High-precision driving control method and system for dexterous hand

Abstract

The application relates to the technical field of robot driving control, and discloses a high-precision driving control method and system for a dexterous hand, which comprises the following steps: collecting end voltage, phase current, rotor angular velocity and external contact force signal of a microminiature hollow cup motor; obtaining a coil resistance estimation value and an inverse electromotive force coefficient estimation value based on the end voltage and the phase current by using an extended Kalman filtering algorithm; mapping the coil resistance estimation value into a viscous friction coefficient evolution value, and mapping the external contact force signal into a Coulomb friction torque correction value; combining the rotor angular velocity to calculate a total friction compensation torque; and normalizing a total electromagnetic torque target value after superposition by using the inverse electromotive force coefficient estimation value to generate a driving current instruction. The application solves the problem of a thermal compensation blind area of a micro driving module without a temperature sensor, and realizes high-precision force and position control of the dexterous hand under complex working conditions with variable temperature and variable load.

Description

Technical Field

[0001] This invention relates to the field of robot drive control technology, specifically to a high-precision drive control method and system for a dexterous hand. Background Technology

[0002] Multi-DOF dexterous hands, as key components for robots to perform precise operations, employ extremely small hollow cored motors in conjunction with micro-precision reducers as joint drive modules to achieve high power density output within a limited mechanical space. During actual operation, the extremely small hollow cored motors, due to their very low heat capacity, experience rapid temperature rise during frequent starts and stops or continuous torque output. Because the internal space of the micro-drive module is relatively narrow, it is difficult to integrate physical temperature sensors within the micro-drive module to directly measure the real-time temperature.

[0003] Existing dexterous hand drive control technologies employ fixed-parameter PID control algorithms or friction compensation strategies based on static models. These traditional control strategies neglect the impact of multi-physics coupling effects on micro-drive modules. Specifically, existing control strategies cannot detect changes in the winding resistance of extremely small coreless motors as temperature increases, nor can they detect the hydrodynamic characteristics of the lubricating grease viscosity decreasing with increasing temperature inside micro-precision reducers.

[0004] When the temperature of a miniature hollow cup motor changes, the viscosity coefficient of the grease inside the micro precision reducer will drift. However, existing control strategies lack a temperature feedback mechanism and can only estimate based on parameters at room temperature, resulting in distortion of the friction compensation torque.

[0005] Furthermore, when a dexterous hand performs a grasping task, the joint ends are subjected to dynamically changing external contact forces. These external contact forces directly act on the gear transmission mechanism of the miniature precision reducer, altering the pressure distribution of the gear meshing backlash and consequently causing nonlinear fluctuations in the Coulomb friction torque.

[0006] Existing control methods cannot establish a dynamic mapping relationship between external contact force and internal friction torque. More seriously, the magnetic properties of the permanent magnets in miniature coreless motors decay with increasing temperature, leading to a decrease in the motor's torque constant. Without real-time correction of the back electromotive force coefficient, even if the controller outputs the theoretically correct current command, the actual electromagnetic torque output by the miniature coreless motor will be lower than expected. Due to the dynamic time-varying characteristics of thermodynamic, fluid dynamic, and electromagnetic parameters, existing dexterous hand-driven control systems struggle to maintain high-precision force-position control performance under complex operating conditions of varying temperature and load.

[0007] Therefore, this invention proposes a high-precision drive control method and system for dexterous hands to overcome the shortcomings of existing technologies. Summary of the Invention

[0008] To address the shortcomings of existing technologies, this invention provides a high-precision drive control method and system for dexterous hands, solving the problem that existing dexterous hand drive control systems struggle to maintain high-precision force and position control performance under complex working conditions of varying temperature and load.

[0009] To achieve the above objectives, the present invention provides the following technical solution:

[0010] The first aspect of this invention provides a high-precision drive control method for a dexterous hand, comprising the following steps:

[0011] The system performs high-frequency acquisition and preprocessing of status signals, synchronously locks the terminal voltage, phase current, rotor angular velocity, and external contact force signals of the end force sensor of the miniature hollow cup motor within the same control cycle, and uses a digital low-pass filter to filter out noise interference, ensuring strict alignment of multi-physics parameters on the time axis.

[0012] Real-time online observation of motor electrical parameters is performed. Based on the discretized state-space model of a miniature hollow cup motor, the extended Kalman filter algorithm is used to process the terminal voltage and the phase current. Without the need for an external temperature sensor, the estimated value of the coil resistance as a function of temperature and the estimated value of the back electromotive force coefficient characterizing the magnetic field strength of the magnet are separated and observed in real time.

[0013] The dynamic evolution of cross-domain parameters is performed to construct coupled evolution paths of thermal and fluid dynamics and coupled evolution paths of load and structural dynamics. In the coupled evolution path of thermal and fluid dynamics, the thermal conduction characteristics between the miniature hollow cup motor and the micro precision reducer are used to take the estimated value of the coil resistance as a temperature index. Combined with the thermal conduction hysteresis compensation logic, the evolution value of the viscosity friction coefficient of the lubricating grease inside the micro precision reducer at the current temperature is calculated.

[0014] In the coupled evolution path of load and structural dynamics, the change in gear meshing backlash of the micro precision reducer is calculated based on the external contact force signal, and the external contact force signal is mapped to the Coulomb friction torque correction value.

[0015] Perform full-parameter friction observation and compensation torque generation. Input the rotor angular velocity, the evolution value of the viscous friction coefficient, and the correction value of the Coulomb friction torque into the nonlinear friction state observer based on the LuGre model, and calculate the total friction compensation torque in real time, including the fluid viscous resistance compensation component and the mechanical dry friction resistance compensation component.

[0016] Perform thermal drift compensation and final command synthesis to obtain the nominal required torque output by the position control loop. Algebraically superimpose the nominal required torque with the total friction compensation torque to obtain the target value of the total electromagnetic torque. Normalize the target value of the total electromagnetic torque using the estimated value of the back electromotive force coefficient to automatically compensate for the torque constant decay caused by motor temperature rise and generate the final drive current command.

[0017] Furthermore, the method also includes a sensorless force sensing mechanism based on current observation. When a fault is detected in the end force sensor, the external contact force acting on the joint end is estimated by using the back electromotive force coefficient estimate, the total friction compensation torque, the phase current, and the rotor angular velocity through a dynamic inverse algorithm. The estimated external contact force is then fed back to correct the Coulomb friction model to achieve fault-tolerant control of the system.

[0018] Furthermore, in order to improve computational efficiency, the dynamic evolution of the cross-domain parameters adopts a combination of table lookup and linear interpolation, and the evolution value of the viscous friction coefficient is quickly obtained through a pre-stored resistance and viscosity coefficient mapping table.

[0019] A second aspect of the present invention provides a high-precision drive control system for a dexterous hand, comprising:

[0020] The status signal acquisition and preprocessing module is configured to trigger synchronous acquisition commands, acquire and filter the terminal voltage, phase current, rotor angular velocity and external contact force signals of the end force sensor of the miniature hollow cup motor, and provide a time-aligned data basis for multi-physics field decoupling.

[0021] The motor electrical parameter observation module is configured to run an extended Kalman filter algorithm. Based on the real-time data of the terminal voltage and the phase current, it dynamically calculates the estimated values ​​of the coil resistance and the back electromotive force coefficient, thereby realizing online monitoring of the internal thermal impedance characteristics and magnetic field characteristics of the motor.

[0022] The cross-domain parameter evolution module is configured to establish a mapping relationship from electrical domain parameters to mechanical domain parameters and fluid domain parameters. Based on the temperature information represented by the coil resistance estimate, the cross-domain parameter evolution module evolves the viscous friction coefficient evolution value of the internal lubricating grease of the micro precision reducer. At the same time, based on the load information represented by the external contact force signal, it evolves the Coulomb friction torque correction value of the micro precision reducer.

[0023] The full-parameter friction observation module is configured to run a nonlinear friction state observer. It comprehensively considers the rotor angular velocity, the evolution value of the viscous friction coefficient, and the correction value of the Coulomb friction torque to calculate the total friction compensation torque that can offset the variable temperature viscous resistance and the variable load dry friction force.

[0024] The thermal drift compensation and final command synthesis module is configured to perform torque superposition and thermal attenuation correction calculations. It uses the estimated value of the back electromotive force coefficient as a divisor to correct the superimposed torque target and generate a final drive current command that can overcome the thermal attenuation of the motor body and the nonlinear friction of the reducer.

[0025] The technical solution provided by this invention reconstructs the electrical and physical model of a miniature hollow cup motor in real time using an extended Kalman filter algorithm. It uses changes in coil resistance to sense the thermal state of the system and changes in the back electromotive force coefficient to sense the attenuation of magnetic properties. Through a cross-domain parameter evolution mechanism, it establishes a dynamic correlation between electrical parameters and fluid viscosity state, as well as between external contact force and mechanical friction state, thus solving the problem of thermal compensation in micro drive modules when space is limited and temperature sensors cannot be installed.

[0026] This also solves the problem of decreased control accuracy caused by friction model parameter drift in precision reducers under varying temperature and load conditions. The synthesis of the final drive command fully considers the thermal decay effect of the motor torque constant, realizing high-precision force-position hybrid control of dexterous hand joints in complex thermodynamic environments.

[0027] This invention provides a high-precision drive control method and system for a dexterous hand. It has the following beneficial effects:

[0028] 1. This invention, by performing dynamic evolution of cross-domain parameters, utilizes the thermal conduction characteristics between a miniature hollow cup motor and a micro precision reducer to map the estimated coil resistance value to the viscous friction coefficient evolution value of the lubricating grease inside the micro precision reducer, and maps the external contact force signal to the Coulomb friction torque correction value. This solves the problem of thermal state perception blind spot caused by the inability to install temperature sensors in the micro drive module in a very small space, and realizes dual dynamic decoupling and compensation of fluid viscous resistance and mechanical dry friction resistance, thereby improving the force and position control accuracy of the dexterous hand under complex working conditions of variable temperature and variable load.

[0029] 2. In the thermal drift compensation and final command synthesis steps, this invention uses the back electromotive force (EMF) coefficient estimate obtained in real time through an extended Kalman filter algorithm to normalize the total electromagnetic torque target value. Since the back EMF coefficient estimate reflects the physical characteristic of the permanent magnet's magnetic flux density decreasing with increasing temperature, using the back EMF coefficient estimate as a divisor to generate the final drive current command automatically increases the current amplitude to compensate for the decrease in torque constant caused by temperature rise in the miniature hollow cup motor. This ensures that the actual mechanical torque output by the dexterous hand joint strictly matches the nominal torque required by the control system.

[0030] 3. This invention is equipped with a sensorless force sensing extension scheme based on current observation. When the end force sensor fails, the external contact force acting on the end of the joint is estimated by using the back electromotive force coefficient estimate, total friction compensation torque, phase current and rotor angular velocity through a dynamic inverse algorithm. The converged multiphysics parameter model is used to maintain the dynamic correction capability of the Coulomb friction torque, avoiding the divergence of the control system or the loss of control of the dexterous hand due to sensor failure, and enhancing the robustness and reliability of the dexterous hand drive control system. Attached Figure Description

[0031] Figure 1 This is a flowchart of the method of the present invention;

[0032] Figure 2 This is a system block diagram of the present invention.

[0033] Legend

[0034] 1. Miniature drive module; 11. Miniature hollow cup motor; 12. Miniature precision reducer; 2. Multi-dimensional sensing unit; 21. Position encoder; 22. End force sensor; 23. Current acquisition circuit; 24. Voltage acquisition circuit; 3. Microcontroller; 4. Power drive circuit; 5. Status signal acquisition and preprocessing module; 6. Motor electrical parameter observation module; 7. Cross-domain parameter evolution module; 71. Thermal and fluid parameter mapping module; 72. Load and structural parameter mapping module; 8. Full-parameter friction observation module; 9. Thermal drift compensation and final command synthesis module; 91. Drive torque synthesis module; 92. Position control loop; 93. Current closed-loop controller; 94. Trajectory planning module. Detailed Implementation

[0035] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] See attached document Figure 2 This invention provides a high-precision drive control system for a dexterous hand. The system hardware architecture mainly consists of a microcontroller 3, a power drive circuit 4, a micro drive module 1, and a multi-dimensional sensing unit 2. The high-precision drive control system for the dexterous hand executes drive logic based on the evolution of multi-physics parameters to achieve high-precision force-position hybrid control of the dexterous hand's end effector.

[0037] The core computing unit of the high-precision drive control system for the dexterous hand is a microcontroller 3. The microcontroller 3 uses a high-performance processor with a floating-point arithmetic unit. The microcontroller 3 is connected to the power drive circuit 4 via electrical wiring. The output of the power drive circuit 4 is connected to the micro-drive module 1 via wires. The micro-drive module 1 is the power source for the dexterous hand joints. The micro-drive module 1 is mechanically connected to the dexterous hand joint linkages, used to drive the fingers of the dexterous hand to produce flexion and extension movements.

[0038] The micro drive module 1 adopts an integrated design, comprising a miniature hollow cup motor 11 and a miniature precision reducer 12. The miniature hollow cup motor 11, as the actuator converting electrical energy into mechanical energy, features low rotational inertia and low inductance. There is direct rigid physical contact between the stator housing of the miniature hollow cup motor 11 and the housing of the miniature precision reducer 12. This rigid physical contact constitutes a heat conduction path for transferring heat from the miniature hollow cup motor 11 to the miniature precision reducer 12. The miniature precision reducer 12 is filled with grease. The viscosity characteristics of the grease change non-linearly with the internal temperature of the miniature precision reducer 12. The Joule heat generated by the miniature hollow cup motor 11 is transferred to the miniature precision reducer 12 through the heat conduction path, causing the temperature of the grease to fluctuate synchronously with the coil temperature of the miniature hollow cup motor 11.

[0039] The output shaft of the miniature hollow cup motor 11 is coaxially coupled to the input end of the micro precision reducer 12. The output shaft of the micro precision reducer 12 is connected to the dexterity hand joint linkage. The micro precision reducer 12 adopts a micro planetary gear reducer or a micro harmonic reducer. When subjected to external load compression, the gear meshing backlash inside the micro precision reducer 12 undergoes microscopic deformation, thereby changing the Coulomb friction characteristics of the micro precision reducer 12.

[0040] The multi-dimensional sensing unit 2 includes a position encoder 21, an end effector force sensor 22, a current acquisition circuit 23, and a voltage acquisition circuit 24. The position encoder 21 is mounted at the tail of the miniature hollow cup motor 11. The position encoder 21 is used to acquire the rotation angle signal of the rotor of the miniature hollow cup motor 11 in real time. The position encoder 21 transmits the rotation angle signal to the microcontroller 3 via a signal line. The end effector force sensor 22 is mounted at the fingertip of the dexterous hand or at the end of the joint linkage of the dexterous hand. The end effector force sensor 22 is used to acquire the contact force signal when the dexterous hand comes into contact with an external object. The end effector force sensor 22 transmits the contact force signal to the microcontroller 3.

[0041] A current acquisition circuit 23 is connected in series in the armature circuit of the miniature hollow cup motor 11. The current acquisition circuit 23 is used to detect the phase current flowing through the coil of the miniature hollow cup motor 11 in real time. A voltage acquisition circuit 24 is connected in parallel to the input terminal of the miniature hollow cup motor 11. The voltage acquisition circuit 24 is used to detect the terminal voltage applied across the miniature hollow cup motor 11 in real time. The analog signal output terminals of the current acquisition circuit 23 and the voltage acquisition circuit 24 are respectively connected to the analog-to-digital conversion interface of the microcontroller 3. The microcontroller 3 uses the acquired phase current and terminal voltage to construct an electrical physical model of the miniature hollow cup motor 11 based on Kirchhoff's voltage law.

[0042] To clarify the basis for the microcontroller 3's calculation of the physical state of the miniature hollow cup motor 11, the dynamic electrical characteristics of the miniature hollow cup motor 11 follow the following voltage balance equation: ;

[0043] In the voltage balance equation: Defined as the instantaneous terminal voltage value of the miniature hollow cup motor 11 acquired by the voltage acquisition circuit 24; Defined as the coil resistance of the miniature hollow cup motor 11, the coil resistance is a physical quantity that changes with temperature; Defined as the instantaneous value of the phase current flowing through the miniature hollow cup motor 11, collected by the current acquisition circuit 23; The coil inductance is defined as that of the miniature hollow cup motor 11, and the coil inductance is considered to be constant. Defined as the differential of the phase current with respect to time; Defined as the back electromotive force coefficient of the miniature hollow cup motor 11, the back electromotive force coefficient is affected by the temperature of the magnet. Defined as the rotor angular velocity of the miniature hollow cup motor 11, the rotor angular velocity is obtained by differentiating the signal collected by the position encoder 21.

[0044] The microcontroller 3 is configured to observe the values ​​of coil resistance and back electromotive force coefficient in real time based on the voltage balance equation and the extended Kalman filter algorithm, and map the value of coil resistance to the viscous state of the lubricating grease inside the micro precision reducer 12, thereby realizing online reconstruction of the system's thermal and hydrodynamic parameters without the need for an external temperature sensor.

[0045] See attached document Figure 2The logical architecture of the high-precision drive control system for the dexterous hand defines the data flow and processing logic within the system. The control system's logical architecture runs within the microcontroller 3 in the form of software algorithms or firmware. To achieve high-precision hierarchical control, the microcontroller 3 is logically divided into five core functional modules: a status signal acquisition and preprocessing module 5, a motor electrical parameter observation module 6, a cross-domain parameter evolution module 7, a full-parameter friction observation module 8, and a thermal drift compensation and final command synthesis module 9.

[0046] The status signal acquisition and preprocessing module 5 is responsible for the synchronization, locking, and noise reduction of the underlying sensor data. The motor electrical parameter observation module 6 is responsible for reconstructing the internal physical model of the motor based on current and voltage data. The cross-domain parameter evolution module 7 is responsible for establishing the parameter mapping relationship between the electric domain, thermal domain, fluid domain, and mechanical domain. The full-parameter friction observation module 8 is responsible for calculating the nonlinear friction compensation amount in real time. The thermal drift compensation and final command synthesis module 9 is responsible for integrating the data from each stage to generate the final drive current command.

[0047] The control system's logical architecture is configured with two main data links that process in parallel: a motion control link for generating motion commands and a parameter observation link for real-time correction of model parameters.

[0048] The motion control link begins at the trajectory planning module 94. The trajectory planning module 94 generates desired joint position commands based on the dexterous hand's task. These desired joint position commands are transmitted to the position control loop 92. The position control loop 92 calculates the deviation between the desired joint position commands and the actual joint position signals fed back by the position encoder 21. Based on the deviation calculation result, the position control loop 92 outputs the nominal required torque. The nominal required torque is the raw torque command that does not yet include friction compensation and thermal attenuation correction.

[0049] The parameter observation link and the motion control link operate synchronously. Microcontroller 3 uses an extended Kalman filter observer to process the phase current signal fed back from current acquisition circuit 23 and the terminal voltage signal fed back from voltage acquisition circuit 24. Based on the electrical physical model of the miniature hollow cup motor 11, the extended Kalman filter observer calculates the estimated coil resistance and back electromotive force coefficient in real time. The estimated coil resistance is transmitted to the thermal and fluid parameter mapping module 71. The estimated back electromotive force coefficient is transmitted to the drive torque synthesis module 91.

[0050] The thermal and fluid parameter mapping module 71 stores the viscosity-temperature characteristic curve data of the lubricating grease in the miniature precision reducer 12. The module uses the received coil resistance estimate as a temperature index variable. Based on the coil resistance estimate, the module calculates the evolution value of the viscous friction coefficient inside the miniature precision reducer 12 at the current moment. The evolution value of the viscous friction coefficient characterizes the fluid damping characteristics of the lubricating grease at the current temperature. The evolution value of the viscous friction coefficient is transmitted to the nonlinear friction state observer.

[0051] The load and structural parameter mapping module 72 receives the contact force signal fed back by the end force sensor 22. Based on the contact force signal, the load and structural parameter mapping module 72 calculates the change in gear meshing backlash of the miniature precision reducer 12. The load and structural parameter mapping module 72 maps the contact force signal to a Coulomb friction torque correction value. The Coulomb friction torque correction value characterizes the dry friction characteristics of the miniature precision reducer 12 under load. The Coulomb friction torque correction value is transmitted to the nonlinear friction state observer.

[0052] The nonlinear friction state observer simultaneously receives the rotor angular velocity signal, the evolution value of the viscous friction coefficient, and the correction value of the Coulomb friction torque from the miniature hollow cup motor 11. The nonlinear friction state observer runs a dynamic friction calculation algorithm based on the LuGre model. The nonlinear friction state observer calculates the total friction compensation torque. The total friction compensation torque includes dual compensation components for fluid viscous drag and mechanical dry friction drag. The total friction compensation torque is transmitted to the drive torque synthesis module 91.

[0053] The drive torque synthesis module 91 is the convergence node of the control system logic architecture. It receives the nominal demand torque from the motion control link, the total friction compensation torque from the nonlinear friction state observer, and the estimated back EMF coefficient from the extended Kalman filter observer. Based on the torque-current conversion relationship, the drive torque synthesis module 91 generates the final reference current command.

[0054] The driving torque synthesis module 91 follows the following driving synthesis equation when generating the reference current command:

[0055] ;

[0056] In the driving synthesis equation: Defined as the first Each control cycle outputs a reference current command to the current loop; Defined as the nominal required torque output by position control loop 92; Defined as the total friction compensation torque output by the nonlinear friction state observer; Defined as the back EMF coefficient estimate output by the extended Kalman filter observer, the back EMF coefficient estimate is numerically equivalent to the motor torque constant at the current temperature.

[0057] The microcontroller 3 inputs the calculated reference current command to the current closed-loop controller 93. The current closed-loop controller 93 controls the on / off state of the power drive circuit 4 through pulse width modulation technology. The power drive circuit 4 outputs a drive voltage to the miniature hollow cup motor 11, thereby driving the dexterous hand joint to complete high-precision force-position control actions.

[0058] See attached document Figure 1 To obtain accurate input data, the status signal acquisition and preprocessing module 5 is configured to perform a high-frequency acquisition and preprocessing method for status signals. The status signal acquisition and preprocessing module 5 performs this method within each fixed control cycle.

[0059] At the start of the control cycle, microcontroller 3 triggers a synchronous acquisition command. This command controls the current acquisition circuit 23, voltage acquisition circuit 24, position encoder 21, and end effector force sensor 22 to synchronously lock onto the current physical state values. Synchronous acquisition ensures that state data from different physical dimensions are strictly aligned on the time axis. This time axis alignment eliminates decoupling errors in multi-physics parameters caused by sampling timing deviations.

[0060] Voltage acquisition circuit 24 detects the instantaneous voltage value at the input terminal of the miniature hollow cup motor 11. Voltage acquisition circuit 24 converts the instantaneous voltage value into digital terminal voltage raw data. Current acquisition circuit 23 detects the instantaneous current value flowing through the armature winding of the miniature hollow cup motor 11. Current acquisition circuit 23 converts the instantaneous current value into digital phase current raw data. Position encoder 21 reads the current mechanical angle of the rotor of the miniature hollow cup motor 11. Position encoder 21 outputs the current mechanical angle as digital rotor position raw data. End effector force sensor 22 measures the contact pressure between the end effector and the external environment. End effector force sensor 22 outputs the contact pressure as digital external contact force raw data.

[0061] Microcontroller 3 receives raw terminal voltage data, raw phase current data, raw rotor position data, and raw external contact force data. Microcontroller 3 uses a digital low-pass filter to filter these raw data. The digital low-pass filter removes high-frequency electromagnetic interference noise and sensor measurement white noise. The digital low-pass filter outputs the filtered terminal voltage signal, the filtered phase current signal, and the filtered external contact force signal.

[0062] Microcontroller 3 calculates the rotor angular velocity of the miniature hollow cup motor 11 based on the filtered raw rotor position data. To obtain a low-noise, real-time rotor angular velocity, microcontroller 3 employs a backward differential algorithm combined with a moving average filtering strategy. Microcontroller 3 determines the current rotor angular velocity based on the following angular velocity calculation equation:

[0063] ;

[0064] In the equation for calculating angular velocity: Defined as the first The rotor angular velocity of the miniature hollow cup motor 11 is calculated in one control cycle; Defined as the first Each control cycle consists of raw rotor position data acquired by position encoder 21 and filtered; Defined as the first Each control cycle consists of raw rotor position data acquired by position encoder 21 and filtered; Defined as the sampling time interval for the microcontroller 3 to execute the control loop.

[0065] The microcontroller 3 stores the calculated rotor angular velocity, filtered terminal voltage signal, filtered phase current signal, and filtered external contact force signal in its random access memory. The data stored in the random access memory will serve as input variables for subsequent motor electrical parameter observation steps, cross-domain parameter evolution steps, and friction torque observation steps.

[0066] See attached document Figure 1 After signal acquisition is completed, the motor electrical parameter observation module 6 executes a real-time online observation method for motor electrical parameters. This method utilizes the internal algorithm logic of the microcontroller 3 to separate and estimate the coil resistance and back electromotive force coefficient, which vary with temperature, during the operation of the miniature hollow cup motor 11. The motor electrical parameter observation module 6 achieves synchronous observation of the coil resistance and back electromotive force coefficient by executing an extended Kalman filter algorithm.

[0067] The microcontroller 3 first establishes a discretized state-space model of the miniature hollow cup motor 11. The discretized state-space model defines the phase current flowing through the miniature hollow cup motor 11, the coil resistance, and the back electromotive force coefficient as state variables. To estimate the slowly varying parameter of the coil resistance, the discretized state-space model assumes that the derivative of the coil resistance with respect to time is zero. Similarly, to estimate the slowly varying parameter of the back electromotive force coefficient, the discretized state-space model assumes that the derivative of the back electromotive force coefficient with respect to time is zero. The microcontroller 3 uses the terminal voltage as the input to the discretized state-space model and the phase current as the observed output.

[0068] Microcontroller 3 executes a prediction step and an update step within each control cycle. In the prediction step, microcontroller 3 predicts the prior state value for the current control cycle based on the state estimate from the previous control cycle and the current terminal voltage input. In the update step, microcontroller 3 calculates the Kalman gain matrix. The Kalman gain matrix is ​​used to determine the confidence weights between the predicted and actual measured values. Microcontroller 3 uses the deviation between the actual phase current value fed back by current acquisition circuit 23 and the predicted phase current value, combined with the Kalman gain matrix, to correct the prior state value, thereby obtaining the posterior state estimate for the current control cycle.

[0069] Through the aforementioned iterative calculation process, microcontroller 3 outputs the converged coil resistance estimate and back electromotive force (EMF) coefficient estimate in real time. The coil resistance estimate characterizes the current impedance characteristics of the winding of the miniature hollow cup motor 11, which is linearly positively correlated with the winding temperature. Therefore, the coil resistance estimate is used by subsequent control logic as a numerical basis for determining the internal thermal state of the system. The back EMF coefficient estimate characterizes the current magnetic field strength characteristics of the miniature hollow cup motor 11, which is negatively correlated with the magnet temperature. The back EMF coefficient estimate is used by subsequent control logic to correct the torque output command.

[0070] To clarify the mathematical iterative process of the extended Kalman filter algorithm, microcontroller 3 performs calculations based on the following discretized state observation equations:

[0071] ;

[0072] In the discretized state observation equations: Defined as the first The prior prediction value of the phase current for each control cycle; Defined as the first The posterior estimate of the phase current for each control cycle; Defined as the sampling time interval; Defined as the coil inductance constant of the miniature hollow cup motor 11; Defined as the first Terminal voltage measurement value for each control cycle; Defined as the first The posterior estimate of the coil resistance for each control cycle; Defined as the first The posterior estimate of the back electromotive force coefficient for each control cycle; Defined as the first Rotor angular velocity per control cycle; Defined as the first The prior prediction value of the coil resistance for each control cycle; Defined as the first The prior prediction of the back electromotive force coefficient for each control cycle.

[0073] Microcontroller 3 uses the state variables predicted by the above equations, combined with the current measurement error at the current moment, and applies the update rule of the standard extended Kalman filter to finally output the first... Estimated coil resistance value per control cycle and back electromotive force coefficient estimates .

[0074] See attached document Figure 1 After acquiring the electrical parameters, the cross-domain parameter evolution module 7 executes the dynamic evolution mechanism of the cross-domain parameters. This mechanism addresses the drift of physical parameters in the micro-precision reducer 12 under varying temperature and load conditions. Upon acquiring the estimated coil resistance and external contact force signal, the microcontroller 3 immediately initiates the dynamic evolution mechanism. This mechanism includes both a thermo-hydrodynamic coupling evolution path and a load-structural dynamic coupling evolution path.

[0075] The cross-domain parameter evolution module 7 internally includes a thermal-fluid parameter mapping subunit and a load-structural parameter mapping subunit. In the coupled thermal-fluid dynamics evolution path, the microcontroller 3 utilizes the thermal conduction characteristics between the miniature hollow cup motor 11 and the micro-precision reducer 12 to map the estimated coil resistance value to the rheological state of the lubricating grease inside the micro-precision reducer 12. Since the stator of the miniature hollow cup motor 11 is in close contact with the housing of the micro-precision reducer 12, the increase in the estimated coil resistance value directly reflects the increase in the overall temperature of the module. The increase in the overall temperature of the module causes the dynamic viscosity of the lubricating grease inside the micro-precision reducer 12 to decrease exponentially. Based on pre-calibrated thermal sensitivity parameters, the microcontroller 3 calculates and converts the estimated coil resistance value into the current viscous friction coefficient evolution value.

[0076] To accurately describe the above heat-fluid mapping relationship, microcontroller 3 executes the following viscosity coefficient evolution equation:

[0077] ;

[0078] In the viscosity coefficient evolution equation: Defined as the first The evolution value of the viscous friction coefficient of the micro precision reducer 12 obtained by calculation of each control cycle; Defined as the inherent viscous friction coefficient of the miniature precision reducer 12 at the nominal reference temperature; Defined as the heat and fluid sensitivity constant, which is determined by the physical properties of the grease; Defined as the first The estimated coil resistance value input in each control cycle; Defined as the cold resistance measurement value of the miniature hollow cup motor 11 at the nominal reference temperature.

[0079] In the coupled evolution path of load and structural dynamics, the microcontroller 3 uses data fed back from the end effector force sensor 22 to correct the dry friction model of the micro-precision reducer 12. When the dexterous hand performs a grasping task, the external contact force acts in the opposite direction on the micro-precision reducer 12 through the transmission chain. The presence of the external contact force increases the meshing tightness of the planetary gears or flexures inside the micro-precision reducer 12, which in turn leads to a nonlinear increase in the Coulomb friction torque of the mechanical transmission part. The microcontroller 3 reads the external contact force signal and calculates the current Coulomb friction torque threshold of the micro-precision reducer 12 in real time based on the contact mechanics model.

[0080] To accurately describe the above load-structure mapping relationship, microcontroller 3 executes the following Coulomb friction correction equation:

[0081] ;

[0082] In the Coulomb friction correction equation: Defined as the first The Coulomb friction torque correction value of the micro precision reducer 12 obtained by calculation in one control cycle; Defined as the basic Coulomb friction torque of the micro precision reducer 12 under no-load conditions; Defined as the load influence factor, the load influence factor characterizes the proportional coefficient by which external load is converted into internal friction increment; Defined as the first The absolute value of the external contact force signal acquired in each control cycle.

[0083] The microcontroller 3 simultaneously outputs the calculated evolution value of the viscous friction coefficient and the correction value of the Coulomb friction torque to the nonlinear friction state observer. The real-time update of the evolution value of the viscous friction coefficient and the correction value of the Coulomb friction torque ensures that the nonlinear friction state observer can always follow the dynamic changes of the physical characteristics of the micro precision reducer 12, thereby eliminating the prediction deviation of the fixed parameter model under hot or heavy load conditions.

[0084] See attached document Figure 1Based on the evolved parameters, the full-parameter friction observation module 8 executes the full-parameter friction observation and compensation torque generation method, which runs in the nonlinear friction state observer. The nonlinear friction state observer aims to solve the technical problem of the difficulty in accurately predicting the nonlinear friction torque of the micro-precision reducer 12 under complex working conditions. The microcontroller 3 inputs the rotor angular velocity, the evolution value of the viscous friction coefficient, and the correction value of the Coulomb friction torque obtained from the aforementioned steps as real-time inputs to the nonlinear friction state observer.

[0085] A nonlinear friction state observer constructs the tribodynamic equations of the micro-precision reducer 12 based on the LuGre dynamic friction model. The LuGre dynamic friction model introduces an internal state variable to simulate the average elastic deformation of the micro-hairs between the two contact surfaces. During the operation of the micro-precision reducer 12, the internal state variable dynamically changes with the rotor angular velocity. The nonlinear friction state observer first updates the Stribeck effect function based on the Coulomb friction torque correction value. The Stribeck effect function determines the critical characteristics when the micro-hairs slip. Using the real-time updated Coulomb friction torque correction value, the nonlinear friction state observer can accurately capture the nonlinear modulation effect of external load changes on the dry friction characteristics of the micro-precision reducer 12.

[0086] The nonlinear friction state observer calculates the total friction compensation torque by combining the evolution value of the viscous friction coefficient. The evolution value of the viscous friction coefficient directly affects the viscous friction component in the tribodynamic equation. Using the real-time updated evolution value of the viscous friction coefficient, the nonlinear friction state observer can accurately reflect the phenomenon of grease thinning caused by heat conduction due to the increase in motor coil temperature. In this way, the nonlinear friction state observer achieves dual decoupling and compensation for changes in mechanical side "thermal and fluid" parameters and "load and structural" parameters.

[0087] To quantitatively describe the above physical process and calculate the final frictional torque value that needs to be compensated, the nonlinear frictional state observer executes the following frictional torque observation equation:

[0088] ;

[0089] In the equation for observing frictional torque: Defined as the first The total friction compensation torque output by the nonlinear friction state observer for each control cycle; Defined as the microscopic contact stiffness coefficient of the contact surface of the 12 gears in a micro precision reducer; Defined as the first The average deformation state variable of the micro-hair on the contact surface of the micro-precision reducer in each control cycle; Defined as the micro damping coefficient of the contact surface of the 12 gears in a micro precision reducer; Defined as the first The rotor angular velocity of the miniature hollow cup motor 11 in one control cycle; Defined as the absolute value of the rotor angular velocity; Defined as the Stribeck effect function, the magnitude of the Stribeck effect function is determined in real time by the Coulomb friction torque correction value; Defined as the first The viscous friction coefficient evolution value is input for each control cycle, and the viscous friction coefficient evolution value contains thermal and fluid coupling information.

[0090] The nonlinear friction state observer updates the value of the average deformation state variable in real time using a numerical integration algorithm. The nonlinear friction state observer outputs the calculated total friction compensation torque to the drive torque synthesis module 91. This total friction compensation torque is then superimposed as a feedforward component into the final control command to counteract the nonlinear resistance generated inside the micro-precision reducer 12 due to temperature and load variations.

[0091] See attached document Figure 1 After completing all observations and compensation calculations, the thermal drift compensation and final command synthesis module 9 executes the thermal drift compensation and final command synthesis method, which is based on the output stage of multi-physics coupled drive control logic. The microcontroller 3 uses the thermal drift compensation and final command synthesis method to generate the final current control command applied to the miniature hollow cup motor 11. The microcontroller 3 integrates the nominal demand torque output from the position control loop 92, the total friction compensation torque output from the nonlinear friction state observer, and the estimated back EMF coefficient output from the extended Kalman filter algorithm into the drive torque synthesis module 91.

[0092] The thermal drift compensation and final command synthesis module 9 integrates a driving torque synthesis unit. The thermal drift compensation and final command synthesis module 9 first performs a torque superposition operation. The microcontroller 3 algebraically sums the nominal required torque and the total friction compensation torque. The nominal required torque is calculated by the position control loop 92 based on the joint position error, and represents the theoretical dynamic torque required to drive the dexterous hand joint movement. The total friction compensation torque is calculated by the nonlinear friction state observer, and represents the additional torque required to overcome the viscous resistance and Coulomb friction within the micro-precision reducer 12 at the current moment. The result of the algebraic summation constitutes the target value of the total electromagnetic torque to be output from the shaft end of the miniature hollow cup motor 11.

[0093] Microcontroller 3 then performs a thermal drift compensation division operation. Microcontroller 3 reads the current-moment back EMF coefficient estimate from the extended Kalman filter algorithm. In SI units, the torque constant of the miniature coreless motor 11 is numerically exact equal to the back EMF coefficient. Since the miniature coreless motor 11 uses permanent magnet excitation, the magnetic flux density of the permanent magnet undergoes irreversible or reversible decay with increasing temperature. This decay directly leads to a decrease in the torque constant of the miniature coreless motor 11. Microcontroller 3 uses the real-time updated back EMF coefficient estimate as a divisor to normalize the total electromagnetic torque target value.

[0094] By using the real-time back EMF coefficient estimate as the denominator, the microcontroller 3 can automatically increase the amplitude of the current command to compensate for the decrease in torque output capability caused by motor temperature rise. This division operation ensures that the actual mechanical torque output of the miniature coreless motor 11 strictly matches the total electromagnetic torque target value expected by the control system, thereby achieving decoupled control of the motor's thermal characteristics.

[0095] To clarify the generation logic of the final current command, microcontroller 3 calculates the final drive current command based on the following current synthesis equation:

[0096] ;

[0097] In the current composition equation: Defined as the first Each control cycle is calculated by microcontroller 3 and sent to the final drive current command of the current loop; Defined as the first Each control cycle is input by the nominal demand torque from the position control loop 92; Defined as the first The total friction compensation torque is input by the nonlinear friction state observer for each control cycle; Defined as the first Each control cycle is fed by an extended Kalman filter algorithm to estimate the back EMF coefficient, which is used to characterize the motor torque constant at the current temperature.

[0098] The microcontroller 3 sends the calculated final drive current command to the power drive circuit 4 via a digital-to-analog converter interface or a pulse width modulation module. The power drive circuit 4 adjusts the duty cycle of the voltage output to the miniature hollow cup motor 11 according to the final drive current command, thereby controlling the actual current flowing through the motor coil and driving the dexterous hand joint to achieve high-precision position and force control.

[0099] To support the physical validity of the aforementioned control logic, this control system employs a thermal conduction hysteresis analysis method for ultra-miniature modules as its theoretical basis. This method is used to reveal and quantify the temporal dynamic characteristics of heat transfer within the micro-drive module 1. The ultra-miniature hollow cup motor 11 and the micro-precision reducer 12 adopt a compact axial series structure. This compact axial series structure creates a low-thermal-resistance thermal conduction path between the stator housing of the ultra-miniature hollow cup motor 11 and the housing of the micro-precision reducer 12.

[0100] When the drive current flows through the armature winding of the miniature hollow cup motor 11, Joule heating is generated in the armature winding. This Joule heating causes the temperature of the armature winding to rise rapidly within a short period. The heat generated by the armature winding is conducted to the miniature precision reducer 12 through the air gap and stator assembly. The miniature precision reducer 12 consists of metal gears, bearings, and filled grease. The miniature precision reducer 12 has a defined specific heat capacity and mass. The specific heat capacity and mass of the miniature precision reducer 12 cause the overall temperature change of the miniature precision reducer 12 to lag behind the temperature change of the armature winding.

[0101] The microcontroller 3 constructs a heat conduction hysteresis model based on lumped-parameter thermal network theory. The heat conduction hysteresis model considers the armature winding of the miniature hollow cup motor 11 as a heat source node. The micro-precision reducer 12 is considered as a heated node. The heat source node and the heated node are connected through equivalent thermal resistance and equivalent thermal capacity. The heat conduction hysteresis model is used to quantitatively describe the time delay effect in the process of temperature transfer from the armature winding to the micro-precision reducer 12.

[0102] To accurately characterize the time delay effect of heat transfer, the heat conduction hysteresis analysis method follows the following thermodynamic differential equation:

[0103] ;

[0104] In the differential equation of thermodynamics: Defined as the rate of change of internal temperature of the miniature precision reducer 12 over time; Defined as the equivalent thermal conductivity resistance between the ultra-small hollow cup motor 11 and the micro precision reducer 12; Defined as the equivalent heat capacity of the miniature precision reducer 12; Defined as The armature winding temperature of the miniature hollow cup motor 11 is obtained by linear derivation from the estimated coil resistance. Defined as The internal average temperature of the micro precision reducer 12 determines the viscosity of the lubricating grease.

[0105] The microcontroller 3 uses the thermal time constant derived from the thermodynamic differential equation to perform phase compensation on the mapping logic from coil resistance to viscous friction coefficient. Phase compensation eliminates the error in predicting friction torque caused by thermal conduction hysteresis, ensuring that the evolution value of viscous friction coefficient can accurately reflect the current lubrication state of the micro precision reducer 12.

[0106] Furthermore, this control system employs a rheological property analysis method based on the micro-friction interface, which reveals the dynamic behavior evolution mechanism of the internal grease of the micro-precision reducer 12 under the influence of a temperature field. Micro-scale gaps exist in the meshing area between the planetary gears and the internal gear ring of the micro-precision reducer 12. Grease fills these gaps and forms a hydrodynamic lubrication film. This hydrodynamic lubrication film functions to transmit shear stress and isolate the metal surfaces during high-speed gear operation.

[0107] When the internal temperature of a miniature module increases, the thermal energy of the grease molecules increases. This increase in thermal energy leads to an increase in the mean free path between grease molecules. The increased mean free path weakens the van der Waals forces between molecules. This weakening of van der Waals forces manifests at the macroscopic physical level as a decrease in the dynamic viscosity of the grease. This decrease in dynamic viscosity alters the laminar shear resistance within the hydrodynamic lubrication film.

[0108] The microcontroller 3 utilizes the Andreid viscosity theory model to describe the exponential decay of the dynamic viscosity of lubricating grease with temperature. The Andreid viscosity theory model explains the physical essence of the viscous friction coefficient evolution equation in the aforementioned cross-domain parameter evolution mechanism from a molecular dynamics perspective. To quantitatively characterize this physical process, the rheological properties analysis method of the micro-friction interface follows the following viscosity-temperature physical equation:

[0109] ;

[0110] In the physical equations of viscosity and temperature: Defined as grease at absolute temperature The dynamic viscosity at the following levels; Defined as the base viscosity of grease under extreme high temperature conditions; Defined as the flow activation energy of grease molecules, the flow activation energy characterizes the minimum energy required for molecules to overcome the potential barrier and undergo relative displacement. Defined as the Boltzmann constant; Defined as the absolute temperature inside the micro precision reducer 12, the absolute temperature is indirectly derived from the estimated value of the coil resistance.

[0111] The viscous drag torque generated by the hydrodynamic lubrication film is directly proportional to the dynamic viscosity. Based on Newton's law of internal friction, the microcontroller 3 converts the calculated dynamic viscosity into shear stress acting on the gear surface. This shear stress generates a viscous friction torque on the drive shaft of the micro-precision reducer 12, hindering motion. By solving the viscosity-temperature physical equations in real time, the control system can explain and predict the inevitability of the decrease in damping characteristics of the micro-precision reducer 12 under high-temperature conditions from a microscopic physical perspective, thus providing solid physical theoretical support for the nonlinear friction state observer.

[0112] As an extension of the control system, this embodiment of the invention also provides a distributed extended implementation scheme based on a multiphysics coupling drive control method. This distributed extended implementation scheme is applied to a dexterous hand joint control system with multiple degrees of freedom. The dexterous hand joint control system includes a main communication controller and multiple node microcontrollers 3. Each node microcontroller 3 independently controls a miniature hollow cup motor 11. The main communication controller synchronously sends position commands to the multiple node microcontrollers 3 via a controller area network bus.

[0113] Multiple node microcontrollers 3 execute the aforementioned state signal acquisition steps, motor electrical parameter observation steps, cross-domain parameter evolution steps, and friction compensation steps in parallel. This parallel execution mechanism ensures that each joint of the dexterous hand can independently handle its own temperature drift and friction torque variations. This independent processing mechanism eliminates inconsistencies in thermal state between different joints caused by differences in load and heat dissipation conditions.

[0114] To enhance system reliability, this embodiment of the invention also provides a sensorless force sensing extension scheme based on current observation, which serves as a redundant backup mechanism in case of a failure of the end-effector force sensor 22. The microcontroller 3 monitors the external contact force signal output by the end-effector force sensor 22 in real time. The microcontroller 3 determines whether the voltage amplitude of the external contact force signal exceeds a preset safe voltage range. When the voltage amplitude of the external contact force signal exceeds the preset safe voltage range, the microcontroller 3 determines that the end-effector force sensor 22 has failed.

[0115] After determining that the end effector force sensor 22 has malfunctioned, microcontroller 3 immediately switches to a sensorless force sensing mode. In this mode, microcontroller 3 uses the converged back electromotive force coefficient estimate and the total friction compensation torque to estimate the external contact force acting on the joint end effector through an inverse dynamics algorithm. This estimation method relies on the accurate multiphysics parameter model established in the preceding steps.

[0116] To achieve accurate estimation of external contact forces, microcontroller 3 executes the following contact force observation equation:

[0117] ;

[0118] In the contact force observation equation: Defined as the first The external contact force at the joint end estimated in each control cycle; Defined as the effective reduction radius of the miniature precision reducer 12; Defined as the first The back electromotive force coefficient is estimated by the extended Kalman filter algorithm input for each control cycle; Defined as the first Each control cycle is determined by the actual phase current value measured by the current acquisition circuit 23; Defined as the first The total friction compensation torque is input by the nonlinear friction state observer for each control cycle; Defined as the total moment of inertia of the rotor of the ultra-small hollow cup motor 11 and the transmission components of the micro precision reducer 12; Defined as the first The rotor angular velocity of the miniature hollow cup motor 11 in one control cycle; Defined as the first The rotor angular velocity of the miniature hollow cup motor 11 in one control cycle; Defined as the sampling time interval.

[0119] The microcontroller 3 feeds back the estimated external contact force at the joint end to the coupled evolution path of load and structural dynamics. The establishment of the feedback loop ensures that the control system can still maintain the ability to dynamically correct for changes in Coulomb friction torque even if physical sensors fail.

[0120] To optimize computational efficiency, this embodiment of the invention also provides a parameter evolution acceleration scheme based on a lookup table, which reduces the computational load on the microcontroller 3. The microcontroller 3's non-volatile memory pre-stores a resistance-viscosity coefficient mapping table. This table discretizes the range of coil resistance variation into several resistance intervals. Each resistance interval corresponds to a pre-calibrated viscosity friction coefficient evolution value.

[0121] Microcontroller 3 acquires the current estimated coil resistance value in each control cycle. It uses a binary search algorithm to retrieve the resistance interval to which the estimated coil resistance value belongs from the resistance-viscosity coefficient mapping table. Microcontroller 3 then reads the corresponding viscous friction coefficient evolution value for that resistance interval. For coil resistance estimates located between the boundary points of two resistance intervals, microcontroller 3 uses linear interpolation to calculate the final viscous friction coefficient evolution value. This combination of table lookup and interpolation replaces the complex natural exponential function calculation, thereby shortening the single-cycle execution time of the control algorithm.

Claims

1. A high-precision drive control method for a dexterous hand, characterized in that, Includes the following steps: The high-frequency acquisition and preprocessing of the execution status signal are used to obtain the terminal voltage, phase current, rotor angular velocity and external contact force signal of the end force sensor (22) of the miniature hollow cup motor (11); Real-time online observation of motor electrical parameters is performed, and the coil resistance estimate and back electromotive force coefficient estimate are obtained based on the observed terminal voltage and phase current using the extended Kalman filter algorithm. The dynamic evolution of cross-domain parameters is performed, and the estimated value of the coil resistance is mapped to the viscous friction coefficient evolution value of the internal grease of the micro precision reducer (12) connected to the miniature hollow cup motor (11) based on the coupled evolution path of thermal and fluid dynamics. The external contact force signal is mapped to the Coulomb friction torque correction value of the micro precision reducer (12) based on the coupled evolution path of load and structural dynamics. Perform full-parameter friction observation and compensation torque generation, and calculate the total friction compensation torque based on the rotor angular velocity, the evolution value of the viscous friction coefficient, and the correction value of the Coulomb friction torque; Perform thermal drift compensation and final command synthesis to obtain the nominal demand torque output by the position control loop (92), and generate the final drive current command based on the nominal demand torque, the total friction compensation torque and the back electromotive force coefficient estimate. Specifically, the coupled evolution path of thermal and fluid dynamics includes: Utilizing the thermal conductivity between the miniature hollow cup motor (11) and the micro precision reducer (12), the estimated coil resistance is used as a temperature index variable; Based on the thermal time constant derived from the thermodynamic differential equation, phase compensation is performed on the mapping logic from the estimated value of the coil resistance to the evolution value of the viscous friction coefficient to eliminate the prediction error caused by the thermal conduction hysteresis. Based on the predetermined thermal sensitivity parameters and the mapping logic after phase compensation, the estimated value of the coil resistance is converted into the evolution value of the viscous friction coefficient at the current sampling time through calculation. The specific steps for performing full-parameter friction observation and compensation torque generation include: The friction dynamics equations of the micro precision reducer (12) are constructed based on the LuGre dynamic friction model; Update the Stribeck effect function in the tribological equations based on the Coulomb friction torque correction value; The total friction compensation torque is calculated by combining the evolution value of the viscous friction coefficient and the Stribeck effect function. The total friction compensation torque includes dual compensation components for fluid viscous resistance and mechanical dry friction resistance. The specific steps for performing thermal drift compensation and final instruction synthesis include: Perform a torque superposition operation to algebraically sum the nominal required torque and the total friction compensation torque to obtain the target value of the total electromagnetic torque that needs to be output at the shaft end of the miniature hollow cup motor (11); Perform thermal drift compensation division, use the estimated back electromotive force coefficient as the divisor to normalize the total electromagnetic torque target value, and calculate the final drive current command.

2. The high-precision drive control method for a dexterous hand according to claim 1, characterized in that, The specific steps for high-frequency acquisition and preprocessing of the execution status signal include: At the sampling time, a synchronous acquisition command is triggered to lock the terminal voltage, the phase current, the original rotor position data corresponding to the rotor angular velocity acquired by the position encoder (21), and the external contact force signal; The terminal voltage, the phase current, and the external contact force signal are filtered using a digital low-pass filter. The rotor angular velocity of the miniature hollow cup motor (11) is calculated based on the original rotor position data after filtering by using a backward differential algorithm combined with a moving average filtering strategy.

3. The high-precision drive control method for a dexterous hand according to claim 1, characterized in that, The specific steps for performing real-time online observation of the motor's electrical parameters include: A discretized state-space model of the miniature hollow cup motor (11) is established, wherein the phase current, the coil resistance and the back electromotive force coefficient are defined as state variables in the discretized state-space model. Perform a prediction step to predict the prior state value at the current sampling time based on the state estimate at the previous sampling time and the terminal voltage at the current sampling time; The update step is performed by using the deviation between the phase current at the current sampling time and the predicted phase current value obtained in the prediction step, combined with the Kalman gain matrix, to correct the state prior value, and output the converged estimated value of the coil resistance and the estimated value of the back electromotive force coefficient.

4. The high-precision drive control method for a dexterous hand according to claim 1, characterized in that, The specific evolution path of the coupling between load and structural dynamics includes: Read the external contact force signal and calculate the Coulomb friction torque threshold of the micro precision reducer (12) at the current sampling time based on the contact mechanics model; The external contact force signal is mapped to the Coulomb friction torque correction value.

5. The high-precision drive control method for a dexterous hand according to claim 1, characterized in that, It also includes a sensorless force sensing extension scheme based on current observation: The external contact force signal output by the end force sensor (22) is monitored in real time to determine whether the voltage amplitude of the external contact force signal exceeds the preset safe voltage range. After determining that the end force sensor (22) has malfunctioned, switch to sensorless force sensing mode; In the sensorless force sensing mode, the microcontroller (3) uses the back electromotive force coefficient estimate, the total friction compensation torque, the phase current and the rotor angular velocity to estimate the external contact force acting on the joint end through the inverse dynamics algorithm, and feeds back the estimated external contact force to the load and structural dynamics coupling evolution path to replace the external contact force signal.

6. The high-precision drive control method for a dexterous hand according to claim 1, characterized in that, It also includes a parameter evolution acceleration scheme based on table lookup: A resistance-viscosity coefficient mapping table is pre-established, which discretizes the range of change of the coil resistance into several resistance intervals; A binary search algorithm is used to retrieve the resistance interval to which the estimated coil resistance value belongs in the resistance-viscosity coefficient mapping table, and the evolution value of the viscosity friction coefficient corresponding to the resistance interval is read. For the estimated coil resistance value located between the boundary points of the two resistance intervals, the final evolution value of the viscous friction coefficient is calculated using linear interpolation.

7. A high-precision drive control system for a dexterous hand, characterized in that, The high-precision drive control method for a dexterous hand according to any one of claims 1-6 includes: The status signal acquisition and preprocessing module (5) is used to perform high-frequency acquisition and preprocessing of status signals to obtain the terminal voltage, phase current, rotor angular velocity and external contact force signal of the end force sensor (22) of the miniature hollow cup motor (11). The motor electrical parameter observation module (6) is used to obtain the estimated value of coil resistance and the estimated value of back electromotive force coefficient based on the observation of the terminal voltage and the phase current using the extended Kalman filter algorithm. The cross-domain parameter evolution module (7) is used to map the estimated value of the coil resistance to the viscous friction coefficient evolution value of the internal grease of the micro precision reducer (12) connected to the micro-miniature hollow cup motor (11) based on the coupled evolution path of thermal and fluid dynamics, and to map the external contact force signal to the Coulomb friction torque correction value of the micro precision reducer (12) based on the coupled evolution path of load and structural dynamics. The full-parameter friction observation module (8) is used to calculate the total friction compensation torque based on the rotor angular velocity, the evolution value of the viscous friction coefficient and the correction value of the Coulomb friction torque; The thermal drift compensation and final command synthesis module (9) is used to obtain the nominal demand torque output by the position control loop (92) and generate the final drive current command based on the nominal demand torque, the total friction compensation torque and the back electromotive force coefficient estimate.

Images