A disturbance observer-based anti-interference control system for a robot arm servo system

Abstract

The application discloses a kind of mechanical arm servo system anti-interference control systems based on disturbance observer, including hybrid disturbance observer module, anti-interference control law module and adaptive frequency suppression module;The hybrid disturbance observer module is used to classify estimation to the multi-source disturbance in mechanical arm servo system, and it includes dynamic disturbance observer, nonlinear extended state observer and double disturbance observer;The anti-interference control law module is based on inversion control framework;The adaptive frequency suppression module is used to suppress mechanical resonance and flexible vibration, including self-optimizing notch filter and fuzzy vibration suppressor.The application is estimated by multi-observer collaborative, and the suppression capability of multi-source heterogeneous interference is higher than single observer scheme, uses modular design, each observer and control module can be independently configured and debugged, is convenient in different types of mechanical arm transplantation and application, with good engineering practicability.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial robots, and in particular to a mechanical arm servo system anti-interference control system based on a disturbance observer. BACKGROUND

[0002] As the core equipment in the fields of modern industrial automation and precision manufacturing, the control precision and anti-interference ability of the servo system of a mechanical arm directly determine the performance of the whole system. However, the mechanical arm faces a complex environment with multiple source disturbances in actual operation, which includes two categories of internal disturbances and external disturbances.

[0003] The internal disturbances are mainly caused by the inherent physical characteristics and model defects of the system itself, including unmodeled dynamics and parameter uncertainties, nonlinear friction, actuator saturation and dead zone, etc. The external disturbances come from the interaction between the mechanical arm and the environment and the uncontrollable factors of the external environment, including load changes, environmental interaction forces, aerodynamic disturbances, etc.

[0004] In the prior art, although the traditional proportional-integral-derivative control, linear active disturbance rejection control and robust control method based on linear matrix inequality are effective in specific working conditions, they have obvious limitations. Most of these methods are based on the design idea of linear time-invariant, trying to use a fixed and single controller structure to cope with the complex disturbance environment of nonlinearity, time-varying and multiple frequency bands, resulting in limited anti-interference effect. SUMMARY

[0005] Therefore, the present application provides a mechanical arm servo system anti-interference control system based on a disturbance observer to solve the problems in the prior art.

[0006] In order to achieve the above purpose, the present application provides the following technical scheme:

[0007] A mechanical arm servo system anti-interference control system based on a disturbance observer, comprising a hybrid disturbance observer module, an anti-interference control law module and an adaptive frequency suppression module;

[0008] The hybrid disturbance observer module is used for classifying and estimating the multiple source disturbances in the mechanical arm servo system, and comprises a dynamic disturbance observer, a nonlinear extended state observer and a double disturbance observer;

[0009] The anti-interference control law module is based on the backstepping control framework and integrates the barrier Lyapunov function and the disturbance feedforward compensation mechanism;

[0010] The adaptive frequency suppression module is used for suppressing mechanical resonance and flexible vibration, and comprises a self-optimizing notch filter and a fuzzy vibration suppressor;

[0011] The mixed disturbance observer module, anti-interference control law module and adaptive frequency suppression module are cooperated through confidence weighted data fusion method to generate optimal overall disturbance estimation and control output.

[0012] Further, the dynamic disturbance observer is constructed based on a deep deterministic policy gradient algorithm, a state space of the dynamic disturbance observer includes motor position error, speed error, accelerometer reading and pressure sensor reading, and an Actor-Critic network structure is adopted to learn disturbance dynamic characteristics.

[0013] Further, the nonlinear extended state observer adopts a nonlinear function to construct, and unifies the internal uncertainty and external disturbance of the system as an extended state, and realizes gradual estimation and rapid convergence through nonlinear gain design.

[0014] Further, the double disturbance observer includes a modelable disturbance observer and a non-modelable disturbance observer, which are respectively used for estimating modelable external disturbance and unknown non-modelable disturbance, and realizing separate observation of multiple source and different types of disturbance.

[0015] Further, the anti-interference control law module includes a disturbance feedforward compensation unit and a feedback control unit, wherein the feedback control unit adopts an adaptive fuzzy PID controller, and is designed based on a barrier Lyapunov function to ensure that the system state does not violate the preset constraint.

[0016] Further, the self-tending optimal notch filter adopts a parameterized notch filtering and frequency iterative correction mechanism, and automatically adjusts the notch center frequency by comparing the variances of signals in different frequency bands to track the change of the system resonance frequency in real time.

[0017] Further, the fuzzy vibration suppressor is based on a fuzzy inference system, adjusts vibration suppression strategy parameters in real time according to vibration characteristics, and indirectly reduces the vibration amplitude of the end of the manipulator by suppressing the speed fluctuation of the servo system.

[0018] Further, the sensor system includes a motor end encoder, a load end encoder, a strain torque sensor, a three-axis accelerometer and a pressure sensor, and is used to collect the running state data of the manipulator.

[0019] Further, the specific calculation of the confidence weighted data fusion method is as follows:

[0020] The total disturbance estimation fusion formula is:

[0021] ;

[0022] The general formula for calculating the confidence weight is:

[0023] ;

[0024] The observer weight normalization formula is:

[0025]

[0026] wherein, represents the total disturbance estimation value at time k; represents the disturbance estimation value of the i th observer at time k; w i (k) represents the confidence weight of the i th observer at time k; k is the time / iteration step; is a weight adjustment coefficient, a preset constant, controls the contribution proportion of each index to the weight (non-negativity needs to be met); R i (k) represents the residual index; S i (k) represents the stability index; C i (k) represents the consistency index; H i (k) represents the health index; j=1,2,3, represents the summation index, the sum of the original weights of all observers, as the normalization denominator.

[0027] The application has the following advantages: the application has higher suppression capability of multi-source heterogeneous interference than a single observer scheme through the multi-observer cooperative estimation architecture, and the trajectory tracking error is reduced in the scenario where wind disturbance and load mutation exist at the same time.

[0028] The system adopts modular design, and each observer and control module can be independently configured and debugged, facilitating transplantation and application on different types of mechanical arms, and having good engineering practicability.

[0029] Other features and advantages of the present application will be described in the following description, and some will become apparent from the description, or will be understood by those skilled in the art from the description. DETAILED DESCRIPTION

[0030] In order to more intuitively illustrate the prior art and the present application, exemplary drawings are given below. It should be understood that the specific shapes, structures shown in the drawings should not be regarded as limiting conditions in the implementation of the present application; for example, based on the technical concept disclosed in the present application and the exemplary drawings, those skilled in the art can easily make routine adjustments or further optimization to some units (components) such as increase / decrease / attribute division, specific shape, positional relationship, connection mode, size ratio relationship, etc.

[0031] Figure 1 A signal flow diagram of a disturbance observer-based mechanical arm servo system anti-interference control system is provided for an embodiment of the present application. DETAILED DESCRIPTION ​​​​

[0032] The present application is further explained in the following detailed description of embodiments of the application, with reference to the following figures, in which like reference numerals refer to like elements throughout. It will be obvious to those skilled in the art that the described embodiments are not the only ways of implementing the present application and should not be used to limit the true scope of the present application. Other embodiments of the present application will be apparent to those skilled in the art from consideration of the specification and practice of the application disclosed herein. It is intended that the present application embrace all such embodiments, and that the patent be not limited to any features described, or any specific embodiments described herein. It is intended that the specification and figures be considered as exemplary only, with the true scope and spirit of the application being indicated by the following claims.

[0033] Please refer to Figure 1 A disturbance observer based anti-disturbance control system for robot manipulator servo system, including a hybrid disturbance observer module, an anti-disturbance control law module and an adaptive frequency rejection module.

[0034] The hybrid disturbance observer module is composed of three specialized observers, namely a dynamic disturbance observer, a nonlinear extended state observer and a dual disturbance observer.

[0035] The dynamic disturbance observer adopts the deep deterministic policy gradient (DDPG) algorithm, and its state space design contains multi-dimensional information such as motor position error, speed error, accelerometer reading and pressure sensor reading. Both the actor network and the critic network adopt a three-layer fully connected neural network structure, and the number of hidden layer nodes is set to 256 and 128 respectively. The observer is updated every 10 milliseconds, and its disturbance estimation performance is continuously optimized through an online learning mechanism. The unique feature of this observer is its ability to learn the dynamic patterns of complex nonlinear disturbances, enabling predictive estimation of non-continuous sudden disturbances and thus providing the possibility for advanced compensation control.

[0036] The state space of the dynamic disturbance observer is defined as: where e θ is the motor position error; e ω is the motor speed error; a acc is the accelerometer reading; and F press is the pressure sensor reading.

[0037] The disturbance estimation output formula is:

[0038] ;

[0039] where is the converged optimal policy parameter; f DDQ is the DDPG mapping function; and S is the input state vector.

[0040] The nonlinear extended state observer (NESO) adopts innovative nonlinear function design to specifically cope with system model uncertainty, unmodeled dynamics, and low-frequency slowly varying disturbances. The observer unifies system internal uncertainty and external disturbances as "extended states", and realizes asymptotic estimation and fast convergence through carefully designed nonlinear gains. Its mathematical expression adopts a special fal function form, which has a high gain to quickly eliminate static error when the estimation error is small, and the gain is saturated to avoid overshoot when the estimation error is large. This nonlinear characteristic makes it have better dynamic performance than traditional linear observers while ensuring global stability.

[0041] The state equation of the nonlinear extended state observer is:

[0042] ;

[0043] where, is the system state estimation, is the total disturbance estimation; is the extended state; b is the control gain coefficient; u is the control input; β i is the adaptive nonlinear gain function; h3(t) is an additional injection signal for exciting the observer; fal(e,a,δ) is a nonlinear saturation function; ε1 is the first-level observation error; ε2 is the second-level observation error; 、 are the nonlinear power parameters of the first and second levels, respectively; 、 are the linear interval thresholds of the first and second levels, respectively.

[0044] The total disturbance output formula is:

[0045] .

[0046] The double disturbance observer (DIO) estimates the modelable external disturbance and unknown unmodeled disturbance through two parallel sub-observers; the modelable disturbance observer is based on the Luenberger observer structure and is specifically designed to handle periodic or deterministic disturbances with known mathematical models; the unmodeled disturbance observer uses nonlinear design based on auxiliary variables to estimate all unknown disturbances. This separation estimation architecture enables the system to implement targeted compensation strategies for different types of disturbances, significantly improving anti-disturbance efficiency.

[0047] The modelable disturbance observer of the double disturbance observer is:

[0048] ;

[0049] where, L1 is the observer bandwidth; d modThe expected disturbance is derived for a known physical model.

[0050] The non-modeled disturbance observer:

[0051] ;

[0052] Wherein, y is the system output; L2 is the robustness enhancement coefficient; C, G, H are observation matrices.

[0053] The synthesized disturbance output is:

[0054] .

[0055] The anti-disturbance control law module is designed based on the inversion control framework, integrates the barrier Lyapunov function and the disturbance feedforward compensation mechanism, and forms a complete control strategy.

[0056] The control law design adopts a step-by-step recursive method; first, the position tracking error is defined, and the first Lyapunov function is constructed to design the virtual control quantity to ensure the stability of the position loop; then, the speed tracking error is defined, and the innovative barrier Lyapunov function (BLF) is introduced to strictly ensure that the speed error does not violate the preset constraint boundary; this design based on BLF enables the system to guarantee tracking performance while effectively preventing actuator saturation and system oscillation caused by excessive compensation.

[0057] The disturbance feedforward compensation unit is responsible for converting the multi-source disturbance estimation value output by the hybrid disturbance observer into a feedforward compensation torque; the disturbance feedforward compensation unit adopts a confidence-weighted data fusion algorithm, dynamically adjusts the weights of the outputs of different observers according to the estimation accuracy of each observer and the current system operating state, generates the optimal overall disturbance estimation value, and calculates the feedforward compensation amount accordingly.

[0058] The feedback control unit adopts an adaptive fuzzy PID controller, which can adjust control parameters in real time according to system state; the feedback control unit combines the reasoning ability of fuzzy logic and the simplicity of PID control, adjusts the proportional, integral, and differential parameters online, so that the system can maintain good control performance under different operating conditions.

[0059] The adaptive frequency suppression module includes:

[0060] The self-optimizing notch filter adopts a parameterized notch filter and a frequency iterative correction mechanism, automatically adjusts the notch center frequency by comparing the variances of signals in different frequency bands, and real-time tracks the system resonance frequency change;

[0061] The fuzzy vibration suppressor is based on a fuzzy inference system, adjusts vibration suppression strategy parameters in real time according to vibration characteristics, and indirectly reduces the amplitude of the end of the robotic arm by suppressing the speed fluctuation of the servo system.

[0062] The mixed disturbance observer module, the anti-interference control law module and the adaptive frequency suppression module are cooperated through data fusion with confidence weighting to generate optimal overall disturbance estimation and control output.

[0063] In addition, the hardware platform of the system adopts a distributed architecture combining an industrial PC and an FPGA; the industrial PC is responsible for running upper-layer algorithms including a mixed disturbance observer, an anti-interference control law and an adaptive frequency suppression algorithm; and the FPGA is responsible for real-time servo control and signal processing to ensure the hard real-time performance of the control system.

[0064] The sensor system includes multiple types of sensors: high-precision encoders for position and speed detection, strain torque sensors for joint torque measurement, three-axis accelerometers for vibration monitoring, and pressure sensors for wind resistance disturbance sensing.

[0065] The working flow of the system follows strict timing control; in each control cycle (typical value: 100 microseconds), the FPGA first synchronously collects data from all sensors and performs preprocessing; then, the data are simultaneously sent to the three disturbance observers for parallel processing.

[0066] The dynamic disturbance observer focuses on high-frequency dynamic disturbance estimation, the nonlinear extended state observer is responsible for total disturbance estimation, and the dual disturbance observer performs disturbance separation estimation; the outputs of the observers are weighted and integrated in the data fusion unit to generate a comprehensive disturbance estimation.

[0067] The total disturbance estimation fusion formula is:

[0068] ;

[0069] The general formula for calculating the confidence weight is:

[0070] ;

[0071] The observer weight normalization formula is:

[0072] ;

[0073] wherein, represents the total disturbance estimation value at time k; represents the disturbance estimation value of the i th observer at time k; w i (k) represents the confidence weight of the i th observer at time k; k is the time / iteration step; 、 、 、 is a weight adjustment coefficient, a preset constant, and controls the contribution proportion of each index to the weight (must satisfy non-negativity); R i(k) represents a residual index; S i (k) represents a stability index; C i (k) represents a consistency index; H i (k) represents a health index; j = 1, 2, 3, represents the summation index, summing the original weights of all observers as the normalization denominator.

[0074] Based on the fused disturbance estimation, the anti-disturbance control law module calculates the feedforward compensation and feedback control, while the adaptive frequency suppression module generates the vibration suppression signal; all these control components are integrated in the control quantity synthesis unit to form the anti-disturbance control command, which is sent to the servo driver through the field bus.

[0075] Embodiment:

[0076] Taking a six-degree-of-freedom industrial robot as the implementation object, the system is verified under the actual working conditions of the automobile body-in-white welding production line; the robot end carries a 5kg welding gun, and performs high-precision welding tasks in an environment with periodic airflow and instantaneous load impact.

[0077] In terms of parameter tuning, the system adopts a hierarchical tuning strategy; first, the initial parameters of the self-optimizing notch filter are set by identifying the system resonance characteristics through offline frequency domain analysis; then, the observer bandwidth is tuned through step response experiments to ensure a balance between estimation speed and noise suppression; finally, the optimal control gain is determined through parameter scanning, and online fine-tuning is performed in actual operation.

[0078] After continuous running tests, the system exhibits excellent performance; compared with the traditional PID controller, the trajectory tracking accuracy is improved, the peak error is reduced during the instant welding of the welding gun; the vibration suppression effect is improved, the residual vibration amplitude of the robot end after high-speed motion is stopped is reduced, and the stabilization time is shortened; at the same time, due to smoother control, the average energy consumption of the servo system is reduced.

[0079] The above only describes the preferred embodiments of the present application and should not be used to limit the present application, any modification, equivalent replacement and improvement made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A disturbance observer based anti-disturbance control system for a robot arm servo system, characterized by, The hybrid disturbance observer module, the anti-disturbance control law module and the adaptive frequency suppression module are cooperated through a confidence weighted data fusion method to generate optimal overall disturbance estimation and control output. The confidence weighted data fusion method is calculated as follows: The total disturbance estimation fusion formula is: The confidence weight calculation general formula is: The observer weight normalization formula is: The dynamic disturbance observer is constructed based on a deep deterministic policy gradient algorithm, and its state space includes motor position error, speed error, accelerometer reading and pressure sensor reading. The nonlinear extended state observer adopts a nonlinear function to construct, and unifies the internal uncertainty and external disturbance of the system as an extended state, and realizes gradual estimation and rapid convergence through nonlinear gain design. ； The double disturbance observer includes a modelable disturbance observer and a non-modelable disturbance observer, which are used to estimate modelable external disturbance and unknown non-modelable disturbance respectively, and realize separate observation of multiple source and different types of disturbance. ； The anti-disturbance control law module includes a disturbance feedforward compensation unit and a feedback control unit, wherein the feedback control unit adopts an adaptive fuzzy PID controller and is designed based on a barrier Lyapunov function to ensure that the system state does not violate the preset constraints. ； wherein, denotes the total disturbance estimate at time k; denotes the disturbance estimate of the i-th observer at time k; w i (k) denotes the confidence weight of the i-th observer at time k; k is the time / iteration step; , , , is a weight adjustment coefficient, a pre-set constant, controlling the contribution proportion of each index to the weight; R i (k) denotes the residual index; S i (k) denotes the stability index; C i (k) denotes the consistency index; H i (k) denotes the health index; j = 1, 2, 3, denotes the summation index, summing the original weights of all observers as the normalization denominator.

2. The disturbance rejection control system for a robot manipulator servo system based on a perturbation observer according to claim 1, characterized in that, The self-optimizing notch filter adopts a parameterized notch filtering and frequency iterative correction mechanism, automatically adjusts the notch center frequency by comparing the variances of signals in different frequency bands, and real-time tracks the change of system resonance frequency.

3. The disturbance rejection control system for a robot manipulator servo system based on a perturbation observer according to claim 1, wherein The fuzzy vibration suppressor is based on a fuzzy inference system, adjusts vibration suppression strategy parameters in real time according to vibration characteristics, and indirectly reduces the vibration amplitude of the end of the manipulator by suppressing the speed fluctuation of the servo system.

4. The disturbance rejection control system for a robot manipulator servo system based on a perturbation observer according to claim 1, wherein The sensor system includes motor end encoder, load end encoder, strain torque sensor, three-axis accelerometer and pressure sensor, which is used to collect the running state data of the manipulator.

5. The disturbance rejection control system for a robot manipulator servo system based on a perturbation observer according to claim 1, wherein ​ 6. The disturbance rejection control system for a robot manipulator servo system based on a perturbation observer according to claim 1, wherein ​ 7. The disturbance rejection control system for a robot manipulator servo system based on a perturbation observer according to claim 1, wherein ​ 8. The disturbance rejection control system for a robot manipulator servo system based on a perturbation observer according to claim 1, wherein, ​

Images