Industrial robot control method and device based on Hall rocker control

Abstract

The present invention relates to the field of rocker control technology, and discloses an industrial robot control method and device based on Hall rocker control, wherein the method comprises: amplifying and filtering the analog signal of the dual-axis rocker position collected by the Hall sensor to obtain the digital quantity of the rocker position; constructing an interactive multi-model set including a uniform motion model, a uniform acceleration motion model and a Jerk motion model, and establishing an interactive multi-model state space expression of the interactive multi-model set; performing Markov chain analysis and model weight balance point analysis to obtain the initial model conversion probability matrix; calculating the target model probability distribution; performing an extended Kalman filter operation on the interactive multi-model set to obtain the robot joint motion state quantity; performing a quintic polynomial interpolation operation based on the robot joint motion state quantity to generate a robot joint motion control sequence. The method improves the accuracy of the rocker control in a steady-state working mode while ensuring rapid response capability.

Description

Technical Field

[0001] The present invention relates to the technical field of rocker control, and in particular to an industrial robot control method and device based on Hall rocker control. Background Art

[0002] Traditional industrial robot control methods mainly rely on preset programs and lack real-time interaction capabilities, making it difficult to meet the flexible operation requirements under complex working conditions. In actual applications, operators often need to make real-time adjustments and precise control of robots, which requires the control system to be able to quickly respond to human-machine interaction instructions.

[0003] The motion characteristics of industrial robots are significantly nonlinear and time-varying, and there are multiple motion mode switches in the control process. The existing single model control method is difficult to accurately describe the dynamic characteristics of the robot in different motion states, resulting in insufficient control accuracy. Especially when the robot needs to switch motion states frequently, traditional control algorithms are difficult to ensure the smoothness of state transitions and the continuity of control. Summary of the invention

[0004] The present invention provides an industrial robot control method and device based on Hall rocker control, which are used to improve the accuracy of rocker control in a steady-state working mode while ensuring rapid response capability.

[0005] In a first aspect, the present invention provides an industrial robot control method based on Hall rocker control, the industrial robot control method based on Hall rocker control comprising:

[0006] The analog signal of the dual-axis rocker position collected by the Hall sensor is amplified and filtered to obtain the digital value of the rocker position;

[0007] Based on the digital quantity of the rocker position, an interactive multi-model set including a uniform motion model, a uniform acceleration motion model and a Jerk motion model is constructed, and an interactive multi-model state space expression of the interactive multi-model set is established;

[0008] Performing Markov chain analysis and model weight balance point analysis on the interactive multi-model state space expression to obtain an initial model conversion probability matrix;

[0009] Calculating the instantaneous position change rate and the instantaneous acceleration value according to the digital quantity of the rocker position, and calculating the weight adjustment coefficient according to the initial model conversion probability matrix to obtain the target model probability distribution;

[0010] According to the probability distribution of the target model, an extended Kalman filter operation is performed on the interactive multi-model set to obtain the robot joint motion state quantity;

[0011] A fifth-order polynomial interpolation operation is performed based on the robot joint motion state quantity to generate a robot joint motion control sequence.

[0012] In a second aspect, the present invention provides an industrial robot control device based on Hall rocker control, the industrial robot control device based on Hall rocker control comprising:

[0013] The acquisition module is used to amplify and filter the analog signal of the dual-axis rocker position acquired by the Hall sensor to obtain the digital value of the rocker position;

[0014] A construction module, used to construct an interactive multi-model set including a uniform motion model, a uniform acceleration motion model and a Jerk motion model based on the digital quantity of the rocker position, and establish an interactive multi-model state space expression of the interactive multi-model set;

[0015] An analysis module, used for performing Markov chain analysis and model weight balance point analysis on the interactive multi-model state space expression to obtain an initial model conversion probability matrix;

[0016] A calculation module, used to calculate the instantaneous position change rate and the instantaneous acceleration value according to the digital quantity of the rocker position, and calculate the weight adjustment coefficient according to the initial model conversion probability matrix to obtain the target model probability distribution;

[0017] A filtering module, used to perform an extended Kalman filter operation on the interactive multi-model set according to the target model probability distribution to obtain the robot joint motion state quantity;

[0018] A generation module is used to perform a quintic polynomial interpolation operation based on the robot joint motion state quantity to generate a robot joint motion control sequence.

[0019] The third aspect of the present invention provides an industrial robot control device based on Hall rocker control, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the industrial robot control device based on Hall rocker control executes the above-mentioned industrial robot control method based on Hall rocker control.

[0020] A fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned industrial robot control method based on Hall rocker control.

[0021] In the technical solution provided by the present invention, a dual-axis Hall sensor structure and a signal processing method are adopted to realize high-precision detection of the rocker position, and the noise and zero drift in the sensor signal are effectively suppressed through multi-stage filtering and compensation algorithms, thereby improving the accuracy of position detection. An interactive multi-model set including a uniform motion model, a uniform acceleration motion model and a Jerk motion model is constructed, which can comprehensively describe the different motion states of the robot and ensure the accuracy of state estimation. Through Markov chain analysis and model weight balance point analysis, a dynamic model conversion mechanism is established to solve the problem of smooth transition when the robot switches between different motion states. The steady-state enhanced IMM algorithm is introduced to process the rocker control signal, which improves the control accuracy of the system in the steady-state working mode while ensuring the rapid response capability. The extended Kalman filter and the state interactive fusion strategy are adopted to effectively handle the nonlinear characteristics of the system and improve the accuracy of state estimation. The trajectory planning method based on quintic polynomial interpolation ensures the continuity and smoothness of the joint motion trajectory and reduces mechanical vibration. Through hierarchical signal processing and control strategies, the whole process control from rocker input to robot execution is realized.

[0022] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0023] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 Schematic diagram of an embodiment of an industrial robot control method based on Hall rocker control in an embodiment of the present invention;

[0025] Figure 2 Schematic diagram of an embodiment of an industrial robot control device based on Hall rocker control in an embodiment of the present invention;

[0026] Figure 3 The figure is a schematic diagram of an embodiment of an industrial robot control device based on Hall rocker control in an embodiment of the present invention. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. 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.

[0028] The terms "including" and "having" and any variations thereof mentioned in the embodiments of the present invention are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device end including a series of steps or units is not limited to the listed steps or units, but may optionally include other steps or units that are not listed, or may optionally include other steps or units that are inherent to these processes, methods, products or device ends.

[0029] To facilitate understanding of this embodiment, firstly, a method for controlling an industrial robot based on Hall rocker control disclosed in an embodiment of the present invention is described in detail. Figure 1 As shown, the method comprises the following steps:

[0030] 101. Amplify and filter the analog signal of the dual-axis rocker position collected by the Hall sensor to obtain a digital value of the rocker position;

[0031] It is understandable that the execution subject of the present invention may be an industrial robot control device based on Hall rocker control, or a terminal or a server, which is not specifically limited here. The embodiment of the present invention is described by taking a server as the execution subject as an example.

[0032] Specifically, the Hall sensor is installed in the X-axis and Y-axis directions at the bottom of the rocker. The Hall effect principle is used to collect the analog signal of the rocker position by detecting the change in magnetic field strength generated by the permanent magnet during the movement of the rocker. The position change of the permanent magnet will cause the magnetic field components in different directions to change. The Hall sensor can convert this change into an electrical signal to form an analog signal output of the dual-axis rocker. The dual-axis rocker position analog signal is amplified by a low-noise operational amplifier to enhance the amplitude of the signal so that it can fully utilize its dynamic range during analog-to-digital conversion. Operational amplification can improve signal strength and effectively suppress noise interference in the signal. In order to remove high-frequency noise and power supply interference, filtering processing is used to optimize the signal. The filtering technology includes a low-pass filter and a band-pass filter to obtain an amplified and filtered dual-axis rocker signal. The amplified and filtered dual-axis rocker signal is converted into an analog-to-digital conversion, and the analog signal of the dual-axis rocker is converted into a digital signal to obtain a dual-axis rocker digital signal. The coordinate system transformation matrix operation is performed on the dual-axis rocker digital signal to convert it into spatial position data. Since there is a nonlinear relationship or coordinate system difference between the signal collected by the Hall sensor and the physical motion state of the rocker, a mathematical transformation method is used to map it to a unified three-dimensional coordinate system so that it can be directly used for robot kinematics and dynamics control. In the process of coordinate system transformation, the parameters of the transformation matrix are optimized in combination with the actual physical model and calibration data to improve the transformation accuracy. Zero drift compensation is performed on the rocker spatial position data. During long-term use, the Hall sensor and rocker structure will cause the zero position to shift due to factors such as temperature change, mechanical wear or circuit drift. If not compensated, it will cause errors or even instability in the control system. The zero drift compensation process is based on a preset calibration value or a real-time adaptive algorithm. By comparing the current rocker position data with the reference zero point, the compensation amount is calculated and the original data is corrected to obtain more accurate compensated position data. Based on the compensated position data, a mapping relationship between the rocker position and the control amount is established. The mapping relationship is obtained by fitting experimental data or model derivation, and defines how the rocker position change is mapped to the motion parameters of the robot joint, such as position, velocity or acceleration. Through mapping, the compensated position data is converted into rocker control data. The joystick control data is fused with the compensated position data. A filtering algorithm (such as Kalman filtering or weighted average method) is used to reduce the impact of noise and uncertainty on the data, while ensuring the system's ability to respond to emergencies, and obtain the digital value of the joystick position.

[0033] 102. Construct an interactive multi-model set including a uniform motion model, a uniform acceleration motion model and a Jerk motion model based on the digital quantity of the joystick position, and establish an interactive multi-model state space expression of the interactive multi-model set;

[0034] Specifically, based on the digital quantity of the joystick position, the initial state data including position, velocity and acceleration are established through analysis and deduction. When constructing the uniform motion model, based on the initial state data, the position and velocity are selected as state variables, and the acceleration is assumed to be zero to construct the state transfer equation. This equation describes how the state of the system evolves over time under the condition of no acceleration. Its mathematical form is expressed as a linear change of the state variable. For example, the position at the next moment is determined by the current velocity and time step, while the velocity remains constant. With such a setting, the uniform motion model can describe the dynamic characteristics of the joystick under the condition of constant velocity motion. In order to describe the acceleration or deceleration motion characteristics of the joystick, a uniform acceleration motion model is constructed. Based on the initial state data, the position, velocity and acceleration are selected as state variables, and the jerk is assumed to be zero. The state transfer equation reflects the relationship between the position and velocity over time under the condition of constant acceleration. For example, the position at the next moment is not only related to the current velocity and time step, but also needs to consider the influence of acceleration. Similarly, the velocity at the next moment is determined by the current acceleration. The model can accurately capture the dynamic behavior of the joystick in uniform acceleration or uniform deceleration motion and is applicable to a wider range of motion states. In order to improve the adaptability to complex motion states, a Jerk motion model is constructed based on the initial state data. Position, velocity, acceleration and jerk are selected as state variables, and the state transition equation contains higher-order dynamic change characteristics. In this model, jerk becomes the core variable to describe the system state, and the changes in position and velocity are further derived through its influence on acceleration. The model can describe the motion characteristics of the joystick under non-constant acceleration conditions, especially when dynamic mutations or rapid adjustments occur. The Jerk model can provide higher accuracy and dynamic adaptability. Based on the above uniform motion model, uniform acceleration motion model and Jerk motion model, they are combined to generate an interactive multi-model set. In the interactive multi-model set, each model corresponds to a different motion state, and can switch between multiple models according to the actual motion characteristics of the joystick. In order to ensure the accuracy and stability of the model, a system noise covariance matrix is ​​set for each model. The system noise covariance matrix is ​​used to describe the uncertainty in the model state transition, and its size reflects the degree of trust in the model error. Generally, the noise covariance of the uniform motion model is small because its state changes are relatively stable, while the covariance of the Jerk model is larger to accommodate more dynamic changes. At the same time, an observation equation is established based on the digital quantity of the joystick position, and a mapping relationship is established between the model state space and the observation space. The observation equation describes how the actual position, speed and other measurement data of the joystick are associated with the model state variables through the observation matrix. The setting of the observation matrix needs to be combined with the characteristics of the Hall sensor, the acquisition accuracy of the digital quantity of the joystick position and specific mathematical relationships. Through this mapping, the internal state of the model is matched with the data actually measured by the sensor.In order to improve the accuracy and robustness of the model, the observation matrix is ​​subjected to noise analysis to obtain the measurement noise matrix. The measurement noise matrix is ​​used to describe the uncertainty introduced by factors such as sensor error and external interference in the acquisition of the digital quantity of the joystick position. Through experimental calibration or statistical methods, the measurement noise can be quantified and its influence range can be expressed in matrix form. The interactive multi-model set, the system noise covariance matrix, the observation matrix and the measurement noise matrix are combined to establish a complete interactive multi-model state space expression.

[0035] 103. Perform Markov chain analysis and model weight equilibrium point analysis on the interactive multi-model state space expression to obtain the initial model conversion probability matrix;

[0036] Specifically, the Markov analysis is performed on the state space expression of the interactive multi-model, and the switching rules between the models are extracted by counting the transfer frequencies between the interactive multi-model sets. The calculation of the model transfer frequency is based on the continuous state observation data, and the transfer of the uniform motion model, uniform acceleration motion model and Jerk motion model in different time steps is recorded in combination with the dynamic changes of the digital quantity of the rocker position. The model transfer probability data is obtained by normalizing these transfer frequencies. The chain structure analysis of the model transfer probability data is performed. The transfer relationship between the models is described by constructing a first-order Markov chain model. The first-order Markov chain assumes that the future state of the system is only related to the current state and has nothing to do with the past historical state. This simplified model shows good computational efficiency and applicability in complex motion control. Based on this, the transfer relationship between the uniform motion model, uniform acceleration motion model and Jerk motion model is mapped into a Markov transfer sequence. The Markov transfer sequence is a set of ordered probability values ​​that represent the possibility of each model switching from the current state to other states. These sequences provide a basic framework for multi-model switching while capturing the dynamic changes of the system. According to the Markov transition sequence, the posterior probability of each model is calculated to quantify the priority of each model under given conditions. The calculation of the posterior probability is based on the Bayesian formula, combining the prior probability of the model and the current observation data to obtain a posterior probability value that is more in line with the actual dynamic characteristics. These posterior probabilities reflect the applicability of different models in the current state and provide input data for further weight balance analysis. In the weight balance analysis, the weight balance point is determined by weighted calculation of the posterior probability values ​​of each model. The weight balance point is a key indicator that indicates the position where the relative importance of each model reaches a dynamic balance under a specific state. The weight change is calculated based on the model weight balance point. The weight change refers to the change range between each model when the model weight is dynamically adjusted. When calculating the weight change, the dynamic characteristics of the system and the change trend of the digital quantity of the joystick position are considered to generate the weight adjustment parameters. These parameters are used to update the state transition equation and calculate the state transition probability. The state transition probability is obtained by dynamically adjusting the conditional probability during the model transition process, reflecting the instant response capability of model switching. In this process, the weight adjustment parameters directly affect the distribution of state transition probabilities, ensuring the adaptability and robustness of the system to external disturbances. In order to ensure the accuracy and stability of the transition probability, the generated transition probability sequence is subjected to error analysis. Error analysis quantifies the deviation in the transition probability by comparing the historical data of model transitions with the current observation data, and optimizes it. During the optimization process, the least squares method or filtering algorithm is used to reduce the influence of random noise and error accumulation, and the optimized probability results are obtained. The optimized probability results are organized into a matrix form to generate the initial model transition probability matrix.The initial model transition probability matrix is ​​the core parameter of the multi-model interaction system, which describes the transition rules and dynamic relationships between models in the form of a matrix. Each element of the matrix represents the probability of transitioning from one model state to another model state.

[0037] 104. Calculate the instantaneous position change rate and the instantaneous acceleration value according to the digital quantity of the joystick position, and calculate the weight adjustment coefficient according to the initial model conversion probability matrix to obtain the target model probability distribution;

[0038] Specifically, the sequence of the digital quantity of the rocker position is rearranged to construct a sampling sequence of a fixed time window. The digital quantity of the rocker position is divided into equally spaced sampling points in time order to form a continuous time window sequence. The instantaneous position change rate is calculated by performing position difference operation on the time window sampling sequence. The position difference operation is based on the position change of every two adjacent sampling points, and accurately converts the change of the rocker motion state in time into the instantaneous velocity. The acceleration calculation is performed on the instantaneous position change rate to obtain the instantaneous acceleration value. The acceleration calculation is achieved by performing a difference operation on the velocity sequence, and its result reflects the dynamic change characteristics of the rocker in the current time window. Based on the initial model transition probability matrix, a state transition equation group is established. The state transition equation group describes the transfer relationship between each model in the interactive multi-model set in matrix form, covering the possibility and conditional probability of the system from one motion state to another. In this process, the structural characteristics of the initial model transition probability matrix are used to obtain the steady-state transfer characteristic quantity of each model through matrix solution operation. The steady-state transfer characteristic quantity reflects the relative dominance and dynamic stability of each model under long-term operation conditions. The weight mapping is performed on the steady-state transfer feature quantity. By combining the feature quantity with the time-related parameters, a weight change curve is generated to reflect the priority change trend of each model in dynamic switching. Based on the weight change curve, the weight adjustment coefficient is generated by calculating the optimal attenuation parameter. The calculation of the optimal attenuation parameter takes into account the dynamic characteristics of the model, the current system noise, and the changes in real-time input data to ensure the smoothness and adaptability of the weight adjustment process. The probability distribution of the instantaneous position change rate, the instantaneous acceleration value, and the weight adjustment coefficient is calculated. The probability value of each model is calculated by comprehensively considering the characteristics of the position change rate and the acceleration value, as well as the dynamic influence of the weight adjustment coefficient. The probability value of each model reflects the applicability and priority of the model as the target model at the current time step. These probability values ​​are represented by constructing a probability distribution matrix, and the dynamic distribution relationship between different models is reflected in the form of a matrix. In order to ensure the mathematical consistency of the probability distribution and the stability of the system, the probability distribution matrix is ​​regularized. The regularization process makes the sum of the probability values ​​of all models equal to 1 through normalization operation, and generates the final target model probability distribution.

[0039] 105. According to the probability distribution of the target model, the extended Kalman filter operation is performed on the interactive multi-model set to obtain the robot joint motion state quantity;

[0040] Specifically, based on the probability distribution of the target model, the mixed probability matrix is ​​calculated for each model in the interactive multi-model set. The initial mixed state quantity is generated by weighted averaging the current state estimate of each model and its weight in the probability distribution of the target model. The initial mixed state quantity is regarded as the comprehensive state description of multiple motion models at the current time step, reflecting the overall estimation of the system for the current motion state. The nonlinear state equation of each model in the interactive multi-model set is linearized by the Jacobian matrix. Based on the nonlinear dynamic equation of each model, the state equation is expanded by the first order Taylor near the current state point by calculating the Jacobian matrix to obtain the linearized state equation. The linearized state equation is substituted into the prediction equation group for one-step prediction calculation. Prediction calculation is the first step of the extended Kalman filter, which uses the current state estimate and the system dynamic equation to predict the state value of the system at the next moment. In this way, the state prediction value of each model in the interactive multi-model set is obtained, which represents the future state of the system under different model assumptions. At the same time, for each state prediction value, observation prediction calculation is performed, and the sensor output data, i.e., observation prediction quantity, is predicted using the observation equation of the system. Calculate the difference between the observed prediction and the actual observed data, i.e., the measurement innovation. The measurement innovation describes the deviation between the predicted state and the actual observation, and is the basis for the filter adjustment. Perform covariance analysis on the measurement innovation to quantify the size and direction of the prediction error and generate the measurement innovation covariance matrix. Based on this matrix, the Kalman gain is calculated using the optimal criterion of the Kalman filter. The Kalman gain is the core parameter of the filter and determines the weight distribution of the predicted value and the observed data in the state update. By optimizing the Kalman gain, the error of the state estimation is minimized and the filtering performance of the system is improved. The optimal gain coefficient is weightedly combined with the state prediction value of each model to calculate the state correction. The state correction is a correction to the predicted value, which is used to adjust the state estimation of the system to the actual observed data. The corrected state estimation combines the dynamic characteristics of the prediction model and the real-time nature of the actual observation, reflecting the dynamic adaptability of the extended Kalman filter to nonlinear systems. At the same time, based on the state correction, the state combination weights of each model in the interactive multi-model set are calculated. The state combination weight determines the contribution of each model at the current moment by performing a comprehensive calculation on the target probability distribution, state estimation error and weight adjustment parameters of each model. The motion state of the robot joint is obtained by fusion calculation of the state correction and the state combination weight. The joint motion state includes dynamic parameters such as joint position, velocity and acceleration.

[0041] 106. Perform quintic polynomial interpolation operation based on the robot joint motion state quantity to generate the robot joint motion control sequence.

[0042] Specifically, the boundary conditions are calculated based on the robot joint motion state obtained by the extended Kalman filter, including the initial position, final position, initial velocity, final velocity, initial acceleration, and final acceleration. These boundary conditions are the basis of the quintic polynomial trajectory planning, which can fully describe the starting and ending states of the motion, as well as the dynamic characteristics during the transition process. Through the boundary condition data, clear constraints are established for trajectory planning. The coefficient matrix of the quintic polynomial is constructed based on these boundary conditions. The form of the quintic polynomial trajectory equation is: . The coefficients are obtained by substituting the boundary conditions into . These coefficients can determine the shape of the entire motion trajectory, including the continuous changes of position, velocity and acceleration within the planning time interval. By matrixing the boundary conditions, the trajectory polynomial coefficients are solved and the polynomial trajectory is ensured to meet the specified initial and terminal state requirements. After obtaining the trajectory polynomial coefficients, the position coordinates of the trajectory points within the planning time interval are calculated to generate an initial trajectory point sequence. These trajectory points are discrete representations of the robot's motion trajectory at different time points, reflecting the spatial position distribution of the robot during the entire motion process. In order to ensure the smoothness of the trajectory, the velocity difference of adjacent trajectory points is analyzed for continuity. If the velocity difference is found to exceed the set threshold range, the trajectory point is smoothed, the position of the trajectory point is adjusted or the time step is reallocated to generate velocity-continuous trajectory data. Based on the velocity-continuous trajectory data, the trajectory acceleration is segmented and the continuity constraint is calculated to generate an acceleration-continuous trajectory curve. The acceleration-continuous trajectory curve is sampled at equal time intervals to generate motion discrete data, so that the motion controller can track the trajectory at a fixed clock cycle, thereby avoiding motion errors or control delays caused by time inequality. The motion discrete data represents the position and state of the robot in the workspace in Cartesian space. In order to convert the motion discrete data into specific joint commands, the joint coordinates of the trajectory points are transformed. Through forward and inverse kinematics solutions, the trajectory points in Cartesian space are mapped to the joint space. The forward solution operation converts the robot joint angle into the position of the end effector, while the inverse solution operation derives the joint angle from the position of the end effector. Through this step, the joint angle command corresponding to each time step is generated. In order to meet the clock synchronization requirements of the control system, the joint angle command is distributed regularly to ensure that the joint angle command can be consistent with the sampling period of the robot controller, so as to achieve precise control of the motion. These distributed commands form the robot joint motion control sequence, and each item of the control sequence represents the joint state of the robot at a specific time step.

[0043] In the embodiment of the present invention, a dual-axis Hall sensor structure and a signal processing method are used to achieve high-precision detection of the rocker position, and the noise and zero drift in the sensor signal are effectively suppressed through multi-stage filtering and compensation algorithms, thereby improving the accuracy of position detection. An interactive multi-model set including a uniform motion model, a uniform acceleration motion model and a Jerk motion model is constructed, which can fully describe the different motion states of the robot and ensure the accuracy of state estimation. Through Markov chain analysis and model weight balance point analysis, a dynamic model conversion mechanism is established to solve the problem of smooth transition when the robot switches between different motion states. The steady-state enhanced IMM algorithm is introduced to process the rocker control signal, which improves the control accuracy of the system in the steady-state working mode while ensuring rapid response capability. The extended Kalman filter and state interactive fusion strategy are used to effectively handle the nonlinear characteristics of the system and improve the accuracy of state estimation. The trajectory planning method based on quintic polynomial interpolation ensures the continuity and smoothness of the joint motion trajectory and reduces mechanical vibration. Through hierarchical signal processing and control strategies, the whole process control from rocker input to robot execution is realized.

[0044] In a specific embodiment, the process of executing step 101 may specifically include the following steps:

[0045] The Hall sensors are set at the X-axis and Y-axis directions at the bottom of the joystick, and the joystick position analog signals are collected by detecting the change of the magnetic field strength of the permanent magnet to obtain the dual-axis joystick position analog signals;

[0046] Performing low-noise operational amplifier amplification and filtering processing on the dual-axis rocker position analog signal to obtain an amplified and filtered dual-axis rocker signal, and performing analog-to-digital conversion on the amplified and filtered dual-axis rocker signal to obtain a dual-axis rocker digital signal;

[0047] A coordinate system transformation matrix operation is performed according to the dual-axis joystick digital signal to obtain the joystick spatial position data, and zero-point drift compensation is performed on the joystick spatial position data to obtain compensated position data;

[0048] A mapping relationship between the joystick position and the control amount is established based on the compensated position data to obtain the joystick control data, and the joystick control data is fused with the compensated position data to obtain the digital amount of the joystick position.

[0049] Specifically, the Hall sensor is precisely installed at the bottom of the joystick in the X-axis and Y-axis directions, and the position signal of the joystick in the two-axis direction is collected by detecting the change in the magnetic field strength of the permanent magnet. The core principle of the Hall sensor is to generate a voltage signal proportional to the change in the magnetic field strength caused by the change in the magnetic field strength caused by the movement of the permanent magnet with the joystick. Assume that the magnetic field strength in the X-axis and Y-axis directions is respectively and , the analog voltage signal output by the Hall sensor is expressed as:

[0050] ;

[0051] in, and are the output voltages of the Hall sensor in the X-axis and Y-axis directions, and is the sensitivity coefficient of the Hall sensor. In order to improve the signal quality, these analog signals are amplified by low-noise operational amplifiers. The amplifier increases the weak Hall sensor output signal to a range suitable for analog-to-digital conversion. The amplified signal is expressed as:

[0052] ;

[0053] in, and is the gain of the amplifier. The amplified signal is filtered. The high-frequency noise and electromagnetic interference are removed by a low-pass filter, and the filtered signal is smoother, which is expressed as and The filtered signal is converted into a digital signal through analog-to-digital conversion to obtain the digital signal of the dual-axis joystick and . Assume that the resolution of the analog-to-digital converter is (such as 12 or 16 bits), the full-scale voltage is , then the expression of the digital signal is:

[0054]

[0055] in, Indicates rounding operation. Digital signal and Represents the discrete position data of the joystick in the X-axis and Y-axis directions. In order to map these data to actual space coordinates, coordinate system transformation matrix operations are performed. Assume that the physical displacement range of the joystick is , the digital signal is mapped to the physical coordinates through linear transformation and :

[0056] ;

[0057] What you get and It is the actual position of the joystick in two-dimensional space. Because the Hall sensor and the system itself are affected by temperature drift or mechanical offset, the zero point position changes, which affects the calculated spatial position. and Perform zero point drift compensation. Suppose the zero point offset is and , compensated position data and It is expressed as:

[0058] ;

[0059] The compensated position data can more accurately reflect the actual physical position of the joystick, eliminating the error caused by the offset. and , establish the mapping relationship between the joystick position and the control amount. Assume that the robot control amount is directly determined by the joystick position, and the control amount is determined by the function If the control quantity is the joint velocity and , using proportional relationship for mapping:

[0060] ;

[0061] in is a proportional coefficient that reflects the sensitivity of the joystick position to the control quantity. The control data of the joystick is fused with the compensated position data to generate the final digital position of the joystick. The data fusion process is achieved through weighted averaging or filtering methods. For example, the current data and historical data are combined through time-weighted filtering to generate the final digital position of the joystick:

[0062] ;

[0063] in, and is the location data of the last moment, is the fusion weight factor, and its value range is [0, 1]. and It is the digital value of the joystick position at the current moment.

[0064] In a specific embodiment, the process of executing step 102 may specifically include the following steps:

[0065] Establish initial state data including position, speed and acceleration based on the digital quantity of the joystick position;

[0066] According to the initial state data, the state transfer equation with position and velocity as state variables and acceleration as zero is set to construct a uniform motion model; according to the initial state data, the state transfer equation with position, velocity and acceleration as state variables and jerk as zero is set to construct a uniform acceleration motion model; according to the initial state data, the state transfer equation with position, velocity, acceleration and jerk as state variables is set to construct a Jerk motion model;

[0067] Generate an interactive multi-model set based on the uniform motion model, uniform acceleration motion model and Jerk motion model, set the corresponding system noise covariance matrix of the interactive multi-model set, establish an observation equation for the digital quantity of the joystick position, and obtain the observation matrix by establishing a mapping relationship from the state vector to the observation space;

[0068] The measurement noise matrix is ​​obtained by performing noise analysis on the observation matrix, and the interacting multi-model set, the system noise covariance matrix, the observation matrix and the measurement noise matrix are combined into the interacting multi-model state space expression.

[0069] Specifically, basic dynamic information is obtained through the position signal of the joystick, including position, velocity, and acceleration. Assume that the digital value of the joystick position is , which means that at discrete time steps The joystick position data at the time. The speed is calculated by the time difference method. and acceleration :

[0070] ;

[0071] in, is the sampling time interval, and are the position and velocity values ​​of the previous time step respectively. The initial state data consists of position ,speed and acceleration Composition, defined as the state vector:

[0072] ;

[0073] When constructing a uniform motion model, it is assumed that the acceleration , the state variables only include position and velocity. The state transition equation describes the change of the system from the current state to the next time step state, and the expression is:

[0074] ;

[0075] in, is the state transition matrix, is the input matrix, is the control input (usually zero), is the process noise, which indicates the uncertainty of the system. For the uniform acceleration motion model, the state variables are expanded to position, velocity and acceleration, assuming that the jerk (rate of change of acceleration) is zero. The state transfer equation is:

[0076] ;

[0077] in, is the state transition matrix, For the Jerk motion model, the state variables are expanded to position, velocity, acceleration and jerk, and the state transfer equation is:

[0078] ;

[0079] in, Based on the above three models, an interactive multi-model set is generated, where each model is used to describe the behavior of the joystick under different dynamic conditions. In order to make these models adapt to the actual noise situation, a corresponding system noise covariance matrix is ​​set for each model. , describing the degree of random error in the model. Assume that the variance of process noise on each state variable is , the system noise covariance matrix is ​​in diagonal matrix form:

[0080] ;

[0081] In order to associate the digital value of the joystick position with the state vector, an observation equation is established. The observation equation describes the mapping relationship between the state vector and the observation space, assuming that the observation value only contains the position of the joystick , the observation equation is:

[0082] ;

[0083] in, is the observed value, is the observation matrix, is the measurement noise. In order to analyze the impact of measurement noise, noise analysis is performed on the observation matrix to obtain the measurement noise covariance matrix Assume that the measurement noise is Gaussian distributed and its variance is , then the measurement noise covariance matrix is:

[0084]

[0085] The interacting multi-model set, the system noise covariance matrix, the observation matrix, and the measurement noise covariance matrix are combined to form the interacting multi-model state space expression:

[0086] ;

[0087] .

[0088] In a specific embodiment, the process of executing step 103 may specifically include the following steps:

[0089] Perform Markov analysis on the state space expression of the interacting multi-model and calculate the transition frequency between the interacting multi-model sets to obtain the model transition probability data;

[0090] The chain structure analysis of the model transfer probability data is carried out, and the transfer relationship between the uniform motion model, uniform acceleration motion model and Jerk motion model is established through the first-order Markov chain model to obtain the Markov transfer sequence.

[0091] The posterior probability is calculated according to the Markov transition sequence to obtain the posterior probability value of each model, and the posterior probability value of each model is weighted and balanced to obtain the model weight balance point;

[0092] The weight change amount is calculated based on the model weight balance point to obtain the weight adjustment parameter, and the weight adjustment parameter is input into the state transfer equation to calculate the state transfer probability to obtain the transfer probability sequence;

[0093] The error analysis of the transition probability sequence is performed to obtain the probability optimization results, which are then organized into a matrix form to generate the initial model transition probability matrix.

[0094] Specifically, the dynamic behaviors between the interactive multi-model sets are counted and modeled. Assume that the interactive multi-model set consists of three models, namely the uniform motion model, the uniform acceleration motion model and the Jerk motion model. These models describe the state evolution of the system under different dynamic conditions. By performing Markov analysis on the state space expression of the model, the transition frequency between models is quantified to obtain the model transition probability data. For the transition frequency between each model, it is counted by recording the state changes of the model between consecutive time steps. Assume Indicates that from the model Transfer to Model The total number of state transitions is Model transition probability Defined as a model Transfer to Model The conditional probability is calculated as:

[0095] ;

[0096] in, Satisfy the normalization condition By analyzing the data from multiple experiments, the model transition probability matrix is ​​constructed. , each element of which Indicates that from the model To Model The transition probability of the system is . Based on the transition probability data, the chain structure analysis is performed. Using the first-order Markov chain assumption, that is, the next state of the system is only related to the current state and has nothing to do with the past state, the transition relationship between the uniform motion model, the uniform acceleration motion model and the Jerk motion model is established. Assume that the current state of the model is , the next state is , the expression of Markov transition sequence is:

[0097] ;

[0098] Under the assumption of a first-order Markov chain, the transfer sequence of the model is calculated, and its steady-state distribution is further derived. satisfy:

[0099] ;

[0100] in, Representation Model The long-term probability of possession, is the transition probability matrix. By solving the equations, we can obtain the steady-state probability distribution of each model. The posterior probability value is calculated based on the Markov transition sequence. Assume that the observed joystick position data is , the posterior probability Represents the possibility of the current model given the observed data, calculated by the Bayesian formula:

[0101]

[0102] in, is the likelihood function, which means that in the model The probability of the observed data appearing is is the prior probability. The posterior probability result is used in the weight balance operation to determine the dynamic weight of each model. The weight balance point is obtained by optimizing the balance of the model weights at the current moment. Assume that the weight is ,satisfy , through balancing operations, ensure that the dynamic distribution of the weights of each model is reasonable, reflecting its importance in the current state. Based on the weight balance point, calculate the weight change. The weight change is defined as the adjustment amplitude of the weight, which is used to dynamically update the transition probability. Let the weight adjustment parameter be , the calculation formula is:

[0103] ;

[0104] in, is the adjustment factor that controls the sensitivity of weight adjustment. The state transition equation is input to update the transition probability and calculate the transition probability sequence. In order to verify the accuracy of the transition probability sequence, error analysis is performed. Error analysis evaluates the quality of the transition probability by calculating the deviation between the predicted state and the actual state. The error is defined as for:

[0105] ;

[0106] in, is the predicted state, is the observed state. Through statistical analysis of the error, the parameters of the transition probability are optimized. The optimized transition probability sequence is organized into a matrix form to generate the initial model transition probability matrix. Taking the three models as an example, the initial model transition probability matrix is:

[0107]

[0108] In a specific embodiment, the process of executing step 104 may specifically include the following steps:

[0109] The digital quantities of the joystick position are rearranged to obtain a time window sampling sequence, and a position difference operation is performed on the time window sampling sequence to obtain an instantaneous position change rate;

[0110] Perform acceleration calculation on the instantaneous position change rate to obtain the instantaneous acceleration value;

[0111] Based on the initial model conversion probability matrix, a state transition equation group is established, and the steady-state probability of each model in the interactive multi-model set is solved by matrix calculation to obtain the steady-state transfer characteristic quantity;

[0112] Perform weight mapping on the steady-state transfer characteristic quantity to obtain a weight change curve, and calculate the optimal attenuation parameter based on the weight change curve to obtain a weight adjustment coefficient;

[0113] The probability distribution of the instantaneous position change rate, instantaneous acceleration value and weight adjustment coefficient is calculated to obtain the probability value of each model. Based on the probability value of each model, a probability distribution matrix is ​​constructed and regularized to obtain the probability distribution of the target model.

[0114] Specifically, the sequence of the digital quantity of the joystick position is rearranged. is a discrete time step By choosing a fixed-length time window , the most recent The sampling points are arranged to form a time window sampling sequence This sequence captures the movement trend of the joystick in a short period of time. The instantaneous position change rate (i.e., speed) is calculated by position difference operation. Using the adjacent position points in the time window, the speed The calculation formula is:

[0115]

[0116] in, is the sampling time interval, and The digital values ​​of the joystick position at the current moment and the previous moment respectively. By performing the above calculations on each pair of adjacent points in the time window, a set of velocity values ​​is generated to describe the instantaneous motion state of the joystick. Acceleration is calculated based on the instantaneous position change rate Acceleration is the rate of change of velocity with respect to time, and is obtained by performing a difference operation on the velocity series again:

[0117]

[0118] in, and are the speed values ​​at the current and previous moments respectively. Through calculation, the dynamic information of the joystick movement is obtained, including instantaneous speed and instantaneous acceleration. Based on the above dynamic data and the initial model conversion probability matrix , establish a group of state transition equations to analyze the steady-state probability of each model in the interactive multi-model set. Assuming that the initial model set includes uniform motion model, uniform acceleration motion model and Jerk motion model, the steady-state probability distribution vector is , satisfying the following matrix equation:

[0119] ;

[0120] in, is the initial model transition probability matrix, whose elements Indicates that from the model To Model By solving this set of equations, we can obtain the steady-state probability value of each model. , reflecting the dynamic stability of the model under long-term operation. The steady-state probability value is weighted and a weight change curve is generated. The weight mapping process is implemented by introducing a nonlinear function to map the steady-state probability to a time-dependent dynamic weight. For example, an exponential function is used to represent the weight change:

[0121] ;

[0122] in, is the attenuation coefficient, which controls the decay rate of the weight over time. The optimal attenuation parameter is calculated through the weight change curve. , so that the smoothness and real-time performance of model switching are balanced. Based on the optimal attenuation parameter, the weight adjustment coefficient is calculated , used to adjust the weight distribution between models:

[0123] ;

[0124] in, is an adjustment factor that reflects the sensitivity of weight changes to model switching. The adjusted weights are used to combine the instantaneous position change rate and the instantaneous acceleration value to calculate the probability distribution. In order to calculate the probability value of each model, the instantaneous velocity, acceleration and weight adjustment coefficient are combined to construct a probability density function. For example, for model , its probability value It is expressed as:

[0125] ;

[0126] in, It is in the model The joint distribution of velocity and acceleration is Represents the model parameters. Through normalization, ensure that the sum of all model probability values ​​is 1. Construct the probability distribution matrix based on the probability values ​​of each model:

[0127] ;

[0128] By regularizing the probability distribution matrix, the final target model probability distribution is obtained. The purpose of regularization is to eliminate the influence of abnormal data and make the probability distribution matrix more stable.

[0129] In a specific embodiment, the process of executing step 105 may specifically include the following steps:

[0130] Based on the probability distribution of the target model, the mixed probability matrix of each model in the interactive multi-model set is calculated to obtain the initial mixed state quantity;

[0131] The initial mixed state quantity is linearized by Jacobian matrix, and the nonlinear state equation of each model in the interactive multi-model set is expanded by first-order Taylor to obtain the linearized state equation;

[0132] Substitute the linearized state equation into the prediction equation group for one-step prediction calculation to obtain the state prediction value of each model in the interactive multi-model set, and perform observation prediction calculation on the state prediction value of each model to obtain the observation prediction amount;

[0133] Perform difference calculation on the observed prediction and the actual observed data to obtain the measurement innovation covariance, and calculate the Kalman gain based on the measurement innovation covariance to obtain the optimal gain coefficient;

[0134] The optimal gain coefficient and the state prediction value of each model are weightedly combined to obtain the state correction value, and the state combination weight of each model in the interactive multi-model set is calculated based on the state correction value to obtain the robot joint motion state value.

[0135] Specifically, the mixed probability matrix of each model in the interactive multi-model set is calculated through the target model probability distribution. Assume that the interactive multi-model set includes models, and the state variables of each model are , the corresponding target probability distribution is . Initial mixed state quantity Expressed as a weighted average of all model states:

[0136] ;

[0137] in, is an initial estimate of the overall state of the system, It is The target probability of the model satisfies The initial mixed state is linearized by the Jacobian matrix. Since the models in the interactive multi-model set may be nonlinear, their state equations are expressed as:

[0138] ;

[0139] in, is the nonlinear state transfer function, is the control input, is the process noise. In order to facilitate prediction and correction, the state equation is expanded by first order Taylor and linearized to obtain:

[0140] ;

[0141] in, is the Jacobian matrix of the state equation, which describes the local linearized relationship between the state and the input. Through this linearization process, the complex dynamics of the nonlinear system is transformed into a manageable linear model. Substitute the linearized state equation into the prediction equation group for one-step prediction. The goal of one-step prediction is to use the current state and control input to estimate the state value at the next moment. For the first The state prediction value of the model is:

[0142] ;

[0143] in, is the control input matrix. The state prediction values ​​of all models are weighted to obtain the overall state prediction of the system:

[0144] ;

[0145] When performing observation prediction calculations on the predicted values, the observation equation of the system is used. Assume that the observation equation is:

[0146] ;

[0147] in, is the observation matrix, is the observation noise. The predicted observation value (i.e., the observation prediction amount) is:

[0148] ;

[0149] Perform difference calculation between the observed predicted value and the actual observed data to obtain the new measurement information:

[0150] ;

[0151] The measurement innovation reflects the deviation between the predicted value and the actual observation and is used to update the state estimate. It represents the statistical characteristics of the new information, and its calculation formula is:

[0152] ;

[0153] in, is the covariance matrix of the predicted state, is the observation noise covariance matrix. Based on the measurement innovation covariance, the Kalman gain is calculated:

[0154] ;

[0155] The Kalman gain is used to combine the predicted value and the actual observation to generate the optimal correction. The state estimate is updated by weighted combination of the optimal gain coefficient and the state prediction value:

[0156] ;

[0157] The system state combination weight is calculated through the modified state of each model and the corresponding target probability value:

[0158] ;

[0159] After normalization, the weights are used to weightedly combine the state correction values ​​of all models to obtain the final robot joint motion state:

[0160] .

[0161] In a specific embodiment, the process of executing step 106 may specifically include the following steps:

[0162] The boundary conditions of the robot joint motion state are calculated to obtain the polynomial boundary condition data, and the coefficient matrix is ​​constructed for the polynomial boundary condition data. The position, velocity and acceleration constraints are substituted into the fifth-order polynomial equation to solve and obtain the trajectory polynomial coefficients.

[0163] Based on the trajectory polynomial coefficients, the position coordinates of the trajectory points within the planned time interval are calculated to obtain the initial trajectory point sequence, and the speed continuity analysis is performed on the initial trajectory point sequence. The speed difference between adjacent trajectory points is controlled within the threshold range for smoothing operation to obtain speed continuous trajectory data;

[0164] Based on the velocity continuous trajectory data, the trajectory acceleration is calculated with piecewise continuity constraints to obtain the acceleration continuous trajectory curve, and the acceleration continuous trajectory curve is sampled at equal time intervals to obtain motion discrete data;

[0165] The motion discrete data is transformed into joint coordinates, and the Cartesian space trajectory points are mapped to the joint space for forward and inverse solution operations to obtain the joint angle commands. The joint angle commands are then distributed in a timely manner according to the control system clock synchronization requirements to obtain the robot joint motion control sequence.

[0166] Specifically, the boundary condition data of the polynomial trajectory is determined according to the starting and ending states. Assume that the robot joint motion state includes the initial position , initial velocity , initial acceleration , Target location , Target speed and target acceleration , these conditions define the boundary constraints of the trajectory. The goal is to generate a quintic polynomial that describes the motion trajectory through these conditions. The trajectory equation of the quintic polynomial is expressed as:

[0167] ;

[0168] in, It's time The joint position under are the coefficients of the polynomial to be solved. Substitute the boundary conditions of position, velocity and acceleration into the quintic polynomial and its derivatives to form a set of constraint equations. The velocity and acceleration are The first derivative of and the second derivative :

[0169] ;

[0170] ;

[0171] Substitution and (the planning duration of the trajectory), forming the following constraints:

[0172]

[0173] The above constraints are written in matrix form:

[0174] ;

[0175] in, is the coefficient matrix, generated by substituting the boundary conditions into the polynomial; is the unknown coefficient vector; is the known boundary condition vector. By solving the linear equations, we get the coefficients of the quintic polynomial. Based on the obtained trajectory polynomial coefficients, we discretize the planning time interval and calculate the joint position at each time point. . Set planning time Divided into equal intervals sampling points, the sampling time is ,in The calculation formula of trajectory points is:

[0176] ;

[0177] in, is the sampling time point; is the joint position corresponding to the sampling time point. The velocity continuity analysis is performed on the generated trajectory point sequence. The velocity difference between adjacent points is calculated:

[0178]

[0179] if If the set threshold is exceeded, the trajectory points need to be adjusted or the time intervals need to be redistributed to smooth the velocity distribution and ensure that the velocity changes continuously throughout the trajectory. Based on the velocity smoothed trajectory, the acceleration is constrained to be piecewise continuous. Acceleration is calculated by second-order difference:

[0180]

[0181] By adjusting the discontinuous acceleration segments, a smooth acceleration curve is generated. After smoothing, the trajectory curve is sampled at equal time intervals to generate motion discrete data, including joint positions. ,speed , acceleration The discrete data are transformed into joint coordinates to map the trajectory from Cartesian space to joint space. The position of the end effector (Cartesian coordinates) is transformed into joint coordinates through the forward and inverse kinematics equations of the robot. Convert to joint angle :

[0182] ;

[0183] in, is the position information of the end effector; is the corresponding joint angle. According to the clock synchronization requirements of the control system, the generated joint angle commands are distributed regularly according to the sampling time to form the robot joint motion control sequence.

[0184] = The above describes the industrial robot control method based on Hall rocker control in the embodiment of the present invention. The following describes the industrial robot control device based on Hall rocker control in the embodiment of the present invention. Please refer to Figure 2 In one embodiment of the present invention, an industrial robot control device based on Hall rocker control includes:

[0185] The acquisition module 201 is used to amplify and filter the analog signal of the dual-axis rocker position acquired by the Hall sensor to obtain a digital value of the rocker position;

[0186] A construction module 202 is used to construct an interactive multi-model set including a uniform motion model, a uniform acceleration motion model and a Jerk motion model based on the digital quantity of the joystick position, and establish an interactive multi-model state space expression of the interactive multi-model set;

[0187] An analysis module 203 is used to perform Markov chain analysis and model weight balance point analysis on the interactive multi-model state space expression to obtain an initial model conversion probability matrix;

[0188] The calculation module 204 is used to calculate the instantaneous position change rate and the instantaneous acceleration value according to the digital quantity of the joystick position, and calculate the weight adjustment coefficient according to the initial model conversion probability matrix to obtain the target model probability distribution;

[0189] The filtering module 205 is used to perform an extended Kalman filter operation on the interactive multi-model set according to the target model probability distribution to obtain the robot joint motion state quantity;

[0190] The generation module 206 is used to perform a quintic polynomial interpolation operation based on the robot joint motion state quantity to generate a robot joint motion control sequence.

[0191] Through the cooperation of the above components, the dual-axis Hall sensor structure and signal processing method are adopted to achieve high-precision detection of the rocker position, and the noise and zero drift in the sensor signal are effectively suppressed through multi-level filtering and compensation algorithms, thereby improving the accuracy of position detection. An interactive multi-model set including uniform motion model, uniform acceleration motion model and Jerk motion model is constructed, which can fully describe the different motion states of the robot and ensure the accuracy of state estimation. Through Markov chain analysis and model weight balance point analysis, a dynamic model conversion mechanism is established to solve the problem of smooth transition when the robot switches between different motion states. The steady-state enhanced IMM algorithm is introduced to process the rocker control signal, which improves the control accuracy of the system in the steady-state working mode while ensuring rapid response capability. The extended Kalman filter and state interactive fusion strategy are adopted to effectively deal with the nonlinear characteristics of the system and improve the accuracy of state estimation. The trajectory planning method based on quintic polynomial interpolation ensures the continuity and smoothness of the joint motion trajectory and reduces mechanical vibration. Through hierarchical signal processing and control strategies, the whole process control from rocker input to robot execution is realized.

[0192] above Figure 2 The industrial robot control device based on Hall rocker control in the embodiment of the present invention is described in detail from the perspective of modular functional entities. The industrial robot control device based on Hall rocker control in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0193] Figure 3 3 is a schematic diagram of the structure of an industrial robot control device based on Hall rocker control provided by an embodiment of the present invention. The industrial robot control device 300 based on Hall rocker control may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 310 (for example, one or more processors) and a memory 320, and one or more storage media 330 (for example, one or more mass storage device terminals) storing application programs 333 or data 332. Among them, the memory 320 and the storage medium 330 can be short-term storage or permanent storage. The program stored in the storage medium 330 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the industrial robot control device 300 based on Hall rocker control. Furthermore, the processor 310 can be configured to communicate with the storage medium 330, and execute a series of instruction operations in the storage medium 330 on the industrial robot control device 300 based on Hall rocker control to implement the steps of the above-mentioned industrial robot control method based on Hall rocker control.

[0194] The industrial robot control device 300 based on Hall rocker control may also include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input and output interfaces 360, and / or one or more operating systems 331, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will appreciate that Figure 3 The structure of the industrial robot control device based on Hall rocker control shown does not constitute a limitation of the industrial robot control device based on Hall rocker control provided by the present invention, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.

[0195] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the industrial robot control method based on Hall rocker control.

[0196] 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.

[0197] 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 invention is essentially or the part that contributes to the prior art or the whole 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 a computer 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 invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0198] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. 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 invention.

Claims

1. An industrial robot control method based on Hall rocker control, characterized in that: The method comprises: The analog signal of the dual-axis rocker position collected by the Hall sensor is amplified and filtered to obtain the digital value of the rocker position; Based on the digital quantity of the rocker position, an interactive multi-model set including a uniform motion model, a uniform acceleration motion model and a Jerk motion model is constructed, and an interactive multi-model state space expression of the interactive multi-model set is established; Performing Markov chain analysis and model weight balance point analysis on the interactive multi-model state space expression to obtain an initial model transition probability matrix; specifically comprising: performing Markov analysis on the interactive multi-model state space expression, and calculating the transfer frequency between the interactive multi-model sets to obtain model transfer probability data; performing chain structure analysis on the model transfer probability data, establishing the transfer relationship between the uniform motion model, the uniform acceleration motion model and the Jerk motion model through a first-order Markov chain model, and obtaining a Markov transfer sequence; performing posterior probability calculation according to the Markov transfer sequence to obtain the posterior probability value of each model, and performing weight balance operation on the posterior probability value of each model to obtain the model weight balance point; performing weight change calculation based on the model weight balance point to obtain weight adjustment parameters, and inputting the weight adjustment parameters into the state transfer equation to perform state transfer probability calculation to obtain a transfer probability sequence; performing error analysis on the transfer probability sequence to obtain a probability optimization result, and organizing the probability optimization result into a matrix form to generate an initial model transition probability matrix; Calculating the instantaneous position change rate and the instantaneous acceleration value according to the digital quantity of the rocker position, and calculating the weight adjustment coefficient according to the initial model conversion probability matrix to obtain the target model probability distribution; According to the probability distribution of the target model, an extended Kalman filter operation is performed on the interactive multi-model set to obtain the robot joint motion state quantity; A fifth-order polynomial interpolation operation is performed based on the robot joint motion state quantity to generate a robot joint motion control sequence.

2. The industrial robot control method based on Hall rocker control according to claim 1, characterized in that: The amplifying and filtering the dual-axis rocker position analog signal collected by the Hall sensor to obtain the rocker position digital quantity includes: The Hall sensors are set at the X-axis and Y-axis directions at the bottom of the joystick, and the joystick position analog signals are collected by detecting the change of the magnetic field strength of the permanent magnet to obtain the dual-axis joystick position analog signals; Performing low-noise operational amplifier amplification and filtering processing on the dual-axis rocker position analog signal to obtain an amplified and filtered dual-axis rocker signal, and performing analog-to-digital conversion on the amplified and filtered dual-axis rocker signal to obtain a dual-axis rocker digital signal; Performing coordinate system transformation matrix operation according to the dual-axis joystick digital signal to obtain joystick spatial position data, and performing zero drift compensation on the joystick spatial position data to obtain compensated position data; A mapping relationship between a joystick position and a control amount is established based on the compensated position data to obtain joystick control data, and the joystick control data is fused with the compensated position data to obtain a digital amount of the joystick position.

3. The industrial robot control method based on Hall rocker control according to claim 2 is characterized in that: The constructing of an interactive multi-model set including a uniform motion model, a uniform acceleration motion model and a Jerk motion model based on the digital quantity of the rocker position, and establishing an interactive multi-model state space expression of the interactive multi-model set includes: Establishing initial state data including position, speed and acceleration based on the digital quantity of the rocker position; According to the initial state data, the state transfer equation with position and velocity as state variables and acceleration as zero is set to construct a uniform motion model; according to the initial state data, the state transfer equation with position, velocity and acceleration as state variables and jerk as zero is set to construct a uniform acceleration motion model; according to the initial state data, the state transfer equation with position, velocity, acceleration and jerk as state variables is set to construct a Jerk motion model; Generate an interactive multi-model set based on the uniform motion model, the uniform acceleration motion model and the Jerk motion model, set the corresponding system noise covariance matrix of the interactive multi-model set, establish an observation equation for the digital quantity of the joystick position, and obtain an observation matrix by establishing a mapping relationship from a state vector to an observation space; Noise analysis is performed on the observation matrix to obtain a measurement noise matrix, and the interacting multi-model set, the system noise covariance matrix, the observation matrix and the measurement noise matrix are combined into an interacting multi-model state space expression.

4. The industrial robot control method based on Hall rocker control according to claim 1, characterized in that: The step of calculating the instantaneous position change rate and the instantaneous acceleration value according to the digital quantity of the rocker position, and calculating the weight adjustment coefficient according to the initial model conversion probability matrix to obtain the target model probability distribution includes: Rearranging the digital quantities of the rocker position to obtain a time window sampling sequence, and performing a position difference operation on the time window sampling sequence to obtain an instantaneous position change rate; Performing acceleration calculation on the instantaneous position change rate to obtain an instantaneous acceleration value; Establishing a state transition equation group based on the initial model conversion probability matrix, performing matrix solution operation on the steady-state probability of each model in the interactive multi-model set, and obtaining a steady-state transfer characteristic quantity; Performing weight mapping on the steady-state transfer characteristic quantity to obtain a weight change curve, and performing optimal attenuation parameter calculation based on the weight change curve to obtain a weight adjustment coefficient; The instantaneous position change rate, the instantaneous acceleration value and the weight adjustment coefficient are subjected to probability distribution calculation to obtain the probability values ​​of each model, and a probability distribution matrix is ​​constructed based on the probability values ​​of each model and regularized to obtain the target model probability distribution.

5. The industrial robot control method based on Hall rocker control according to claim 4 is characterized in that: The method of performing an extended Kalman filter operation on the interactive multi-model set according to the target model probability distribution to obtain the robot joint motion state quantity includes: Based on the target model probability distribution, a mixed probability matrix is ​​calculated for each model in the interactive multi-model set to obtain an initial mixed state quantity; Performing Jacobian matrix linearization processing on the initial mixed state quantity, performing first-order Taylor expansion on the nonlinear state equation of each model in the interactive multi-model set, and obtaining a linearized state equation; Substituting the linearized state equation into the prediction equation group to perform a one-step prediction calculation to obtain the state prediction value of each model in the interactive multi-model set, and performing observation prediction calculation on the state prediction value of each model to obtain an observation prediction amount; Performing a difference operation between the observed prediction and the actual observed data to obtain a measurement innovation covariance, and performing a Kalman gain calculation based on the measurement innovation covariance to obtain an optimal gain coefficient; The optimal gain coefficient and the state prediction value of each model are weightedly combined to obtain a state correction amount, and the state combination weight of each model in the interactive multi-model set is calculated based on the state correction amount to obtain the robot joint motion state amount.

6. The industrial robot control method based on Hall rocker control according to claim 5, characterized in that: The method of performing a quintic polynomial interpolation operation based on the robot joint motion state quantity to generate a robot joint motion control sequence comprises: Perform boundary condition calculation on the robot joint motion state quantity to obtain polynomial boundary condition data, perform coefficient matrix construction operation on the polynomial boundary condition data, substitute position, velocity and acceleration constraints into the quintic polynomial equation to solve, and obtain trajectory polynomial coefficients; Based on the trajectory polynomial coefficients, position coordinate calculation is performed on the trajectory points within the planned time interval to obtain an initial trajectory point sequence, and speed continuity analysis is performed on the initial trajectory point sequence, and the speed difference between adjacent trajectory points is controlled within a threshold range for smoothing calculation to obtain speed continuous trajectory data; Based on the velocity continuous trajectory data, performing piecewise continuity constraint calculation on the trajectory acceleration to obtain an acceleration continuous trajectory curve, and performing equal time interval sampling on the acceleration continuous trajectory curve to obtain motion discrete data; The motion discrete data is transformed into joint coordinates, and the Cartesian space trajectory points are mapped to the joint space for forward and inverse solution operations to obtain joint angle commands, and the joint angle commands are distributed in a timely manner according to the control system clock synchronization requirements to obtain the robot joint motion control sequence.

7. An industrial robot control device based on Hall rocker control, characterized in that: Used to execute the industrial robot control method based on Hall rocker control as described in any one of claims 1 to 6, the device comprises: The acquisition module is used to amplify and filter the analog signal of the dual-axis rocker position acquired by the Hall sensor to obtain the digital value of the rocker position; A construction module, used to construct an interactive multi-model set including a uniform motion model, a uniform acceleration motion model and a Jerk motion model based on the digital quantity of the rocker position, and establish an interactive multi-model state space expression of the interactive multi-model set; An analysis module, used for performing Markov chain analysis and model weight balance point analysis on the interactive multi-model state space expression to obtain an initial model conversion probability matrix; A calculation module, used to calculate the instantaneous position change rate and the instantaneous acceleration value according to the digital quantity of the rocker position, and calculate the weight adjustment coefficient according to the initial model conversion probability matrix to obtain the target model probability distribution; A filtering module, used to perform an extended Kalman filter operation on the interactive multi-model set according to the target model probability distribution to obtain the robot joint motion state quantity; A generation module is used to perform a quintic polynomial interpolation operation based on the robot joint motion state quantity to generate a robot joint motion control sequence.

8. An industrial robot control device based on Hall rocker control, characterized in that: The industrial robot control device based on Hall rocker control includes: a memory and at least one processor, wherein instructions are stored in the memory; The at least one processor calls the instruction in the memory to enable the industrial robot control device based on Hall rocker control to execute the industrial robot control method based on Hall rocker control as described in any one of claims 1-6.

9. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the industrial robot control method based on Hall rocker control as described in any one of claims 1 to 6 is implemented.

Images