Self-adaptive sliding mode variable structure control method and system for robot

Abstract

The invention relates to the technical field of robot control, in particular to a self-adaptive sliding mode variable structure control method and system for a robot, and the method comprises the steps: obtaining sensor data of a motion motor of the robot and a trajectory tracking controller of a corresponding control system, carrying out the time sequence decomposition of the sensor data, and carrying out the feature extraction in combination with a decomposition result, the method comprises the following steps: obtaining a feature group of sensor data, and determining a buffeting influence coefficient of the sensor data by combining the feature group of the sensor data and a buffeting condition of a state point on a sliding mode surface, thereby obtaining an adjustment coefficient of an approaching factor in a trajectory tracking controller so as to adjust the approaching factor in the trajectory tracking controller. And rapid and accurate tracking of the expected trajectory by the robot motion motor is realized. According to the invention, the cooperative improvement of the control precision and the dynamic performance is realized, so that the robot can still keep a good motion control effect in a complex and changeable working environment.

Description

Technical Field

[0001] This invention relates to the field of robot control technology, and specifically to an adaptive sliding mode variable structure control method and system for robots. Background Technology

[0002] The rapid development of embodied intelligence technology has driven the rapid innovation of robot intelligent control technology. As robots perceive and interact with their environment, they need to precisely control the robot's motion motors. DC torque motors are often chosen for use in position servo systems and low-speed servo systems because they have a large torque coefficient, fast response speed, and small torque and speed fluctuations. However, DC torque motors are easily affected by friction, vibration, and sudden load changes caused by the transmission mechanism during operation. Therefore, the method of using sliding mode variable structure control for robot motion control has emerged.

[0003] In existing sliding mode variable structure control methods, although improving the reaching law can suppress chattering and improve convergence speed to a certain extent, the controller parameters usually cannot effectively adapt to the actual operating state of the robot and external disturbances. As a result, when the robot's working environment or load conditions change, the controller often cannot effectively suppress chattering, which further leads to insufficient motion control accuracy of the robot. Summary of the Invention

[0004] This invention provides an adaptive sliding mode variable structure control method and system for robots to solve existing problems.

[0005] The adaptive sliding mode variable structure control method and system for robots of the present invention adopts the following technical solution: One embodiment of the present invention provides an adaptive sliding mode variable structure control method for robots, the method comprising the following steps: The sensor data of the robot's motion motors and the trajectory tracking controller of the corresponding control system are acquired, and the trajectory tracking controller contains a convergence factor; The sensor data of the motion motor is decomposed in time series, and feature extraction is performed based on the data change characteristics in the decomposition results to obtain the feature group of the sensor data. By combining the characteristic groups of sensor data and the chattering of the state points of the control system on the sliding surface, the chattering influence coefficient of each sensor data is determined. Based on the chattering influence coefficient of all sensor data, the adjustment coefficient of the approach factor in the trajectory tracking controller is obtained. The approach factor in the trajectory tracking controller is adjusted using an adjustment coefficient, and the motion motor of the robot is controlled by the trajectory tracking controller.

[0006] Optionally, the specific method for performing time-series decomposition on the sensor data of the motion motor includes: The EEMD signal decomposition algorithm is used to perform mode decomposition on arbitrary sensor data, and the quantity control function is determined during the mode decomposition process. Based on the numerical change trend of the quantity control function, several IMF component data corresponding to the sensor data are obtained.

[0007] Optionally, the method for obtaining the quantity control function is as follows: For any sensor data, the number of noise additions K of the EEMD signal decomposition algorithm is preset. All noise addition processes and the EMD decomposition process of the sensor data after noise addition are processed in parallel. Based on the fluctuation characteristics of the IMF component data decomposed in each parallel process, the quantity control function under the mode decomposition process of sensor data is obtained.

[0008] Optionally, the specific method for obtaining the quantity control function includes: Any sensor data is denoted as target sensor data. K white noise sequences with a mean of 0 and different standard deviations are obtained and added to the target sensor data respectively to obtain K noisy target sensor data. For any noisy target sensor data, iterative EMD decomposition is performed on the noisy target sensor data. Each decomposition yields IMF component data of the corresponding order. During the iterative EMD decomposition of all noisy target sensor data, the data points in the corresponding IMF component data at each order are analyzed for fluctuation and similarity to construct the quantity control function at the corresponding order.

[0009] Optionally, the specific method for obtaining K white noise sequences with a mean of 0 and different standard deviations includes: Obtain the standard deviation of all extreme points in the target sensor data, and then... As the range of standard deviation values ​​for the K white noise sequences, within the range of standard deviation values, As the initial value, every We take values ​​to represent the standard deviation of all elements in each white noise sequence, thus obtaining K white noise sequences with a mean of 0 and different standard deviations; where The standard deviation of all extreme points in the target sensor data. These are preset range parameters.

[0010] Optionally, the specific method for obtaining several IMF component data corresponding to the target sensor data based on the numerical change trend of the quantity control function includes: A two-dimensional rectangular coordinate system is constructed, with the numerical value corresponding to the quantity control function as the vertical axis and the order of the IMF component data as the horizontal axis. The quantity control function values ​​of the IMF component data at each order of the target sensor data are mapped to the two-dimensional rectangular coordinate system to form a corresponding scatter plot, denoted as the quantity control function scatter plot of the target sensor data. The scatter plot of the quantity control function of the target sensor data is curve-fitted using the least squares method. The inflection point in the curve-fitting result of the quantity control function scatter plot is obtained using the elbow method. The order closest to the inflection point is taken as the order of the IMF component data of the target sensor data, denoted as the target order of the target sensor data. Thus, the IMF component data of the target sensor data is obtained through the EEMD signal decomposition algorithm, and the order of the IMF component data of the target sensor data is the corresponding target order.

[0011] Optionally, the specific method for obtaining the feature groups of the sensor data is as follows: For each IMF component data of the target sensor data, the instantaneous frequency sequence and amplitude envelope sequence of the IMF component data are calculated using the Hilbert-Huang transform. Based on the instantaneous frequency sequence and amplitude envelope sequence, the average frequency, frequency bandwidth and energy ratio of the IMF component data are calculated. The array formed by the average frequency, frequency bandwidth and energy ratio of the IMF component data is used as the feature group of the IMF component data. The feature group of all IMF component data of the target sensor data is obtained.

[0012] Optionally, the method for determining the chattering influence coefficient of each sensor data point by combining the feature group of sensor data and the chattering situation of the control system's state points on the sliding surface includes: The sliding surface deviation is acquired in real time, a window is constructed for the current moment, and the current moment is taken as the rightmost moment of the window. The absolute values ​​of the sliding surface deviations at all moments within the window are accumulated and used as the chattering index for the current moment. The IMF component data with the largest energy proportion in the feature group corresponding to the IMF component data of the target sensor data is taken as the main IMF component data of the target sensor data. The feature group of continuously acquired main IMF component data is obtained, and the absolute mean of the cross-correlation coefficients between all elements in the feature group and the chattering index is calculated and used as the chattering influence coefficient of the target sensor data.

[0013] Optionally, the method for obtaining the adjustment coefficient of the approach factor in the trajectory tracking controller based on the jitter influence coefficient of all sensor data includes: A preset reference jitter level is set, the ratio of the reference jitter level to the jitter index at the current moment is obtained, and the ratio is input into the sigmoid function for normalization. The normalization result is used as the jitter deviation value at the current moment. An adjustment coefficient is calculated by combining the vibration impact coefficient and the vibration deviation value at the current moment. The adjustment coefficient is positively correlated with both the vibration impact data and the vibration deviation value.

[0014] An adaptive sliding mode variable structure control system for a robot includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of any one of the adaptive sliding mode variable structure control methods for a robot.

[0015] The beneficial effects of the technical solution of the present invention are as follows: After acquiring sensor data, by performing time-series decomposition and feature extraction on the data, key feature groups reflecting changes in system state are effectively identified, providing a reliable basis for the adaptive adjustment of subsequent control parameters. On this basis, combined with the chattering situation of the sliding surface state, the influence of different sensor data on system chattering is further evaluated. This process enables the controller to dynamically perceive external disturbances and internal state changes according to the actual working conditions, and then generate reasonable adjustment coefficients to optimize the approach factor in the trajectory tracking controller. The adaptive adjustment of the approach factor significantly enhances the robustness of the system to uncertain disturbances and load fluctuations, suppressing the chattering phenomenon commonly found in traditional sliding mode control, and ensuring the speed and stability of system response. Therefore, the embodiments of the present invention achieve a synergistic improvement in control accuracy and dynamic performance, enabling the robot to maintain good motion control performance in complex and ever-changing working environments. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating the steps of an adaptive sliding mode variable structure control method for a robot according to the present invention. Figure 2 A schematic diagram of the deviation curve between the state point and the sliding surface provided in one embodiment of the present invention; Figure 3 This is a structural block diagram of an adaptive sliding mode variable structure control system for robots according to the present invention. Detailed Implementation

[0018] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an adaptive sliding mode variable structure control method and system for robots proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0020] The following description, in conjunction with the accompanying drawings, details a specific scheme for an adaptive sliding mode variable structure control method and system for robots provided by the present invention.

[0021] Please see Figure 1 The diagram illustrates a flowchart of an adaptive sliding mode variable structure control method for a robot according to an embodiment of the present invention. The method includes the following steps: Step S001: Obtain sensor data of the robot's motion motors and the corresponding trajectory tracking controller of the control system.

[0022] It should be noted that in the process of controlling the robot's motion motors using the sliding mode variable structure control method, a sliding surface is typically designed first based on the robot's motor system to achieve the desired dynamic performance on the sliding surface. Then, a corresponding control law is constructed based on the sliding surface to ensure that the control system's state point can reach and remain on the sliding surface within a finite time. The designed control law is then applied to the robot's motor control system, adjusting the control signal in real time to cope with changes in system parameters and external disturbances. However, during robot movement, changes in the external environment affect the motor load, which intensifies the system's back-and-forth movement near the sliding surface. This leads to reduced accuracy in motion control and jitter, specifically manifested in the system's state point moving back and forth near the sliding surface, resulting in chattering. Figure 2 The diagram shows the deviation curve between the state point and the sliding surface, where the horizontal axis represents time and the vertical axis represents the deviation value. The curve shows that the deviation value fluctuates in the early stages, resulting in chattering. Therefore, to further reduce chattering during sliding mode variable structure control of the robot's motion motors, this embodiment of the invention utilizes the working environment and parameters of the robot's motion motors to enable the controller parameters to adaptively adjust to external disturbances, thereby improving the motion control effect of the robot.

[0023] Specifically, in order to implement the adaptive sliding mode variable structure control method for robots proposed in this embodiment, it is first necessary to collect the working environment data and working parameter data of the robot's motion motors. The specific process is as follows: First, several types of sensors are used to acquire sensor data of the robot's motion motors in real time, and the sensor data is divided into working environment data and working parameter data.

[0024] The aforementioned sensors include: vibration sensors, electromagnetic field strength sensors, current sensors, voltage sensors, and torque sensors.

[0025] The working environment data includes: vibration data and electromagnetic field strength data.

[0026] The operating parameter data includes: current data, voltage data, and torque data.

[0027] It should be noted that in this embodiment of the invention, the data acquisition frequency of the sensor is set to 200Hz. In other embodiments, it can be adjusted according to the actual situation. This embodiment of the invention does not impose any specific limitations.

[0028] Then, the acquired working environment data and working parameter data are standardized and filtered using a Gaussian filter.

[0029] Finally, using the method provided in "Torque Motor Terminal Sliding Mode Tracking Control Based on Approach Law, Journal of Huazhong University of Science and Technology (Natural Science Edition), Wang Yan, Lu Yiyun, Ge Yunwang", the trajectory tracking controller of the robot's motion motor is obtained and the parameters in the trajectory tracking controller are preset. and The initial value of the parameter and Let them be denoted as the first approaching factor and the second approaching factor. The first approaching factor and the second approaching factor are collectively referred to as approaching factors. The approaching law term corresponding to the first approaching factor in the trajectory tracking controller is denoted as the first approaching term, and the approaching law term corresponding to the second approaching factor in the trajectory tracking controller is denoted as the second approaching term.

[0030] It should be noted that, in the embodiments of this invention, the first approach factor is based on the "Torque Motor Terminal Sliding Mode Trajectory Tracking Control Based on Approach Law, Journal of Huazhong University of Science and Technology (Natural Science Edition), Wang Yan, Lu Yiyun, Ge Yunwang". Second convergent factor The value of is preset with the first convergence factor. Second convergent factor The initial values ​​are 5 and 20 respectively, that is , The embodiments of the present invention are not specifically limited and can be adjusted according to the actual situation.

[0031] It should also be noted that the motion motor described in this embodiment of the invention is a DC torque motor.

[0032] Thus, the working environment data and working parameter data of the robot's motion motor are obtained through the above method.

[0033] Step S002: Perform time-series decomposition on the sensor data of the motion motor, and extract features by combining the data change characteristics in the decomposition results to obtain the feature group of the sensor data.

[0034] It should be noted that during the robot's movement driven by the motion motors, the trajectory tracking controller analyzes the motor's trajectory to control the motion motors. To reduce chattering when the control system's state point approaches the sliding surface, the paper "Torque Motor Terminal Sliding Trajectory Tracking Control Based on the Approach Law, Journal of Huazhong University of Science and Technology (Natural Science Edition), Wang Yan, Lu Yiyun, Ge Yunwang" proposes an improvement based on a combination of exponential and power-law approach laws. This allows the control system's state point to reach the sliding surface quickly and smoothly. However, due to changes in the motor's working environment and the resulting workload during actual robot movement, the actual control effect on the motor cannot be fully realized. Sufficient stability and accuracy are required. Therefore, in the process of motion control of the robot, the actual working environment of the motor should also be analyzed. This allows the controller parameters to adaptively adjust to external disturbances during the control of the motor using the trajectory tracking controller. In addition, since any mechanical system has specific natural frequencies and mode shapes during operation, which constitute the physical modes of the system, the working environment of the motor has certain modal characteristics during the robot's perception and interaction with the outside world. It is these modal characteristics that affect the working condition of the motor. Therefore, this embodiment of the invention selects to perform modal analysis on the acquired working environment data and working parameter data of the motion motor to determine the corresponding modal characteristics.

[0035] Specifically, in step S201, modal decomposition is performed on the sensor data of the motion motor to obtain several IMF component data corresponding to the sensor data.

[0036] It should be noted that robots interact with their environment using specific interaction modes. For example, when a robot moves a heavy object, it exhibits a low-frequency, large-amplitude load change mode; when crossing small obstacles, it exhibits a medium-frequency, short-pulse impact mode; and when moving on a soft surface, it exhibits a high-frequency, damped oscillation adaptation mode. These interaction modes are transmitted to the motor through the mechanical transmission system, forming specific current and voltage waveform characteristics and affecting the motor's torque, thereby influencing the chattering degree of sliding mode control. It is evident that during the process of a robot system being driven by a motion motor, multiple physical fields from both mechanical and electrical aspects are involved, forming multi-physical field modes. That is, mechanical vibration leads to changes in electrical load, which further changes mechanical performance, resulting in changes in vibration characteristics. It is this coupling that forms a complex feedback loop, generating specific composite modal characteristics. Furthermore, it is these modal characteristics that determine the robot's behavior pattern in a specific environment. This is also the fundamental reason why this embodiment of the invention chooses to perform modal analysis on the working environment data and working parameter data of the motion motor to facilitate the subsequent implementation of adaptive sliding mode variable structure control.

[0037] As a preferred embodiment, the method for performing modal decomposition on the sensor data of the motion motor to obtain several modal component data corresponding to the sensor data includes: First, the EEMD signal decomposition algorithm is used to perform mode decomposition on arbitrary sensor data, and the quantity control function is determined during the mode decomposition process.

[0038] It should be noted that the EEMD (Ensemble Empirical Mode Decomposition) signal decomposition algorithm is an existing empirical mode decomposition algorithm, and therefore will not be described in detail in this embodiment of the invention.

[0039] As a preferred embodiment, the method for obtaining the quantity control function is as follows: For any sensor data, the preset number of noise additions in the EEMD signal decomposition algorithm is... All noise-adding processes and the EMD decomposition process of the sensor data after noise addition are processed in parallel. Based on the fluctuation characteristics of the IMF component data decomposed in each parallel processing, the quantitative control function under the mode decomposition process of sensor data is obtained.

[0040] It should be noted that, in the embodiments of the present invention, the number of noise additions K in the EEMD signal decomposition algorithm is preset to 20 based on experience, which can be adjusted according to the actual situation. The embodiments of the present invention do not impose specific limitations.

[0041] As an optional embodiment, the specific method for obtaining the quantity control function includes: denoting any sensor data as target sensor data; obtaining K white noise sequences with a mean of 0 and different standard deviations; adding the white noise sequences to the target sensor data respectively to obtain K noisy target sensor data; for any noisy target sensor data, performing iterative EMD decomposition on the noisy target sensor data; obtaining IMF component data of the corresponding order in each decomposition; and constructing a quantity control function of the corresponding order based on the fluctuation degree and similarity of data points in the IMF component data of each order during the iterative EMD decomposition of all noisy target sensor data of the target sensor data.

[0042] As an optional embodiment, the acquisition The method for generating white noise sequences with a mean of 0 and different standard deviations includes: obtaining the standard deviation of all extreme points in the target sensor data, and then... As the range of standard deviation values ​​for the K white noise sequences, within the range of standard deviation values, As the initial value, every We take values ​​and use them as the standard deviation of all elements in each white noise sequence to obtain... There are several white noise sequences with a mean of 0 and different standard deviations; among them The standard deviation of all extreme points in the target sensor data. These are preset range parameters.

[0043] It should be noted that when the EEMD signal decomposition algorithm decomposes data, the purpose of adding white noise to the sensor data is to eliminate the influence of randomness by adding noise multiple times, thus avoiding mode aliasing in the subsequent IMF component data. However, if the amplitude of the added white noise is too large, it will overwhelm the sensor data; if it is too small, it will not suppress mode aliasing. Therefore, this embodiment of the invention selects the standard deviation of the extreme points in the target sensor data as the range of standard deviation values ​​for the white noise sequence when adding white noise to the target sensor data. This ensures that the IMF component data obtained after subsequent decomposition can effectively contain the modal information of the robot's motion motors in the corresponding working environment or working parameters. This allows for subsequent feature analysis using this modal information to further adjust the relevant parameters in the trajectory tracking controller of the motion motors, reduce the chattering phenomenon of the control system's state points on the sliding surface, and improve the control accuracy of the motion motors. In addition, in this embodiment of the invention, a range parameter is preset based on experience. The value is 1 / 2, which can be adjusted according to the actual situation. This embodiment of the invention does not impose specific limitations.

[0044] As an optional embodiment, the specific calculation method of the quantity control function is as follows: ;in, The first part representing the target sensor data The value of the quantity control function for the IMF component data of the order; The first part representing the target sensor data The first noisy sensor data Standard deviation of IMF component data of order; The first part representing the target sensor data The first noisy sensor data The mean of the normalized cross-correlation coefficients between the order IMF component data and other corresponding IMFs of order K-1; This indicates the amount of sensor data after adding noise to the target sensor data.

[0045] It should be noted that the quantitative adjustment function value reflects the degree to which the IMF component data of a corresponding order contains modal feature information. The smaller the value of the quantitative adjustment function, the less modal feature information the IMF component data of the corresponding order contains, and the higher the probability of stopping further EMD decomposition at that order. Used to measure the The degree of fluctuation of the IMF under different noise disturbances; This is used to reflect the similarity of corresponding modal components under different noise environments; in addition, since the noise sequence is used to add white noise to the sensor data, the number of noise sequences is equal to the number of sensor data obtained after adding noise to any sensor data, that is... and They are equal in numerical value.

[0046] It should be noted that, since the sensor data of the motion motor consists of different aspects, namely the mechanical level and the electrical level, corresponding to working environment data and working parameter data respectively, and since the dynamic characteristics of the system reflected by different physical quantities are fundamentally different, and the correlation between different modal components and the control target is different, the order of the IMF component data of the sensor data that needs to be acquired when using the modal characteristics of different sensor data for subsequent adjustment of the trend factor is different. In addition, in the process of modal decomposition of data by the EEMD signal decomposition algorithm, the number of IMF component data obtained is preset by humans, so it may not effectively reflect modal information in some modal components, thus resulting in a certain amount of information redundancy. Therefore, in this embodiment of the invention, a quantity control function is constructed to control the decomposition process of the EEMD signal decomposition algorithm. By determining the quantity control function value of the IMF component data at different orders of sensor data, the computational load of acquiring IMF component data and the accuracy of subsequent adjustment of the trend factor are balanced, so that the system can more accurately identify the modal characteristics of the current working environment and then adaptively adjust the sliding mode control parameters (i.e., the first trend factor). Second trend factor This effectively suppresses chattering while maintaining system stability, thereby improving the accuracy and adaptability of robot motion control.

[0047] Then, based on the numerical change trend of the quantity control function, several IMF component data corresponding to the sensor data are obtained.

[0048] As an optional embodiment, the specific method for obtaining several modal component data corresponding to the working environment data and working parameter data based on the quantity control function includes: constructing a two-dimensional rectangular coordinate system, using the value corresponding to the quantity control function as the vertical axis of the two-dimensional rectangular coordinate system, using the order of the IMF component data as the horizontal axis of the two-dimensional rectangular coordinate system, mapping the quantity control function values ​​of the IMF component data at each order of the target sensor data to the two-dimensional rectangular coordinate system to form a corresponding scatter plot, denoted as the quantity control function scatter plot of the target sensor data; performing curve fitting on the quantity control function scatter plot of the target sensor data using the least squares method, using the elbow method to obtain the inflection point in the curve fitting result of the quantity control function scatter plot, and using the order closest to the inflection point as the order of the IMF component data of the target sensor data, denoted as the target order of the target sensor data, thereby obtaining the IMF component data of the target sensor data through the EEMD signal decomposition algorithm, and the order of the IMF component data of the target sensor data is the corresponding target order.

[0049] Step S202: Based on the modal features corresponding to the data changes in all IMF component data corresponding to the sensor data, extract the feature groups of the sensor data.

[0050] As an optional embodiment, the specific method for obtaining the feature group is as follows: for each IMF component data of the target sensor data, the instantaneous frequency sequence and amplitude envelope sequence of the IMF component data are calculated using the Hilbert-Huang transform; the average frequency, frequency bandwidth, and energy ratio of the IMF component data are calculated based on the instantaneous frequency sequence and amplitude envelope sequence; and the array formed by the average frequency, frequency bandwidth, and energy ratio of the IMF component data is used as the feature group of the IMF component data, thereby obtaining the feature group of all IMF component data of the target sensor data.

[0051] Thus, the feature set of sensor data is obtained through the above method.

[0052] Step S003: Combine the characteristic groups of sensor data and the chattering situation of the state points of the control system on the sliding surface to determine the chattering influence coefficient of each sensor data. Based on the chattering influence coefficient of all sensor data, obtain the adjustment coefficient of the approach factor in the trajectory tracking controller.

[0053] It should be noted that, by analyzing the modal characteristics of the working environment data and working parameter data reflected in the working environment data due to the influence of the working environment during the process of the motion motor driving the robot, it is necessary to further determine the impact of different modal characteristics on system chattering. In this way, the adjustment degree of the first and second approach factors in the trajectory tracking controller can be constructed. This adjustment is made to adjust the specific degree of dominance of the first and second approach terms in the trajectory tracking controller when the system approaches or moves away from the sliding surface, thereby further reducing the chattering phenomenon that occurs in the actual motion control process and improving the motion driving effect of the motion motor on the robot.

[0054] Specifically, in step S301, the chattering influence coefficient of each sensor data is determined by combining the characteristic group of the sensor data and the chattering situation of the state point of the control system on the sliding surface.

[0055] As an optional embodiment, the specific method for obtaining the chattering influence coefficient is as follows: real-time acquisition of sliding surface deviation, construction of a window at the current moment and taking the current moment as the rightmost moment of the window, and accumulation of the absolute values ​​of the sliding surface deviation at all moments within the window as the chattering index at the current moment; taking the IMF component data with the largest energy proportion in the feature group corresponding to the IMF component data of the target sensor data as the main IMF component data of the target sensor data, acquiring the feature group of continuously acquired main IMF component data, and calculating the mean absolute value of the cross-correlation coefficients between all elements in the feature group and the chattering index as the chattering influence coefficient of the target sensor data.

[0056] It should be noted that, in this embodiment of the invention, the window length is preset to 100 based on experience, but it can be adjusted in other embodiments. This embodiment of the invention does not impose any specific limitations.

[0057] It should be further noted that the chatter index directly reflects the chatter intensity. The larger the value of the chatter index, the more frequently the system crosses near the sliding surface, and the more severe the control signal jitter. In addition, in this embodiment of the invention, since the energy proportion best characterizes the modal influence, the feature group of the IMF component data with the highest energy proportion in the sensor data is selected as the input feature. The cross-correlation quantifies the synchronicity between modal energy change and chatter intensity. The larger the absolute value of the cross-correlation coefficient, the more significant the influence of the modal characteristics of the sensor data on chatter.

[0058] It should be noted that the chattering influence coefficient reflects the coupling strength between modal characteristics and system dynamics. That is, when the working environment excites a specific mode (such as the interaction between the robot and the ground during the robot's movement, causing high-frequency vibration of the motion motor), the chattering influence coefficient automatically increases, triggering the adaptive adjustment of the subsequent approach factor, thereby specifically suppressing the chattering caused by that mode.

[0059] Step S302: Based on the jitter impact coefficient of all sensor data, obtain the adjustment coefficient of the approach factor in the trajectory tracking controller.

[0060] First, a reference jitter level is preset, the ratio of the reference jitter level to the jitter index at the current moment is obtained, and the ratio is input into the sigmoid function for normalization. The normalization result is used as the jitter deviation value at the current moment.

[0061] It should be noted that, in this embodiment of the invention, a reference chatter level is calibrated and preset through an interference-free experiment. In other embodiments, adjustments can be made based on actual circumstances; if the jitter index at the current moment... If the value is 0, it indicates that the chattering is slight, and the approach factor can be increased to improve the response speed; while when When the threshold is reached, it indicates severe chattering, requiring a reduction in the convergence factor to suppress chattering. This represents the jitter index at the current moment.

[0062] Then, by combining the jitter impact coefficient and the jitter deviation value at the current moment, an adjustment coefficient is calculated. The adjustment coefficient is positively correlated with both the jitter impact data and the jitter deviation value.

[0063] As an optional embodiment, the specific method for calculating the adjustment coefficient is as follows: in , Indicates the current time period of the first... One adjustment coefficient; This represents the chattering deviation value at the current moment. hour, and ;when hour, and ;when hour, ; Indicates the first The jitter impact coefficient of individual sensor data; This indicates the amount of sensor data.

[0064] It should be noted that, Provides adjustment direction and range; The adjustment coefficient represents the sensitivity to chattering, reflecting the influence of environmental modes. A larger adjustment coefficient increases the approach factor, resulting in a faster system response, which is suitable for stable environments. Conversely, a smaller adjustment coefficient decreases the approach factor, weakening abrupt changes in the control signal, which is suitable for environments with severe chattering.

[0065] Thus, the adjustment coefficient is obtained through the above method.

[0066] Step S004: Adjust the approach factor in the trajectory tracking controller using the adjustment coefficient, and control the robot's motion motors through the trajectory tracking controller.

[0067] It should be noted that this step applies the adjustment coefficient to the trajectory tracking controller, updates the convergence factor in real time, and forms an environment-adaptive sliding mode control law. Its core is to suppress sliding surface chattering by dynamically adjusting the dominance of the first and second convergence terms, while maintaining system stability. Compared to traditional fixed-parameter sliding mode control, this embodiment of the invention can automatically switch control strategies when the environment changes abruptly. When a high-frequency adaptive mode is detected, the first convergence term is weakened to smooth the control signal; when the system is stable, the second convergence term is strengthened to improve the response speed.

[0068] Specifically, first, obtain the adjustment coefficient at the current moment. and The approach factor in the trajectory tracking controller is updated, and the updated approach factor is substituted into the trajectory tracking controller to obtain the control signal.

[0069] As an optional embodiment, the specific calculation method for the approach factor in the updated trajectory tracking controller is as follows: ,in and These are the preset initial values ​​corresponding to the first and second convergence factors.

[0070] It should be noted that the update of the convergence factor is directly related to the environmental mode; for example, when hour, This reduces the intensity of the first approach term and decreases high-frequency chattering.

[0071] Then, the control signal is input into the driver of the motion motor. The driver adjusts the voltage of the DC torque motor according to the control signal, driving the robot to execute the desired motion trajectory.

[0072] Finally, steps S001 to S004 are repeated with a preset control cycle to form a closed-loop adaptive control.

[0073] It should be noted that, in this embodiment of the invention, the driver for the motion motor is a PWM speed control module. The closed-loop mechanism ensures that the controller continuously tracks environmental changes. For example, when the robot moves from hard ground to sand, high-frequency vibration modes are identified. Automatic reduction and smoothing of control signals prevent jitter in motion. In addition, in this embodiment of the invention, the control cycle is preset to 5ms based on experience, but it can be adjusted according to the actual situation. This embodiment of the invention does not impose specific limitations.

[0074] The above steps complete the control process of the robot's motion motors using the adaptive sliding mode variable structure control method.

[0075] Please see Figure 3 The illustration shows an adaptive sliding mode variable structure control system for a robot provided by an embodiment of the present invention, including a memory 302, a processor 301, and a computer program 3021 stored in the memory 302 and executable on the processor. When the processor 301 executes the computer program 3021, it implements steps S001 to S004 of the adaptive sliding mode variable structure control method for a robot.

[0076] Furthermore, in an optional embodiment, the memory 302 described above may include read-only memory and random access memory, and provide instructions and data to the processor. The memory 302 may also include non-volatile random access memory. For example, the memory may also store device type information.

[0077] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An adaptive sliding mode variable structure control method for robots, characterized in that, The method includes the following steps: The sensor data of the robot's motion motors and the trajectory tracking controller of the corresponding control system are acquired, and the trajectory tracking controller contains a convergence factor; The sensor data of the motion motor is decomposed in time series, and feature extraction is performed based on the data change characteristics in the decomposition results to obtain the feature group of the sensor data. By combining the characteristic groups of sensor data and the chattering of the state points of the control system on the sliding surface, the chattering influence coefficient of each sensor data is determined. Based on the chattering influence coefficient of all sensor data, the adjustment coefficient of the approach factor in the trajectory tracking controller is obtained. The approach factor in the trajectory tracking controller is adjusted using an adjustment coefficient, and the motion motor of the robot is controlled by the trajectory tracking controller.

2. The adaptive sliding mode variable structure control method for robots according to claim 1, characterized in that, The specific method for performing time-series decomposition of the sensor data of the motion motor includes: The EEMD signal decomposition algorithm is used to perform mode decomposition on arbitrary sensor data, and the quantity control function is determined during the mode decomposition process. Based on the numerical change trend of the quantity control function, several IMF component data corresponding to the sensor data are obtained.

3. The adaptive sliding mode variable structure control method for robots according to claim 2, characterized in that, The method for obtaining the quantity control function is as follows: For any sensor data, the number of noise additions K of the EEMD signal decomposition algorithm is preset. All noise addition processes and the EMD decomposition process of the sensor data after noise addition are processed in parallel. Based on the fluctuation characteristics of the IMF component data decomposed in each parallel process, the quantity control function under the mode decomposition process of sensor data is obtained.

4. The adaptive sliding mode variable structure control method for robots according to claim 3, characterized in that, The specific method for obtaining the quantity control function includes: Any sensor data is denoted as target sensor data. K white noise sequences with a mean of 0 and different standard deviations are obtained and added to the target sensor data respectively to obtain K noisy target sensor data. For any noisy target sensor data, iterative EMD decomposition is performed on the noisy target sensor data. Each decomposition yields IMF component data of the corresponding order. During the iterative EMD decomposition of all noisy target sensor data, the data points in the corresponding IMF component data at each order are analyzed for fluctuation and similarity to construct the quantity control function at the corresponding order.

5. The adaptive sliding mode variable structure control method for robots according to claim 4, characterized in that, The specific method for obtaining K white noise sequences with a mean of 0 and different standard deviations is as follows: Obtain the standard deviation of all extreme points in the target sensor data, and then... As the range of standard deviation values ​​for the K white noise sequences, within the range of standard deviation values, As the initial value, every We take values ​​to represent the standard deviation of all elements in each white noise sequence, thus obtaining K white noise sequences with a mean of 0 and different standard deviations; where The standard deviation of all extreme points in the target sensor data. These are preset range parameters.

6. The adaptive sliding mode variable structure control method for robots according to claim 4, characterized in that, The specific method for obtaining several IMF component data corresponding to the target sensor data based on the numerical change trend of the quantity control function includes: A two-dimensional rectangular coordinate system is constructed, with the numerical value corresponding to the quantity control function as the vertical axis and the order of the IMF component data as the horizontal axis. The quantity control function values ​​of the IMF component data at each order of the target sensor data are mapped to the two-dimensional rectangular coordinate system to form a corresponding scatter plot, denoted as the quantity control function scatter plot of the target sensor data. The scatter plot of the quantity control function of the target sensor data is curve-fitted using the least squares method. The inflection point in the curve-fitting result of the quantity control function scatter plot is obtained using the elbow method. The order closest to the inflection point is taken as the order of the IMF component data of the target sensor data, denoted as the target order of the target sensor data. Thus, the IMF component data of the target sensor data is obtained through the EEMD signal decomposition algorithm, and the order of the IMF component data of the target sensor data is the corresponding target order.

7. The adaptive sliding mode variable structure control method for robots according to claim 6, characterized in that, The specific method for obtaining the feature groups of the sensor data is as follows: For each IMF component data of the target sensor data, the instantaneous frequency sequence and amplitude envelope sequence of the IMF component data are calculated using the Hilbert-Huang transform. Based on the instantaneous frequency sequence and amplitude envelope sequence, the average frequency, frequency bandwidth and energy ratio of the IMF component data are calculated. The array formed by the average frequency, frequency bandwidth and energy ratio of the IMF component data is used as the feature group of the IMF component data. The feature group of all IMF component data of the target sensor data is obtained.

8. The adaptive sliding mode variable structure control method for robots according to claim 1, characterized in that, The method for determining the chattering influence coefficient of each sensor data point by combining the feature group of sensor data and the state point of the control system on the sliding surface includes: The sliding surface deviation is acquired in real time, a window is constructed for the current moment, and the current moment is taken as the rightmost moment of the window. The absolute values ​​of the sliding surface deviations at all moments within the window are accumulated and used as the chattering index for the current moment. The IMF component data with the largest energy proportion in the feature group corresponding to the IMF component data of the target sensor data is taken as the main IMF component data of the target sensor data. The feature group of continuously acquired main IMF component data is obtained, and the absolute mean of the cross-correlation coefficients between all elements in the feature group and the chattering index is calculated and used as the chattering influence coefficient of the target sensor data.

9. The adaptive sliding mode variable structure control method for robots according to claim 8, characterized in that, The method for obtaining the adjustment coefficient of the approach factor in the trajectory tracking controller based on the jitter influence coefficient of all sensor data includes: A preset reference jitter level is set, the ratio of the reference jitter level to the jitter index at the current moment is obtained, and the ratio is input into the sigmoid function for normalization. The normalization result is used as the jitter deviation value at the current moment. An adjustment coefficient is calculated by combining the vibration impact coefficient and the vibration deviation value at the current moment. The adjustment coefficient is positively correlated with both the vibration impact data and the vibration deviation value.

10. An adaptive sliding mode variable structure control system for a robot, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of an adaptive sliding mode variable structure control method for a robot as described in any one of claims 1 to 9.

Images