Earthmoving vehicle seat motion anti-pinch control method based on kalman filtering

Abstract

The present application relates to the technical field of vehicle control, in particular to a soil engineering vehicle seat movement anti-pinch control method based on Kalman filtering, comprising: the method acquires vehicle bus telemetry parameters and sensor feedback, classifies working conditions and determines transient physical impact, and actively suspends electromechanical adjustment and amplifies observation disturbance parameters to shield distorted electrical signals in strong impact. At the same time, engine speed and motor sensor parameters are input into the first channel Kalman filter unit, and adaptive subtraction is carried out based on the vibration model to extract pure motor background operation parameters. Then, the parameters and disturbance parameters are input into the second channel Kalman filter unit for forward multi-step deduction, and the average value of the seat equivalent pinch physical resistance prediction and the anti-pinch early warning trigger upper limit after the future period are output, and braking or reverse rotation is performed in advance when the physical threshold is exceeded. The present application avoids anti-pinch false triggering under complex vibration and realizes active safety protection.

Description

Technical Field

[0001] This invention relates to the field of vehicle control technology, specifically to a method for preventing seat movement pinching in earthmoving vehicles based on Kalman filtering. Background Technology

[0002] Existing electric seat control systems for engineering vehicles typically employ fixed-threshold motor current monitoring for anti-pinch protection. However, earthmoving vehicles face extremely harsh physical environments during actual operation, with the vehicle body frequently subjected to alternating interferences from irregular terrain bumps, transient pulses in the hydraulic system, and high-frequency engine vibrations. These intense external environmental loads are directly coupled to the seat system through the mechanical transmission structure, causing significant wide-frequency nonlinear fluctuations and distortions in the real-time drive current of the seat adjustment motor. Traditional fixed current thresholds or simple low-pass filtering anti-pinch algorithms cannot accurately distinguish between a slow current rise caused by a real human body clamping the seat and a transient current surge caused by external road bumps and engine resonance in a complex, high-noise background. This severe data crosstalk leads to two major dilemmas in the practical application of existing technologies: first, the algorithm is too sensitive, resulting in frequent false triggers and severely interfering with the driver's normal operation; second, setting the threshold too high causes the anti-pinch system to fail, failing to respond in time when a real pinching danger occurs, posing a significant safety hazard.

[0003] To address this, a method for preventing seat pinching in earthmoving vehicles based on Kalman filtering is proposed. Summary of the Invention

[0004] This invention aims to provide a method for preventing seat movement and pinching in earthmoving vehicles based on Kalman filtering. By separating vibration interference through a dual-channel filtering architecture and performing forward multi-step deduction of electromechanical state, it completely solves the problems of false triggering and response lag in the seat anti-pinch system under harsh working conditions, and realizes feedforward active physical protection.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A method for preventing seat pinching in earthmoving vehicles based on Kalman filtering includes: The system acquires vehicle bus telemetry parameters and electromechanical sensor feedback, classifies road vibration conditions and determines whether the transient physical impact load on the seat mechanical structure exceeds the limit, obtains the current vehicle operating condition category and impact indicator, and simultaneously switches the process disturbance covariance matrix of the Kalman filter unit in the anti-pinch controller; and when the system determines that the mechanical impact is strong, it actively suspends the electromechanical adjustment function of the seat motor and amplifies the parameters of the observed disturbance covariance matrix to the preset hardware extreme value to shield the distorted sensor signal flow caused by strong vibration. The mechanical speed parameters of the engine and the electromechanical sensing parameters of the seat motor are collected and input into the first channel Kalman filter unit to perform online tracking and adaptive subtraction compensation of the base vibration, and extract the base operating parameters of the motor after removing the vehicle body vibration. Based on the motor's baseline operating parameters and the updated process disturbance covariance matrix and observation disturbance covariance matrix at the current moment, forward multi-step deduction and parameter evolution calculation of the electromechanical state are performed to obtain the predicted mean value of the seat's equivalent clamping physical resistance after N electromechanical control cycles, as well as the corresponding upper limit for triggering the anti-pinch warning. The upper limit of the anti-pinch warning trigger is compared with the preset absolute anti-pinch safety physical threshold of the seat structure to generate an electromechanical anti-pinch trigger command and drive the seat motor to execute it in advance.

[0006] Preferably, obtaining the current vehicle operating condition category and impact indicator specifically includes: The vehicle bus telemetry parameters are physical quantities such as engine speed, main hydraulic pump pressure and hydraulic oil temperature read from the vehicle chassis control bus. The electromechanical sensor feedback quantities are the three-axis acceleration components of the base output by the multi-axis inertial measurement unit and the attitude deflection angle output by the seat internal angle encoder. The sliding time-domain fluctuation of the three-axis acceleration components of the base is calculated to extract the physical characteristics of vibration energy; the vehicle bus telemetry parameters and the physical characteristics of vibration energy are synchronously recombined during the sampling period using a time-series alignment mechanism; the synchronously recombined control parameters are input into the operating condition mapping unit, and the operating condition category of the current vehicle is output. A spatial attitude transformation matrix is ​​constructed using the attitude deflection angle. The three-axis acceleration components of the base are decomposed into the physical motion axis direction of the seat guide rail to obtain the axial inertial acceleration component. The initial static physical load of the seat motor is obtained and the occupant mass coefficient is calculated. The axial inertial acceleration component is multiplied by the occupant mass coefficient to calculate the equivalent inertial impact along the seat guide rail direction. The equivalent inertial impact is compared with the preset equivalent upper limit of inertial impact. When the equivalent inertial impact exceeds the upper limit of the equivalent inertial impact, a hardware-level strong impact flag is triggered; and after the equivalent inertial impact falls back to the preset mechanical safety hysteresis range, the strong impact flag is revoked.

[0007] Preferably, the process disturbance covariance matrix of the Kalman filter unit in the switching anti-pinch controller; and when a strong mechanical impact is detected, the electromechanical adjustment function of the seat motor is actively suspended, and the parameters of the observed disturbance covariance matrix are amplified to a preset hardware extreme value to shield the distorted sensor signal flow caused by strong vibration, specifically including: Establish an electromechanical mapping table corresponding to operating condition categories and process disturbance parameters; according to the operating condition category, retrieve the corresponding target process disturbance parameters from the electromechanical mapping table and update the process disturbance covariance matrix of the Kalman filter unit; when the impact flag is triggered, cut off the power drive circuit of the seat motor and activate the mechanical self-locking brake; in response to the impact flag, replace the diagonal variance parameter of the observation disturbance covariance matrix of the Kalman filter unit with a preset hardware extreme value; after the impact flag is revoked, restore the parameter configuration of the observation disturbance covariance matrix.

[0008] Preferably, the application performs online tracking and adaptive subtraction compensation of the base vibration based on the physical vibration model of the base under known engine excitation, specifically including: Using the engine's mechanical speed parameter, the fundamental frequency of mechanical vibration transmitted from the frame to the seat and the corresponding structural harmonic frequencies are calculated. Combining the pre-extracted natural frequencies and physical damping coefficients of the seat's mechanical structure, the fundamental frequency of mechanical vibration and the structural harmonic frequencies are used as input parameters to construct a physical vibration model of the base that includes electromechanical amplitude variables and phase variables. The physical vibration model of the base is then input into the first channel Kalman filter unit to reconstruct the state transition matrix of the electromechanical system.

[0009] Preferably, the extraction of the motor's baseline operating parameters after removing vehicle body vibration specifically includes: The physical observation deviation is extracted by subtracting the look-ahead vibration disturbance estimate from the previous electromechanical sampling cycle and the electromechanical sensing parameters of the seat motor. Based on the physical observation deviation, the phase variables inside the physical vibration model of the base are corrected through feedback, and the corrected vibration disturbance estimate for the current cycle is output. The vibration disturbance estimate is compensated by subtraction between the electromechanical sensing parameters of the seat motor and the vibration disturbance estimate. The electromechanical parameters obtained after the compensation operation are used as the background operating parameters of the motor.

[0010] Preferably, obtaining the predicted average of the equivalent clamping physical resistance of the seat after a preset number of electromechanical control cycles, and the corresponding upper limit for triggering the anti-pinch warning, specifically includes: The electromechanical dynamics equations of the seat DC motor are constructed with the seat guide rail position, mechanical adjustment speed and equivalent clamping physical resistance as state variables, and the state transition matrix of the electromechanical system is extracted. The dynamic control gain is calculated using the motor's background operating parameters, the updated process disturbance covariance matrix and the observation disturbance covariance matrix at the current moment, and the posterior physical state estimate of the state variables and the corresponding posterior error covariance matrix are solved. Starting from the posterior physical state estimate, a preset number of forward state transition operations are performed using the electromechanical system state transition matrix to calculate the predicted mean of the equivalent clamping physical resistance of the seat for the corresponding future period. Using the electromechanical system state transition matrix, the transpose of the electromechanical system state transition matrix, and the process disturbance covariance matrix, a preset number of state evolution accumulation calculations are performed on the posterior error covariance matrix to obtain the prediction error covariance matrix for the corresponding future period. Extract the diagonal elements of the equivalent clamping physical resistance in the prediction error covariance matrix and calculate the standard deviation of the physical resistance; add the predicted mean of the equivalent clamping physical resistance of the seat to the standard deviation of the physical resistance by a preset multiple to obtain the upper limit of the anti-pinch warning trigger.

[0011] Preferably, the step of generating the electromechanical anti-pinch trigger command and driving the seat motor to perform mechanical braking or reverse force relief action in advance specifically includes: Obtain the pre-calibrated static safety mechanical threshold of the base; extract the first-order time derivative of the motor's base operating parameters, and use the first-order time derivative to dynamically and physically adjust and compensate the static safety mechanical threshold of the base to obtain the absolute anti-pinch safety physical threshold for the current cycle. The upper limit of the anti-pinch warning trigger is compared with the absolute anti-pinch safety physical threshold; when the upper limit of the anti-pinch warning trigger is greater than the absolute anti-pinch safety physical threshold, an electromechanical anti-pinch trigger command is generated; in response to the electromechanical anti-pinch trigger command, the current seat motor pulse width modulation control signal is cut off; the same side bridge arm of the seat motor drive circuit is connected to form a short circuit to perform motor braking, or a drive signal with reverse polarity is applied to control the seat motor to perform a reverse yielding action.

[0012] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. By constructing a physical vibration model of the base, online tracking and adaptive subtraction compensation of the base vibration are performed in the first channel filtering unit. This allows for the accurate extraction of pure motor base operating parameters from sensor signals contaminated by strong turbulence and resonance, thereby reducing the interference of nonlinear noise on anti-pinch judgment from the source.

[0013] 2. By utilizing the second-channel filtering unit to perform forward multi-step extrapolation of the electromechanical state, the system proactively calculates the predicted average value of the equivalent clamping physical resistance and the upper limit of the anti-pinch warning trigger after multiple control cycles. This proactive prediction mechanism provides the system with a valuable anti-pinch response time window, enabling the motor to brake or unload and reverse before causing actual physical crushing damage.

[0014] 3. Deep coordination of working condition classification, hardware-level physical suspension and dual-channel filtering: When encountering extreme impact, the system physically cuts off the drive circuit and self-locks, while at the mathematical level it maximizes the observation of disturbance parameters to isolate distorted signals, and is supplemented by dynamic switching of process parameters. This coordination mechanism effectively avoids seat anti-pinch failure and algorithm crash. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the steps of the anti-pinch control method for the seat movement of earthmoving vehicles based on Kalman filtering according to the present invention. Figure 2 This is a schematic diagram of the process for obtaining the upper limit of the anti-pinch warning trigger in this invention; Figure 3 This is a flowchart illustrating the actions to be performed after the anti-pinch warning of the present invention is triggered. Detailed Implementation

[0016] The technical solutions of 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.

[0017] Please see Figures 1 to 3 This invention provides a method for preventing seat movement pinching in earthmoving vehicles based on Kalman filtering, referring to... Figure 1 Step-by-step flowchart Figure 2 A flowchart illustrating the process of obtaining the upper limit of the anti-pinch warning trigger, and Figure 3 A flowchart illustrating the actions to be performed after the anti-pinch warning is triggered; The technical solution of the present invention is as follows: The system acquires vehicle bus telemetry parameters and electromechanical sensor feedback, classifies road vibration conditions and determines whether the transient physical impact load on the seat mechanical structure exceeds the limit, obtains the current vehicle operating condition category and impact indicator, and simultaneously switches the process disturbance covariance matrix of the Kalman filter unit in the anti-pinch controller; and when the system determines that the mechanical impact is strong, it actively suspends the electromechanical adjustment function of the seat motor and amplifies the parameters of the observed disturbance covariance matrix to the preset hardware extreme value to shield the distorted sensor signal flow caused by strong vibration. The mechanical speed parameters of the engine and the electromechanical sensing parameters of the seat motor are collected and input into the first channel Kalman filter unit to perform online tracking and adaptive subtraction compensation of the base vibration, and extract the base operating parameters of the motor after removing the vehicle body vibration. Based on the motor's baseline operating parameters and the updated process disturbance covariance matrix and observation disturbance covariance matrix at the current moment, forward multi-step deduction and parameter evolution calculation of the electromechanical state are performed to obtain the predicted mean value of the seat's equivalent clamping physical resistance after N electromechanical control cycles, as well as the corresponding upper limit of the anti-pinch warning trigger. The upper limit of the anti-pinch warning trigger is compared with the preset absolute anti-pinch safety physical threshold of the seat structure to generate an electromechanical anti-pinch trigger command and drive the seat motor to execute it in advance.

[0018] Example 1: When hydraulic excavators operate in complex terrain in the field, the anti-pinch control of the seat guide rail motor faces the triple challenge of wide-frequency vibration interference from the base, uncertainty of multi-condition dynamic models, and distortion of strong impact signals. No single static threshold scheme can suppress malfunctions while ensuring clamping safety. This embodiment provides an anti-pinch control method for the seat movement of earthmoving vehicles based on Kalman filtering. It should be noted that the present invention has performed normalization processing before numerical calculation to prevent inconsistencies in dimensions.

[0019] First, based on the multi-channel real-time parameters of the chassis control bus, the current operating condition category is identified and strong mechanical impact events are detected. Using this as an index, the Kalman filter process disturbance covariance matrix is ​​dynamically switched. During strong impacts, a dual mechanism of hardware circuit breaking and extreme value replacement of observed parameters is used to shield distorted sensor signals. Then, based on the dual-index amplitude spectrum of engine speed and torque, the phase reference of the crankshaft top dead center synchronization frame, and the offline calibrated natural frequency of the seat, a multi-order harmonic base physical vibration model is constructed, and the state transition matrix is ​​reconstructed. After electromechanical transmission coupling mapping transformation, it is adaptively subtracted from the motor sensor parameters to extract and eliminate vibration interference. The system analyzes the background operating parameters of the motor after disturbance; then, using a 3D electromechanical state space as a framework, it completes Kalman filtering prediction and updating, and forward extrapolates the predicted mean of the equivalent clamping physical resistance and the confidence upper limit based on the error covariance to a preset number of control cycles in the future, forming a dynamic adaptive anti-pinch warning trigger upper limit; finally, it extracts the rate of change of the background operating parameters after double-level smoothing, dynamically lowers the static safety mechanical threshold of the base, and compares the predicted trigger upper limit with the dynamic safety threshold in real time. Once exceeded, an electromechanical anti-pinch trigger command is generated, driving the seat motor to complete braking or reverse retraction actions in advance before detecting physical clamping.

[0020] Specifically, the calibration of the absolute baseline values ​​of the diagonal parameters of the process disturbance covariance matrix is ​​the absolute core that determines the balance between sensitivity and filtering robustness of the Kalman filter system under different operating conditions. The physical meaning of the diagonal elements in this matrix is ​​a quantified representation of the variance of internal uncertainties that were not precisely mathematically modeled in the system's difference model. These uncertainties mainly originate from difficult-to-observe implicit physical processes such as the mechanical backlash of transmission gears, the nonlinear frictional changes caused by long-term wear of the lead screw, and the physical attenuation of magnetic flux of the motor as the operating temperature increases.

[0021] To obtain the baseline process disturbance parameter set of the system under the most basic idling standby condition, the calibration process introduced a variance statistical algorithm based on steady-state prediction residuals (such as Allan's analysis of variance). During factory bench calibration, the drive seat continuously reciprocated in a stationary external environment without external clamping. High-precision external flange torque sensors and laser displacement gauges were used to synchronously collect the actual physical load torque and position derivatives, and the time series of deviations between the pure prediction derivation results of the state equations and the actual physical values ​​was calculated. Statistical analysis was performed on this deviation series to extract the covariance characteristics of each dimension. Under the typical physical constraints of a general industrial-grade DC motor and ball screw drive platform, the parameters of the reasonably calibrated idling baseline process disturbance covariance matrix are typically on the following order of magnitude: the variance of the corresponding position state variables is approximately per control cycle. square meters; the variance of the corresponding velocity state variable is approximately Square meters per square second; the variance of the corresponding equivalent clamping physical resistance is approximately Square Newton-meters. After determining the absolute scale for this extremely low-noise environment, the parameters for other higher noise levels, such as flat ground, off-road, and broken ground conditions, are then dynamically amplified using the rigorously measured baseline value as the base, according to a multiplier (such as a gradient of 3 to 5 times) corresponding to the severity of the corresponding working conditions. This baseline plus dynamic scaling mechanism ensures that the filter gains the ability to suppress sudden mechanical shocks without losing the basic steady-state Kalman filter smoothing properties.

[0022] Further, obtaining the current vehicle operating condition category and impact indicator specifically includes: the vehicle bus telemetry parameters are physical quantities such as engine speed, main hydraulic pump pressure, and hydraulic oil temperature read from the vehicle chassis control bus; the electromechanical sensor feedback quantities are the base three-axis acceleration components output by the multi-axis inertial measurement unit and the attitude deflection angle output by the seat internal angle encoder; the sliding time-domain fluctuation of the base three-axis acceleration components is calculated to extract the physical characteristics of vibration energy; using a time-series alignment mechanism, the vehicle bus telemetry parameters and the physical characteristics of vibration energy are synchronously recombined during the sampling period; the synchronously recombined control parameters are input into the operating condition mapping unit to output the current vehicle operating condition category; and the system utilizes the... A spatial attitude transformation matrix is ​​constructed based on the attitude deflection angle. The three-axis acceleration components of the base are decomposed into the physical motion axis direction of the seat guide rail to obtain the axial inertial acceleration component. The initial static physical load of the seat motor is obtained and the occupant mass coefficient is calculated. The axial inertial acceleration component is multiplied by the occupant mass coefficient to calculate the equivalent inertial impact along the seat guide rail direction, which is used for strong impact state identification and sensor signal protection mode triggering. The equivalent inertial impact is compared with a preset equivalent upper limit of inertial impact. When the equivalent inertial impact is greater than the equivalent upper limit of inertial impact, a hardware-level strong impact flag is triggered. The strong impact flag is revoked after the equivalent inertial impact falls back to the preset mechanical safety hysteresis range.

[0023] The core engineering objective of the algorithm for obtaining the initial static physical load of the seat motor and calculating the occupant mass coefficient is to completely eliminate the physical interference of uneven spatial distribution of static friction and rotor starting inertia on the measurement of the vertical component of gravity. The complete closed-loop implementation steps of this calculation process are as follows: First, the extremely low duty cycle of the drive motor during sweeping must be strictly constrained within a critical physical duty cycle range that just overcomes the maximum static friction of the guide rail, allowing the motor to generate a continuous creeping displacement with no dead zone. In a typical 24-volt vehicle power supply system, this duty cycle is usually between 2% and 5% of the motor's rated maximum drive pulse width. Under this weak electromagnetic driving force, the rotor angular acceleration of the motor is extremely small, effectively preventing false additional load torque caused by the acceleration inertia due to Newton's second law from being mixed into the calculation logic.

[0024] Secondly, to eliminate nonlinear spatial fluctuations in local frictional force along the guide rail caused by uneven grease lubrication or machining assembly tolerances, the sweeping process must be forcibly set at the software level to include complete forward and reverse two-way continuous motion. During the entire closed-loop process of moving forward to the end of the physical limit and turning back, the controller needs to trigger high-frequency current sampling at fixed small spatial displacement intervals along the way to ensure that the total number of effective current samples within a single physical stroke is not less than 50.

[0025] Subsequently, the control module performs an arithmetic summation and division by 2 on the corresponding elements of the forward and reverse creep current sampling sequences within the same spatial physical location range in the memory. Since the work done by the constant Coulomb friction force is strictly opposite in the opposite directions of system translation, this bidirectional averaging operation automatically cancels out most of the physical components of the guide rail friction resistance directly related to the direction of motion in the mathematical formula. The remaining pure net value truly represents the steady-state equivalent support load generated by the vertical gravity component along the tilting axis of the lead screw. Finally, this net current value is combined with the motor torque constant multiplied by the reduction ratio and lead conversion coefficient, and then divided by the gravitational acceleration constant to obtain a reliable and friction-loss-free reference for the total mass coefficient of the occupants.

[0026] Specifically, taking a hydraulic excavator as an example, the anti-pinch controller is integrated into the cab chassis electronic control unit in the form of an independent microcontroller chip. The chip is connected to three hardware channels via the on-chip peripheral bus: the chassis control bus transceiver module, the serial acquisition interface of the multi-axis inertial measurement unit, and the quadrature decoding circuit of the seat encoder. When each channel is triggered by its own hardware interrupt, it writes the latest data into an independent first-in-first-out buffer register in the static random access memory area on the controller chip. The downstream calculation modules read the data according to the register address in the corresponding calculation cycle. All data flows on the internal bus of the chip without external transmission, ensuring that data acquisition and processing are completed in a closed loop on the same computing entity from the hardware architecture perspective.

[0027] The multi-axis inertial measurement unit continuously outputs three numerical values ​​in the base coordinate system: the front-to-back axial acceleration component, the left-to-right axial acceleration component, and the vertical axial acceleration component. Each value is quantized in signed 16-bit integer format by the on-chip analog-to-digital converter and sequentially pushed into its corresponding independent circular buffer queue. The unit is milligrams, and the range is set to ±16 times the gravitational acceleration. The three raw values ​​are used directly as the data input of the downstream sliding time-domain fluctuation calculation module without any pre-filtering. The write pointer of the circular queue automatically increments with each new sample, and the read pointer is moved as needed by the calculation module in each cycle.

[0028] The sliding time-domain fluctuation calculation module is triggered and executed within the timer interrupt of each controller cycle. It takes the latest 50 sampling points in each of the three circular queues of the front-back axis, left-right axis, and vertical axis as the current calculation window. The arithmetic mean of the 50 values ​​in each channel is calculated first. Then, the difference between each value and the mean is squared. The 50 squared results are accumulated and divided by 50 to obtain the discrete acceleration metric value of that channel in the current window. Finally, the three discrete metric values ​​are added one by one to obtain a comprehensive scalar vibration energy characteristic value. This value is written to the vibration energy characteristic register in single-precision floating-point format, with the unit being square meters per fourth power second. The physical meaning is the statistical quantitative characterization of the composite vibration intensity of the base in three-dimensional space within the current sliding window. The output of this register is directly used as one of the data inputs of the timing alignment module.

[0029] The chassis control bus pushes message frames to the bus transceiver module at a 100 Hz broadcast cycle. After the transceiver module completes the frame decoding, it writes the three physical quantities—engine speed, main hydraulic pump pressure, and hydraulic oil temperature—into their respective dedicated bus data registers. The effective update cycle for each frame is 10 ms, which is strictly consistent with the overwrite cycle of the vibration energy characteristic register. The timing alignment module executes at the end of the timer interrupt service routine in the same cycle. Using the hardware counter timestamp of the interrupt as the sole reference, it synchronously reads the current valid values ​​of the three bus data registers and the current updated value of the vibration energy characteristic register, totaling four physical quantities. These are then concatenated and written into the operating condition feature vector register in a fixed field order of engine speed, hydraulic pump pressure, hydraulic oil temperature, and vibration energy characteristic value. The four-element vector in this register is the data input for each operation cycle of the operating condition mapping unit. All four components strictly belong to the same cycle beat on the time axis, eliminating cross-cycle misalignment.

[0030] The core of the working condition mapping unit is a multi-condition branch judgment rule table. This rule table is built and burned once during the offline bench calibration stage before the system leaves the factory. The construction process is as follows: The calibration engineer operates the same model of hydraulic excavator in a standard working area to stably execute 5 representative working conditions in sequence. For each working condition, the system maintains a stable state and continuously collects working condition feature vector samples for no less than 5 minutes. The host computer software records all samples with timestamps and simultaneously labels the working condition category. After the collection is completed, the host computer calculates the 5000-fold weighting of each of the four physical quantity dimensions for the sample set of each working condition. The % quantile and 95th quantile serve as the lower and upper boundaries of the effective numerical range for the corresponding dimension for this working condition. A judgment rule for this working condition is composed of 8 boundary parameters from 4 dimensions plus 1 working condition category number, totaling 9 fields. A total of 5 rules are formed for 5 working conditions. After being arranged from high to low priority according to the hydraulic breaker impact operation, off-road terrain driving, normal excavation operation, flat ground transfer driving, and idling standby, they are burned into the Flash configuration sector of the controller in a fixed-length structure array format, thus completing the construction of the working condition mapping unit.

[0031] To ensure the universal applicability and high recognition accuracy of the multidimensional rule table for the working condition mapping unit, this paper provides quantitative reference values ​​for the boundaries of key dimension parameters under different typical working conditions, as well as scientific scaling and calibration criteria. In typical 20-ton medium-sized hydraulic excavator applications, there are significant numerical clustering differences in the decision interval boundaries of the five working conditions. The specific decision parameter domain feature baselines are shown in Table 1.

[0032] Table 1: Reference Boundary Table for Multidimensional Judgment Interval Mapping of Typical Earthmoving Machinery Working Conditions

[0033] For vehicle platforms with different tonnages or hydraulic architectures, the above parameters cannot be directly hard-coded as absolute values. Instead, they should be adjusted proportionally based on the rated maximum speed and the maximum set pressure of the relief valve as indicated on the engine and hydraulic system nameplates for that vehicle model. Furthermore, to verify the statistical sufficiency of the 5% to 95% percentile boundaries extracted during the initial calibration, the final solidification of the rule table must meet the cross-validation standard: that is, the number of offline steady-state samples collected by the calibration engineer must ensure that, during hold-out validation, the diagonal classification accuracy of the confusion matrix for all five operating conditions on the test set reaches a physical confidence threshold of over 95%. If this confidence level is not reached, it indicates insufficient sample diversity, and the bench test duration under specific operating conditions must be extended to expand the coverage of boundary samples.

[0034] Furthermore, during the online operation of the working condition mapping unit, the controller reads the 4-element real-time vector in the working condition feature vector register during each cycle interruption. It then compares the four components of the vector with the corresponding interval boundaries of the current rule in the Flash rule table, prioritizing them from high to low priority. If all four components fall within the interval corresponding to the current rule, the working condition category number of the rule is written into the working condition category register, and the comparison of subsequent rules is terminated. If all five rules fail, the off-road terrain working condition number is written into the working condition category register as a conservative default value. The output of the working condition category register serves as the trigger input for the process disturbance covariance matrix parameter switching module, and the Kalman parameters are synchronously updated in the next module.

[0035] Based on the actual operating characteristics of hydraulic excavators, the physical descriptions of five working conditions are as follows: Condition 1, idling standby, corresponds to the engine being in the low idle speed range, the hydraulic pump maintaining the minimum standby pressure, and vibration energy continuously below the stable baseline; the anti-pinch parameter is switched to the most sensitive setting. Condition 2, flat terrain transfer, corresponds to the engine being in the medium travel speed range, the hydraulic pump supplying pressure to the travel motor, and vibration energy in the range of slight periodic fluctuations. Condition 3, normal excavation operation, corresponds to the engine being in the high rated power speed range, the hydraulic pump pressure periodically rising to the excavation load range with the bucket movement rhythm, and vibration energy in the range of medium pulse amplitude. Condition 4, off-road terrain travel, corresponds to the effective travel speed and the hydraulic pump pressure continuously being in the high load travel range, and vibration energy exceeding the upper limit for flat terrain travel. Condition 5, hydraulic breaker impact operation, corresponds to the hydraulic pump pressure being stably maintained in the high-pressure breaking range, and vibration energy continuously exceeding the peak baseline; this is the scenario with the highest risk of false triggering of the anti-pinch mechanism, and the diagonal parameter of the process disturbance covariance matrix in the corresponding rule table is set to the maximum allowable value.

[0036] The seat's internal angle encoder outputs two components: pitch and roll angle, at a sampling rate of 200 Hz. The two values ​​are written to the encoder register in signed 16-bit integer format. After reading the values, the controller generates a 3x3 rotation matrix around the horizontal axis using the pitch angle as a parameter, and then generates a 3x3 rotation matrix around the vertical axis using the roll angle as a parameter. The two matrices are multiplied sequentially to obtain the spatial attitude transformation matrix from the base coordinate system to the seat guide rail motion axis coordinate system. This matrix is ​​written to the attitude matrix register in single-precision floating-point format. This matrix is ​​automatically recalculated with each encoder update to track the real-time attitude changes of the seat.

[0037] The controller reads the latest triaxial acceleration values ​​from the acceleration loop buffer queue to form a 3D column vector. This vector is then multiplied by the 3x3 spatial attitude transformation matrix stored in the attitude matrix register. The component corresponding to the longitudinal motion axis of the seat guide rail is extracted from the resulting vector and written into the axial impact register. The unit is meters per second squared. This operation accurately projects the three-dimensional inertial load on the base onto the physical motion direction of the guide rail, eliminating the interference of lateral and vertical acceleration on the determination of longitudinal clamping force. The output of the axial impact register serves as the input to the inertial impact equivalent calculation module.

[0038] During the power-on initialization phase, the controller drives the seat motor to perform a full-stroke low-speed reciprocating scan using a pulse width modulation signal with an extremely low duty cycle. After continuously collecting the drive current several times at each guide rail position, the average value is taken. The average value at each position is multiplied by the electrical characteristic coefficient of the motor winding to convert it into the equivalent static support force at that position. Then, the support force is divided by the gravitational acceleration to obtain the equivalent unloaded mass. The total mass coefficient is obtained by superimposing the upper limit of the preset standard weight range for occupants (e.g., 100 kg) and stored in the mass coefficient register in single-precision floating-point format. During online operation, the controller performs a multiplication operation between the current axial inertial acceleration component of the axial impact register and the total mass coefficient of the mass coefficient register. The resulting equivalent inertial impact quantity is written to the impact force register in Newtons for the threshold comparison module to read.

[0039] The controller compares the current value of the impact force register with the pre-calibrated and fixed equivalent upper limit of inertial impact in the Flash configuration sector (e.g., 2800 Newtons; this boundary is derived from the equivalent value converted from the peak value of the critical impact load corresponding to the occupant dummy model being compressed beyond the industry safety contact limit during the bench mechanical calibration phase before the system leaves the factory). When the value of the impact force register exceeds this boundary for the first time, the controller sets the impact flag register to a logic high level and simultaneously triggers a hardware interrupt to notify the anti-pinch main task to immediately... In response, to prevent the flag register from repeatedly flipping due to slight fluctuations in the impact force near the trigger boundary, the controller simultaneously maintains a hysteresis cancellation threshold below the trigger boundary (e.g., 1900 Newtons, derived from the design requirements of the anti-pinch system for mechanical fall-back buffering) and a continuous satisfaction counter (this value can be selected in actual situations). Only when the impact force register value is below the hysteresis cancellation threshold for several consecutive cycles and the counter reaches the preset threshold, will the controller clear the impact flag register and notify the main task to cancel the strong impact state, forming a complete hysteresis closed-loop determination of strong impact entry and exit.

[0040] By organically combining the vibration energy characteristics of the sliding window with the hierarchical rule table mapping unit jointly constructed by the multi-dimensional bus telemetry of the hydraulic excavator and the bench calibration, and in conjunction with the attitude projection decomposition and the dual-threshold hysteresis impact judgment link, real-time online identification of five typical working conditions and reliable judgment of strong impact events are realized. This provides an accurate working condition context for subsequent dual-channel Kalman filter adaptive parameter switching and avoids the risk of cross-working condition misjudgment of the anti-pinch system under fixed parameter configuration.

[0041] Furthermore, the process disturbance covariance matrix of the Kalman filter unit in the switching anti-pinch controller is adjusted; and when a strong mechanical impact is detected, the electromechanical adjustment function of the seat motor is actively suspended, and the parameters of the observed disturbance covariance matrix are amplified to a preset hardware extreme value to shield the distorted sensor signal flow caused by strong vibration. Specifically, this includes: establishing an electromechanical mapping table corresponding to the operating condition category and the process disturbance parameters; retrieving the corresponding target process disturbance parameters from the electromechanical mapping table according to the operating condition category, and updating the process disturbance covariance matrix of the Kalman filter unit; when the impact flag is triggered, cutting off the power drive circuit of the seat motor and activating the mechanical self-locking brake; responding to the impact flag, replacing the diagonal variance parameter of the observed disturbance covariance matrix of the Kalman filter unit with a preset hardware extreme value; and restoring the parameter configuration of the observed disturbance covariance matrix after the impact flag is revoked.

[0042] Specifically, the Kalman filter unit of the anti-pinch controller runs on the main computing core of the microcontroller chip. The current effective parameters of the process disturbance covariance matrix and the observation disturbance covariance matrix are stored in single-precision floating-point format in two independent parameter register groups in the on-chip static random access memory area. The electromechanical mapping table entries and the hardware extrema of the observation disturbance are fixed in read-only form in the on-chip Flash configuration sector. Before the start of each Kalman operation cycle, the controller first reads the current effective values ​​of the two covariance matrices from the parameter register group, and then enters the filtering iteration calculation to ensure that any matrix parameter switching takes effect immediately in the next operation cycle, and there is no timing misalignment between parameter update and filtering operation.

[0043] The electromechanical mapping table is burned into the Flash configuration sector in the form of a 5-row fixed-length structure array. Each row of the structure contains one working condition category number field and three single-precision floating-point number fields. The three floating-point number fields correspond to the three independent element values ​​on the diagonal of the process disturbance covariance matrix in the 3D electromechanical dynamics equation with the seat guide rail position, adjustment speed and equivalent clamping physical resistance as state variables. Off-diagonal elements are fixed to zero and are not stored in the mapping table to save storage space. Based on the bench calibration results of the system model error intensity under the five working conditions of the hydraulic excavator, the diagonal element combinations corresponding to the five working conditions are arranged in a gradient increasing order. The idle no-load standby working condition (Category 1) corresponds to the smallest three element values, and the hydraulic breaker impact operation working condition (Category 5) corresponds to the largest three element values. The ratio between the same element values ​​of adjacent working condition categories is set to, for example, 3 to 5 times, to quantitatively cover the physical differences in the uncertainty of the electromechanical system dynamics model under each working condition.

[0044] Within each cycle interrupt service routine, the controller reads the current operating condition number from the operating condition category register. Using this number as a row index, it locates the corresponding structure row in the Flash electromechanical mapping table and sequentially reads out three target process disturbance parameters. The controller then writes the three current stored values ​​at the corresponding positions in the process disturbance covariance matrix parameter register group one by one with the read target values, thus completing the update of the process disturbance covariance matrix for this cycle. In the next Kalman operation cycle, the filter calculation module reads the updated three diagonal elements according to the register address and substitutes them into the forward extrapolation calculation of the prediction error covariance matrix, enabling the filter to match the uncertainty estimate of the system state to the physical characteristics of the current operating condition in real time.

[0045] When the impact flag register is set to logic high, the controller sends a low-level blocking signal to the hardware enable pin of the seat motor drive circuit through the on-chip general-purpose output port. The gate drive signals of all power switching transistors in the drive circuit are then blocked by hardware, and the power circuit between the motor winding and the DC power supply bus is completely disconnected, so the motor loses its driving capability at the electrical level. At the same time, the controller continuously applies excitation current to the electromagnetic coil of the mechanical self-locking braking mechanism through another general-purpose output port. The electromagnetic force presses the brake pin into the positioning locking groove on the side of the guide rail, locking the guide rail in the current position at the mechanical level. This prevents the seat from displacing due to residual electrical drive or from non-command slippage caused by external impact load during the strong impact. The two hardware actions are executed sequentially within the same interrupt, and the total response delay does not exceed one control cycle.

[0046] Furthermore, in response to the logic high level of the impulse flag register, the controller, within the same interrupt service routine, replaces the current stored values ​​of all diagonal elements in the observed disturbance covariance matrix parameter register group with preset hardware extreme values ​​pre-stored in the Flash configuration sector (e.g., taking values ​​of magnitude reaching a certain level). (The fixed positive value), the off-diagonal elements remain unchanged at zero; after the replacement is completed, when the Kalman filter unit calculates the Kalman gain with the maximum observation perturbation covariance matrix in subsequent operation cycles, the calculation results of each element of the gain matrix approach zero. The filter actually completely ignores the real-time measurements from the current sensor, speed sensor and position encoder, and only relies on the electromechanical system state transition matrix to perform pure model prediction and deduction, thus mathematically cutting off the propagation path of distorted sensor electrical signals to state estimates during strong vibration.

[0047] After the impact flag register is cleared, the controller reads the current valid operating condition number from the operating condition category register in the next cycle interrupt service routine. Using this number as an index, it reads the normal operating parameter group of the observation disturbance covariance matrix pre-calibrated and fixed for this operating condition from the Flash configuration sector. The diagonal element values ​​of this parameter group are written one by one to the observation disturbance covariance matrix parameter register group, replacing the hardware extreme values ​​written during the impact. At the same time, the controller cancels the low-level blocking signal of the motor drive circuit enable pin and cuts off the excitation current of the mechanical self-locking braking mechanism. The brake pin is released from the locking slot under the action of the reset spring. The seat motor can receive the pulse width modulation drive signal again in the first control cycle after recovery. The Kalman filter unit synchronously restores the normal fusion update capability of the real-time measurement. All state quantities switch back from pure prediction mode to measurement auxiliary mode.

[0048] The process disturbance covariance matrix is ​​adaptively switched periodically by period through the electromechanical mapping table index. The distorted sensing signal during the strong impact is isolated synchronously by the dual mechanism of hardware circuit breaking and extreme value replacement of observation parameters. After the impact subsides, the filtering parameters are automatically restored according to the operating conditions, thus completely eliminating the risk of algorithm state collapse under extreme operating conditions.

[0049] Furthermore, the application performs online tracking and adaptive subtraction compensation of the base vibration based on the physical vibration model of the base with known engine excitation. Specifically, it includes: using the engine mechanical speed parameter to calculate the fundamental frequency of mechanical excitation transmitted from the frame to the seat and the corresponding structural harmonic frequencies; combining the pre-extracted natural frequencies and physical damping coefficients of the seat mechanical structure, and using the fundamental frequency of mechanical excitation and the structural harmonic frequencies as input parameters, constructing a physical vibration model of the base that includes electromechanical amplitude variables and phase variables; and inputting the physical vibration model of the base into the first channel Kalman filter unit to reconstruct the state transition matrix of the electromechanical system.

[0050] Specifically, taking a hydraulic excavator as an example, the first channel Kalman filter unit runs on the main computing core of the anti-pinch controller microcontroller chip. The calculation results of the excitation fundamental frequency and harmonic frequencies, the amplitude mapping query results, the phase reference timestamp, the calibration values ​​of the natural frequency and damping coefficient, and the state transition matrix reconstructed in each control cycle are all written into independent register groups in the on-chip static random access memory in single-precision floating-point format. Each computing module reads and writes in the order of register addresses within the corresponding cycle of operation, without passing through external storage. All data flows complete the closed loop inside the chip.

[0051] The engine mechanical speed parameter is read via the chassis control bus in units of speed per minute, updated periodically, and stored in the speed register in single-precision floating-point format. After reading this value, the controller calculates the number of periodic combustion excitations received by the base per unit time based on the engine cylinder number and stroke type parameters pre-stored in the Flash configuration sector (hydraulic excavators are usually equipped with 4-stroke diesel engines, with the number of cylinders being, for example, 6 cylinders), and writes it into the excitation base frequency register in Hertz. Then, the second to fourth harmonic frequencies are calculated sequentially at 2, 3, and 4 times the excitation base frequency and written into the four harmonic frequency registers respectively. All frequency values ​​are recalculated synchronously with each speed update to track the impact of engine speed fluctuations on the excitation frequency in real time. The outputs of the excitation base frequency register and each harmonic frequency register serve as the input data source for the vibration model construction module and the state transition matrix reconstruction module.

[0052] Furthermore, engine speed alone cannot uniquely determine the base vibration amplitude because changes in engine load at the same speed significantly alter the peak combustion pressure, thus changing the base vibration intensity. Therefore, the system pre-stores a two-dimensional vibration amplitude mapping map in the Flash configuration sector. This map is generated by engineers during the factory bench calibration phase by measuring and recording the base vibration acceleration amplitude under different combinations of engine speeds and loads. The rows are divided into speed increments of, for example, 100 revolutions per minute, and the columns are based on the percentage of engine output torque read via the chassis control bus. The system divides the spectrum into increments of 10%, with each cell storing the amplitude calibration value of the fundamental frequency component of the base excitation under the corresponding combination. During online operation, the controller uses the current value of the speed register and the real-time torque percentage reading as dual indices to locate four adjacent cells in the spectrum and perform bilinear interpolation to obtain the estimated value of the fundamental frequency amplitude under the current working condition, which is then written into the amplitude estimation register. The amplitude values ​​of each harmonic are multiplied by the estimated fundamental frequency amplitude according to the pre-stored step-by-step attenuation ratio coefficient in Flash and then written into the corresponding harmonic amplitude registers as real-time initialization inputs for the amplitude variables of each harmonic in the vibration model.

[0053] The phase source of the base vibration also needs to be determined independently; relying solely on rotational speed cannot establish an absolute phase reference. The controller continuously monitors the first cylinder compression top dead center synchronization message frame sent by the engine control unit at a fixed position in each engine working cycle via the chassis control bus. When the hardware interrupt is triggered upon receiving the frame, the controller records the current count value of the high-resolution hardware timer on the chip and stores this count value in the phase reference timestamp register, which serves as the vibration phase time zero point for the current cycle. Within each Kalman operation cycle, the controller reads the difference between the current hardware timer count value and the phase reference timestamp, multiplies this time difference by the current value of the excitation fundamental frequency register, converts it into a radian phase angle, obtains the phase angle of the fundamental frequency component at the current moment, and writes it into the fundamental frequency phase register. The phase angles of each harmonic are calculated sequentially as integer multiples of the fundamental frequency phase angle and written into the corresponding harmonic phase registers. The outputs of all phase registers serve as the real-time initialization reference for each harmonic phase variable of the vibration model.

[0054] The natural frequencies and physical damping coefficients of the seat's mechanical structure are extracted once during the vibration characteristic calibration stage before the system leaves the factory: The calibration engineer fixes a broadband accelerometer to the seat guide rail mounting plate and applies broadband transient excitation to multiple points of the base structure with a standard force hammer in a static bench environment. The host computer synchronously collects the excitation force signal and the base acceleration response signal and performs frequency domain transformation to obtain the amplitude-frequency characteristic curve of the base acceleration response. The frequency corresponding to each maximum amplitude point in the curve is taken as the structural natural frequency of the corresponding order. The frequency width corresponding to the point where the amplitude drops to 0.707 times the peak value on both sides of each maximum point is divided by twice the natural frequency of that order as the physical damping coefficient of the corresponding order. The above process is repeated for no less than 5 sets of independent excitation response data and the arithmetic mean is taken. The final natural frequencies and damping coefficients of each order are written into the inherent characteristic parameter area of ​​the Flash configuration sector in single-precision floating-point format. When the controller is powered on and initialized, all parameters are read from this area and loaded into the inherent characteristic register group of the on-chip static random access memory for later use.

[0055] Specifically, the physical vibration model of the base is constructed using a multi-harmonic superposition structure. Taking the typical vibration characteristics of a hydraulic excavator base as an example, the model takes three engine excitation frequency components: the fundamental frequency component and the second and third harmonic components, as well as one structural resonance response component corresponding to the natural frequency of the seat structure. Each frequency component is represented by two state variables, which are the cosine amplitude coefficient and the sine amplitude coefficient projected onto the motor sensing axis, respectively. The four components form a total of eight state variables, which are stored in the vibration model state vector register in column vector format. The amplitude variables of the three engine excitation components are initialized with the current values ​​of the corresponding harmonic amplitude registers, and the phase variables are initialized with the current values ​​of the corresponding harmonic phase registers. The amplitude and phase variables of the structural resonance component are initialized with the parameters of the inherent characteristic register group. The physical meaning of the above eight state variables is the contribution of each frequency component to the vibration superposition interference caused by the motor sensing parameters, and the unit is consistent with the motor current sensing parameters.

[0056] Based on the current values ​​of the excitation fundamental frequency register, harmonic frequency registers, and inherent characteristic register group, as well as the length of the periodic control cycle, the controller calculates the rotational mapping coefficients of the cosine and sine terms of the state variables of each frequency component between two adjacent control cycles. The rotational mapping coefficients of each component are filled into the corresponding 2-row, 2-column sub-block of the electromechanical system state transition matrix. The coefficients of the structural resonance component sub-block are also superimposed with the amplitude attenuation factor determined by the natural frequency and damping coefficient. Finally, they are combined to form a complete 8-row, 8-column state transition matrix, which is written into the state transition matrix register group in single-precision floating-point format. This matrix is ​​recalculated after each effective update of the excitation fundamental frequency register to ensure that the state propagation direction of the filter is always synchronized with the physical evolution law of the actual vibration when the engine speed changes. The reconstructed state transition matrix register group serves as the direct data input for the prediction step of the first channel Kalman filter unit.

[0057] Using the dual-index amplitude spectrum of rotational speed and torque and the phase reference of the top dead center synchronous frame as the deterministic source of amplitude and phase for the vibration model, and combining the offline calibrated natural frequency and damping coefficient, a multi-order harmonic state transition matrix that can be dynamically reconstructed according to the working conditions is constructed, so that the first channel Kalman filter has a convergence basis for accurately tracking the vibration law of the base at the physical level.

[0058] Furthermore, the extraction of the motor's baseline operating parameters after removing vehicle body vibration specifically includes: performing a subtraction operation between the look-ahead vibration interference estimate from the previous electromechanical sampling period and the electromechanical sensing parameters of the seat motor to extract the physical observation deviation; based on the physical observation deviation, correcting the phase variables within the physical vibration model of the base through feedback, and outputting the corrected vibration interference estimate for the current period; performing a subtraction compensation operation between the electromechanical sensing parameters of the seat motor and the vibration interference estimate; and using the electromechanical parameters obtained after the compensation operation as the motor's baseline operating parameters.

[0059] Specifically, the above processing flow runs on the main computing core of the anti-pinch controller microcontroller chip, with one processing cycle. The estimated forward vibration interference, physical observation deviation, corrected vibration interference, and motor background operating parameters are all stored in single-precision floating-point format in the independent function register group of the on-chip static random access memory area. The interrupt service routine of each cycle reads and writes in sequence according to the register address order. All data flows complete the closed loop inside the chip without passing through external storage.

[0060] Before performing the difference operation, the system needs to address the issue of dimensional consistency between the mechanical vibration of the base and the electrical signal from the motor sensor. This is because the base vibration model outputs physical quantities of mechanical displacement or force, while the motor sensor parameters are current sampling values, and the two cannot be directly subtracted algebraically. To address this, the controller pre-stores three physical parameters of the seat transmission mechanism—mechanical reduction ratio, lead screw lead, and transmission efficiency—in the Flash configuration sector. Before performing the difference operation in each processing cycle, the system first calls the electromechanical transmission coupling mapping module. Using the estimated mechanical vibration of the base output by the vibration model from the previous cycle as input, and combining the reduction ratio, lead screw lead, and transmission efficiency parameters, the mechanical vibration of the base is equivalently converted into the load resistance fluctuation at the motor output shaft. This is then further converted into an expected value of equivalent current disturbance consistent with the dimensions of the motor current sensor parameters and written into the look-ahead vibration interference equivalent current register. The output of this register is the look-ahead vibration interference estimate that can be directly used in subsequent difference operations, with units consistent with the motor current sensor parameters.

[0061] The source of the look-ahead vibration disturbance estimate is the stored value written to the look-ahead vibration disturbance equivalent current register at the end of the previous cycle. The seat motor electromechanical sensing parameter is the real-time current sampling value output by the current sampling module in the current cycle. Both data are stored in corresponding registers in amperes. After the controller reads the current values ​​of the two registers, it performs a subtraction operation, subtracting the look-ahead vibration disturbance estimate from the current sampling value. The difference obtained is the physical observation deviation, which is written to the physical observation deviation register. The physical meaning of this deviation is: the residual between the current measured current of the sensor and the vibration superposition current predicted by the model in the previous cycle. It includes the tracking error component caused by the phase accumulation drift of the vibration model. For example, when the engine speed fluctuates slightly in the previous control cycle, a deviation residual of about 10% to 20% of the rated disturbance amplitude will appear in the physical observation deviation register. This residual is the effective information source for drive phase correction.

[0062] The current value of the physical observation deviation register is used as a feedback signal input to the phase correction module of the physical vibration model of the base. This module calculates the phase angle correction increment for the current cycle based on the sign and amplitude of the deviation, using the phase correction step gain pre-stored in the Flash configuration sector as a proportional coefficient. This increment is then superimposed on the current value of the phase register corresponding to each frequency component of the vibration model, completing the feedback correction of the phase variable within each harmonic component. The corrected phase variable and the current amplitude variable in the amplitude register are used together as input to the vibration model output module. This module uses the corrected phase and current amplitude of each harmonic component as parameters, and outputs the corrected vibration interference estimate for the current cycle after conversion by the electromechanical transmission coupling mapping module. This estimate is written into the corrected vibration interference estimate register, and the value of this register is synchronously overwritten into the look-ahead vibration interference equivalent current register, providing an updated look-ahead estimate for the difference operation in the next cycle, forming a cycle-by-cycle closed-loop phase adaptive tracking mechanism.

[0063] Furthermore, the phase correction step gain within the phase correction module is not an empirical scalar that can be arbitrarily set. Its determination method and value range must be subject to strict mathematical constraints based on the system's discrete sampling period and the engine's current excitation fundamental frequency to ensure a balance between the convergence stability and dynamic response speed of the adaptive tracking loop. The physical meaning of this gain is the maximum phase adjustment radians that can be triggered by a unit current observation deviation within a single control cycle. To prevent high-frequency phase oscillations caused by overcorrection, the calibration baseline value of this gain should be directly proportional to the time span of the system control cycle and the current mechanical excitation angular frequency. During the offline vehicle bench test calibration phase before delivery, calibration engineers used a host computer to record the transient physical process data of the engine accelerating rapidly from idle speed to rated full load, and injected trial step gain values ​​of different gradients into the controller. Through a traversal comparison algorithm, the gain point that minimizes the root mean square value of the current compensation residual during transient acceleration over five consecutive control cycles, and whose phase angle does not experience continuous periodic fluctuations during steady-state constant speed operation, was selected as the optimal configuration parameter. Based on the rotational dynamics characteristics of typical earthmoving machinery engines, the reasonable operating range of this gain is usually strictly limited to 5% to 20% of the product of the control cycle and the engine excitation angular frequency. Within this range, the system can both keep up with speed changes and phase slippage caused by load variations at a sufficiently fast rate and effectively suppress spurious phase jitter caused by random sensor measurement noise during steady-state operation.

[0064] The current value of the corrected vibration interference estimate register and the current value of the seat motor electromechanical sensing parameter register are read into the subtraction compensation module at the same time. The corrected vibration interference estimate is subtracted from the motor current sampling value, and the difference is written into the motor base operating parameter register. The value in this register is the net current value that truly reflects the load state of the motor body after removing all equivalent current disturbance components caused by the base vibration from the motor sensing parameters within the current control cycle. The unit is amperes, and the dimension is consistent with the original motor current sensing parameter. This value serves as the input data source for the anti-pinch detection module and the position adjustment module and is directly used in subsequent control cycles without carrying the interference components introduced by the base vibration.

[0065] By using electromechanical transmission coupling mapping transformation, the dimensions of the vibration estimate are made consistent with those of the sensing parameters. The phase variable is corrected by closed-loop feedback of physical observation deviation, and the phase drift of the vibration model is eliminated cycle by cycle. Finally, the baseline operating parameters that truly reflect the load state of the motor body are output, effectively suppressing the interference of base vibration on the anti-pinch judgment.

[0066] Furthermore, obtaining the predicted mean of the equivalent clamping physical resistance of the seat after a preset number of electromechanical control cycles, and the corresponding upper limit for triggering the anti-pinch warning, specifically includes: constructing the electromechanical dynamics equation of the seat DC motor with the seat guide rail position, mechanical adjustment speed, and equivalent clamping physical resistance as state variables, and extracting the electromechanical system state transition matrix; calculating the dynamic control gain using the motor's baseline operating parameters, the updated process disturbance covariance matrix at the current moment, and the observation disturbance covariance matrix, and solving for the posterior physical state estimate of the state variables and the corresponding posterior error covariance matrix; using the posterior physical state estimate as the initial point, and using the electromechanical system state transition matrix... Perform a preset number of forward state transition operations to calculate the predicted mean of the equivalent clamping physical resistance of the seat for the corresponding future period; use the electromechanical system state transition matrix, the transpose of the electromechanical system state transition matrix, and the process disturbance covariance matrix to perform a preset number of state evolution accumulation calculations on the posterior error covariance matrix to obtain the prediction error covariance matrix for the corresponding future period; extract the diagonal elements corresponding to the equivalent clamping physical resistance in the prediction error covariance matrix and calculate the standard deviation of the physical resistance; add a preset multiple of the standard deviation of the physical resistance to the predicted mean of the equivalent clamping physical resistance of the seat to obtain the upper limit of the anti-pinch warning trigger.

[0067] Specifically, all the above operations are performed on the main computing core of the anti-pinch controller microcontroller chip. Each electromechanical control cycle is a period. The state transition matrix, posterior physical state estimation vector, posterior error covariance matrix, prediction mean, prediction error covariance matrix and anti-pinch warning trigger upper limit are all stored in a group of independent function registers in the on-chip static random access memory in single-precision floating-point format. The interrupt service routine reads and writes the data sequentially according to the register address order in each cycle. All data completes a closed-loop flow inside the chip without passing through external storage.

[0068] Specifically, the process of constructing this three-dimensional electromechanical equation is as follows: The armature circuit equations of the DC motor, Newton's equations of motion of the guide rail transmission mechanism, and the quasi-static modeling assumptions of the clamping resistance need to be established simultaneously. First, the kinematic basis is established, and the guide rail position variable is strictly defined as the first-order integral relationship of the mechanical adjustment speed with respect to time. Second, the electromechanical coupling mechanism is introduced, and the back electromotive force voltage component determined by the mechanical adjustment speed and gear reduction ratio is subtracted from the armature voltage balance equation. In the Newton rotational dynamics equation, the external clamping load is physically converted to the motor shaft end through the lead screw and reduction ratio, and substituted as the equivalent clamping physical resistance torque that resists the rotor motion. Finally, given that the equivalent clamping physical resistance has extremely slow static change characteristics when there is no sudden foreign object intervention, it is independently modeled as a quasi-static physical quantity that follows a discrete random walk process (i.e., the transmission coefficient of its own historical state is set to 1 in the state transition matrix, and uncertain sudden changes are attributed to process disturbances), thus completing the complete construction from continuous physical laws to a discrete state space transition matrix.

[0069] The system uses three state variables: seat rail position, mechanical adjustment speed, and equivalent clamping physical resistance. These variables are constructed using the armature circuit equations of the DC motor, Newton's equations of motion for the rail transmission mechanism, and quasi-static modeling assumptions for clamping resistance. The physical meaning of the rail position variable is the linear displacement of the rail corresponding to the motor output shaft after being converted via the lead screw at the current moment, in millimeters. The adjustment speed variable is the first-order rate of change of the rail displacement with respect to time, in millimeters per second. The equivalent clamping physical resistance variable is the equivalent resistance torque of the external clamping load on the rail after being converted to the motor shaft end, in Newton-meters. It is close to zero under normal non-clamping conditions and shows a significant positive jump when foreign object clamping occurs. The above three state variables are stored in a 3D state vector register group in column vector format. The initial values ​​are assigned by the position encoder reading, speed estimate, and resistance initialization zero value when the system is powered on.

[0070] Using the periodic control cycle as the time step, based on the physical evolution relationship of each state variable in the electromechanical dynamics equation between two adjacent control cycles, the linear influence coefficient of each state variable on each state variable in the next cycle is calculated. All coefficients are arranged into a 3x3 electromechanical system state transition matrix and written into the state transition matrix register group in single-precision floating-point format. This matrix is ​​a constant matrix under the condition that the system parameters (armature resistance, lead screw, reduction ratio, etc.) remain unchanged. It is only triggered to recalculate when the physical parameters of the transmission mechanism change. In daily operation, it is directly read and used from the state transition matrix register group.

[0071] In each Kalman filter update step, the controller takes the updated process disturbance covariance matrix and the prior error covariance matrix written in the previous cycle as inputs. Following the standard calculation logic of Kalman gain, it performs joint calculations on the 3×3 prior error covariance matrix, the 1×3 observation matrix, and the observation disturbance covariance matrix to obtain a 3D Kalman gain vector, which is written to the gain vector register group in single-precision floating-point format. The current valid parameters of the process disturbance covariance matrix come from the process disturbance covariance matrix parameter register group updated according to the operating condition category in the previous implementation. The parameters of the observation disturbance covariance matrix come from the corresponding register group. The two sets of parameters have been switched according to the aforementioned mechanism under strong impact conditions. Here, the current values ​​of the registers are directly read and substituted into the calculation without additional judgment.

[0072] The current value of the gain vector register group and the current value of the motor's baseline operating parameter register (i.e., the net current sample value after vibration compensation, in amperes) are jointly input into the posterior state update module. This module first multiplies the state transition matrix with the posterior state estimation vector of the previous cycle to obtain a 3D prior state estimation vector, which is then written into the prior state estimation register group. Subsequently, the prior state estimation vector is mapped to the observation space using the observation matrix (i.e., the linear combination component corresponding to the motor current is extracted) and subtracted from the motor's baseline operating parameters to obtain a scalar innovation value, which is written into the innovation register. Finally, the Kalman gain vector is multiplied with the innovation value and superimposed onto the prior state estimation vector to obtain a 3D posterior physical state estimation vector, which is written into the posterior state estimation register group. The third element is the posterior estimate of the equivalent clamping physical resistance for the current cycle, in Newton-meters (Nm). For example, this value is approximately 0.05 Nm during normal no-load regulation, and can jump to, for example, over 0.8 Nm when foreign object clamping occurs.

[0073] The posterior error covariance matrix is ​​calculated from the gain vector, the observation matrix, and the prior error covariance matrix according to the standard update logic. It is written into the posterior error covariance matrix register group in 3×3 single-precision floating-point matrix format. The diagonal element in the 3rd row and 3rd column of this matrix represents the uncertainty variance of the current period's estimation of the equivalent clamping physical resistance, which physically reflects the degree of dispersion of the resistance estimation error under the current operating condition.

[0074] Specifically, the observation matrix is ​​defined as a one-dimensional row vector with eight columns. Each element in this row vector represents a comprehensive mapping coefficient that transforms the mechanical displacement amplitude of the corresponding frequency component (cosine or sine) into the equivalent motor current through electromechanical coupling. This comprehensive mapping coefficient is derived from the lead screw, transmission mechanism reduction ratio, mechanical transmission efficiency, and DC motor torque constant through algebraic physical conversion. In the eight-dimensional state vector, two adjacent variables represent the cosine and sine amplitudes of the same frequency component, respectively. Therefore, in the row vector of the observation matrix, the elements at odd and even positions are numerically set to the same scaling factor of the comprehensive mapping coefficient, ensuring that orthogonal mechanical vibration components of different phases can be projected onto the scalar observation space of the motor stator current with equal weight. During the measurement update phase of each Kalman control cycle, the controller calculates the system's predicted observation value of the current superimposed vibration current by multiplying this one-row, eight-column observation matrix with the eight-dimensional prior state estimation column vector, and then subtracts it from the actual motor current sensing parameters to obtain physical information.

[0075] Using the current 3D vector of the a posteriori physical state estimation register group as the initial point, the controller continuously performs a preset number of forward state transition operations (e.g., 5 times, corresponding to forward prediction for the next 5 cycles) within the same cycle interrupt. Each operation multiplies the current state vector by the state transition matrix on the left, and the product result is written to the temporary state vector register. This process is repeated until all preset number of operations are completed. Finally, the third element in the temporary state vector register is the predicted mean of the equivalent clamping physical resistance after the preset number of control cycles. This value is written to the resistance prediction mean register in single-precision floating-point format, with the unit being Newton-meters.

[0076] The underlying electromechanical dynamics model of the second-channel Kalman filter unit is deeply dependent on the physical parameters of the DC motor. Specifically, the controller inputs the motor's armature resistance, back EMF coefficient, mechanical reduction ratio, and transmission efficiency as fixed physical constants into the system. In the discretized state transition matrix, the matrix elements describing the effect of velocity on position integral are directly determined by the sampling time step of the control cycle; the matrix elements describing the self-coupling attenuation of back EMF on velocity are determined by the quotient of the back EMF coefficient and the armature resistance; and the equivalent clamping physical resistance variable, due to its extremely slow static change characteristics when no external foreign object clamping occurs, is assumed in the model to be a quasi-static physical quantity following a discrete random walk process. Therefore, in the state transition matrix, the main diagonal elements corresponding to the historical state transmission of the clamping resistance itself are strictly set to a value of 1, and the possible abrupt changes and uncertainties are completely quantified in the process disturbance covariance matrix.

[0077] Furthermore, the observation matrix used in the second channel aims to map the three-dimensional electromechanical state vector to the scalar background current of the motor. Therefore, this observation matrix is ​​designed as a 1x3 row vector. Since the pure spatial position variable does not directly generate dynamic current consumption in steady state, its corresponding first row element is set to 0. For the mechanical adjustment speed variable in the second dimension, which reflects the linear motion speed of the guide rail, when the DC motor rotates to drive the seat, the rotor cutting the magnetic field lines generates a back electromotive force (EMF) opposite to the direction of the input drive voltage. The magnitude of this back EMF directly depends on the motor speed, which is determined by the linear mechanical adjustment speed of the guide rail and the gear reduction ratio. The presence of the back EMF will offset part of the power supply armature voltage, thereby proportionally reducing the actual dynamic current through the motor. This section clarifies the specific calculation logic for the second row element of the observation matrix (i.e., the coefficient mapping the velocity state to the current observation value): this element should be configured as a negative value. In the specific algebraic conversion, the back EMF constant calibrated at the motor's factory is first multiplied by the system's mechanical reduction ratio to obtain the equivalent back EMF generated per unit linear velocity. Then, this product is divided by the fixed resistance value of the motor's armature winding. Finally, a negative sign is added before the result to obtain the element's value. This rigorous derivation ensures that the Kalman filter can accurately calculate and subtract the natural current decay component caused by the seat's movement speed during measurement updates. The equivalent clamping resistance is directly expressed as the rotor load torque and linearly converted into the drive current; its corresponding third row element is configured as the product of the reciprocal of the motor torque constant and the mechanical reduction ratio. This specific topology of the observation matrix ensures that the filter correctly separates the velocity dynamic response component and the actual clamping resistance abrupt change component from the current sensing fluctuations during the update step.

[0078] Furthermore, in parallel with the forward state transition operation, the controller starts from the current value of the a posteriori error covariance matrix register group and performs the same number of prediction error covariance matrix recursive operations as the forward prediction. Each recursion is performed by left-multiplying the state transition matrix by the current covariance matrix, then right-multiplying by the transpose of the state transition matrix, and finally superimposing the process disturbance covariance matrix. The result is written to the temporary covariance matrix register, and the loop iterates until all preset number of iterations are completed. Finally, the diagonal element of the 3rd row and 3rd column in the temporary covariance matrix register is the variance value of the prediction uncertainty of the equivalent clamping physical resistance at the future preset period time. The square root of this variance value is taken and written to the resistance prediction standard deviation register, with the unit being Newton-meters. This standard deviation increases with the increase of the working condition disturbance intensity. For example, under the impact condition of a hydraulic breaker, its value is about 3 to 5 times that under the idling condition.

[0079] The current values ​​of the resistance prediction standard deviation register and the resistance prediction mean register are simultaneously read into the anti-pinch warning upper limit calculation module. The product of the predicted mean plus a preset multiple and the predicted standard deviation (the preset multiple is fixed in the Flash configuration sector with the factory calibration value, for example, 3 times to cover the 99.7% probability distribution range of resistance fluctuation under normal working conditions) is used as the anti-pinch warning trigger upper limit. This upper limit value is written into the anti-pinch warning trigger upper limit register, in Newton-meters. The current value of this register is read by the anti-pinch judgment module at the end of each cycle and compared with the current posterior equivalent clamping physical resistance estimate. Once the posterior estimate exceeds the trigger upper limit, a stop command is issued to the anti-pinch execution module.

[0080] In the operating logic of the second-channel Kalman filter unit, the prediction step N of the forward state transition operation is the optimal integer obtained by solving the double envelope constraint equation jointly imposed by the hardware electrical and mechanical physical response hysteresis limit of the control system and the mathematical filter prediction divergence boundary.

[0081] First, the absolute lower limit constraint of the prediction step number N must be determined: this lower limit must ensure that the total advance warning time physical window derived by the filter forward derivation is strictly and always greater than the physical time of the comprehensive hardware action delay of the entire electromechanical system. This unavoidable delay time consists of two rigid physical time consumptions in series: one is the electrical response delay of the microcontroller cutting off the pulse width modulation signal and establishing a short-circuit braking cycle path in the power bridge arm after receiving the software instruction, and the other is the mechanical braking deceleration time required for the seat DC motor rotor and its coupled gear reduction group to attenuate their rotational kinetic energy to a complete stop through inertia after being subjected to the short-circuit electromagnetic reverse braking torque. Let this comprehensive delay time obtained by physical bench measurement be a constant value. Dividing this constant value by the duration of a single electromechanical discrete control cycle and rounding the result up gives the absolute safety lower limit benchmark for the prediction step number.

[0082] Secondly, an absolute upper limit constraint on the number of prediction steps N must be determined: Since forward multi-step extrapolation is essentially an open-loop error accumulation mathematical process lacking closed-loop correction from subsequent real sensor observation data, the prediction error covariance calculated by the covariance matrix recursive equation will rapidly expand as the number of iteration steps N increases. When the iterative extrapolation reaches a certain future cycle, if the predicted physical standard deviation of the corresponding clamping resistance is amplified to a level sufficient to mask the safety margin between the normal unloaded mechanical friction fluctuation baseline and the dangerous clamping safety threshold, the extrapolated upper limit value will completely lose any engineering value for anti-pinch trigger judgment due to its low mathematical confidence. Calibration engineers need to establish the limit number of cycles corresponding to this error variance divergence boundary through offline state-space simulation analysis during the factory design phase, using this as the absolute prohibition upper limit of N.

[0083] During the actual microcontroller firmware compilation and solidification, the optimal forward prediction step number N is selected at a precise point that closely follows the lower limit reference and has appropriate redundancy of one to two discrete control cycles. This scientifically rigorous dual-envelope selection mechanism not only reserves 100% sufficient physical give-off time space for the underlying motor hardware to perform full-power emergency braking, but also mathematically minimizes the random false alarms caused by the divergence of system uncertainties introduced by over-model extrapolation, thus achieving a theoretical balance between Kalman feedforward safety protection and electromechanical control.

[0084] Using a 3D electromechanical state space as a framework, the prediction and update closed loop of Kalman filtering is fully realized. The sum of the forward prediction mean of the equivalent clamping physical resistance and the recursively obtained standard deviation is used as the upper limit of the adaptive anti-pinch warning trigger, so that the threshold dynamically adapts with the intensity of the working condition disturbance, which can effectively eliminate the false alarm and missed alarm problem of fixed threshold under strong vibration conditions.

[0085] Furthermore, the step of generating an electromechanical anti-pinch trigger command and driving the seat motor to perform mechanical braking or reverse unloading actions in advance specifically includes: obtaining a pre-calibrated static safety mechanical threshold of the base; extracting the first-order time derivative value of the motor's baseline operating parameters, and using the first-order time derivative value to dynamically and physically lower the static safety mechanical threshold of the base to obtain the absolute anti-pinch safety physical threshold of the current cycle; comparing the upper limit of the anti-pinch warning trigger with the absolute anti-pinch safety physical threshold; generating an electromechanical anti-pinch trigger command when the upper limit of the anti-pinch warning trigger is greater than the absolute anti-pinch safety physical threshold; responding to the electromechanical anti-pinch trigger command, cutting off the current seat motor pulse width modulation control signal; and conducting the same-side bridge arm of the seat motor drive circuit to form a short-circuit loop to perform motor braking, or applying a drive signal of opposite polarity to control the seat motor to perform a reverse yielding action.

[0086] Specifically, all the above-mentioned calculations and hardware control actions are performed on the main computing core and on-chip general-purpose output ports of the anti-pinch controller microcontroller chip. The control cycle is based on a period. The static safety mechanical threshold of the base, the smoothed background operating parameters, the smoothed rate of change value, the dynamic adjustment compensation amount, the absolute anti-pinch safety physical threshold, and the anti-pinch trigger instruction flag are all stored in independent function register groups in the on-chip static random access memory area in single-precision floating-point format or logic level format. Each module completes reading and writing sequentially in the same cycle interrupt service routine according to the register address order. All data completes closed-loop flow inside the chip without passing through external storage.

[0087] The static safety mechanical threshold of the base is determined once during the mechanical boundary calibration stage before the system leaves the factory: The calibration engineer applies a gradually increasing standard load to the seat guide rail clamping mechanism in a static bench environment, and simultaneously records the equivalent resistance torque converted to the motor shaft end. The equivalent resistance torque corresponding to the point when the guide rail transmission mechanism begins to show irreversible deformation or the contact force on the occupant limb model (a physical dummy model that meets industry safety standards) exceeds the safety limit is taken as the static safety boundary. After taking the arithmetic mean of no less than 5 sets of independent pressure data, the final calibration value is burned into the safety threshold parameter area of ​​the Flash configuration sector in single-precision floating-point format, with the unit being Newton-meters. When the controller is powered on and initialized, it is read from this area and loaded into the base static safety mechanical threshold register for later use. For example, when the calibration result is 1.2 Newton-meters, the fixed value stored in this register is 1.2.

[0088] Before extracting the first-order time derivative value from the motor's baseline operating parameters, a digital low-pass smoother is needed to suppress high-frequency noise in the baseline operating parameters. This is because directly performing differential operations on the original sampled values ​​will drastically amplify the high-frequency random electrical noise in the sensor, causing severe fluctuations in the subsequently calculated dynamic compensation value, which in turn leads to irregular jumps in the absolute anti-pinch safety physical threshold and false triggering of the anti-pinch command. The digital low-pass smoother is implemented using a first-order exponential weighted average algorithm. The smoothing coefficient is fixed in the Flash configuration sector with the factory calibration value (for example, 0.15, which means the current smoothed output value is the sum of 0.85 times the smoothed value of the previous cycle and 0.15 times the original value of the current baseline operating parameter). In each cycle, the current net current value is read from the motor's baseline operating parameter register, and after weighted calculation with this smoothing coefficient, the result is written to the smoothed baseline operating parameter register. The output value is in amperes. This output serves as the input data source for subsequent differential operations, replacing the original sampled values ​​in the rate of change calculation.

[0089] The current value of the smoothed baseline operating parameter register is subtracted from the historical value written to the register in the previous cycle (the controller temporarily stores the old value in the historical smoothed parameter register before writing a new value each time), and then divided by the cycle control cycle length to obtain the first-order time derivative value of the baseline operating parameter, in amperes per second. This difference result is written to the original rate of change register. Subsequently, to further suppress the short-term noise spikes remaining after the difference operation, the controller applies a sliding time window averaging algorithm to the output of the original rate of change register. The arithmetic mean is calculated using the historical values ​​of the original rate of change of the most recent five control cycles continuously stored in the on-chip static random access memory (maintained in a cyclic overwrite manner, with the oldest record updated each cycle). The mean result is written to the smoothed rate of change register, still in amperes per second. The output value of this register is the first-order time derivative smoothed value finally used for dynamic adjustment compensation calculation. Its physical meaning is the smoothed rate of change of motor load at the current moment. A positive value represents the load accelerating upward, while a zero or negative value represents the load stabilizing or decreasing.

[0090] The current value of the smooth change rate register is input to the dynamic adjustment compensation calculation module. This module multiplies the smooth change rate by the electromechanical sensitivity compensation coefficient pre-stored in the Flash configuration sector. The product is the dynamic adjustment compensation amount, which is written to the compensation amount register in Newton-meters. The compensation amount is only used in subsequent adjustment calculations when the smooth change rate is positive (i.e., the load is continuously increasing). When the smooth change rate is zero or negative, the compensation amount is forcibly clamped to zero to prevent the reverse effect of lowering the anti-pinch sensitivity due to threshold adjustment when the load decreases. This clamping judgment is executed sequentially within the same interrupt, and the result is overwritten and written to the compensation amount register.

[0091] Specifically, to ensure the absolute physical safety of the first-order differential dynamic down-adjustment mechanism and prevent frequent false triggering caused by overcompensation, the final determination of the electromechanical sensitivity compensation coefficient must strictly rely on a real-vehicle mechanical bench calibration procedure that conforms to international automotive-grade safety standards. The specific calibration process and physical constraints are as follows: The calibration process requires the use of a simulated dummy's limb resistance gauge that conforms to automotive anti-pinch safety industry standards, or a mechanical spring clamp with a calibrated linear stiffness coefficient, placed vertically in the expected clamping danger zone along the seat guide rail adjustment path. The test platform software first controls the seat motor to actively and continuously impact the aforementioned spring clamp with a stepped, incremental closed-loop adjustment speed sequence.

[0092] During each controlled impact physical process, the external high-speed data acquisition card must synchronously record two critical timestamps with microsecond-level precision: the first is the dangerous penetration moment when the measured physical compressive force of the resistance gauge first exceeds the legally mandated safe contact mechanical limit (e.g., 100 Newtons); the second is the moment when the anti-pinch controller completes Kalman deduction and sends a pulse width modulation cutoff signal to the drive axle, thereby triggering the electrical response of the motor braking. The total system response delay is strictly defined as the complete physical time period from the first sharply rising abrupt edge of the current differential rate to the actual physical activation of the braking hardware, the complete elimination of rotor kinetic energy, and the occurrence of reverse backlash.

[0093] The calculation logic of the electromechanical sensitivity compensation coefficient essentially lies in finding an optimal slope value in a multi-dimensional space. Through a nonlinear inverse fitting algorithm, the calibration system iteratively optimizes this compensation coefficient, ensuring that the dynamically adjusted compensation amount calculated from the first-order time derivative, covering all verification speed test levels (low, medium, and high), precisely offsets the additional destructive torque generated by the motor due to massive inertial slippage during the entire total response delay period. Simultaneously, this fitting optimization process must be subject to a mandatory underlying safety margin constraint: regardless of the external clamping speed, the absolute downward displacement of the safety threshold caused by the calculated compensation coefficient must not fall below the absolute minimum margin required to distinguish the anti-pinch program from the random noise amplitude of the motor's inherent no-load friction. Only coefficient values ​​calculated under this stringent physical testing process, encompassing boundary conditions, can be considered legitimate and safe data, and ultimately securely burned into the non-volatile memory of the control chip.

[0094] The fixed calibration value in the static safety mechanical threshold register of the base and the current value in the compensation register are simultaneously read into the absolute anti-pinch safety physical threshold calculation module. The absolute anti-pinch safety physical threshold for the current cycle is obtained by subtracting the compensation amount from the static threshold and written to the absolute threshold register, with the unit being Newton-meters (Nm). To prevent the absolute threshold from being lowered to an unreasonably negative or excessively low value due to excessive compensation amount under extreme working conditions, the controller performs a lower limit clamp on the calculation result before writing. The lower boundary is the minimum allowable threshold (e.g., 50% of the static safety mechanical threshold, i.e., 0.6 Nm) that is factory-fixed in Flash. If the calculated value is lower than this lower boundary, the lower boundary value is forcibly written to ensure that the absolute anti-pinch safety physical threshold is always within the effective physical safety range. For example, during normal low-speed adjustment, the smooth change rate is close to zero, and the absolute threshold is approximately equal to the static calibration value of 1.2 Nm. During rapid adjustment, the smooth change rate is larger, and the absolute threshold can be lowered to, for example, 1.0 Nm to tighten the anti-pinch judgment boundary in advance.

[0095] The absolute threshold register and the current value written to the anti-pinch warning trigger upper limit register in the previous implementation are simultaneously read into the anti-pinch trigger comparison module to perform a single scalar size comparison operation. When the read value of the anti-pinch warning trigger upper limit is greater than the read value of the absolute anti-pinch safety physical threshold, the comparison module sets the on-chip anti-pinch trigger instruction flag register to a logic high level and generates an electromechanical anti-pinch trigger instruction. Otherwise, the flag remains at a logic low level. The comparison result is overwritten and updated once in each cycle to ensure that the anti-pinch judgment responds in real time to the relative change between the resistance prediction value and the safety threshold.

[0096] After the anti-pinch trigger instruction flag register is set to logic high, the controller immediately writes a forced stop instruction with a duty cycle of zero to the pulse width modulation signal generation module within the same interrupt service routine. The modulation signal on the pulse width modulation output pin is cut off within the current control cycle, the gate drive signal of the power switch in the motor drive circuit is synchronously cleared to zero, and the active drive current loop in the motor winding is disconnected. Subsequently, the controller selects the subsequent action according to the anti-pinch response mode configuration word (the value is either braking mode or yielding mode, written by the product model at the factory) pre-stored in the Flash configuration sector. If the configuration word indicates braking mode, the controller simultaneously sends a turn-on instruction to the two power switches on the same side of the upper and lower bridge arms of the motor drive circuit through the on-chip general-purpose output port. After both switches are turned on simultaneously, a low voltage is formed at both ends of the motor winding through the internal drive circuit. In the impedance short-circuit loop, the back electromotive force generated by the motor during inertial rotation drives the braking current to circulate within the short-circuit loop, forming a braking torque opposite to the direction of rotation. The guide rail decelerates to a standstill, preventing it from continuing to move in the clamping direction. If the configuration word indicates the retraction mode, the controller writes a reverse polarity enable flag and a preset retraction duty cycle value (e.g., 40% of the rated drive duty cycle) to the pulse width modulation signal generation module. The drive circuit switches to the reverse bridge arm conduction combination, and a reverse current is supplied to the motor windings. The motor outputs a reverse torque to drive the guide rail to retract away from the clamped object. The retraction action continues until the position encoder detects that the guide rail has retracted a preset distance (e.g., 5 mm), at which point the controller cancels the reverse drive command. During the retraction process, the anti-pinch trigger command flag remains high to ensure that the retraction action is not overwritten or interrupted by the normal adjustment logic.

[0097] By introducing a two-stage noise reduction mechanism of exponential weighted smoothing and sliding window averaging before and after the differential operation, the smooth physical continuity of the dynamically adjusted compensation amount is ensured. The anti-pinch command is driven by real-time comparison of the predicted trigger upper limit and the dynamic adaptive safety threshold, which is compatible with two hardware execution paths: braking and reversal. This achieves early response to rapid clamping events and effective suppression of false triggering caused by sensor noise.

[0098] This embodiment achieves a highly sensitive early response to seat rail clamping events under all working conditions of a hydraulic excavator through the synergistic effect of four mechanisms: working condition adaptive filter parameter switching, multi-order harmonic vibration active compensation, forward resistance probability prediction, and two-level smoothing dynamic threshold. At the same time, it effectively suppresses anti-pinch malfunctions caused by vibration noise and strong impacts, significantly improving the safety and reliability of the system under extreme working conditions compared to a fixed threshold scheme.

[0099] Example 2: In the case of a hydraulic excavator operating under combined travel and hydraulic working conditions (i.e., the machine is simultaneously traveling on the ground and operating in conjunction with the hydraulic working device), the vibration excitation experienced by the seat base originates from two independent mechanical vibration sources: the engine combustion excitation and the travel drive hydraulic motor. The excitation frequencies of the two sources vary independently with their respective rotational speeds, and there is an irregular beat frequency superposition effect in the time domain. If a single-source vibration model based solely on the engine excitation frequency is used for compensation, the vibration component introduced by the travel motor will continuously mix into the motor's baseline operating parameters as residual interference. When the travel speed is close to the engine speed, low-frequency synthetic beat vibrations will occur, causing periodic false fluctuations in the Kalman estimate of the equivalent clamping physical resistance. As a result, the upper limit of the anti-pinch warning trigger repeatedly approaches the safety threshold, leading to unnecessary false braking actions. This embodiment refines the two steps of vibration model construction and baseline parameter extraction for the above-mentioned combined vibration source scenario.

[0100] Within each cycle interrupt service routine, in addition to reading the engine mechanical speed from the chassis control bus, the controller synchronously reads the output speed parameter of the travel hydraulic motor (stored in the travel motor speed register in units of speed per minute). Based on the travel motor cylinder number and displacement parameters pre-stored in the Flash configuration sector, the controller calculates the hydraulic pulsation excitation fundamental frequency generated by the travel motor at the current speed and writes it into the travel excitation fundamental frequency register in Hertz. Simultaneously, the controller calculates its second and third harmonic frequencies and writes them into the travel harmonic frequency register group respectively. All frequencies are recalculated synchronously as the travel motor speed is updated. The controller and the engine excitation frequency register group are stored independently and do not overwrite each other.

[0101] The physical vibration model of the base expands the travel motor excitation channel on the basis of the original multi-harmonic components based on the engine excitation frequency. It adds three travel excitation frequency components: the fundamental frequency component of travel excitation and the second and third harmonic components. Each component is represented by two state variables, which are the cosine term amplitude coefficient and the sine term amplitude coefficient, respectively. The three travel excitation components add a total of 6 state variables. After merging with the original 8 state variables, the state vector is expanded to 14 dimensions and written into the extended vibration model state vector register group in column vector format. The amplitude variables of each order of travel excitation are initialized from the travel motor excitation amplitude mapping spectrum pre-stored in Flash (with travel motor speed and hydraulic system pressure as dual indices, and the calibration method is the same as the engine amplitude spectrum). The phase reference is the reference synchronization frame sent by the travel motor control unit on the chassis control bus at a fixed position in each working cycle. The reception time is recorded in the travel phase reference timestamp register. The phase angle of each order of travel harmonics is written into the travel harmonic phase register group according to the same calculation logic as the engine phase reference.

[0102] The state transition matrix is ​​then expanded to 14 rows and 14 columns. The engine excitation component sub-block and the travel excitation component sub-block each occupy an independent sub-matrix region on the diagonal. The cross-coupling element between the two is fixed to zero, reflecting the physical fact that the two vibration sources are mechanically independent. After each effective update of the engine excitation base frequency register or the travel excitation base frequency register, the controller only recalculates the sub-matrix sub-block occupied by the corresponding vibration source and overwrites the corresponding position written to the state transition matrix register group. The non-updated sub-blocks keep the stored value of the previous cycle unchanged, avoiding unnecessary full matrix recalculation overhead. The output of the reconstructed 14 rows and 14 columns state transition matrix register group serves as the direct data input for the prediction step of the extended first channel Kalman filter unit.

[0103] In the background operating parameter extraction stage, the look-ahead vibration disturbance estimate output by the extended vibration model is expanded from the original 8-dimensional state vector mapping output to a complete 14-dimensional state vector mapping output. The electromechanical transmission coupling mapping module performs conversion and summation on all 14 state variable components to obtain a comprehensive look-ahead vibration disturbance estimate that includes the equivalent current disturbance contribution of all frequency components of engine excitation and travel motor excitation. This estimate is written into the look-ahead vibration disturbance equivalent current register. The subsequent subtraction operation process is completely consistent with Example 1. The combined interference of the two vibration sources is subtracted from the motor sensing parameters in one step to output the net current background operating parameter that truly reflects the load state of the motor body.

[0104] The electromechanical transmission coupling mapping module must execute a rigorous three-level cross-domain physical conversion logic and perform normalization processing to ensure dimensional consistency. The first level is the translational conversion from kinematics to mechanics: The controller first extracts the total equivalent translational mass of the seat system stored in the internal register. This mass includes the self-weight of the seat's mechanical structure and the adaptively calculated occupant mass. This total mass is multiplied by the estimated axial mechanical vibration acceleration output based on the vibration model, and then superimposed with the pre-calibrated Coulomb friction constant of the guide rail. This calculates the transient physical impact linear force dynamically acting on the guide rail lead screw. The second level is the transmission conversion from linear mechanics to rotational torque: The above linear force is multiplied by the physical lead of the guide rail lead screw, and then divided by a preset double value of pi, as well as the mechanical transmission efficiency coefficient and gear reduction ratio. This accurately converts the translational linear force acting on the guide rail into the rotational resistance torque fluctuation acting on the output shaft of the seat's DC motor. The third stage involves the final conversion from rotational torque to electromagnetic parameters: the controller uses the DC motor's electromagnetic torque constant, pre-measured and fixed offline at the factory, to divide the converted rotational resistance torque by this torque constant. During this process, the controller also performs a slight nonlinear down-adjustment compensation based on the current power supply bus voltage read from the bus to overcome magnetic circuit saturation effects, ultimately outputting the expected equivalent current disturbance value that is completely consistent with the current sensor's sampling dimension. This rigorous end-to-end conversion ensures that mechanical vibration interference is accurately and losslessly mapped to electrical current fluctuation components, providing a theoretically unbiased physical basis for subsequent dimensional subtraction.

[0105] Furthermore, when the difference between the travel motor speed and the engine excitation fundamental frequency is less than the beat frequency detection threshold (e.g., 2 Hz) pre-stored in Flash, the controller sets the beat frequency warning flag register to a logic high level and simultaneously replaces the process disturbance covariance parameters of each order component of the travel excitation with a beat frequency enhancement coefficient (e.g., 8 times the normal value) to improve the adaptive tracking bandwidth of the Kalman filter in this frequency band, accelerate the convergence speed of the amplitude and phase variables of the travel excitation components in the beat frequency range, and prevent the decrease in the accuracy of the synthetic interference estimation when the frequencies of the two vibration sources are close.

[0106] This embodiment expands the base vibration model from a single excitation source to a dual excitation source independent channel structure of engine and walking motor. It uses a 14-dimensional state vector to synchronously track the full-frequency components of the two vibration sources and synthesize the equivalent compensation amount. This fundamentally eliminates the periodic interference of the beat frequency superposition of the two vibration sources on the anti-pinch resistance estimation under the walking compound operation condition. Without changing the overall anti-pinch control architecture, the system's adaptability to compound working condition vibration scenarios is extended from a single excitation source to a multi-excitation source coupled scenario.

[0107] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for preventing seat pinching in earthmoving vehicles based on Kalman filtering, characterized in that, include: The system acquires vehicle bus telemetry parameters and electromechanical sensor feedback, classifies road vibration conditions and determines whether the transient physical impact load on the seat mechanical structure exceeds the limit, obtains the current vehicle operating condition category and impact indicator, and simultaneously switches the process disturbance covariance matrix of the Kalman filter unit in the anti-pinch controller; and when the system determines that the mechanical impact is strong, it actively suspends the electromechanical adjustment function of the seat motor and amplifies the parameters of the observed disturbance covariance matrix to the preset hardware extreme value to shield the distorted sensor signal flow caused by strong vibration. The mechanical speed parameters of the engine and the electromechanical sensing parameters of the seat motor are collected and input into the first channel Kalman filter unit to perform online tracking and adaptive subtraction compensation of the base vibration, and extract the base operating parameters of the motor after removing the vehicle body vibration. Based on the motor's baseline operating parameters and the updated process disturbance covariance matrix and observation disturbance covariance matrix at the current moment, forward multi-step deduction and parameter evolution calculation of the electromechanical state are performed to obtain the predicted mean value of the seat's equivalent clamping physical resistance after N electromechanical control cycles, as well as the corresponding upper limit for triggering the anti-pinch warning. The upper limit of the anti-pinch warning trigger is compared with the preset absolute anti-pinch safety physical threshold of the seat structure to generate an electromechanical anti-pinch trigger command and drive the seat motor to execute it in advance.

2. The method for preventing seat movement pinching in earthmoving vehicles based on Kalman filtering according to claim 1, characterized in that, Obtain the current vehicle operating condition category and impact indicator, specifically including: The vehicle bus telemetry parameters are physical quantities such as engine speed, main hydraulic pump pressure and hydraulic oil temperature read from the vehicle chassis control bus. The electromechanical sensor feedback quantities are the three-axis acceleration components of the base output by the multi-axis inertial measurement unit and the attitude deflection angle output by the seat internal angle encoder. The sliding time-domain fluctuation of the three-axis acceleration components of the base is calculated to extract the physical characteristics of vibration energy; the vehicle bus telemetry parameters and the physical characteristics of vibration energy are synchronously recombined during the sampling period using a time-series alignment mechanism; the synchronously recombined control parameters are input into the operating condition mapping unit, and the operating condition category of the current vehicle is output. A spatial attitude transformation matrix is ​​constructed using the attitude deflection angle. The three-axis acceleration components of the base are decomposed into the physical motion axis direction of the seat guide rail to obtain the axial inertial acceleration component. The initial static physical load of the seat motor is obtained and the occupant mass coefficient is calculated. The axial inertial acceleration component is multiplied by the occupant mass coefficient to calculate the equivalent inertial impact along the seat guide rail direction. The equivalent inertial impact is compared with the preset equivalent upper limit of inertial impact. When the equivalent inertial impact exceeds the upper limit of the equivalent inertial impact, a hardware-level strong impact flag is triggered; and after the equivalent inertial impact falls back to the preset mechanical safety hysteresis range, the strong impact flag is revoked.

3. The method for preventing seat movement pinching in earthmoving vehicles based on Kalman filtering according to claim 1, characterized in that, The process of switching the Kalman filter unit in the anti-pinch controller involves a disturbance covariance matrix; and when a strong mechanical impact is detected, the electromechanical adjustment function of the seat motor is actively suspended, and the parameters of the observed disturbance covariance matrix are amplified to a preset hardware extreme value to shield the distorted sensor signal flow caused by strong vibration. Specifically, this includes: Establish an electromechanical mapping table corresponding to operating condition categories and process disturbance parameters; according to the operating condition category, retrieve the corresponding target process disturbance parameters from the electromechanical mapping table and update the process disturbance covariance matrix of the Kalman filter unit; when the impact flag is triggered, cut off the power drive circuit of the seat motor and activate the mechanical self-locking brake; in response to the impact flag, replace the diagonal variance parameter of the observation disturbance covariance matrix of the Kalman filter unit with the hardware extremum; after the impact flag is revoked, restore the parameter configuration of the observation disturbance covariance matrix.

4. The method for preventing seat movement pinching in earthmoving vehicles based on Kalman filtering according to claim 1, characterized in that, The application of a physical vibration model of the base based on known engine excitation for online tracking and adaptive subtraction compensation of base vibration includes: Using the engine's mechanical speed parameter, the fundamental frequency of mechanical vibration transmitted from the frame to the seat and the corresponding structural harmonic frequencies are calculated. Combining the pre-extracted natural frequencies and physical damping coefficients of the seat's mechanical structure, the fundamental frequency of mechanical vibration and the structural harmonic frequencies are used as input parameters to construct a physical vibration model of the base that includes electromechanical amplitude variables and phase variables. The physical vibration model of the base is then input into the first channel Kalman filter unit to reconstruct the state transition matrix of the electromechanical system.

5. The method for preventing seat movement pinching in earthmoving vehicles based on Kalman filtering according to claim 4, characterized in that, The specific parameters for extracting the base operating parameters of the motor after removing vehicle body vibration include: The physical observation deviation is extracted by subtracting the look-ahead vibration disturbance estimate from the previous electromechanical sampling cycle and the electromechanical sensing parameters of the seat motor. Based on the physical observation deviation, the phase variables inside the physical vibration model of the base are corrected through feedback, and the corrected vibration disturbance estimate for the current cycle is output. The vibration disturbance estimate is compensated by subtraction between the electromechanical sensing parameters of the seat motor and the vibration disturbance estimate. The electromechanical parameters obtained after the compensation operation are used as the background operating parameters of the motor.

6. The method for preventing seat movement pinching in earthmoving vehicles based on Kalman filtering according to claim 1, characterized in that, The predicted average of the equivalent clamping physical resistance of the seat after N electromechanical control cycles is obtained, along with the corresponding upper limit for triggering the anti-pinch warning, specifically including: The electromechanical dynamics equations of the seat DC motor are constructed with the seat guide rail position, mechanical adjustment speed and equivalent clamping physical resistance as state variables, and the state transition matrix of the electromechanical system is extracted. The dynamic control gain is calculated using the motor's background operating parameters, the updated process disturbance covariance matrix and the observation disturbance covariance matrix at the current moment, and the posterior physical state estimate of the state variables and the corresponding posterior error covariance matrix are solved. Starting from the posterior physical state estimate, a preset number of forward state transition operations are performed using the electromechanical system state transition matrix to calculate the predicted mean of the equivalent clamping physical resistance of the seat for the corresponding future period. Using the electromechanical system state transition matrix, the transpose of the electromechanical system state transition matrix, and the process disturbance covariance matrix, the posterior error covariance matrix is ​​accumulated by state evolution to obtain the prediction error covariance matrix for the corresponding future period. Extract the diagonal elements of the equivalent clamping physical resistance in the prediction error covariance matrix and calculate the standard deviation of the physical resistance; add the predicted mean of the equivalent clamping physical resistance of the seat to the standard deviation of the physical resistance by a preset multiple to obtain the upper limit of the anti-pinch warning trigger.

7. The method for preventing seat movement pinching in earthmoving vehicles based on Kalman filtering according to claim 1 or 6, characterized in that, Generate an electromechanical anti-pinch trigger command and drive the seat motor to perform mechanical braking or reverse force relief actions in advance, specifically including: Obtain the pre-calibrated static safety mechanical threshold of the base; extract the first-order time derivative of the motor's base operating parameters, and use the first-order time derivative to dynamically and physically adjust and compensate the static safety mechanical threshold of the base to obtain the absolute anti-pinch safety physical threshold for the current cycle. The upper limit of the anti-pinch warning trigger is compared with the absolute anti-pinch safety physical threshold; when the upper limit of the anti-pinch warning trigger is greater than the absolute anti-pinch safety physical threshold, an electromechanical anti-pinch trigger command is generated; in response to the electromechanical anti-pinch trigger command, the current seat motor pulse width modulation control signal is cut off; the same side bridge arm of the seat motor drive circuit is connected to form a short circuit to perform motor braking, or a drive signal with reverse polarity is applied to control the seat motor to perform a reverse yielding action.

Images