Search

CN-121979110-A - Robot anti-interference control method based on neural network interference observer

CN121979110ACN 121979110 ACN121979110 ACN 121979110ACN-121979110-A

Abstract

The invention discloses a robot anti-interference control method based on a neural network interference observer, which specifically comprises the following steps of firstly establishing a robot dynamics model containing external interference and constructing a state space model, secondly establishing a fixed time interference observer based on a radial basis neural network to carry out external interference estimation on the robot dynamics model, and thirdly designing a fixed time sliding mode controller based on an interference estimation result to realize fixed time control under the condition of external interference. According to the robot anti-interference control method based on the neural network interference observer, accurate and rapid estimation of interference is achieved by utilizing the fixed time interference observer, convergence of a system state in fixed time is achieved by combining a fixed time sliding mode controller, and feedforward compensation and feedback suppression of the interference are combined, so that the method has the advantages of being simple in implementation mode, high in interference estimation accuracy and high in response speed.

Inventors

  • LIU XIANG

Assignees

  • 东莞理工学院

Dates

Publication Date
20260505
Application Date
20251209

Claims (8)

  1. 1. The robot anti-interference control method based on the neural network interference observer is characterized by comprising the following steps of: Step one, a robot dynamics model containing external interference is established, and a state space model is established; Step two, a fixed time interference observer based on a radial basis function neural network is established, and external interference estimation is carried out on a robot dynamics model; and thirdly, designing a fixed time sliding mode controller based on an interference estimation result to realize fixed time control under the condition of external interference.
  2. 2. The robot antijamming control method based on the neural network jamming observer as set forth in claim 1, wherein the expression of the robot dynamics model is: ; Wherein, the Is a joint position vector, And Respectively are Is a first time derivative and a second time derivative of (c), Is an inertial matrix of the robot and is provided with a plurality of sensors, Is a coriolis force and centripetal force matrix, The force vector of the gravity is used to determine, Is the vector of friction force and other disturbance moment, To control the moment.
  3. 3. The robot antijamming control method based on the neural network jamming observer as set forth in claim 2, wherein the expression of the state space model is: ; Wherein the method comprises the steps of , , And Respectively state variables And With respect to the first derivative of time, Is the total disturbance moment vector under the state space model, Representing the inertial matrix of the robot under the state space model, Representation of Is used for the inverse matrix of (a), Representing the matrix coefficients of coriolis and centripetal forces in the state space model, To control the moment.
  4. 4. The method for controlling robot anti-interference based on neural network interference observer according to claim 3, wherein the calculation formula of the interference moment vector is: ; In the formula, Is the total disturbance moment vector under the state space model, Representing the inertial matrix of the robot under the state space model, Representation of Is used for the inverse matrix of (a), Is a gravity vector under a state space model, Is the friction force and other disturbance moment vector under the state space model.
  5. 5. The method for controlling robot anti-interference based on neural network interference observer according to claim 4, wherein the expression of the fixed time interference observer is: ; In the formula, , , And Respectively state variables And With respect to the first derivative of time, Is to Is used for the estimation of (a), Is the total disturbance moment vector under the state space model, Is to Is used for the estimation of (a), As a radial basis vector weight matrix, Is to Is used for the estimation of (a), Is to Is used for the first-order time derivative of (a), Representing the inertial matrix of the robot under the state space model, Representation of Is used for the inverse matrix of (a), Representing the matrix coefficients of coriolis and centripetal forces in the state space model, In order to control the moment of force, 、 、 、 And Respectively, is a positive constant, and the positive constant, Is a state of Estimate of (2) Is expressed as , 、 And Respectively error vectors 1 St, 2 nd and 3 rd dimensional elements of (c), Is shown as such and is to be understood, , Represented as Superscript "T" means a transpose operation, superscript " AND " "All are exponent powers respectively satisfying And , As an absolute value of the absolute value, As a function of the sign of the symbol, , , wherein, 、 And Respectively is vector The superscript "T" indicates the transpose operation for the 1, 2, and 3-dimensional elements of (2).
  6. 6. The method for controlling robot anti-interference based on neural network interference observer according to claim 5, wherein the radial basis function neural network has Neurons, the first The radial basis function expression of the individual neurons is: ; In the formula, In order to input the vector(s), Is the first The center vector of each neuron is defined by, Is the first The width of the individual neurons and the number of the individual neurons, The superscript "T" represents a transpose operation as a natural exponential function.
  7. 7. The method for controlling robot anti-interference based on neural network interference observer according to claim 6, wherein the expression of the fixed time sliding mode controller is: ; ; ; In the formula, Is a sliding-mode surface, And Is the normal number of the two groups of the, 、 And Respectively state vectors The superscript "T" indicates the transpose operation, And (3) with Are all exponentials of the exponent and , , As an absolute value of the absolute value, As a sign function.
  8. 8. The method for controlling robot anti-interference based on neural network interference observer according to claim 7, wherein the expression of the fixed time anti-interference controller is: ; In the formula, And The normal numbers are respectively.

Description

Robot anti-interference control method based on neural network interference observer Technical Field The invention relates to the technical field of robot control, in particular to a robot anti-interference control method based on a neural network interference observer. Background Robots are used as important remote control and automation tools for modern social development, and are widely applied to various fields such as industrial manufacture, medical surgery, nuclear energy maintenance, disaster relief, aerospace exploration and the like. Compared with manual operation, the robot not only can replace human beings to complete high-risk tasks in dangerous or complex environments, but also can improve the working efficiency and the execution precision, so that the control performance of the robot is directly related to the reliability and the safety of task completion. However, during actual operation, robots are often affected by internal and external multi-source disturbances, such as joint friction, gear backlash, load variations, sensor noise, external environmental disturbances, etc. These disturbances are transmitted to the system body through the robot kinematic coupling, which easily causes joint motion deviations, resulting in misalignment of the end effector, degrading the overall control performance of the robot. Because of the strong nonlinear approximation capability, the neural network is introduced into the fields of interference modeling and observer design in recent years, and can realize high-precision interference estimation under the condition of unknown interference mechanism or incomplete data. However, most of the existing methods rely on asymptotic convergence characteristics, and cannot realize rapid and effective interference compensation within a limited time, so that the dynamic performance and stability of the robot in a complex environment are difficult to comprehensively guarantee. Therefore, how to design a fixed-time anti-interference control method based on a neural network interference observer to realize rapid estimation and accurate compensation of multi-source interference becomes a key technical problem that needs to be broken through in the field of robot control. Currently, studies on the problem of robot fixed time control based on neural network observers are still relatively limited. The China patent application 202110413699.4 proposes a self-adaptive fixed time control method of a mechanical arm under the triggering of an event, which can realize the system stability within fixed time, reduce the resource waste in the control process to a certain extent and improve the tracking speed and the tracking precision. However, the method cannot effectively introduce an interference observer, has insufficient treatment on interference, and therefore has limitation in further improving control precision, and the Chinese patent application 202510445207.8 discloses a control method and a control system for a double-link mechanical arm based on a neural network observer, which utilize learning capability of reinforcement learning and combine a gradient descent method to design an adaptive update rate of an Actor-Critic neural network, so that an approximate controller can approximate to an actual controller. However, this scheme is not combined with a fixed time control method, and it is difficult to achieve a rapid control effect. Based on the retrieval of the data, the traditional method can be seen to be either lack of effective introduction of an interference observer, so that the anti-interference capability is limited, or can not be fused with a fixed time control strategy, and the quick and accurate control of a robot system can not be realized. Disclosure of Invention Aiming at the defects of the prior art, the invention provides a robot anti-interference control method based on a neural network interference observer, which solves the problems in the background art. The invention aims to realize the aim by adopting the following technical scheme that the robot anti-interference control method based on the neural network interference observer comprises the following steps: Step one, a robot dynamics model containing external interference is established, and a state space model is established; Step two, a fixed time interference observer based on a radial basis function neural network is established, and external interference estimation is carried out on a robot dynamics model; and thirdly, designing a fixed time sliding mode controller based on an interference estimation result to realize fixed time control under the condition of external interference. The invention is further arranged that the expression of the robot dynamics model is as follows: Wherein, the Is a joint position vector,AndRespectively areIs a first time derivative and a second time derivative of (c),Is an inertial matrix of the robot and is provided with a plurality of sensors,Is a coriolis force a