S1, acquiring position and velocity vectors of each joint of the robotic arm based on multiple sensors; calculating the system position error based on the desired position; S2, calculating the normalized error based on the performance functions of each joint of the robotic arm and the position error; obtaining the unconstrained position error vector based on a bijective transform function; and obtaining the unconstrained velocity error vector; S3, calculating the fixed-time sliding mode variable vector; S4, obtaining the total uncertainty function based on the fixed-time sliding mode variable vector, position vector, velocity vector, and the current control law of the robotic arm; approximating the total uncertainty function based on a hybrid kernel-width neural network to obtain an estimated total uncertainty value; S5, obtaining the adaptive update law based on actual weights; calculating the update control law; S6, calculating the dynamic model based on the velocity vector, position vector, and update control law.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of robot control technology, and in particular to a method for controlling the performance of a non-affine fault robotic arm with a fixed-time preset. Background Technology

[0002] With the deepening development of industrial automation and intelligent manufacturing, robotic arms are widely used in intelligent manufacturing equipment industries such as industrial robots, special-purpose robots, service and consumer robots, additive manufacturing equipment manufacturing, precision assembly, human-robot collaboration, and aerospace. In these applications, the actual control performance of robotic arms is inevitably affected by unknown nonlinear dynamics, external time-varying disturbances, and actuator failures. Among these, actuator failures are particularly critical, as they can be caused by factors such as mechanical wear and electrical aging, leading to a severe decline in system performance or even instability.

[0003] Most existing fault-tolerant control methods assume that faults are affine, meaning they enter the control channel in a linearly parameterized form. However, in real-world systems, complex actuator dynamics and internal nonlinear mechanisms can cause fault effects to depend nonlinearly on system state and control input, resulting in non-affine faults. These non-affine faults introduce stronger nonlinear coupling. Furthermore, to improve control accuracy, pre-defined performance control has proven to be an effective framework for ensuring both transient and steady-state performance of tracking errors.

[0004] The following technical problems exist in the existing technology: Existing fault-tolerant control methods are rarely based on non-affine systems, and they also fail to simultaneously address the issues of fixed-time convergence and robustness to non-affine faults. Summary of the Invention

[0005] To address the shortcomings of existing fault-tolerant control methods, which rarely utilize non-affine approaches and fail to simultaneously resolve fixed-time convergence and robustness against non-affine faults, this invention provides a fixed-time preset performance control method for a robotic arm with non-affine fault capabilities. The control method includes: S1. The robotic arm is equipped with multiple sensors, and the position vectors of each joint of the robotic arm are obtained based on the multiple sensors. and velocity vector Based on the position vector The system position error is calculated based on the expected position. S2. Calculate the normalized error based on the performance functions of each joint of the robotic arm and the position error; obtain the unconstrained position error vector based on the bijective transformation function and the normalized error; obtain the unconstrained velocity error vector based on the unconstrained position error vector. S3. Obtain a fixed-time sliding mode variable vector based on the unconstrained position error vector and the unconstrained velocity error vector; S4. Based on the fixed-time sliding mode variable vector and the position vector The velocity vector The total uncertainty function is obtained by combining the current control law of the robotic arm; the total uncertainty estimate is obtained by approximating the total uncertainty function based on a hybrid kernel width neural network; S5. Based on the hybrid kernel width neural network and the actual weights, obtain the adaptive update law; based on the adaptive update law, obtain the update control law; S6, Based on the velocity vector The position vector The dynamic model of the robotic arm is calculated using the update control law.

[0006] Optionally, in step S1: The system position error is given by formula (1): (1) in, This refers to the system position error; For the desired position and ; The system position error is in the first stage of the robotic arm. The error components of each joint are given by formula (2): (2) in, Let t be the error component of the system position error at the i-th joint of the robotic arm; Let be the position vector component of the i-th joint at time t; Let be the expected position component of the i-th joint at time t.

[0007] Optionally, in step S2: The performance function is given by formula (3). (3) in, For time t, the first Performance functions for each joint; For the first The performance boundary of each joint at the initial moment, and satisfying ; For the first The performance boundary of each joint at steady state, and satisfying ; The upper bound of the pre-defined fixed convergence time is, and ; The convergence rate adjustment parameter is used, and ; The normalization error is calculated using formula (4): (4) in, Let be the normalized error of the i-th joint at time t, and Satisfying formula (5): (5) in, The first boundary coefficient, The second boundary coefficient, and and Satisfying formula (6): (6) in To allow overshoot parameters, and .

[0008] Optionally, in step S2: The unconstrained position error vector is calculated using formula (7): (7) in, This is the unconstrained position error vector; The unconstrained position variable component at the i-th joint is calculated using formula (8): (8) in, It is a bijective transformation function; The unconstrained velocity error vector is obtained based on formula (9): (9) in, Let be the unconstrained velocity error vector; Γ is a positive definite diagonal matrix, and satisfies Γ = Γ + Γ. ,in Variables that are greater than 0.

[0009] Optionally, in step S3: The fixed-time sliding mode variable vector is given by formula (10): (10) in, Let be a fixed-time sliding mode variable vector, and ,in, The fixed-time sliding mode variable vector is the fixed-time sliding mode variable at the nth joint; Let be the first positive definite diagonal gain matrix, and ; It is the second positive definite diagonal gain matrix, and r is a constant exponent and is greater than 1; It is a vector function; For piecewise nonlinear function vectors; in, For formula (11): (11) in Standard symbolic functions; in, For formula (12): (12) in, For the piecewise nonlinear function vector at the th The components of each joint are calculated using formula (13): (13) in, It is a constant exponent, and ; It is a constant, and .

[0010] Optionally, in step S4: The total uncertainty function is given by formula (14): (14) in, This is the total uncertainty function; This is the input vector of the hybrid kernel width neural network; Let be the inertia matrix of the robotic arm; The Coriolis matrix of the robotic arm; Let be the gravity vector of the robotic arm; For the non-affine fault items of the robotic arm, where The current control law; External disturbances to the robotic arm; To provide an auxiliary reference acceleration, it is calculated using formula (15): (15) in, For the desired acceleration; Calculated using formula (16): (16) in, Calculate using formula (17): ;(17)

[0011] Optionally, in step S4: The total uncertainty function is approximated based on the hybrid kernel width neural network using formula (18): (18) in, This is the ideal weight matrix for the hybrid kernel width neural network; A vector of basis functions; The approximation error of the hybrid kernel width neural network is and ,and < ,in A constant greater than 0; The total uncertainty estimate is calculated using formula (19): (19) in, This is the estimated total uncertainty. These are the actual weights of the hybrid kernel width neural network; in, For formula (20): (20) in, This is the first component of the basis function vector; This is the second component of the basis function vector.

[0012] Optionally, step S4 further includes: S41. Set the time for starting the update of the initial node of the hybrid kernel width neural network to be... ; will be The receptive field center of the hybrid kernel width neural network at time step is set to ; will be The weights of the hybrid kernel width neural network at time step are set as follows: ; S42. Within the center of the receptive field, those closest to the input vector... The center is set as ;in, ; S43. The updated receptive field is calculated using formula (21): ;(twenty one) in, The updated receptive field; To adjust the parameter vector; The mean of the center is calculated using formula (22); ;(twenty two) in, For the The i-th center among the centers; The updated weights are calculated using formula (23): ;(twenty three) in, This is the updated weight.

[0013] Optionally, in step S5: The adaptive update law is calculated according to formula (24): ;(twenty four) in, For the adaptive update law in the th The weight vector components corresponding to each joint; It is a positive definite symmetric gain matrix; It is an adjustable system parameter and is greater than 0. For the first The basis function vector components of each joint For the sliding mode variable in the th The weight of each joint; The update control law is calculated according to formula (25): (25) in, For the update control law; It is the third positive definite diagonal gain matrix; It is the fourth positive definite diagonal gain matrix.

[0014] Optionally, in step S6: The dynamic model is given by formula (26): (26) in, Let be the acceleration vector of each joint of the robotic arm; For the updated non-affine fault items of the robotic arm, This is the update control law.

[0015] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: This invention utilizes a hybrid kernel-wide learning system (BLS) to estimate the total uncertainty of the system, combined with fixed-time sliding mode control and a preset performance function, to ensure the transient and steady-state performance of the system under fault conditions. First, the position and velocity information of the robotic arm are acquired through sensors, and the tracking error is calculated, transforming the constrained error into an unconstrained position variable. Second, a fixed-time sliding surface is designed, and the BLS is used to approximate model uncertainties, non-affine faults, and external disturbances. Finally, update control laws and adaptive update laws are designed to generate the control torque of the robotic arm, achieving high-precision trajectory tracking. This method introduces a wide learning structure to reduce the computational burden, while the fixed-time convergence mechanism ensures that the convergence time is independent of the initial state, enabling the robotic arm to achieve accurate and rapid trajectory tracking under non-affine faults. This invention also considers non-affine actuator faults, external disturbances, and model uncertainties, ensuring that the tracking error converges to the preset performance boundary within a fixed time, meeting the technical needs of intelligent manufacturing equipment industries such as industrial robots, special-purpose robots, service robots, additive manufacturing equipment manufacturing, precision assembly, human-robot collaboration, and aerospace. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0017] Figure 1 This is a flowchart of a fixed-time preset performance control method for a non-affine fault robotic arm provided by an embodiment of the present invention; Figure 2 This is a schematic diagram of the robotic arm joint used in the embodiments of the present invention; Figure 3 This is a schematic diagram of the experimental environment provided in an embodiment of the present invention; Figure 4 This is a schematic diagram illustrating the tracking effect of the controller acting on the joints of the two robotic arms according to an embodiment of the present invention; Figure 5 This is a schematic diagram of the tracking error of the first joint of the two robotic arms by the controller provided in an embodiment of the present invention; Figure 6 This is a schematic diagram of the tracking error of the second joint of the two robotic arms by the controller provided in an embodiment of the present invention. Detailed Implementation

[0018] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0019] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0020] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0021] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0022] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0023] like Figure 1 As shown, this embodiment of the invention provides a fixed-time preset performance control method for a robotic arm with non-affine failure, the control method comprising: S1. The robotic arm is equipped with multiple sensors, and the position vectors of each joint of the robotic arm are obtained based on the multiple sensors. and velocity vector Based on the position vector The system position error is calculated based on the expected position. S2. Calculate the normalized error based on the performance functions of each joint of the robotic arm and the position error; obtain the unconstrained position error vector based on the bijective transformation function and the normalized error; obtain the unconstrained velocity error vector based on the unconstrained position error vector. S3. Obtain a fixed-time sliding mode variable vector based on the unconstrained position error vector and the unconstrained velocity error vector; S4. Based on the fixed-time sliding mode variable vector and the position vector The velocity vector The total uncertainty function is obtained by combining the current control law of the robotic arm; the total uncertainty estimate is obtained by approximating the total uncertainty function based on a hybrid kernel width neural network; S5. Based on the hybrid kernel width neural network and the actual weights, obtain the adaptive update law; based on the adaptive update law, obtain the update control law; S6, Based on the velocity vector The position vector The dynamic model of the robotic arm is calculated using the update control law.

[0024] This invention utilizes a hybrid kernel-wide learning system (BLS) to estimate the total uncertainty of the system, combined with fixed-time sliding mode control and a preset performance function, to ensure the transient and steady-state performance of the system under fault conditions. First, the position and velocity information of the robotic arm are acquired through sensors, and the tracking error is calculated, transforming the constrained error into an unconstrained position variable. Second, a fixed-time sliding surface is designed, and a wide-time learning system is used to approximate model uncertainties, non-affine faults, and external disturbances. Finally, an adaptive update law is designed to calculate the updated control law and generate the control torque of the robotic arm, achieving high-precision trajectory tracking. This method introduces a wide-time learning structure to reduce the computational burden, while the fixed-time convergence mechanism ensures that the convergence time is independent of the initial state, enabling the robotic arm to achieve accurate and rapid trajectory tracking under non-affine faults. This invention also considers non-affine actuator faults, external disturbances, and model uncertainties, ensuring that the tracking error converges to within the preset performance boundary within a fixed time. This meets the technical needs of intelligent manufacturing equipment industries such as industrial robots, special-purpose robots, service robots, additive manufacturing equipment, precision assembly, human-robot collaboration, and aerospace.

[0025] Optionally, in step S1: The system position error is given by formula (1): (1) in, This refers to the system position error; For the desired position and ; The system position error is in the first stage of the robotic arm. The error components of each joint are given by formula (2): (2) in, Let t be the error component of the system position error at the i-th joint of the robotic arm; Let be the position vector component of the i-th joint at time t; Let be the expected position component of the i-th joint at time t.

[0026] Where the desired position There are positive numbers , making any All 。

[0027] Optionally, in step S2: The performance function is given by formula (3). (3) in, For time t, the first Performance functions for each joint; For the first The performance boundary of each joint at the initial moment, and satisfying ; For the first The performance boundary of each joint at steady state, and satisfying ; The upper bound of the pre-defined fixed convergence time is, and ; The convergence rate adjustment parameter is used, and ; The performance function is a fixed-time performance function; The normalization error is calculated using formula (4): (4) in, Let be the normalized error of the i-th joint at time t, and Satisfying formula (5): (5) in, The first boundary coefficient, The second boundary coefficient, and and Satisfying formula (6): (6) in To allow overshoot parameters, and .

[0028] Optionally, in step S2: The unconstrained position error vector is calculated using formula (7): (7) in, This is the unconstrained position error vector; The unconstrained position variable component at the i-th joint is calculated using formula (8): (8) in, It is a bijective transformation function; This transformation has the following properties: when hour ;when hour, ;when hour, Therefore, as long as it is guaranteed Bounded, thus ensuring Thus, the original tracking error, i.e., the error component Strictly meet the preset performance constraints.

[0029] The unconstrained velocity error vector is obtained based on formula (9): (9) in, Let be the unconstrained velocity error vector; Γ is a positive definite diagonal matrix, and satisfies Γ = Γ + Γ. ,in Variables that are greater than 0. Among them, Let be a diagonal matrix function, representing a diagonal matrix composed of the components of vector x as its main diagonal elements.

[0030] Among them, by differentiating formula (8) and according to formula (4), we can obtain formula (29): (29) Among them, variables and variables They are defined as ; The Jacobian of the transformation function and >2.

[0031] in, yes abbreviation, yes The abbreviations all represent the same quantity that changes over time; the independent variable t is omitted in the derivation for brevity, but it is not a newly defined variable.

[0032] The unconstrained velocity error vector can also be expressed as: ;in It is a positive semi-definite diagonal matrix.

[0033] Steps S1 and S2 obtain the actual joint position and the desired equilibrium position using sensors, and calculate the system position error. To meet preset performance requirements, a performance function constraint is applied to the error, and the constrained error is converted into an unconstrained position variable through a bijective transformation. The velocity error is derived based on the transformed variable. Through the above steps, the transformation of the system position error of each joint into an unconstrained position variable is completed, and these are combined to form an unconstrained position error vector. At the same time, an unconstrained velocity error vector, i.e., an intermediate velocity error vector, is obtained, laying the foundation for subsequent sliding surface design.

[0034] Optionally, in step S3: The fixed-time sliding mode variable vector is given by formula (10): (10) in, Let be a fixed-time sliding mode variable vector, and ,in, The fixed-time sliding mode variable vector is the fixed-time sliding mode variable at the nth joint; Let be the first positive definite diagonal gain matrix, and ; It is the second positive definite diagonal gain matrix, and r is a constant exponent and is greater than 1; It is a vector function; For piecewise nonlinear function vectors; The first positive definite diagonal gain matrix and the second positive definite diagonal gain matrix These are used to adjust the weights of the linear and nonlinear terms, respectively; the constant exponent r is used to ensure the fixed-time convergence characteristic of the system. in, For formula (11): (11) in Standard symbolic functions; in, For formula (12): (12) in, For the piecewise nonlinear function vector at the th The components of each joint are calculated using formula (13): (13) in, It is a constant exponent, and ; It is a constant, and .

[0035] The constant exponent is used to enhance the convergence speed of the sliding surface near the origin; the constant Used to avoid when Approaching zero The strange problem arising from the term.

[0036] Fixed-time control has significant advantages in practical applications because its upper bound on convergence time is independent of the initial state of the system.

[0037] The sliding surface designed in step S3 can lay the foundation for the subsequent controller design.

[0038] To address the issues of non-affine actuator failure, external disturbances, and model uncertainty in the system, a hybrid kernel neural network based on a width-learning system is employed to approximate the unknown model and total uncertainty of the robotic arm. Based on the system's error and weight information, a fixed-time convergent adaptive update law is designed, and on this basis, an update control law considering fixed-time fault tolerance is designed.

[0039] Optionally, in step S4: The total uncertainty function is given by formula (14): (14) in, This is the total uncertainty function; This is the input vector of the hybrid kernel width neural network; Let be the inertia matrix of the robotic arm; The Coriolis matrix of the robotic arm; Let be the gravity vector of the robotic arm; For the non-affine fault items of the robotic arm, where The current control law; External disturbances to the robotic arm; To provide an auxiliary reference acceleration, it is calculated using formula (15): (15) in; For the desired acceleration; Calculated using formula (16): (16) in, Calculate using formula (17): ;(17)

[0040] Optionally, in step S4: The total uncertainty function is approximated based on the hybrid kernel width neural network using formula (18): (18) in, This is the ideal weight matrix for the hybrid kernel width neural network; A vector of basis functions; The approximation error of the hybrid kernel width neural network is and ,and < ,in A constant greater than 0; L is the number of nodes in the neural network; ; The total uncertainty estimate is calculated using formula (19): (19) in, This is the estimated total uncertainty. These are the actual weights of the hybrid kernel width neural network; in, For formula (20): (20) in, This is the first component of the basis function vector; This is the second component of the basis function vector.

[0041] The basis function vector in this embodiment A hybrid kernel strategy is employed, combining Gaussian and wavelet kernels to simultaneously handle smoothing uncertainties and abrupt faults. The Gaussian kernel function is defined as follows: In the formula For the first The receptive field center of a Gaussian kernel Its width. The wavelet kernel function is defined as: ;in For mother wavelet functions (such as Mexican hat wavelet). and These are the scale and translation parameters, respectively. For input vectors The Each component.

[0042] Traditional hybrid kernel width neural networks or fuzzy systems, when used to approximate system uncertainties, often face problems such as complex structures, high computational loads, and a lack of theoretical guidance for node selection. Width learning systems, due to their flattened structure and incremental learning capabilities, offer a new approach to handling complex uncertainties; however, research on their application in the fixed-time control of robotic arms with non-affine faults is still insufficient.

[0043] The control method provided by this invention improves the node selection of the hybrid kernel width neural network based on a width learning system. Traditional neural networks may generate a large number of redundant nodes, causing computational burden. Therefore, based on the width learning structure, this invention improves the node selection method of the radial basis function neural network, as follows: Optionally, step S4 further includes: S41. Set the time for starting the update of the initial node of the hybrid kernel width neural network to be... ; will be The receptive field center of the hybrid kernel width neural network at time step is set to ; will be The weights of the hybrid kernel width neural network at time step are set as follows: ; S42. Within the center of the receptive field, those closest to the input vector... The center is set as ;in, Wherein, "closer to the input vector" refers to n centers whose distance from the input vector is less than a preset threshold.

[0044] S43. The updated receptive field is calculated using formula (21): ;(twenty one) in, The updated receptive field; To adjust the parameter vector; The mean of the center is calculated using formula (22); where, ; ;(twenty two) in, For the The i-th center among the centers; The updated weights are calculated using formula (23): ;(twenty three) in, This is the updated weight.

[0045] After the update begins, the receptive field center is defined at all times as shown in formula (29): (29) in, Let be the set of receptive field centers at time t. For updates The senses of time are gathered at the center of the field. The threshold is adjustable. For the number of nodes, The maximum number of nodes is set, and only when... A new node is generated only when the value is greater than the threshold and the number of nodes is less than the maximum number of nodes. This dual constraint of the threshold and the maximum number of nodes makes the generation of nodes in the neural network more reasonable.

[0046] Furthermore, neural networks employing a wide learning node structure perform better in approximating unknown functions, while also effectively reducing the computational burden on the system.

[0047] In summary, in the embodiments provided by the present invention, the hybrid kernel width neural network maintains a single-layer / flat structure without increasing the depth; when the distance between the new input and the existing center is large enough and the number of nodes has not reached the upper limit, a hybrid kernel node is added horizontally, and the initial output weight of the added node is set to 0. Then, the adaptive update law is updated according to the following steps to update the weights.

[0048] That is, in the subsequent steps, an adaptive update law for the neural network weights is designed so that it can approximate the total uncertainty of the unknown robotic arm model online and make the weight estimation error converge within a fixed time.

[0049] Before designing the update law, define As ideal weights, the actual weights of the neural network are defined as follows: .

[0050] Optionally, in step S5: The adaptive update law is calculated according to formula (24): ;(twenty four) in, For the adaptive update law in the th The weight vector components corresponding to each joint ; It is a positive definite symmetric gain matrix. ; It is an adjustable system parameter and is greater than 0. For the first The basis function vector components of each joint For the sliding mode variable in the th The components of each joint; the update law has been proven through subsequent derivation of Lyapunov theory that its weight estimation error converges in a fixed time.

[0051] The update control law is calculated according to formula (25): (25) in, For the update control law; It is the third positive definite diagonal gain matrix; It is the fourth positive definite diagonal gain matrix. .

[0052] Optionally, in step S6: The dynamic model is given by formula (26): (26) in, Let be the acceleration vector of each joint of the robotic arm; For the updated non-affine fault items of the robotic arm, For the update control law, and , , , , , ,and ; ; Wherein, the position vector For the formula; (27): (27) Among them, the to From the position vector of the first joint of the robotic arm to the position vector of the nth joint of the robotic arm; The velocity vector For formula (28): (28) Among them, the to The velocity vector from the first joint of the robotic arm to the velocity vector of the nth joint of the robotic arm.

[0053] In one embodiment of the present invention, a seven-degree-of-freedom robotic arm (Kinova Gen3) is used as the controlled object. The robot is equipped with a high-precision encoder and a speed sensor, which can acquire the position and speed data of each joint in real time. Figure 2 This is a schematic diagram of the robot joints provided in this embodiment, where J1 to J7 are the seven joints of the robot. For hardware safety and to avoid complex coupling, the designed controller will be applied to joints 4 (J4) and 6 (J6) to verify the effectiveness of the algorithm. Figure 3 In the experimental environment shown, the host computer communicates with the robot via Ethernet at a frequency of 1kHz. The host computer generates the desired trajectory and writes the controller algorithm using Python, and performs data acquisition and control algorithm calculation in a real-time environment.

[0054] In step S2, select , and take The performance function is smooth, strictly positive, and monotonically decreasing, and satisfies... ; and when hour, It remains unchanged.

[0055] Furthermore, in step S2, all data processing is performed in Python, and the desired trajectories of joints J4 and J6 are both selected... Therefore, the expected trajectory is chosen as The desired velocity and acceleration can be obtained through analytical differentiation. The performance function parameters are set based on the absolute value of the initial error to ensure that the error at the initial moment is strictly within the boundary.

[0056] In step S3, for the two-joint robotic arm system, the parameters of the time sliding surface are fixed as follows: The above parameters meet the design requirements. , , ,and All are positive definite diagonal gain matrices. In subsequent steps, based on the fixed-time sliding surface S designed in this step, an adaptive update law will be further designed to drive the system state to reach and maintain on the sliding surface, thereby achieving the expected fixed-time tracking performance.

[0057] Further in steps S4 and S5, let , The initial weights of the hybrid kernel width neural network are all set to 0, the threshold value is set to 0.001, the initial number of nodes is set to 121, and the maximum number of nodes is set to 800. The hybrid kernel width neural network updates the weights with an adaptive update law at each sampling. Before reaching the maximum number of nodes, it dynamically increases the number of nodes based on the input, gradually approximating the unknown model and fault characteristics of the robotic arm, thereby applying more precise control to the robotic arm.

[0058] Next, in step S6, the third diagonal gain matrix and the fourth diagonal gain matrix are set to... The updated control law is then obtained. This updated control law is input into the controller to obtain the control torque of the robotic arm, which is used to control the robotic arm to complete the trajectory tracking task.

[0059] In a specific implementation, the robotic arm transmits the acquired data to the host computer, where a controller designed in Python generates the dynamic models and input torques of each joint of the robotic arm. The torques are then transmitted to the robot via Ethernet to achieve the desired control effect.

[0060] Figure 4This is a schematic diagram illustrating the tracking effect of the controller acting on the joints of two robotic arms according to an embodiment of the present invention, where the horizontal axis represents time and the vertical axis represents joint position. Figure 4 In and As can be seen from the curve, the controller achieved fast convergence and high-precision position tracking for both joints, verifying the effectiveness of the control strategy in dynamic trajectory tracking tasks. Figure 5 This is a schematic diagram of the tracking error of the first joint of the two robotic arms by the controller provided in an embodiment of the present invention; Figure 6 This is a schematic diagram illustrating the tracking error of the controller acting on the second joints of the two robotic arms according to an embodiment of the present invention. Figure 5 and Figure 6 The test results show that the proposed method has a faster convergence speed and higher steady-state accuracy than the PPB method in the tracking control of the first and second joints. It can not only quickly eliminate the initial large error, but also maintain the steady-state tracking error at an extremely low level, which is significantly better than the traditional PPB method.

[0061] In this embodiment of the invention, the designed controller is applied to the joints of the robot, achieving good tracking performance even under non-affine actuator failures, and the system's position error can be constrained within a specified range. This invention enhances the robotic arm's ability to handle complex constraints, and in addressing the uncertainty of the robotic arm model, it applies a width-learning node structure to a radial basis function neural network, reducing the system's computational burden. This meets the technical needs of intelligent manufacturing equipment industries such as industrial robots, special-purpose robots, service robots, additive manufacturing equipment, precision assembly, human-robot collaboration, and aerospace.

[0062] Next, the stability of the above control method will be proved using Lyapunov functions: The relationship between actual weights and ideal weights satisfies ; This represents the error between the actual and ideal values. Existing theorems prove that for a control system, if there exists a Lyapunov function satisfying... ,in , , , If the system is stable at a given time, then it is practically stable at a fixed time. A system with fixed-time stability can reach an equilibrium state within a fixed time under different initial conditions. When a robotic arm performs a task, it will inevitably be in different initial positions. Fixed-time stability allows the robotic arm to stabilize within a certain time under different initial states.

[0063] In the arrival phase of the first part, the sliding surface converges in a fixed time, and the constructed Lyapunov function is shown in equation (29): (29) Differentiating formula (29) and substituting it into the controller and adaptive law, and then scaling it using Young's inequality and matrix eigenvalue inequality, we obtain formula (30): (30) The parameters in formulas (29) and (30) are defined as in formulas (31) to (33): (31) (32) (33) in, For design parameters, , Here is the gain matrix. The inertia matrix, , For adaptive law parameters, , These are the coefficients of Young's inequality.

[0064] Therefore, and At a fixed time The convergence time of the immediate neighborhood that converges to the origin satisfies formula (34): (34) In the sliding phase of the second part, the position error converges asymptotically to zero after a fixed time to the neighborhood: once the system state reaches the sliding surface... In the neighborhood of the sliding surface, the motion of the closed-loop system is governed by the sliding surface equation (35): (35) when When, take the Lyapunov function Differentiating, we get formula (36): (36) in , , At a fixed time Converging inward to the set .

[0065] when hour, It converges asymptotically to zero.

[0066] Combining the two phases, the total convergence time of the system satisfies The tracking error of the robotic arm is initially within a fixed time. The sliding mode variable is brought to the neighborhood of the sliding surface within a fixed time. Internal transformation error convergence to The neighborhood eventually converges asymptotically to zero.

[0067] This invention proposes a fixed-time preset performance control strategy for an uncertain robotic arm based on width learning and non-affine fault compensation. First, the position and velocity information of the robotic arm are collected by sensors, and the tracking error is calculated. Second, the constrained error is converted into an unconstrained position variable using a preset performance function, and a fixed-time sliding surface is designed. Then, the total uncertainty of the system (including non-affine faults, external disturbances, and model uncertainties) is approximated by combining a width learning system with a hybrid kernel function. Finally, an adaptive update law is designed to generate the control input for the robotic arm. The designed controller is applied to the trajectory tracking task of the robotic arm, achieving good tracking results and demonstrating strong robustness in handling non-affine actuator faults and model uncertainties. In dealing with the model uncertainty of the robotic arm, the width learning node structure is applied to nonlinear approximation, ensuring approximation accuracy while reducing the computational burden of the system. This method achieves active compensation for non-affine faults, ensuring fixed-time convergence and preset performance, thereby improving the control effect and reliability under complex constraints. It meets the technical needs of intelligent manufacturing equipment industries such as industrial robots, special operation robots, service and consumer robots, additive manufacturing equipment manufacturing, precision assembly, human-machine collaboration, and aerospace.

[0068] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0069] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0070] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0071] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not 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.

[0072] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0073] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0074] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0075] 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; that is, they may 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.

[0076] In addition, 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.

[0077] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0078] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for performance control of a non-affine fault robotic arm with a fixed-time preset, characterized in that, The control method includes: S1. The robotic arm is equipped with multiple sensors, and the position vectors of each joint of the robotic arm are obtained based on the multiple sensors. and velocity vector Based on the position vector The system position error is calculated based on the expected position. S2. Calculate the normalized error based on the performance functions of each joint of the robotic arm and the position error; obtain the unconstrained position error vector based on the bijective transformation function and the normalized error; obtain the unconstrained velocity error vector based on the unconstrained position error vector. S3. Obtain a fixed-time sliding mode variable vector based on the unconstrained position error vector and the unconstrained velocity error vector; S4. Based on the fixed-time sliding mode variable vector and the position vector The velocity vector The total uncertainty function is obtained by combining the current control law of the robotic arm; the total uncertainty estimate is obtained by approximating the total uncertainty function based on a hybrid kernel width neural network; S5. Based on the hybrid kernel width neural network and the actual weights, obtain the adaptive update law; based on the adaptive update law, obtain the update control law; S6, Based on the velocity vector The position vector The dynamic model of the robotic arm is calculated using the update control law.

2. The method for fixed-time preset performance control of a robotic arm with non-affine failure according to claim 1, characterized in that, In step S1: The system position error is given by formula (1): ;(1) in, This refers to the system position error; For the desired position and ; The system position error is in the first stage of the robotic arm. The error components of each joint are given by formula (2): ;(2) in, Let t be the error component of the system position error at the i-th joint of the robotic arm; Let be the position vector component of the i-th joint at time t; Let be the expected position component of the i-th joint at time t.

3. The method for fixed-time preset performance control of a non-affine fault robotic arm according to claim 2, characterized in that, In step S2: The performance function is given by formula (3). ;(3) in, For time t, the first Performance functions for each joint; For the first The performance boundary of each joint at the initial moment, and satisfying ; For the first The performance boundary of each joint at steady state, and satisfying ; The upper bound of the pre-defined fixed convergence time is, and ; The convergence rate adjustment parameter is used, and ; The normalization error is calculated using formula (4): ;(4) in, Let be the normalized error of the i-th joint at time t, and Satisfying formula (5): ;(5) in, The first boundary coefficient, The second boundary coefficient, and and Satisfying formula (6): ;(6) in To allow overshoot parameters, and .

4. The method for fixed-time preset performance control of a non-affine fault robotic arm according to claim 3, characterized in that, In step S2: The unconstrained position error vector is calculated using formula (7): ;(7) in, This is the unconstrained position error vector; The unconstrained position variable component at the i-th joint is calculated using formula (8): ;(8) in, It is a bijective transformation function; The unconstrained velocity error vector is obtained based on formula (9): ;(9) in, Let be the unconstrained velocity error vector; Γ is a positive definite diagonal matrix, and satisfies Γ = Γ + Γ. ,in Variables that are greater than 0.

5. The method for controlling the performance of a non-affine fault robotic arm with a fixed-time preset according to claim 4, characterized in that, In step S3: The fixed-time sliding mode variable vector is given by formula (10): ; (10) in, Let be a fixed-time sliding mode variable vector, and ,in, The fixed-time sliding mode variable vector is the fixed-time sliding mode variable at the nth joint; Let be the first positive definite diagonal gain matrix, and ; It is the second positive definite diagonal gain matrix, and r is a constant exponent and is greater than 1; It is a vector function; For piecewise nonlinear function vectors; in, For formula (11): ;(11) in Standard symbolic functions; in, For formula (12): ;(12) in, For the piecewise nonlinear function vector at the th The components of each joint are calculated using formula (13): ;(13) in, It is a constant exponent, and ; It is a constant, and .

6. The method for controlling the performance of a non-affine fault robotic arm with a fixed-time preset according to claim 5, characterized in that, In step S4: The total uncertainty function is given by formula (14): ; (14) in, This is the total uncertainty function; This is the input vector of the hybrid kernel width neural network; Let be the inertia matrix of the robotic arm; The Coriolis matrix of the robotic arm; Let be the gravity vector of the robotic arm; For the non-affine fault items of the robotic arm, where The current control law; External disturbances to the robotic arm; To provide an auxiliary reference acceleration, it is calculated using formula (15): ;(15) in, For the desired acceleration; Calculated using formula (16): ;(16) in, Calculate using formula (17): ;(17)。 7. The method for controlling the performance of a non-affine fault robotic arm with a fixed-time preset according to claim 6, characterized in that, In step S4: The total uncertainty function is approximated based on the hybrid kernel width neural network using formula (18): ;(18) in, This is the ideal weight matrix for the hybrid kernel width neural network; A vector of basis functions; The approximation error of the hybrid kernel width neural network is and ,and < ,in A constant greater than 0; The total uncertainty estimate is calculated using formula (19): ;(19) in, This is the estimated total uncertainty. These are the actual weights of the hybrid kernel width neural network; in, For formula (20): ;(20) in, This is the first component of the basis function vector; This is the second component of the basis function vector.

8. The method for fixed-time preset performance control of a non-affine fault robotic arm according to claim 1, characterized in that, Step S4 further includes: S41. Set the time for starting the update of the initial node of the hybrid kernel width neural network to be... ; will be The receptive field center of the hybrid kernel width neural network at time step is set to ; will be The weights of the hybrid kernel width neural network at time step are set as follows: ; S42. Within the center of the receptive field, those closest to the input vector... The center is set as ;in, ; S43. The updated receptive field is calculated using formula (21): ;(21) in, The updated receptive field; To adjust the parameter vector; The mean of the center is calculated using formula (22); ;(22) in, For the The i-th center among the centers; The updated weights are calculated using formula (23): ;(23) in, This is the updated weight.

9. A method for controlling the performance of a non-affine fault robotic arm with a fixed-time preset according to claim 7, characterized in that, In step S5: The adaptive update law is calculated according to formula (24): ;(24) in, For the adaptive update law in the th The weight vector components corresponding to each joint; It is a positive definite symmetric gain matrix; It is an adjustable system parameter and is greater than 0. For the first The basis function vector components of each joint For the sliding mode variable in the th The weight of each joint; The update control law is calculated according to formula (25): ;(25) in, For the update control law; It is the third positive definite diagonal gain matrix; It is the fourth positive definite diagonal gain matrix.

10. A method for controlling the performance of a non-affine fault robotic arm with a fixed-time preset according to claim 9, characterized in that, In step S6: The dynamic model is given by formula (26): ; (26) in, Let be the acceleration vector of each joint of the robotic arm; For the updated non-affine fault items of the robotic arm, This is the update control law.