Automation equipment control method and system based on low-power joystick

Abstract

The present application relates to the technical field of equipment control, and discloses an automated equipment control method and system based on a low-power joystick, the method comprising: respectively collecting left-right translation signals and front-back translation signals of the upper and lower rocker arms of the joystick, and converting them into a first electrical signal and a second electrical signal respectively; performing analog-to-digital conversion to obtain a digital signal matrix; performing feature extraction to obtain motion feature data; constructing a multi-dimensional extended state space model to obtain control parameters; performing power mode division, switching between full power mode, energy-saving mode and sleep mode when the displacement meets different threshold conditions, and obtaining a target control instruction; mapping the target control instruction to the control amount of each execution device through a set proportional coefficient and offset, and adjusting the control parameters based on device state feedback information to obtain a multi-device collaborative control result, thereby improving the anti-interference capability and signal processing efficiency of the low-power joystick.

Description

Technical Field

[0001] The present application relates to the field of equipment control technology, and in particular to an automated equipment control method and system based on a low-power joystick. Background Art

[0002] Traditional joystick controllers mainly use contact detection methods of brushes and carbon resistors, and control is achieved by moving the brushes on the carbon resistors to generate different resistance values. However, this method has obvious defects: when using high-hardness carbon resistors, the brushes wear quickly and generate large electrical noise; when using low-hardness carbon resistors, carbon powder easily falls off and adheres to the brushes, affecting electrical performance.

[0003] Current joystick control systems generally have problems with high power consumption and short life. Due to continuous physical contact friction, the brushes and carbon resistance will accelerate wear, resulting in reduced system reliability. At the same time, the existing technology often ignores the coupling relationship between motions when processing the multi-dimensional motion data of the joystick, fails to fully utilize the motion feature information, and affects the control accuracy. In addition, the existing joystick controller lacks effective low-power strategies and device coordination mechanisms in multi-device collaborative control scenarios. The system cannot dynamically adjust the working mode according to the actual usage status, resulting in energy waste; at the same time, when controlling multiple devices, there is a lack of effective collaborative control mechanism between the execution devices, making it difficult to ensure the overall performance of the system. Summary of the invention

[0004] The present application provides a method and system for controlling an automated device based on a low-power joystick, thereby improving the anti-interference capability and signal processing efficiency of the low-power joystick.

[0005] In a first aspect, the present application provides an automation device control method based on a low-power joystick, and the automation device control method based on a low-power joystick includes:

[0006] The left-right translation signal and the front-back translation signal are respectively collected from the upper rocker arm and the lower rocker arm of the rocker, and converted into a first electrical signal and a second electrical signal respectively;

[0007] Performing analog-to-digital conversion and low-pass filtering on the first electrical signal and the second electrical signal to obtain a digital signal matrix;

[0008] Inputting the digital signal matrix into a parallel multi-dimensional joystick attention network for feature extraction, and dynamically adjusting the sampling frequency according to the correlation coefficient of the two-directional motion to obtain motion feature data;

[0009] According to the motion characteristic data, a multi-dimensional extended state space model is constructed using an extended state observer, and a state feedback matrix is ​​updated through online parameter identification to obtain control parameters;

[0010] The control parameters are divided into power modes according to set thresholds, and when the displacement meets different threshold conditions, the full power mode, energy saving mode and sleep mode are switched to obtain target control instructions;

[0011] The target control instruction is mapped to the control quantity of each execution device through the set proportional coefficient and offset, and the control parameters are adjusted based on the device state feedback information to obtain the multi-device collaborative control result.

[0012] The second aspect of the present application provides an automation equipment control system based on a low-power joystick, and the automation equipment control system based on a low-power joystick includes:

[0013] The acquisition module is used to respectively acquire left-right translation signals and front-back translation signals of the upper rocker arm and the lower rocker arm of the rocker, and convert them into first electrical signals and second electrical signals respectively;

[0014] a conversion module, configured to perform analog-to-digital conversion and low-pass filtering on the first electrical signal and the second electrical signal to obtain a digital signal matrix;

[0015] An extraction module, used for inputting the digital signal matrix into a parallel multi-dimensional joystick attention network for feature extraction, and dynamically adjusting the sampling frequency according to the correlation coefficient of the two-directional motion to obtain motion feature data;

[0016] A construction module is used to construct a multi-dimensional extended state space model using an extended state observer according to the motion feature data, and to update a state feedback matrix through online parameter identification to obtain control parameters;

[0017] A switching module, used to divide the control parameters into power modes according to a set threshold value, and switch between full power mode, energy saving mode and sleep mode when the displacement meets different threshold conditions to obtain a target control instruction;

[0018] The adjustment module is used to map the target control instruction to the control quantity of each execution device through the set proportional coefficient and offset, and adjust the control parameters based on the device state feedback information to obtain the multi-device collaborative control result.

[0019] Compared with the prior art, the present application has the following beneficial effects: the use of magnetic induction detection to replace the traditional brush carbon resistance structure eliminates physical contact friction, significantly prolongs the service life of the system, and improves the detection accuracy. By introducing a parallel multi-dimensional rocker attention network, the multi-dimensional extraction and analysis of the rocker motion characteristics are realized, the coupling relationship between the motion data is fully utilized, and the control accuracy is improved. A state space model based on an extended state observer is designed, and accurate observation of the system state and real-time optimization of control parameters are achieved through online parameter identification. An innovative three-level power mode division mechanism is proposed, which automatically switches the working mode according to the use status of the rocker, effectively reducing the overall power consumption of the system. A complete multi-device collaborative control framework is established, and precise control of multiple execution devices is achieved through the mapping mechanism of proportional coefficients and offsets. A dual-channel signal processing structure is adopted, combined with a dynamic sampling strategy, to improve the system's anti-interference ability and signal processing efficiency. By setting a hierarchical state feedback mechanism, an effective combination of fast dynamic response and slow steady-state response is achieved, which improves the dynamic performance of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0021] The structures, proportions, sizes, etc. illustrated in the drawings of this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with this technology. They are not used to limit the conditions under which the present invention can be implemented, and therefore have no substantive technical significance. Any structural modification, change in proportion or adjustment of size, without affecting the effects and purposes that can be achieved by the present invention, should still fall within the scope of the technical contents disclosed by the present invention.

[0022] Figure 1 is a flow chart of a method for controlling an automated device based on a low-power joystick provided by an embodiment of the present invention;

[0023] Figure 2 It is a schematic block diagram of the structure of an automation equipment control system based on a low-power joystick provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0024] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0025] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may also be decomposed, combined or partially merged, so the actual execution order may change according to actual conditions.

[0026] It should also be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0027] It should be further understood that the term "and / or" used in the specification and appended claims of this application refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations. Figure 1 In the embodiment of the present application, an embodiment of the automation equipment control method based on the low power consumption joystick includes:

[0028] Step 100: respectively collecting left-right translation signals and front-back translation signals from the upper rocker arm and the lower rocker arm of the rocker, and converting them into first electrical signals and second electrical signals respectively;

[0029] It is understandable that the execution subject of the present application can be an automation equipment control system based on a low-power joystick, or a terminal or a server, which is not limited here. The present application embodiment is described by taking a server as the execution subject as an example.

[0030] Specifically, the position of the slide bar in the guide groove set at one end of the upper rocker arm is collected, and the slide bar is driven to translate left and right in the slide groove of the first fixed part by the connecting rod, so as to obtain the translation displacement of the first movable part. The cooperation between the guide groove and the slide bar enables the movement of the rocker arm to be smoothly converted into the linear movement of the slide bar. At the same time, a first magnetic block is arranged on the inner side of the first movable part, and the magnetic block is connected to the first translation magnet assembly. Through the synchronous translation of the first magnetic block, a translation signal in the left and right direction is obtained. When the upper rocker arm is displaced, the slide bar moves in the guide groove, driving the connecting rod structure to further push the slide bar. During the process of the slide bar moving left and right in the slide groove of the first fixed part, the first magnetic block connected thereto is directly pushed to move synchronously. Since the position change of the first magnetic block directly reflects the left and right movement state of the upper rocker arm, the translation signal in the left and right direction is obtained by detecting the translation movement of the first magnetic block. Similarly, in the implementation of the lower rocker arm, the position of the slide bar in the guide groove set at one end of the lower rocker arm is collected, and the slide bar is driven to translate back and forth in the slide groove of the second fixed part by the connecting rod, so as to obtain the translation displacement of the second movable part. Through this structure, the movement of the lower rocker arm can be accurately converted into the linear motion of the slider in the second fixed part. A second magnetic block is installed on the inner side of the second movable part, and the magnetic block is connected to the second translation magnet assembly. Similarly, the translation signal in the front-to-back direction is obtained through the synchronous translation of the second magnetic block. The lower rocker arm can drive the connecting rod, the slider and the second magnetic block connected thereto to move synchronously through the positional relationship between the guide groove and the slide bar, so the displacement of the second magnetic block has a direct corresponding relationship with the front-to-back movement of the lower rocker arm. For the left-right translation signal generated by the first magnetic block, the first magnetic induction element installed below the first translation magnet assembly is sampled to obtain a first electrical signal corresponding to the left-right translation. When the first magnetic block moves, the magnetic induction element senses the change in magnetic field strength and converts it into a corresponding electrical signal. The electrical signal directly reflects the left-right translation amount of the upper rocker arm and is used for subsequent control processing. Similarly, for the front-to-back translation signal generated by the second magnetic block, the second magnetic induction element installed below the second translation magnet assembly is sampled to obtain a second electrical signal corresponding to the front-to-back translation. The second magnetic sensing element senses the change in the magnetic field of the second magnetic block when it moves forward and backward, and converts it into an electrical signal representing the displacement in the forward and backward directions, thereby providing an accurate electrical signal representation method for the motion state of the lower rocker arm.

[0031] Step 200, performing analog-to-digital conversion and low-pass filtering on the first electrical signal and the second electrical signal to obtain a digital signal matrix;

[0032] Specifically, the signal strength of the first electrical signal is detected, and the signal strength of the first signal is obtained in real time through a dedicated signal detection circuit. According to the first signal strength, the gain coefficient of the first signal processing channel is dynamically set to obtain a suitable first gain adjustment signal. The gain adjustment is to adjust the amplitude of the signal to a suitable range before entering the analog-to-digital conversion to improve the signal-to-noise ratio and reduce the conversion error. Similarly, the signal strength of the second electrical signal is detected to obtain the second signal strength. According to this strength, the gain coefficient of the second signal processing channel is dynamically set to obtain the second gain adjustment signal. By adjusting the gain coefficient, the changes in different signal strengths can be responded to in real time during the signal processing process, so that the system has better adaptability and robustness. The first gain adjustment signal is input into the analog-to-digital conversion unit of the first signal processing channel, and the first gain adjustment signal is quantized and encoded using a successive approximation analog-to-digital conversion circuit. Successive approximation analog-to-digital conversion is a low-power, high-precision analog-to-digital conversion method that converts an analog signal into a corresponding digital signal by gradually approaching the target voltage to obtain a first digitized signal. The second gain adjustment signal is input into the analog-to-digital conversion unit of the second signal processing channel, and quantized and encoded by a successive approximation analog-to-digital conversion circuit to obtain a second digitized signal. After the analog-to-digital conversion is completed, the first digitized signal is input into the digital filter of the first signal processing channel and processed by a weighted average algorithm. The weighted average algorithm effectively suppresses high-frequency noise by weighted averaging of multiple sampling points, and retains the main components of the signal to obtain a first filtered signal. Similarly, the second digitized signal is input into the digital filter of the second signal processing channel, and high-frequency noise is suppressed by a weighted average algorithm to obtain a second filtered signal. The first filtered signal and the second filtered signal are subjected to synchronous sampling and alignment processing to avoid errors caused by asynchronous sampling time. The first filtered signal and the second filtered signal are combined in data structure and integrated into a unified digital signal matrix.

[0033] Step 300: Input the digital signal matrix into the parallel multi-dimensional joystick attention network for feature extraction, and dynamically adjust the sampling frequency according to the correlation coefficient of the two-directional motion to obtain motion feature data;

[0034] It should be noted that the digital signal matrix is ​​decoupled through the feature separation layer. The feature separation layer consists of four convolutional layers, each of which is followed by a BatchNorm layer and a ReLU activation function to ensure the equalization and nonlinear activation of the data during the convolution process. Through the layer-by-layer convolution of these four layers, the different scale features of the signal are extracted, and the noise interference is effectively reduced to obtain the initial feature signal matrix. The introduction of the BatchNorm layer stabilizes the distribution of the data, and the ReLU activation function increases the nonlinear representation ability of the network. The initial feature signal matrix is ​​input into the parallel multi-dimensional rocker attention network for shunting processing to obtain the displacement feature vector, the velocity feature vector and the acceleration feature vector. The parallel multi-dimensional rocker attention network contains three parallel feature channels, namely the first feature channel, the second feature channel and the third feature channel, which independently process different motion features to ensure the independence and specificity of different features. The separate processing method of the displacement feature vector, the velocity feature vector and the acceleration feature vector enables each feature to be targetedly enhanced in the subsequent feature extraction and enhancement without being interfered by other features. The displacement feature vector is input into the attention network of the first feature channel, which consists of a multi-head self-attention mechanism and a feedforward neural network layer. The multi-head self-attention mechanism captures the intrinsic connection between features by calculating the correlation weights between features, and performs weighted synthesis of features according to the weights, which significantly enhances the representation ability of displacement features. The multi-head self-attention mechanism processes multiple attention heads at the same time, so that the model can focus on different aspects of features, improve the richness and robustness of feature extraction, and obtain enhanced displacement features. The feedforward neural network layer performs nonlinear transformation on the features processed by attention to improve the expressiveness of features. The speed feature vector is input into the temporal attention module in the second feature channel. The temporal attention module uses a sliding window mechanism to calculate local temporal dependencies. By moving the window in the time dimension of the features, the time series dependencies between features are captured, so that the speed feature can better reflect the dynamic characteristics that change over time, and the speed feature with temporal enhancement is obtained. Through the temporal enhancement operation, the control system can more accurately capture the short-term and medium-term speed change laws when analyzing speed changes, thereby improving control accuracy and response speed. The acceleration feature vector is input into the spatial attention module of the third feature channel. The spatial attention module calculates the importance of different spatial positions through adaptive weights, highlights the features of important positions, and suppresses the features of secondary positions. Acceleration features have different spatial distributions. The spatial attention module obtains spatially enhanced acceleration features by weighting the importance of different spatial positions. The enhanced displacement features, time-series enhanced velocity features, and spatially enhanced acceleration features are fused to obtain a fused feature vector.Based on the fused feature vector, the correlation coefficient matrix of the motion in two directions is calculated, and the adjustment factor of the sampling frequency is determined by the ratio of the diagonal elements to the off-diagonal elements. The calculation of the correlation coefficient matrix reveals the motion correlation between different directions. The diagonal elements represent the autocorrelation in each direction, while the off-diagonal elements represent the interaction between different directions. By comparing the diagonal elements and the off-diagonal elements, the independence and coupling degree of the current motion are obtained, and the sampling frequency adjustment factor is determined to dynamically adjust the sampling strategy. The fused feature vector is resampled according to the dynamic sampling strategy, and the sampling frequency is adjusted according to the correlation between the features, so that the sampling frequency is reduced in the case of high correlation to save resources, and the sampling frequency is increased in the case of low correlation to ensure accuracy. The motion feature data is mapped to the feature space through the fully connected layer.

[0035] Step 400: Based on the motion feature data, a multi-dimensional extended state space model is constructed using an extended state observer, and the state feedback matrix is ​​updated through online parameter identification to obtain control parameters;

[0036] Specifically, the motion feature data is input into the first feature separation module, which includes a displacement component extraction unit, a velocity component extraction unit and an acceleration component extraction unit. The first state equation, the second state equation and the third state equation are constructed by these units respectively to obtain a multi-dimensional extended state space model. Error calculation is performed on the displacement state quantity and the velocity state quantity to calibrate the deviation between the model and the actual state. For the displacement state quantity and the velocity state quantity in the multi-dimensional extended state space model, the difference between the actual measured value and the theoretical calculated value is calculated by the first error calculation unit and the second error calculation unit respectively to obtain the first state error matrix and the second state error matrix. The first state error matrix and the second state error matrix are input into the dual-channel Kalman state estimator, which consists of a prediction channel and an update channel. Through alternating iterations, the system state is subjected to multi-stage filtering processing to obtain the first optimal state sequence and the second optimal state sequence. The role of the prediction channel is to predict the current state based on the state of the previous moment, while the update channel corrects the predicted state according to the measurement error. Through multi-stage processing, the system state is estimated more accurately, thereby reducing the impact of noise on the system. An online identification operation is performed on the first optimal state sequence and the second optimal state sequence. Through a parallel parameter identification unit, the parallel parameter identification unit includes a first recursive least squares unit and a second recursive least squares unit, and the first parameter update law and the second parameter update law are respectively constructed to obtain a first model parameter matrix and a second model parameter matrix. The recursive least squares algorithm updates the system parameters in a continuous iterative manner, so that the model can adapt to the change of the state in real time and effectively improve the accuracy of the model. According to the first model parameter matrix and the second model parameter matrix, a dual state observer is constructed, and the dual state observer includes a main state observer and an auxiliary state observer. The gain coefficients of the two groups of observers are calculated by the pole configuration optimization algorithm to obtain the first observer matrix and the second observer matrix. The pole configuration algorithm configures the closed-loop poles of the system appropriately so that the dynamic response characteristics of the system meet the expected requirements. The first and second observer matrices can make the observer more accurately estimate the system state and reduce the observation error through reasonable pole configuration. The first observer matrix and the second observer matrix are respectively orthogonally matched with the corresponding optimal state sequence, and the observation state deviation is calculated by the first error correction unit and the second error correction unit to obtain the first observation error vector and the second observation error vector. Orthogonal matching is to ensure that the relationship between the adjustment direction of the observer gain and the error can minimize the error and more effectively correct the state estimation deviation of the system. The state feedback matrix is ​​updated hierarchically according to the first observation error vector and the second observation error vector. The hierarchical update includes fast dynamic response update and slow steady-state response update, the purpose of which is to respond quickly when the system changes dynamically and remain stable in steady state.The feedback gain coefficients at different time scales are optimized by an adaptive gradient algorithm to obtain a hierarchical state feedback matrix. The hierarchical state feedback matrix is ​​updated separately at different time scales, so that the system can maintain flexible response in dynamic changes and ensure the steady-state performance of the system in the stable stage. The current system state is matrix multiplied with the hierarchical state feedback matrix by the state feedback synthesizer. The state feedback synthesizer includes a dynamic weight allocation mechanism, which dynamically adjusts the feedback weights of different levels according to the response characteristics of the system to obtain the final control parameters.

[0037] Step 500, dividing the control parameters into power modes according to the set thresholds, switching between the full power mode, the energy saving mode and the sleep mode when the displacement meets different threshold conditions, and obtaining the target control instruction;

[0038] Specifically, the left and right displacements and the front and back displacements of the joystick in the control parameters are sampled and detected respectively, and data is collected for 50 consecutive sampling points by the first displacement sampling unit and the second displacement sampling unit to obtain the first displacement sequence and the second displacement sequence. The first displacement sequence and the second displacement sequence are subjected to signal energy analysis by the power calculation unit. The power calculation unit adopts the effective value calculation method of the magnetic induction signal, that is, the energy is calculated based on the effective value of these displacement signals to obtain the power value in the left and right direction and the power value in the front and back direction. The power value in the left and right direction and the power value in the front and back direction are compared with the first threshold value T1. If the power values ​​of 50 consecutive sampling points are all less than the first threshold value T1, the system determines that the current activity of the joystick is in a lower energy state, switches from the full power mode to the energy-saving mode, and obtains the first mode switching signal. The introduction of the energy-saving mode is intended to reduce unnecessary energy consumption. When it is detected that the joystick movement is in a lower power state, the power is automatically reduced to save energy. The power values ​​in the left and right directions and the power values ​​in the front and back directions are compared with the second threshold value T2. If the power values ​​of 200 consecutive sampling points are all less than the second threshold value T2, the system determines that the joystick is almost inactive, switches from the energy-saving mode to the sleep mode, and obtains the second mode switching signal. In the sleep mode, the system will significantly reduce power consumption and only maintain the most basic state monitoring. The first mode switching signal and the second mode switching signal are state synthesized, and the mode switching condition is judged by the first finite state machine to obtain the current working mode identifier. The finite state machine ensures that the system can accurately switch between different modes and maintain the consistency and stability of the working mode through clear logical states and conditional judgments. According to the current working mode identifier, the sampling frequency of the joystick is adjusted. In full power mode, the highest sampling frequency is maintained to ensure fast response and precise control; in energy-saving mode, the sampling frequency is reduced by 50% to maintain basic monitoring of the joystick movement and effectively reduce energy consumption; in sleep mode, the sampling frequency is reduced by 90%, only the most basic monitoring requirements are maintained, thereby maximizing the saving of power resources. Through adjustment, sampling resources are dynamically allocated according to the actual use status of the joystick, and corresponding sampling frequency control signals are generated. The control parameters are resampled according to the sampling frequency control signal to reduce the frequency of the signal and reduce the computational burden. The validity of the control parameters is determined by the second finite state machine to obtain effective control parameters. The effective control parameters are input into the mode controller, and the corresponding control quantities are generated by the control laws in different modes. In full power mode, energy saving mode and sleep mode, the controller decides how to generate the control quantity according to the respective control laws to ensure the operation effect of the system in each state. In order to improve the control effect, the generated control quantity is linearly compensated to reduce the deviation that may be caused by nonlinear factors and ensure the accuracy and reliability of the target control instructions. The target control instructions are obtained.

[0039] Step 600: Map the target control instruction to the control quantity of each execution device through the set proportional coefficient and offset, and adjust the control parameters based on the device state feedback information to obtain the multi-device collaborative control result.

[0040] Specifically, the left and right translation control components in the target control command are mapped with proportional coefficients and compensated with offsets, and the left and right displacement signals of the rocker are mapped into specific device control signals to obtain the left and right device control quantities. Similarly, by mapping the front and back displacement signals with proportional coefficients and compensating with offsets, a mapping relationship in the front and back directions is established to generate the device control quantities in the front and back directions. The function of the proportional coefficient mapping is to scale the input signal to an appropriate range, while the compensation of the offset is used to eliminate the static deviation in the system to ensure that the control quantity finally generated can accurately reflect the expected motion state of the rocker. The left and right device control quantities and the front and back device control quantities are processed, and the system groups them according to different device types, and matches them by looking up the corresponding relationship table between the rocker displacement and the device control quantity to obtain the initial device instruction. The corresponding relationship between the device control quantity and the rocker displacement is realized by table lookup. The purpose of the table lookup is to quickly obtain the initial device instruction that conforms to the current motion state, so that each execution device can obtain the instruction corresponding to the rocker action and realize preliminary synchronous control. The operating parameters of the execution equipment are monitored, including position data acquisition and speed data acquisition. Synchronous sampling is performed based on the sampling period of the magnetic induction signal to obtain the operating status of the equipment. These operating status data reflect the actual operating conditions of each execution equipment, and the deviation is calculated with the theoretical operating parameters. The deviation is dynamically compensated through the load characteristic curve of the equipment to obtain the state correction amount. The load characteristic curve of the equipment is used to describe the operating characteristics of the equipment under different load conditions. By comparing the actual state with the theoretical expectation, the deviation is calculated and dynamically compensated based on the load characteristics, and the control error caused by environmental changes or load changes is corrected to ensure the stability and consistency of the equipment operation. The state correction amount is used to establish a control parameter update formula according to the time series, and the control parameters are corrected by the cumulative error term to obtain the corrected control instructions. The timing analysis of the response of multiple devices is performed on the corrected control instructions. By establishing the action sequence constraint relationship between the devices, a collaborative control sequence is formed to ensure the coordination of the actions of multiple execution devices when executing tasks, and avoid control conflicts or failures caused by timing asynchrony. The collaborative control sequence is distributed to each execution device according to the set control cycle. The synchronization and coordination between multiple devices are ensured by real-time updating of control parameters, and the collaborative control results of multiple devices are obtained. The control cycle is set to ensure that each device can act according to the corresponding instructions at different time nodes. By updating the control parameters of each device in real time, the system can dynamically adapt to the control changes of the joystick and the adjustment of the device operating status, and realize accurate collaborative control of multiple devices.

[0041] In the embodiment of the present application, magnetic induction detection is used to replace the traditional brush carbon resistance structure, which eliminates physical contact friction, significantly extends the service life of the system, and improves the detection accuracy. By introducing a parallel multi-dimensional rocker attention network, multi-dimensional extraction and analysis of the rocker motion characteristics are realized, the coupling relationship between motion data is fully utilized, and the control accuracy is improved. A state space model based on an extended state observer is designed, and accurate observation of the system state and real-time optimization of control parameters are achieved through online parameter identification. An innovative three-level power mode division mechanism is proposed, which automatically switches the working mode according to the use status of the rocker, effectively reducing the overall power consumption of the system. A complete multi-device collaborative control framework is established, and precise control of multiple execution devices is achieved through the mapping mechanism of proportional coefficients and offsets. A dual-channel signal processing structure is adopted, combined with a dynamic sampling strategy, to improve the system's anti-interference ability and signal processing efficiency. By setting a hierarchical state feedback mechanism, an effective combination of fast dynamic response and slow steady-state response is achieved, which improves the dynamic performance of the system.

[0042] In a specific embodiment, the process of executing step 100 may specifically include the following steps:

[0043] The position of the slide bar in the guide groove provided at one end of the upper rocker arm is collected, and the slide bar is driven to translate left and right in the slide groove of the first fixed part through the connecting rod to obtain the translation displacement of the first movable part;

[0044] Connecting the first magnetic block on the inner side of the first movable part to the first translation magnet assembly, and obtaining a left-right translation signal through the synchronous translation of the first magnetic block;

[0045] The position of the slide bar in the guide groove provided at one end of the lower rocker arm is collected, and the slide bar is driven to translate forward and backward in the slide groove of the second fixed part through the connecting rod to obtain the translation displacement of the second movable part;

[0046] Connecting the second magnetic block on the inner side of the second movable part to the second translation magnet assembly, and obtaining a front-to-back translation signal through the synchronous translation of the second magnetic block;

[0047] The left-right translation signal generated by the first magnetic block is sampled by a first magnetic induction element installed below the first translation magnet assembly to obtain a first electrical signal corresponding to the left-right translation;

[0048] The forward and backward translation signal generated by the second magnetic block is sampled by a second magnetic induction element installed below the second translation magnet assembly to obtain a second electrical signal corresponding to the forward and backward translation.

[0049] Specifically, a guide groove is provided at one end of the upper rocker arm, and motion collection is realized by changing the position of the slide bar in the guide groove. When the user operates the rocker, the upper rocker arm drives the slide bar to move along the guide groove, and this motion is transmitted to the slider through the connected connecting rod, so that the slider translates left and right in the slide groove of the first fixed part. The motion state of the slider reflects the displacement of the upper rocker arm. Through this structural design, the rotational motion of the upper rocker arm is effectively converted into the linear motion of the slider, and the translation displacement of the first movable part is obtained. In order to further convert this mechanical motion into an electrical signal, a first magnetic block is provided on the inner side of the first movable part, and the magnetic block is connected to the first translation magnet assembly. When the slider moves left and right, the first magnetic block will also translate synchronously, generating a translation signal in the left and right direction. The signal represents the movement amplitude and direction of the upper rocker arm. The combination of the slider and the magnetic block ensures the synchronous movement of the two through a rigid or flexible connection method to avoid the deviation of motion transmission caused by looseness or other factors. A similar structure is applied to the lower rocker arm to capture its forward and backward movement. A guide groove is also provided at one end of the lower rocker arm, and displacement collection is realized by changing the position of the slide bar. The movement of the lower rocker arm drives the slide bar to move, and this movement is transmitted to the slider through the connecting rod, so that the slider moves forward and backward in the slide groove of the second fixed part. This movement is converted into the translation displacement of the second movable part, and the movement detection of the lower rocker arm is realized in this way. In order to convert this displacement into an electrical signal, a second magnetic block is arranged on the inner side of the second movable part, and the magnetic block is connected to the second translation magnet assembly. The movement of the slider drives the second magnetic block to move synchronously, generating a translation signal in the front and back direction. For the left and right translation signal of the first magnetic block, the first magnetic induction element installed under the first translation magnet assembly is sampled to obtain a first electrical signal corresponding to the left and right translation. Similarly, for the front and back translation signal of the second magnetic block, the second magnetic induction element installed under the second translation magnet assembly is sampled to obtain a second electrical signal corresponding to the front and back translation. The function of the magnetic induction element is to detect the change in magnetic field strength caused by the movement of the magnetic block, thereby generating an electrical signal corresponding to the displacement. This process is based on the Faraday electromagnetic induction principle. When the magnetic block moves, it changes the magnetic flux around the magnetic induction element, thereby generating a voltage in the induction element. The generation of the electrical signal is described by a formula. Assume that the electromotive force generated by the magnetic induction element is , the speed of the magnet during its movement is , the magnetic induction intensity is , the effective length of the sensing element is , then according to Faraday's law of electromagnetic induction, the electromotive force is expressed as:

[0050] ;

[0051] in, Indicates the magnetic induction intensity of the magnetic field, which depends on the material of the magnetic block and the strength of the magnetic pole; It is the effective length of the magnetic sensing element, which is consistent with the length of the part of the magnetic block that moves relatively; is the moving speed of the magnetic block, which represents the moving speed of the slider in the guide groove. This formula shows that the electromotive force generated by the magnetic induction element is proportional to the magnetic induction intensity, the length of the induction element and the moving speed of the magnetic block. When the slider moves faster or the magnetic field intensity is greater, the induced electromotive force will also increase. The electrical signal obtained by sampling is very important in the control system. Assume that the sampling period is , then the displacement in one sampling period is expressed as an integral:

[0052] ;

[0053] in, represents the displacement within the sampling period, is the speed signal. Through this integral operation, the control system obtains the displacement within a sampling period according to the speed signal, and then converts it into the corresponding control instruction. For example, if the speed signal of the upper rocker arm at a certain moment is , the voltage signal generated by the magnetic induction element at this moment is , the control system obtains the total displacement through multiple sampling and integration, thereby determining the operation intention of the joystick. In order to convert the physical signal into a usable electrical signal, signal processing is performed, such as analog-to-digital conversion and filtering. Suppose the digital signal after analog-to-digital conversion is , the electromotive force is , the quantization accuracy is , then the digital signal It is expressed as:

[0054] ;

[0055] in, Represents the quantization accuracy of the analog-to-digital converter, that is, the resolution of converting analog signals into digital signals. The higher the quantization accuracy, the higher the fineness of the electrical signal is retained when it is converted into a digital signal. It is directly input into the digital processing module of the control system as the input of the controller, thereby guiding the next control decision of the system.

[0056] In a specific embodiment, the process of executing step 200 may specifically include the following steps:

[0057] Performing signal strength detection on the first electrical signal to obtain a first signal strength, and dynamically setting a gain coefficient of the first signal processing channel according to the first signal strength to obtain a first gain adjustment signal;

[0058] Performing signal strength detection on the second electrical signal to obtain a second signal strength, and dynamically setting a gain coefficient of the second signal processing channel according to the second signal strength to obtain a second gain adjustment signal;

[0059] Inputting the first gain adjustment signal into the analog-to-digital conversion unit of the first signal processing channel, and quantizing and encoding the first gain adjustment signal through a successive approximation analog-to-digital conversion circuit to obtain a first digitized signal;

[0060] Inputting the second gain adjustment signal into the analog-to-digital conversion unit of the second signal processing channel, and quantizing and encoding the second gain adjustment signal through a successive approximation analog-to-digital conversion circuit to obtain a second digitized signal;

[0061] Inputting the first digitized signal into a digital filter of a first signal processing channel, and performing high-frequency noise suppression on the first digitized signal by a weighted average algorithm to obtain a first filtered signal;

[0062] Inputting the second digitized signal into a digital filter of a second signal processing channel, and performing high-frequency noise suppression on the second digitized signal by a weighted average algorithm to obtain a second filtered signal;

[0063] The first filtered signal and the second filtered signal are subjected to synchronous sampling alignment and data structure combination to obtain a digital signal matrix.

[0064] Specifically, the signal strength of the first electrical signal is detected to determine its amplitude at the current moment. Assume that the instantaneous voltage value of the first electrical signal is , then the signal strength By taking its effective value we get:

[0065] ;

[0066] in, is the length of the integration time window, is the signal at time The instantaneous value on. Represents the overall energy level of the signal. According to the strength of the first signal, the gain coefficient of the first signal processing channel is dynamically adjusted. Let the gain coefficient be , then the dynamic gain adjustment is expressed as:

[0067] ;

[0068] in is a gain adjustment function that adjusts the amplifier gain based on the signal strength. If the signal strength is low, the gain needs to be increased. To amplify the signal; if the signal strength is high, the gain needs to be reduced to prevent signal overload. After gain adjustment, the first gain adjustment signal is:

[0069] ;

[0070] in is the signal after gain adjustment. Similarly, the signal strength of the second electrical signal is detected to obtain the strength of the second signal , and by dynamically setting the gain coefficient of the second signal processing channel To adjust the amplitude of the signal, obtain the second gain adjustment signal:

[0071] ;

[0072] The first gain adjustment signal is input into the analog-to-digital conversion unit of the first signal processing channel, and the first gain adjustment signal is quantized and encoded using a successive approximation analog-to-digital conversion (SAR ADC). Successive approximation analog-to-digital conversion is an efficient analog-to-digital conversion method that gradually approximates the voltage value of the analog signal to convert it into a corresponding digital value. Assume that the first digitized signal after analog-to-digital conversion is , then it is expressed as:

[0073] ;

[0074] Among them, SAR Represents the process of successive approximation analog-to-digital conversion, Indicates the resolution of the analog-to-digital converter, that is, the number of bits used. The higher the The higher the value, the higher the accuracy of the converted digital signal. The same process is applied to the second gain-adjusted signal to obtain the second digitized signal :

[0075] ;

[0076] After completing the analog-to-digital conversion, the first digitized signal and the second digitized signal are filtered to remove the existing high-frequency noise. The first signal is input into the digital filter of the first signal processing channel and filtered using the weighted average algorithm. The weighted average filter can effectively suppress high-frequency noise by weighted averaging multiple sampling points. Suppose the first signal after filtering is , then it is expressed as:

[0077] ;

[0078] in, is the length of the filter window, It is The choice of weight depends on the design of the filter. Uniform weighting (all equal) or non-uniform weighting (such as Gaussian distribution weighting) to meet different filtering requirements. Similarly, for the second digitized signal Perform weighted average filtering to obtain the second filtered signal :

[0079] ;

[0080] The filtered signal and Perform synchronous sampling alignment and data structure combination. The purpose of synchronous sampling alignment is to ensure that at each time point, the samples of the two signals correspond to each other, so as to be combined into a complete state description. Assume and Respectively expressed in time On the first and second filtered signals, the synchronous sampling alignment is expressed as:

[0081] ;

[0082] in, It's in time The digital signal matrix on the image contains data information in two directions: left-right translation and front-back translation. By combining the data at each time point, a complete digital signal matrix is ​​obtained. :

[0083] ;

[0084] Digital signal matrix As input to the subsequent control system, it provides accurate joystick movement information.

[0085] In a specific embodiment, the process of executing step 300 may specifically include the following steps:

[0086] The digital signal matrix is ​​decoupled through the feature separation layer, which contains 4 convolutional layers. Each convolution layer is followed by a BatchNorm layer and a ReLU activation function to obtain the initial feature signal matrix.

[0087] The initial feature signal matrix is ​​input into the parallel multi-dimensional joystick attention network for shunting processing to obtain a displacement feature vector, a velocity feature vector and an acceleration feature vector. The parallel multi-dimensional joystick attention network contains three parallel feature channels, namely the first feature channel, the second feature channel and the third feature channel;

[0088] The displacement feature vector is input into the attention network of the first feature channel. The attention network includes a multi-head self-attention mechanism and a feedforward neural network layer. The enhanced displacement feature is obtained by calculating the correlation weights between the features.

[0089] The speed feature vector is input into the temporal attention module of the second feature channel, and the local temporal dependency is calculated through the sliding window mechanism to obtain the temporally enhanced speed feature;

[0090] The acceleration feature vector is input into the spatial attention module of the third feature channel, and the importance of different spatial positions is calculated through adaptive weights to obtain the spatially enhanced acceleration feature;

[0091] The enhanced displacement features, the time-series enhanced velocity features and the space-enhanced acceleration features are fused to obtain a fused feature vector;

[0092] The correlation coefficient matrix of motion in two directions is calculated based on the fused feature vector, and the sampling frequency adjustment factor is determined by the ratio of diagonal elements to off-diagonal elements to obtain the dynamic sampling strategy. The fused feature vector is resampled according to the dynamic sampling strategy and mapped to the feature space through the fully connected layer to obtain motion feature data.

[0093] Specifically, the digital signal matrix is ​​decoupled through the feature separation layer to extract different motion features of the signal. The feature separation layer contains 4 convolution layers, each of which is followed by a Batch Normalization layer and a ReLU activation function to normalize and activate the feature signal. Assume that the input digital signal matrix is , whose size is ,in represents the length of the time series, represents the number of features. The weight matrix of the first convolutional layer of the feature separation layer is , the convolution operation is expressed by the following formula:

[0094] ;

[0095] in, represents the output matrix after convolution, is the weight matrix of the first convolution kernel, is the bias term, and the symbol * represents the convolution operation. Through the convolution operation, the convolution layer extracts the local features of the input signal. Output The BatchNorm layer is used for standardization to ensure that the distribution of data in the neural network remains stable and to avoid the problem of gradient disappearance or gradient explosion. The output after standardization is expressed as:

[0096] ;

[0097] After the ReLU activation function, nonlinear features can be extracted, and the output after activation is:

[0098] ;

[0099] The function of the ReLU activation function is to set all negative values ​​to 0 and only retain positive values, thereby improving the sparsity and computational efficiency of the model. After four layers of convolution, BatchNorm and ReLU activation, the initial feature signal matrix is ​​obtained. , contains the preliminary features extracted from the input signal, which can effectively represent the local spatial relationship and change trend of the signal. The input is sent to the parallel multi-dimensional joystick attention network for shunting processing to obtain the displacement feature vector, velocity feature vector and acceleration feature vector. The parallel multi-dimensional joystick attention network contains three parallel feature channels, namely the first feature channel, the second feature channel and the third feature channel. Each channel independently processes different types of features, where the first feature channel is used to process displacement features, the second feature channel is used to process velocity features, and the third feature channel is used to process acceleration features. The attention network input to the first feature channel contains a multi-head self-attention mechanism and a feedforward neural network layer. The multi-head self-attention mechanism captures the relationship between features by calculating the correlation weights between features. , and Represent the query, key and value matrices respectively, and the output of multi-head self-attention is expressed as:

[0100] ;

[0101] in, is the dimension of the key vector, and The inner product of represents the similarity between different features, and the softmax function is used to calculate the weights of these similarities. Through calculation, the enhanced displacement feature is obtained. Similarly, the velocity feature vector The temporal attention module is input into the second feature channel. This module calculates local temporal dependencies through a sliding window mechanism. Let the window size be , at each time step The local timing dependency is calculated by the following formula:

[0102] ;

[0103] in, is the weight coefficient within the sliding window, Indicates at time The time-series enhanced speed feature is obtained by sliding window calculation to capture the dependency of the speed signal in the time dimension and obtain the time-series enhanced speed feature. Input to the spatial attention module of the third feature channel, and calculate the importance of different spatial positions through adaptive weights. Suppose the acceleration feature vector at different spatial positions is represented as , the adaptive weight calculation formula is:

[0104] ;

[0105] The weighted acceleration characteristics are:

[0106] ;

[0107] Through this step, the spatial attention module can highlight important position features and obtain spatially enhanced acceleration features. The enhanced displacement features, time-series enhanced velocity features, and spatially enhanced acceleration features are fused to obtain a fused feature vector. , calculate the motion correlation coefficient matrix in two directions (such as left and right and front and back) .set up and Represent the characteristic components in two directions respectively, and the correlation coefficient matrix is ​​calculated by the following formula:

[0108] ;

[0109] in, represents the covariance of the two directional feature components, and Represent the standard deviation of the characteristic component respectively. The diagonal elements of the correlation coefficient matrix represent the autocorrelation in their respective directions, while the off-diagonal elements represent the mutual relationship between the two directions. The sampling frequency adjustment factor is determined by the ratio of the diagonal elements to the off-diagonal elements. :

[0110] ;

[0111] The sampling frequency adjustment factor is used to dynamically adjust the sampling strategy. If the correlation is strong (the ratio is large), the sampling frequency is reduced to save power consumption; if the correlation is weak, the sampling frequency needs to be increased to ensure that sufficient motion information is captured. According to the obtained dynamic sampling strategy, the fused feature vector is resampled and mapped to the feature space through the fully connected layer. Let the weight matrix of the fully connected layer be , the bias is , then the motion feature data after mapping is:

[0112] ;

[0113] The role of the fully connected layer is to map the fused feature vector to a new feature space for subsequent control tasks and prediction applications.

[0114] In a specific embodiment, the process of executing step 400 may specifically include the following steps:

[0115] The motion feature data is input into a first feature separation module, the first feature separation module includes a displacement component extraction unit, a velocity component extraction unit and an acceleration component extraction unit, and a first state equation, a second state equation and a third state equation are respectively constructed by the displacement component extraction unit, the velocity component extraction unit and the acceleration component extraction unit to obtain a multi-dimensional expanded state space model;

[0116] For the displacement state quantity and the velocity state quantity in the multi-dimensional expanded state space model, the difference between the actual measured value and the theoretical calculated value is calculated by the first error calculation unit and the second error calculation unit respectively, so as to obtain the first state error matrix and the second state error matrix;

[0117] The first state error matrix and the second state error matrix are input into a dual-channel Kalman state estimator, the dual-channel Kalman state estimator includes a prediction channel and an update channel, and a multi-stage filtering process is performed on the system state through alternating iteration to obtain a first optimal state sequence and a second optimal state sequence;

[0118] Performing online identification operations on the first optimal state sequence and the second optimal state sequence through a parallel parameter identification unit, wherein the parallel parameter identification unit includes a first recursive least squares unit and a second recursive least squares unit, respectively constructing a first parameter update law and a second parameter update law to obtain a first model parameter matrix and a second model parameter matrix;

[0119] A dual state observer is constructed according to the first model parameter matrix and the second model parameter matrix, wherein the dual state observer includes a main state observer and an auxiliary state observer, and two groups of observer gain coefficients are calculated by a pole placement optimization algorithm to obtain a first observer matrix and a second observer matrix;

[0120] The first observer matrix and the second observer matrix are respectively orthogonally matched with the corresponding optimal state sequence, and the observation state deviation is calculated by the first error correction unit and the second error correction unit to obtain the first observation error vector and the second observation error vector;

[0121] The state feedback matrix is ​​updated hierarchically according to the first observation error vector and the second observation error vector, wherein the hierarchical update includes a fast dynamic response update and a slow steady-state response update, and the feedback gain coefficients of different time scales are optimized by an adaptive gradient algorithm to obtain a hierarchical state feedback matrix;

[0122] The hierarchical state feedback matrix is ​​multiplied by the state feedback synthesizer and the current system state. The state feedback synthesizer includes a dynamic weight allocation mechanism, which dynamically adjusts the feedback weights of different levels according to the system response characteristics to obtain control parameters.

[0123] Specifically, the motion feature data is input into the first feature separation module. This module includes a displacement component extraction unit, a velocity component extraction unit and an acceleration component extraction unit, which are used to extract the corresponding displacement, velocity and acceleration information from the original motion feature data. Assume that the motion feature data is ,in represents time, and after the first feature separation module, the displacement is obtained respectively ,speed and acceleration Component. The first state equation, the second state equation and the third state equation are constructed respectively through the displacement component extraction unit, the velocity component extraction unit and the acceleration component extraction unit. For the displacement component, it is expressed as:

[0124] ;

[0125] in, is the rate of change of displacement, is the state matrix of the system, describing the dynamic relationship between the system states. is the input matrix, is the input control variable. Similarly, the state equations of the velocity component and the acceleration component are:

[0126] ;

[0127] ;

[0128] Through these three state equations, a multi-dimensional extended state space model is obtained, which includes a dynamic description of displacement, velocity and acceleration. Error calculations are performed on the displacement state quantity and velocity state quantity in the multi-dimensional extended state space model respectively. The first error calculation unit and the second error calculation unit are used to perform difference calculations between the actual measured values ​​of displacement and velocity and the theoretical calculated values ​​to obtain the first state error matrix and the second state error matrix . Assume that the actual measurement value is The theoretical calculated value is , then the error matrix is ​​expressed as:

[0129] ;

[0130] Similarly, for the velocity state variable, the error matrix is:

[0131] ;

[0132] The first state error matrix and the second state error matrix The input is sent to the dual-channel Kalman state estimator, which includes a prediction channel and an update channel. The system state is filtered in multiple stages through alternating iterations to obtain the first optimal state sequence. and the second optimal state sequence The prediction channel estimates the current state based on the state at the previous moment, and the update channel corrects the predicted state through the current measurement value. The Kalman prediction process is expressed as:

[0133] ;

[0134] in, Indicates time Based on The predicted value at time, and They are the state matrix and the input matrix respectively. The Kalman update process is:

[0135] ;

[0136] in, is the Kalman gain matrix, is the actual measured value, is the output matrix. Through iterative calculation, the optimal state estimation of the system is obtained. The first optimal state sequence is identified by using the parallel parameter identification unit. and the second optimal state sequence The parallel parameter identification unit includes a first recursive least squares unit and a second recursive least squares unit, which are respectively used to construct a first parameter update law and a second parameter update law to obtain a first model parameter matrix and the second model parameter matrix The recursive least squares algorithm updates parameters using the following formula:

[0137] ;

[0138] in, is the parameter vector, is the gain vector, is the output signal, is the regression vector of the input signal. Through the recursive least squares method, the system can fine-tune the parameters in each iteration to adapt to the change of state. According to the first model parameter matrix and the second model parameter matrix , construct a dual state observer, which includes a main state observer and an auxiliary state observer. The gain coefficients of the two groups of observers are calculated by the pole placement optimization algorithm to obtain the first observer matrix and the second observer matrix The pole configuration optimizes the dynamic performance by adjusting the closed-loop pole position of the system, which can be expressed as:

[0139] ;

[0140] in, is the system state matrix, is the output matrix, is the set of expected poles. In this way, it is ensured that the observer can accurately track the system state. The first observer matrix and the second observer matrix The corresponding optimal state sequence and Perform an orthogonal matching operation, calculate the observation state deviation through the first error correction unit and the second error correction unit, and obtain the first observation error vector and the second observation error vector These observation error vectors are used to correct the system state so that the observer can better track the actual dynamic changes of the system. and the second observation error vector , the state feedback matrix is ​​updated hierarchically, including fast dynamic response update and slow steady-state response update. Fast dynamic response is used to cope with rapid changes in the system, while slow steady-state response is used to ensure the stability of the system. The feedback gain coefficients of different time scales are optimized by the adaptive gradient algorithm to obtain the hierarchical state feedback matrix The adaptive gradient algorithm is expressed by the following formula:

[0141] ;

[0142] in, is the feedback gain matrix, is the learning rate, is the error gradient at the current moment. The state feedback matrix is ​​hierarchically constructed by the state feedback synthesizer and the current system state. Perform matrix multiplication to obtain control parameters. The state feedback synthesizer includes a dynamic weight allocation mechanism to dynamically adjust the feedback weights of different levels according to the response characteristics of the system. , so that the feedback gain can adapt to the system requirements in different working modes:

[0143] ;

[0144] Through the above process, the control parameters for controlling the execution device are finally obtained.

[0145] In a specific embodiment, the process of executing step 500 may specifically include the following steps:

[0146] The left-right displacement and the front-back displacement of the joystick in the control parameters are sampled and detected respectively, and data of 50 consecutive sampling points are collected by the first displacement sampling unit and the second displacement sampling unit to obtain the first displacement sequence and the second displacement sequence;

[0147] The power calculation unit performs signal energy analysis on the first displacement sequence and the second displacement sequence, and the power calculation unit adopts an effective value calculation method of the magnetic induction signal to obtain a power value in the left-right direction and a power value in the front-back direction;

[0148] Compare the left-right power value and the front-back power value with the first threshold value T1. If the power values ​​of 50 consecutive sampling points are all less than the first threshold value T1, switch from the full power mode to the energy-saving mode to obtain a first mode switching signal.

[0149] Compare the power values ​​in the left-right direction and the power values ​​in the front-back direction with the second threshold value T2. If the power values ​​of 200 consecutive sampling points are all less than the second threshold value T2, switch from the energy-saving mode to the sleep mode to obtain a second mode switching signal.

[0150] Performing state synthesis on the first mode switching signal and the second mode switching signal, judging the mode switching condition through the first finite state machine, and obtaining the current working mode identifier;

[0151] The joystick sampling frequency is adjusted according to the current working mode identifier, the highest sampling frequency is maintained in full power mode, the sampling frequency is reduced by 50% in energy saving mode, and the sampling frequency is reduced by 90% in sleep mode, and a sampling frequency control signal is obtained;

[0152] The control parameters are resampled according to the sampling frequency control signal, the validity of the control parameters is determined by the second finite state machine to obtain effective control parameters, and the effective control parameters are input into the mode controller. The corresponding control quantities are generated by the control laws under different modes, and the control quantities are linearly compensated to obtain the target control instructions.

[0153] Specifically, the left-right displacement and the front-back displacement of the joystick in the control parameters are sampled and detected respectively, and the left-right and front-back displacement of the joystick are collected by the first displacement sampling unit and the second displacement sampling unit. Assume that the left-right displacement is , the displacement in the front-to-back direction is , the sampling period is , by collecting 50 consecutive sampling points, the first displacement sequence and the second displacement sequence are obtained, which are and The two displacement sequences obtained by the displacement sampling unit enter the power calculation unit for signal energy analysis. The power calculation unit uses the effective value calculation method of the magnetic induction signal to estimate the power of the joystick movement. , the effective value is expressed as:

[0154] ;

[0155] in, Indicates the effective value, Indicates The displacement value of each sampling point, Indicates the total number of sampling points. In the left and right directions and the front and back directions, the effective value calculation method is used to calculate the left and right power values. And the power value in the front and rear directions :

[0156] ;

[0157] Through this step, the power values ​​in the left and right directions and the front and back directions are obtained, which represent the motion energy of the joystick in these two directions. And the power value in the front and rear directions The first threshold value is set If the power values ​​of 50 consecutive sampling points are all less than the first threshold ,Right now:

[0158] ;

[0159] It is determined that the current joystick motion energy is low, and the full power mode is switched to the energy saving mode, and a first mode switching signal is obtained. Energy saving mode can reduce unnecessary energy consumption. When the joystick moves less, it can save energy by reducing the system power. And the power value in the front and rear directions With the second threshold If the power values ​​of 200 consecutive sampling points are all less than the second threshold ,Right now:

[0160] ;

[0161] The system determines that the joystick is almost stationary, switches from energy-saving mode to sleep mode, and obtains the second mode switching signal. The sleep mode further reduces system power consumption and only retains the most basic status monitoring functions. The first mode switching signal and the second mode switching signal The state synthesizer performs comprehensive processing, and the first finite state machine determines the conditions for mode switching to obtain the current working mode identifier. The function of the finite state machine is to perform logical control on the transition between different modes. According to the different states of the mode switching signal, it determines whether the current system should be in full power mode, energy saving mode or sleep mode. According to the current working mode identification , adjust the joystick sampling frequency. In full power mode, the sampling frequency Keep it at the highest level To ensure that the movement of the joystick can be accurately captured. In energy-saving mode, the sampling frequency is reduced by 50%, that is:

[0162] ;

[0163] In sleep mode, the sampling frequency is further reduced by 90%:

[0164] ;

[0165] The sampling frequency is dynamically adjusted according to the requirements of different modes to achieve a balance between power saving and response, and the corresponding sampling frequency control signal is obtained. The control parameters are resampled according to the sampling frequency control signal to adapt to the current working mode of the system. During the resampling process, the second finite state machine determines the validity of the control parameters and obtains effective control parameters. Effective control parameters are the basis for ensuring that the system can operate normally in different modes. The second finite state machine determines the resampled parameters to ensure that they meet the control requirements and effectiveness. The effective control parameters are input into the mode controller, and the corresponding control quantities are generated through the control laws under different modes. Let the control law be , then the corresponding control quantity is expressed as:

[0166] ;

[0167] in, Is the current working mode identifier, is the control parameter. The control law determines how the control quantity is generated according to different modes (such as full power mode, energy saving mode or sleep mode). After the control quantity is generated, linear compensation is performed on it to reduce the error and nonlinear effect in the control. Let the linear compensation coefficient be , the offset is , then the target control instruction is:

[0168] ;

[0169] The role of linear compensation is to fine-tune the control amount so that the final control command can more accurately reflect the movement intention of the joystick, so that the movement of the device meets the user's expectations.

[0170] In a specific embodiment, the process of executing step 600 may specifically include the following steps:

[0171] The left and right translation control components in the target control command are mapped by proportional coefficients and compensated by offset, and a mapping relationship is established based on the displacement signals in the left and right directions to obtain the left and right device control quantities;

[0172] Perform proportional coefficient mapping and offset compensation on the front-rear translation control component in the target control command, establish a mapping relationship based on the displacement signal in the front-rear direction, and obtain the front-rear direction device control quantity;

[0173] The left and right device control quantities and the front and rear device control quantities are grouped according to the device type, and the initial device instructions are obtained by searching and matching the corresponding relationship table between the joystick displacement and the device control quantity;

[0174] The operating parameters of the executing equipment are monitored through position data collection and speed data collection, and synchronous sampling is performed based on the sampling period of the magnetic induction signal to obtain the equipment operating status, and the deviation between the equipment operating status and the theoretical operating parameters is calculated, and dynamic compensation is performed through the equipment load characteristic curve to obtain the state correction value;

[0175] The state correction amount is used to establish the control parameter update formula according to the time series, and the control parameter is corrected by the cumulative error term to obtain the corrected control instruction. The multi-device response timing analysis is performed on the corrected control instruction, and the coordinated control sequence is obtained by establishing the action sequence constraint relationship between the devices.

[0176] The collaborative control sequence is distributed to each execution device according to the control cycle, and the multi-device collaborative control result is obtained through real-time updating of control parameters.

[0177] Specifically, the left and right translation control components and the front and back translation control components are subjected to proportional coefficient mapping and offset compensation. Assume that the left and right translation components in the target control command are , the forward and backward translation components are The proportional coefficients are and , the offsets are and . The amount of equipment control in the left and right directions Calculated by the following formula:

[0178] ;

[0179] Similarly, the amount of equipment control in the front and rear directions It is expressed as:

[0180] ;

[0181] In the above formula, the proportionality coefficient and The role of is to scale the control component so that it can adapt to the input range of the device, and the offset and It is used to eliminate the deviation of the initial state of the equipment to ensure that the equipment can execute the control instructions from the correct starting point. The mapping relationship is established based on the displacement signals in the left and right directions and the front and back directions to obtain the equipment control quantity in the left and right directions and the front and back directions. and . Group the left and right device control quantities and the front and back device control quantities by device type. In a multi-device control scenario, different devices have their own specific functions. The initial control instructions for each device are determined by searching and matching the corresponding relationship table between the joystick displacement and the device control quantity. Assume that the corresponding relationship between the left and right device control quantity and the front and back device control quantity and device A and device B is determined by the lookup table OK, get the initial device instruction by looking up the table and

[0182] ;

[0183] ;

[0184] This process ensures that each device can obtain the corresponding control instructions according to the operation of the joystick, thereby achieving synchronous multi-device control. In order to ensure that the device maintains the accuracy of the state when executing the control instructions, the operating parameters of the device are monitored in real time, including the collection of position data and speed data. The operating status of each executing device is obtained by synchronous sampling based on the sampling period of the magnetic induction signal. , such as the current position and speed information of the device. Assume that the sampling period is , the sampled position and velocity data are expressed as:

[0185] ;

[0186] in, and Represent the position and velocity data sequences respectively, and They are The position and speed of each sampling point. Based on these data, the deviation between the operating state of the equipment and the theoretical operating parameters is calculated to obtain the deviation:

[0187] ;

[0188] ;

[0189] in, and They are the measured value and the theoretical calculated value, respectively, and the deviation and Reflects the difference between the actual operating state of the equipment and the theoretical expectation. These deviations are dynamically compensated according to the load characteristic curve of the equipment to obtain the state correction value. . Assume the load characteristic curve is a function , then the state correction is expressed as:

[0190] ;

[0191] The load characteristic curve describes the operating characteristics of the equipment under different load conditions. By compensating for the deviation, the control error caused by load changes can be reduced to ensure that the equipment can accurately execute instructions. The state correction amount is used to establish the update formula of the control parameter according to the time series to correct the cumulative error term of the control parameter. Assume that the cumulative error term is , the update formula of the control parameters is:

[0192] ;

[0193] in, is the current control instruction, is the learning rate, which is used to control the update speed of the accumulated error. In this way, the corrected control instructions can more accurately reflect the actual needs of the device. Perform multi-device response timing analysis on the corrected control instructions, and obtain the collaborative control sequence by establishing the action sequence constraint relationship between devices. The action sequence constraint relationship between devices is used to ensure that different devices can work together when performing actions to avoid conflicts. For example, assuming that the action of device A needs to be completed before the action of device B, the following constraint relationship needs to be established:

[0194] ;

[0195] in, and Respectively represent the action start time of device A and device B, is the duration of the action of device A. By using constraints, the actions between devices are ensured to be orderly and coordinated. The collaborative control sequence is distributed to each execution device according to the set control period, and the result of multi-device collaborative control is obtained by real-time updating of control parameters. Assume that the control period is , the cooperative control sequence is , then for each execution device , the control instruction is expressed as:

[0196] ;

[0197] This control method ensures that each device obtains appropriate control instructions within the specified control cycle and adjusts the execution action according to the current state, thereby achieving coordinated control of multiple devices.

[0198] The above describes the automation equipment control method based on the low power consumption joystick in the embodiment of the present application. The following describes the automation equipment control system based on the low power consumption joystick in the embodiment of the present application. Figure 2 In the embodiment of the present application, an embodiment of an automation equipment control system based on a low-power joystick includes:

[0199] The acquisition module 11 is used to respectively acquire left-right translation signals and front-back translation signals of the upper rocker arm and the lower rocker arm of the rocker, and convert them into first electrical signals and second electrical signals respectively;

[0200] The conversion module 12 is used to perform analog-to-digital conversion and low-pass filtering on the first electrical signal and the second electrical signal to obtain a digital signal matrix;

[0201] An extraction module 13 is used to input the digital signal matrix into a parallel multi-dimensional joystick attention network for feature extraction, and dynamically adjust the sampling frequency according to the correlation coefficient of the two-directional motion to obtain motion feature data;

[0202] A construction module 14 is used to construct a multi-dimensional extended state space model using an extended state observer according to the motion feature data, and to update the state feedback matrix through online parameter identification to obtain control parameters;

[0203] The switching module 15 is used to divide the control parameters into power modes according to the set threshold value, and switch between the full power mode, the energy saving mode and the sleep mode when the displacement meets different threshold conditions to obtain the target control instruction;

[0204] The adjustment module 16 is used to map the target control instruction to the control quantity of each execution device through the set proportional coefficient and offset, and adjust the control parameters based on the device state feedback information to obtain the multi-device collaborative control result.

[0205] Through the cooperation of the above components, magnetic induction detection is used to replace the traditional brush carbon resistance structure, eliminating physical contact friction, significantly extending the service life of the system, and improving the detection accuracy. By introducing a parallel multi-dimensional rocker attention network, the multi-dimensional extraction and analysis of the rocker motion characteristics are realized, the coupling relationship between motion data is fully utilized, and the control accuracy is improved. A state space model based on an extended state observer is designed, and accurate observation of the system state and real-time optimization of control parameters are achieved through online parameter identification. An innovative three-level power mode division mechanism is proposed, which automatically switches the working mode according to the use status of the rocker, effectively reducing the overall power consumption of the system. A complete multi-device collaborative control framework is established, and precise control of multiple execution devices is achieved through the mapping mechanism of proportional coefficients and offsets. The dual-channel signal processing structure is adopted, combined with a dynamic sampling strategy, to improve the system's anti-interference ability and signal processing efficiency. By setting a hierarchical state feedback mechanism, an effective combination of fast dynamic response and slow steady-state response is achieved, which improves the dynamic performance of the system.

[0206] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0207] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable an electronic device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.

[0208] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for controlling an automated device based on a low-power joystick, characterized in that: The method comprises: The left-right translation signal and the front-back translation signal are respectively collected from the upper rocker arm and the lower rocker arm of the rocker, and converted into a first electrical signal and a second electrical signal respectively; Performing analog-to-digital conversion and low-pass filtering on the first electrical signal and the second electrical signal to obtain a digital signal matrix; Inputting the digital signal matrix into a parallel multi-dimensional joystick attention network for feature extraction, and dynamically adjusting the sampling frequency according to the correlation coefficient of the two-directional motion to obtain motion feature data; According to the motion characteristic data, a multi-dimensional extended state space model is constructed using an extended state observer, and a state feedback matrix is ​​updated through online parameter identification to obtain control parameters; The control parameters are divided into power modes according to set thresholds, and when the displacement meets different threshold conditions, the full power mode, energy saving mode and sleep mode are switched to obtain target control instructions; The target control instruction is mapped to the control quantity of each execution device through the set proportional coefficient and offset, and the control parameters are adjusted based on the device state feedback information to obtain the multi-device collaborative control result.

2. The automation equipment control method based on low-power joystick according to claim 1, characterized in that: The upper rocker arm and the lower rocker arm of the pair of rockers respectively collect left-right translation signals and front-back translation signals, and convert them into first electrical signals and second electrical signals respectively, including: The position of the slide bar in the guide groove provided at one end of the upper rocker arm is collected, and the slide bar is driven to translate left and right in the slide groove of the first fixed part through the connecting rod to obtain the translation displacement of the first movable part; Connecting the first magnetic block on the inner side of the first movable part to the first translation magnet assembly, and obtaining a left-right translation signal through the synchronous translation of the first magnetic block; The position of the slide bar in the guide groove provided at one end of the lower rocker arm is collected, and the slide bar is driven to translate forward and backward in the slide groove of the second fixed part through the connecting rod to obtain the translation displacement of the second movable part; Connecting the second magnetic block on the inner side of the second movable part to the second translation magnet assembly, and obtaining a front-to-back translation signal through the synchronous translation of the second magnetic block; The left-right translation signal generated by the first magnetic block is sampled by a first magnetic induction element installed under the first translation magnet assembly to obtain a first electrical signal corresponding to the left-right translation; The forward and backward translation signal generated by the second magnetic block is sampled by a second magnetic induction element installed below the second translation magnet assembly to obtain a second electrical signal corresponding to the forward and backward translation.

3. The method for controlling an automated device based on a low-power joystick according to claim 2, characterized in that: The step of performing analog-to-digital conversion and low-pass filtering on the first electrical signal and the second electrical signal to obtain a digital signal matrix includes: Performing signal strength detection on the first electrical signal to obtain a first signal strength, and dynamically setting a gain coefficient of a first signal processing channel according to the first signal strength to obtain a first gain adjustment signal; Performing signal strength detection on the second electrical signal to obtain a second signal strength, and dynamically setting a gain coefficient of a second signal processing channel according to the second signal strength to obtain a second gain adjustment signal; Inputting the first gain adjustment signal into an analog-to-digital conversion unit of a first signal processing channel, and quantizing and encoding the first gain adjustment signal through a successive approximation analog-to-digital conversion circuit to obtain a first digitized signal; Inputting the second gain adjustment signal into an analog-to-digital conversion unit of a second signal processing channel, and quantizing and encoding the second gain adjustment signal through a successive approximation analog-to-digital conversion circuit to obtain a second digitized signal; Inputting the first digitized signal into a digital filter of a first signal processing channel, and performing high-frequency noise suppression on the first digitized signal by a weighted average algorithm to obtain a first filtered signal; Inputting the second digitized signal into a digital filter of a second signal processing channel, and performing high-frequency noise suppression on the second digitized signal by a weighted average algorithm to obtain a second filtered signal; The first filtered signal and the second filtered signal are synchronously sampled and aligned and data structured to obtain a digital signal matrix.

4. The automation equipment control method based on low-power joystick according to claim 3 is characterized in that: The digital signal matrix is ​​input into a parallel multi-dimensional joystick attention network for feature extraction, and the sampling frequency is dynamically adjusted according to the correlation coefficient of the two-directional motion to obtain motion feature data, including: Decoupling the digital signal matrix through a feature separation layer, wherein the feature separation layer includes four convolution layers, each convolution layer is followed by a BatchNorm layer and a ReLU activation function to obtain an initial feature signal matrix; Inputting the initial feature signal matrix into a parallel multi-dimensional joystick attention network for shunting processing to obtain a displacement feature vector, a velocity feature vector and an acceleration feature vector, wherein the parallel multi-dimensional joystick attention network comprises three parallel feature channels, namely a first feature channel, a second feature channel and a third feature channel; Inputting the displacement feature vector into the attention network of the first feature channel, the attention network includes a multi-head self-attention mechanism and a feedforward neural network layer, and obtaining an enhanced displacement feature by calculating the correlation weights between the features; Inputting the speed feature vector into the temporal attention module of the second feature channel, calculating the local temporal dependency through a sliding window mechanism, and obtaining a temporally enhanced speed feature; Inputting the acceleration feature vector into the spatial attention module of the third feature channel, calculating the importance of different spatial positions through adaptive weights, and obtaining a spatially enhanced acceleration feature; Performing feature fusion on the enhanced displacement feature, the temporally enhanced velocity feature, and the spatially enhanced acceleration feature to obtain a fused feature vector; Based on the fused feature vector, a correlation coefficient matrix of motion in two directions is calculated, and a sampling frequency adjustment factor is determined by the ratio of diagonal elements to off-diagonal elements to obtain a dynamic sampling strategy. The fused feature vector is resampled according to the dynamic sampling strategy, and mapped to the feature space through a fully connected layer to obtain motion feature data.

5. The method for controlling an automated device based on a low-power joystick according to claim 4, characterized in that: The method of constructing a multi-dimensional extended state space model using an extended state observer according to the motion feature data and updating a state feedback matrix through online parameter identification to obtain control parameters includes: The motion feature data is input into a first feature separation module, wherein the first feature separation module includes a displacement component extraction unit, a velocity component extraction unit, and an acceleration component extraction unit, and a first state equation, a second state equation, and a third state equation are constructed by the displacement component extraction unit, the velocity component extraction unit, and the acceleration component extraction unit, respectively, to obtain a multi-dimensional expanded state space model; For the displacement state quantity and the velocity state quantity in the multi-dimensional expanded state space model, the difference between the actual measured value and the theoretical calculated value is calculated by the first error calculation unit and the second error calculation unit respectively, so as to obtain the first state error matrix and the second state error matrix; Inputting the first state error matrix and the second state error matrix into a dual-channel Kalman state estimator, wherein the dual-channel Kalman state estimator includes a prediction channel and an update channel, and performing multi-stage filtering processing on the system state through alternating iterations to obtain a first optimal state sequence and a second optimal state sequence; Performing online identification operations on the first optimal state sequence and the second optimal state sequence through a parallel parameter identification unit, wherein the parallel parameter identification unit includes a first recursive least squares unit and a second recursive least squares unit, respectively constructing a first parameter update law and a second parameter update law to obtain a first model parameter matrix and a second model parameter matrix; A dual state observer is constructed according to the first model parameter matrix and the second model parameter matrix, wherein the dual state observer includes a main state observer and an auxiliary state observer, and two groups of observer gain coefficients are calculated by a pole placement optimization algorithm to obtain a first observer matrix and a second observer matrix; The first observer matrix and the second observer matrix are respectively orthogonally matched with the corresponding optimal state sequence, and the observation state deviation is calculated by the first error correction unit and the second error correction unit to obtain the first observation error vector and the second observation error vector; hierarchically updating a state feedback matrix according to the first observation error vector and the second observation error vector, wherein the hierarchical updating includes a fast dynamic response update and a slow steady-state response update, and optimizing feedback gain coefficients of different time scales by an adaptive gradient algorithm to obtain a hierarchical state feedback matrix; The hierarchical state feedback matrix is ​​subjected to matrix multiplication operation by a state feedback synthesizer and the current system state. The state feedback synthesizer includes a dynamic weight allocation mechanism, which dynamically adjusts feedback weights of different levels according to system response characteristics to obtain control parameters.

6. The method for controlling an automation device based on a low-power joystick according to claim 5, characterized in that: The power mode division of the control parameters according to the set threshold value, switching between the full power mode, the energy saving mode and the sleep mode when the displacement meets different threshold conditions, and obtaining the target control instruction includes: The left-right displacement and the front-back displacement of the joystick in the control parameters are sampled and detected respectively, and data are collected at 50 consecutive sampling points by a first displacement sampling unit and a second displacement sampling unit to obtain a first displacement sequence and a second displacement sequence; Performing signal energy analysis on the first displacement sequence and the second displacement sequence by a power calculation unit, wherein the power calculation unit adopts an effective value calculation method of a magnetic induction signal to obtain a left-right direction power value and a front-back direction power value; Compare the left-right power value and the front-back power value with a first threshold value T1, and if the power values ​​of 50 consecutive sampling points are all less than the first threshold value T1, switch from the full power mode to the energy-saving mode to obtain a first mode switching signal; Compare the left-right power value and the front-back power value with a second threshold value T2, and if the power values ​​of 200 consecutive sampling points are all less than the second threshold value T2, switch from the energy-saving mode to the sleep mode to obtain a second mode switching signal; Performing state synthesis on the first mode switching signal and the second mode switching signal, judging the mode switching condition through a first finite state machine, and obtaining a current working mode identifier; The joystick sampling frequency is adjusted according to the current working mode identifier, the highest sampling frequency is maintained in the full power mode, the sampling frequency is reduced by 50% in the energy saving mode, and the sampling frequency is reduced by 90% in the sleep mode, so as to obtain a sampling frequency control signal; The control parameters are resampled according to the sampling frequency control signal, the validity of the control parameters is determined by a second finite state machine to obtain effective control parameters, and the effective control parameters are input into a mode controller, corresponding control quantities are generated by control laws under different modes, and linear compensation is performed on the control quantities to obtain target control instructions.

7. The method for controlling an automated device based on a low-power joystick according to claim 6, characterized in that: The target control instruction is mapped to the control amount of each execution device through the set proportional coefficient and offset, and the control parameter is adjusted based on the device state feedback information to obtain the multi-device collaborative control result, including: Performing proportional coefficient mapping and offset compensation on the left and right translation control components in the target control command, establishing a mapping relationship based on the displacement signals in the left and right directions, and obtaining the left and right device control quantities; Performing proportional coefficient mapping and offset compensation on the front-rear translation control component in the target control command, establishing a mapping relationship based on the displacement signal in the front-rear direction, and obtaining the front-rear direction device control amount; The left-right direction device control amount and the front-back direction device control amount are grouped according to device type, and a corresponding relationship table between the joystick displacement amount and the device control amount is searched and matched to obtain an initial device instruction; The operating parameters of the execution equipment are monitored by collecting position data and speed data, synchronous sampling is performed based on the sampling period of the magnetic induction signal to obtain the equipment operating state, and the deviation between the equipment operating state and the theoretical operating parameters is calculated, and dynamic compensation is performed through the equipment load characteristic curve to obtain the state correction value; The state correction amount is used to establish a control parameter update formula according to a time series, the control parameter is corrected by a cumulative error term to obtain a corrected control instruction, and a multi-device response timing analysis is performed on the corrected control instruction, and a coordinated control sequence is obtained by establishing an action sequence constraint relationship between devices; The collaborative control sequence is distributed to each execution device according to the control period, and the multi-device collaborative control result is obtained through real-time updating of control parameters.

8. An automation equipment control system based on a low-power joystick, characterized in that: Used to execute the automation equipment control method based on the low-power joystick according to claim 1, the automation equipment control system based on the low-power joystick comprises: The acquisition module is used to respectively acquire left-right translation signals and front-back translation signals of the upper rocker arm and the lower rocker arm of the rocker, and convert them into first electrical signals and second electrical signals respectively; a conversion module, configured to perform analog-to-digital conversion and low-pass filtering on the first electrical signal and the second electrical signal to obtain a digital signal matrix; An extraction module, used for inputting the digital signal matrix into a parallel multi-dimensional joystick attention network for feature extraction, and dynamically adjusting the sampling frequency according to the correlation coefficient of the two-directional motion to obtain motion feature data; A construction module is used to construct a multi-dimensional extended state space model using an extended state observer according to the motion feature data, and to update a state feedback matrix through online parameter identification to obtain control parameters; A switching module, used to divide the control parameters into power modes according to a set threshold value, and switch between full power mode, energy saving mode and sleep mode when the displacement meets different threshold conditions to obtain a target control instruction; The adjustment module is used to map the target control instruction to the control quantity of each execution device through the set proportional coefficient and offset, and adjust the control parameters based on the device state feedback information to obtain the multi-device collaborative control result.

Images