Vehicle load estimation method and system based on parallel hybrid architecture

Abstract

The invention provides a vehicle load estimation method and system based on a parallel hybrid architecture, and belongs to the technical field of vehicle load regulation and control. Comprising the steps of obtaining vehicle composite data in real time based on a multi-modal sensor and executing preprocessing; inputting the preprocessed vehicle composite data into a parallel hybrid architecture to predict a load motion state and a load compensation value; carrying out dynamic load separation on the predicted value of the parallel hybrid architecture by adopting an improved EEMD algorithm; fusing the separation result based on federated Kalman filtering to obtain an estimated value of the vehicle load; and the estimated value is verified based on a triple modular redundancy architecture, and then whether a vehicle load fault recovery process is triggered or not is determined. According to the method, through multi-source data fusion and parallel hybrid architecture modeling, a physical equation is combined with machine learning, and high-precision estimation under a static working condition and adaptive optimization of a dynamic working condition can be realized while the cost and hardware dependence are reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of vehicle load control, and in particular relates to a vehicle load estimation method and system based on a parallel hybrid architecture. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Virtual sensor technology has moved from theoretical research to commercial application, particularly in high-end passenger vehicles equipped with air suspension. However, the technology is still in the optimization and expansion phase. In commercial vehicle air suspension scenarios, existing virtual load estimation technology still faces some difficult-to-solve technical challenges, such as: (1) Insufficient adaptability under heavy-load conditions: Existing commercial vehicle solutions are mostly based on pressure sensors or height sensors. In this way, if the load is large (such as a general passenger car with a load exceeding 10 tons), the static error will increase sharply when passing over speed bumps or bumpy roads. The center of gravity offset estimation error will also increase significantly during dynamic cornering, which cannot meet the commercial vehicle load control and suspension stability requirements.

[0004] (2) Difficulty in decoupling multi-axle loads: The calculation of axle load distribution for commercial vehicles with three or more axles relies on complex mechanical models. The existing algorithm based on the two-degree-of-freedom suspension model is difficult to accurately decouple the dynamic loads of each axle, resulting in excessive deviation in the estimation of axle load transfer during braking.

[0005] (3) Poor robustness under adverse working conditions: Commercial vehicles often travel on unpaved roads, and the static load separation error of existing dynamic compensation algorithms (such as wavelet transform) under high-frequency vibration is relatively high, resulting in the inability to effectively suppress interference from scenes such as gravel roads and washboard roads. Summary of the Invention

[0006] To overcome the shortcomings of the above-mentioned existing technologies, the present invention provides a vehicle load estimation method and system based on a parallel hybrid architecture. By integrating multi-source data fusion and parallel hybrid architecture modeling, physical equations are combined with machine learning. This method can achieve high-precision estimation under static conditions and adaptive optimization under dynamic conditions while reducing costs and hardware dependence.

[0007] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: A first aspect of the present invention provides a vehicle load estimation method based on a parallel hybrid architecture.

[0008] A vehicle load estimation method based on a parallel hybrid architecture, comprising: Acquire vehicle composite data in real time based on multimodal sensors; performing preprocessing operations on the obtained vehicle composite data; Inputting the pre-processed vehicle composite data into a parallel hybrid architecture to predict load motion and load compensation values; wherein the parallel hybrid architecture includes a 7-degree-of-freedom vehicle dynamics model with temperature compensation and a data-driven model using a lightweight LSTM network; An improved EEMD algorithm is used to perform dynamic load separation on the predicted value of the parallel hybrid architecture, and static load and dynamic load are separated by adaptive white noise amplitude adjustment; The separation results of static load and dynamic load are fused based on the federated Kalman filter to obtain an estimated value of the vehicle load; the obtained estimated value is verified based on the triple-module redundant architecture, and the triggering of the vehicle load fault recovery process is determined based on the verification result.

[0009] Furthermore, an improved EEMD algorithm is used to perform dynamic load separation on the predicted values ​​of the parallel hybrid architecture, and the static load and dynamic load are separated by adaptive white noise amplitude adjustment, including: first, injecting adaptive amplitude white noise into the original signal; then, extracting the intrinsic mode function through Hilbert transform, and using the kurtosis index to screen out the effective IMF components to reconstruct the static load to achieve dynamic load separation.

[0010] Furthermore, a pre-processing operation is performed on the vehicle composite data, including: performing sliding mean filtering, temperature compensation, wavelet noise reduction and time alignment operations on the vehicle composite data in sequence.

[0011] Furthermore, the 7-DOF vehicle dynamics model includes: vertical, pitch, and roll motions of the sprung mass, vertical motions of four unsprung masses, and tire stiffness and damping characteristic equations; wherein, the model parameters of the 7-DOF vehicle dynamics model are updated through a real-time parameter identification algorithm.

[0012] Furthermore, the structure of the data-driven model using a lightweight LSTM network is: an input layer with 10 nodes, a single hidden layer with 8 nodes, and an output layer with 1 node, and the weights are solidified into DSRAM after INT8 quantization; wherein, the 10 nodes of the input layer include pressure gradient, height change rate, and frequency domain features.

[0013] Furthermore, the separation results of static load and dynamic load are fused based on the federated Kalman filter, including: first, local filters are used to process the pressure sensor subgroup, IMU subgroup, and altitude sensor subgroup respectively; then, the main filter is used to fuse the local results, and the sensor confidence weight factor is introduced into the covariance matrix update formula of the main filter.

[0014] Furthermore, the triple-module redundant architecture includes a main channel, a monitoring channel, and a protection channel; wherein, the main channel is used to execute the estimated value of the vehicle load, the monitoring channel is used to run the 7-degree-of-freedom vehicle dynamics model, and the protection channel is used to implement the rule checker logic. When the result deviation of the three channels exceeds the set threshold, the fault recovery process is triggered.

[0015] A second aspect of the present invention provides a vehicle load estimation system based on a parallel hybrid architecture.

[0016] A vehicle load estimation system based on a parallel hybrid architecture, comprising: The data acquisition module is configured to: acquire vehicle composite data in real time based on multimodal sensors; A preprocessing module is configured to: perform a preprocessing operation on the obtained vehicle composite data; a parallel hybrid architecture module configured to input preprocessed vehicle composite data into the parallel hybrid architecture to predict a load motion state and a load compensation value; wherein the parallel hybrid architecture includes a 7-degree-of-freedom vehicle dynamics model with a temperature compensation term and a data-driven model using a lightweight LSTM network; a dynamic load separation module configured to: perform dynamic load separation on the predicted value of the parallel hybrid architecture using an improved EEMD algorithm, and separate static load and dynamic load by adaptive white noise amplitude adjustment; The vehicle load estimation module is configured to: fuse the separation results of static load and dynamic load based on the federated Kalman filter to obtain an estimated value of the vehicle load; verify the obtained estimated value based on the triple-module redundant architecture, and determine whether to trigger the vehicle load fault recovery process based on the verification result. A third aspect of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps of the vehicle load estimation method based on a parallel hybrid architecture as described in the first aspect of the present invention.

[0017] The fourth aspect of the present invention provides an electronic device, comprising a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps of a vehicle load estimation method based on a parallel hybrid architecture as described in the first aspect of the present invention are implemented.

[0018] One or more of the above technical solutions have the following beneficial effects: (1) The present invention inputs the pre-processed vehicle composite data into a parallel hybrid architecture to predict the load motion state and load compensation value; wherein, the parallel hybrid architecture includes a 7-degree-of-freedom vehicle dynamics model with temperature compensation term and a data-driven model using a lightweight LSTM network. The 7-degree-of-freedom model introduces a real-time identification algorithm for axle load distribution parameters, which can automatically adjust the stiffness matrix of each axle according to the load, solving the mechanical imbalance problem of the traditional two-degree-of-freedom model under heavy load; in the parallel hybrid architecture, the physical model and the lightweight LSTM network are operated in parallel, and the LSTM can learn the nonlinear load distribution characteristics under heavy load to compensate for the error of the physical model under extreme load; at the same time, the multi-sensor fusion algorithm can automatically increase the weight of the pressure sensor under heavy load, and suppress the sensor noise amplification effect caused by heavy load through the federated Kalman filter. Therefore, compared with the existing technology, even when the load is large, whether facing speed bumps, bumpy roads, or sharp turns, it can maintain smaller static errors and dynamic offset errors, thereby meeting the load supervision and suspension stability requirements of commercial vehicles.

[0019] (2) The present invention calculates the vertical displacement of the sprung mass and the motion state of the unsprung mass based on a 7-DOF vehicle dynamics model. The model incorporates a multi-axis dynamic coupling term. By solving the cross-influence of each axle acceleration and suspension displacement in real time, accurate dynamic axle load distribution can be achieved. Simultaneously, the federated Kalman filter is introduced into the axle load decoupling state vector, and the load transfer amount of each axle is corrected in real time using the three-axis angular velocity data of the IMU. This can reduce the delay in axle load transfer estimation during braking. This can avoid the problem of excessive deviation in axle load transfer estimation during braking due to the difficulty in accurately decoupling the dynamic loads of each axle.

[0020] (3) The present invention uses an improved EEMD algorithm to perform dynamic load separation on the predicted values ​​of the parallel hybrid architecture. The EEMD algorithm can effectively separate high-frequency vibration noise from static loads through multiple integrated empirical mode decompositions, and then selects the IMF3-IMF4 low-frequency components to reconstruct the static load to suppress the interference of unpaved roads. At the same time, the sensor priority strategy can automatically increase the weight of the IMU under harsh working conditions and combine the vehicle body acceleration signal to compensate for load fluctuations caused by road excitation in real time. In addition, the lightweight LSTM network input contains 10-dimensional vibration features (such as kurtosis and root mean square value), which can effectively learn the load fluctuation patterns corresponding to different road textures. Compared with existing technologies, it can reduce dynamic errors in extreme scenarios such as washboard roads. Advantages of additional aspects of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0022] Figure 1 This is a flow chart of a vehicle load estimation method based on a parallel hybrid architecture in Embodiment 1 of the present invention.

[0023] Figure 2 This is a flowchart of performing preprocessing operations in embodiment 1 of the present invention.

[0024] Figure 3 This is a schematic diagram of the calibration process among the calibration platform, ECU, and hydraulic press in Example 1 of the present invention.

[0025] Figure 4 This is a fault recovery flow chart when vehicle load fault recovery is triggered in the first embodiment of the present invention. DETAILED DESCRIPTION

[0026] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0027] It should be noted that the terms used herein are for describing particular embodiments only and are not intended to be limiting of exemplary embodiments according to the present invention.

[0028] In the absence of conflict, the embodiments of the present invention and the features thereof may be combined with each other.

[0029] Example 1 This embodiment discloses a vehicle load estimation method based on a parallel hybrid architecture.

[0030] like Figure 1 As shown, a vehicle load estimation method based on a parallel hybrid architecture includes: Step S1: acquiring vehicle composite data in real time based on a multimodal sensor; Step S2: performing preprocessing operations on the obtained vehicle composite data; Step S3: Inputting the pre-processed vehicle composite data into a parallel hybrid architecture to predict the load motion state and load compensation value; wherein the parallel hybrid architecture includes a 7-DOF vehicle dynamics model with a temperature compensation term and a data-driven model using a lightweight LSTM network; Step S4: using an improved EEMD algorithm to perform dynamic load separation on the predicted value of the parallel hybrid architecture, and separating the static load and the dynamic load by adaptively adjusting the amplitude of white noise; Step S5: Based on the federated Kalman filter, the separation results of the static load and the dynamic load are fused to obtain an estimated value of the vehicle load; the estimated value is verified based on the triple-module redundancy architecture, and it is determined whether to trigger the vehicle load fault recovery process based on the verification result.

[0031] Based on the above process, the present invention combines physical equations with machine learning through multi-source data fusion and parallel hybrid architecture modeling. This reduces costs and hardware dependencies while achieving high-precision estimation under static conditions and adaptive optimization under dynamic conditions. To facilitate understanding of the technical solution of the present invention, the specific implementation methods of the technical solution of the present invention are further explained and illustrated below.

[0032] In step S1 , vehicle composite data is acquired in real time based on a multimodal sensor.

[0033] Multimodal sensors collect real-time data such as air spring pressure, suspension height, acceleration, and vehicle speed as composite vehicle data. Specifically, the multimodal sensors consist of a basic sensor network and an extended signal source. The basic sensor network includes a high-linearity pressure sensor, a contactless magnetoresistive height sensor, and a 6-axis MEMS IMU. The high-linearity pressure sensor monitors the internal pressure of the air spring in real time and calculates vehicle load. It can be directly integrated into the quintuple valve or at the air spring chamber interface. The contactless magnetoresistive height sensor accurately measures the relative height of the vehicle body and wheels. It can be installed between the suspension arm and the vehicle body through a linkage mechanism. The 6-axis MEMS IMU monitors the vehicle's three-axis acceleration (X / Y / Z) and three-axis angular velocity (pitch / yaw / roll) to dynamically identify vehicle motion and provide instantaneous feedback for suspension adjustment. It can be installed near the vehicle's center of gravity to minimize measurement errors. The extended signal source collects multiplexed CDC shock absorber current signals and analyzes CAN bus vehicle speed and steering angle data.

[0034] In step S2, a pre-processing operation is performed on the obtained vehicle composite data.

[0035] like Figure 2 As shown, preprocessing operations are performed on the vehicle composite data, including: performing sliding mean filtering, temperature compensation, wavelet noise reduction and time alignment operations on the vehicle composite data in sequence. Specifically: 1) Perform sliding mean filtering on the vehicle composite data to eliminate high-frequency noise. Specifically: First, missing values ​​are addressed by aligning multi-sensor timestamps. Data is then normalized (e.g., pressure in bar, height in mm, acceleration in g) for pre-elimination data processing. Subsequently, a sliding window design is implemented, and filtering is performed by cyclically updating the window data and calculating the mean point by point to eliminate high-frequency noise. It should be noted that any sliding mean filtering method used can be used as long as it can eliminate high-frequency noise; this embodiment does not impose any specific limitations on the implementation method.

[0036] 2) Perform temperature compensation on the vehicle composite data to correct the inherent temperature drift error of the sensor, namely: ; in, Indicates the pressure value after temperature compensation. represents the temperature coefficient; represents the original pressure, Indicates the current temperature. Indicates the reference temperature.

[0037] On this basis, wavelet denoising is performed on the composite vehicle data to preserve effective signal features, and time alignment is performed on the composite vehicle data. It should be noted that wavelet denoising and time alignment are performed solely to provide more accurate and analyzable data for the parallel hybrid architecture. Existing similar technologies can also be used for denoising and time alignment. Therefore, this embodiment does not impose specific limitations on the actual implementation methods.

[0038] In step S3, the pre-processed vehicle composite data is fed into a parallel hybrid architecture to predict the load motion state and load compensation value. The parallel hybrid architecture includes a 7-DOF vehicle dynamics model with temperature compensation and a data-driven model using a lightweight LSTM network. This can be achieved through the following methods: 1) 7-DOF vehicle dynamics model.

[0039] A 7-DOF vehicle dynamics model is used for vehicle posture processing, and a real-time parameter identification algorithm is added to improve the accuracy and precision of vehicle processing data. Specifically, the 7-DOF vehicle dynamics model includes: vertical motion of the sprung mass, pitch, and roll, vertical motion of the four unsprung masses, and tire stiffness and damping characteristic equations; the model parameters of the 7-DOF vehicle dynamics model are updated using a real-time parameter identification algorithm.

[0040] Furthermore, the 7-DOF vehicle dynamics model adopted includes: vertical motion of the vehicle body , body roll motion , vehicle body pitch motion , longitudinal movement , lateral movement , yaw motion and wheel rotation When processing vehicle posture through this model, it specifically includes: establishing the motion equations for each degree of freedom based on Newton's laws of motion and the momentum theorem, such as the vertical dynamics equation of the vehicle body: ; in, For the body quality, The vertical force of each suspension is obtained by combining the wheel speed, acceleration, steering angle and other data collected by the sensor in real time, and solving the motion equation to obtain the vehicle body posture parameters (such as roll angle, pitch angle, vertical displacement, etc.), providing a posture reference for load estimation.

[0041] A recursive least squares (RLS) algorithm is used for real-time parameter identification, and key vehicle parameters (such as suspension stiffness, damping coefficient, and tire stiffness) are updated online. The deviation between the posture parameters predicted by the model and the posture parameters measured by the sensors is used as input, and the parameter values ​​are iteratively corrected through the RLS algorithm to minimize the deviation. A parameter update cycle is set (such as 10ms) to ensure that the parameters can adapt in real time when the vehicle load changes (such as passengers getting on and off, cargo additions and subtractions) or when the operating conditions switch (such as asphalt road / gravel road).

[0042] Compared with low-DOF models (such as 3-DOF), the 7-DOF model can more comprehensively reflect the vehicle's dynamic characteristics and reduce posture calculation errors. Real-time parameter identification can eliminate parameter drift caused by vehicle aging and load changes, ensuring that the model always matches the actual vehicle state and improving the accuracy and precision of subsequent load estimation.

[0043] 2) A data-driven model using a lightweight LSTM network.

[0044] The micro-LSTM network used in this paper is a lightweight model with a single-layer LSTM architecture. Its input feature dimensions include sensor data such as wheel height and acceleration, and its projected output dimensions (directly corresponding to the four-wheel load predictions). The input data dimensions are expressed as [10, 1, 8], where 10 represents the time step (10 frames of historical data), 1 represents the batch size (batch_size=1, suitable for real-time processing), and 8 represents the number of features per frame. During forward propagation, only the results of the last time step are output, discarding the hidden states. The output shape is expressed as [1, 4], corresponding to the real-time four-wheel load estimation. This lightweight design results in only approximately 500 parameters, making it optimized for embedded deployment and providing direct end-to-end output.

[0045] The data-driven model based on the lightweight LSTM network has a 10-node input layer, a single hidden layer with 8 nodes, and a 1-node output layer. The weights are quantized to INT8 and then stored in DSRAM. The 10-node input layer contains pressure gradient, altitude change rate, and frequency domain features. Furthermore, the lightweight model's micro-LSTM network is optimized based on the traditional LSTM structure through the following: 1) Simplified network structure: reducing the number of hidden layers to only one, reducing the number of hidden layer neurons (from the conventional 64-128 to 16-32), and removing redundant gated connections; 2) Quantization compression: using 8-bit integer quantization instead of 32-bit floating-point operations to reduce computational complexity; and 3) Input feature filtering: retaining only sensor data strongly related to load (such as pressure sensor output, vertical acceleration, and vehicle height) to reduce input dimensionality. All other aspects remain unchanged from the traditional LSTM structure.

[0046] By employing a data-driven model based on a lightweight LSTM network, this method reduces computing resource requirements and can run in real time on an onboard embedded chip (such as an MCU), meeting the real-time requirements for vehicle dynamic load estimation. Notably, despite its lightweight structure, this data-driven model retains the LSTM's advantage in modeling time series data (such as dynamic load changes on continuously bumpy roads). Compared to traditional neural networks (such as CNNs), it is more suitable for processing the time-varying nature of loads. Furthermore, it can adapt to the low-power environment of vehicles, avoiding heat generation caused by excessive computing power.

[0047] In step S4, the improved EEMD algorithm is used to perform dynamic load separation on the predicted value of the parallel hybrid architecture, and the static load and the dynamic load are separated by adaptive white noise amplitude adjustment.

[0048] In order to better achieve dynamic load separation, the improved EEMD algorithm is used to perform dynamic load separation on the predicted values ​​of the parallel hybrid architecture, that is, the static load and dynamic load are separated by adaptive white noise amplitude adjustment.

[0049] The improved EEMD algorithm is used to separate the static and dynamic components from the vehicle acceleration signal. Its basic implementation process can be summarized as follows: First, inject adaptive amplitude white noise into the original signal, that is: ; in, represents the amplitude white noise added, represents the working condition adaptive coefficient, Indicates the signal standard deviation.

[0050] Subsequently, the intrinsic mode functions are extracted by Hilbert transform, and the effective IMF components are screened out using the kurtosis index to reconstruct the static load to achieve dynamic load separation.

[0051] As an optional embodiment, during implementation, the original acceleration signal can be perturbed 100 times with Gaussian noise (noise standard deviation 0.1). EMD decomposition is performed after each perturbation to obtain multiple sets of IMF components. Subsequently, the third- and fourth-order IMF components are selected and weighted in a 7:3 ratio to form the static component. Finally, the static component is subtracted from the original signal to obtain the dynamic component. Thus, the present invention improves stability through multiple noise-adding decompositions (traditional EMD produces a single decomposition), intelligently selects specific IMF combinations (3rd + 4th order) rather than simply using all IMFs, and directly separates the components through arithmetic operations, avoiding complex post-processing. In the resulting output, the static component reflects low-frequency information such as vehicle weight distribution, while the dynamic component includes high-frequency information such as road vibration.

[0052] In step S5, the static load and dynamic load separation results are fused based on the federated Kalman filter to obtain an estimated vehicle load value. The estimated value is verified based on the triple-module redundancy architecture, and a decision is made based on the verification result whether to trigger the vehicle load fault recovery process. This can be achieved through the following methods: Step S5-1: Fusing the separation results of the static load and the dynamic load based on the federated Kalman filter to obtain an estimated value of the vehicle load.

[0053] The Federated Kalman Filter (FKF) algorithm is used as the multi-sensor fusion algorithm. It takes each sensor's local state estimate x_local and covariance matrix p_local as input and outputs the fused global state estimate x_est and covariance matrix P_fused. The intermediate fusion process uses the information matrix (the inverse of the covariance matrix) for weighted fusion. The local covariance matrices are first inverted and summed, and then the sum is inverted to obtain the fused covariance matrix. Furthermore, the state estimate fusion uses information-weighted averaging, weighting the local state estimates with the information matrix of each sensor. Finally, the fused information matrix is ​​used for normalization.

[0054] The separation results of static load and dynamic load are fused based on the federated Kalman filter, including: first, using local filters to process the pressure sensor subgroup, IMU subgroup, and altitude sensor subgroup respectively; then, using the main filter to fuse the local results, and introducing the sensor confidence weight factor in the covariance matrix update formula of the main filter , specifically: 1) Local filters process each sensor subgroup.

[0055] Pressure sensor subgroup: This subgroup uses an extended Kalman filter (EKF) to process static loads. It uses tire contact pressure collected by the pressure sensor as input and combines it with the vehicle's stationary posture parameters (roll angle and pitch angle) to estimate static loads (such as the static distribution of occupants and cargo). IMU subgroup: Uses unscented Kalman filtering (UKF) to process dynamic loads. Based on the three-axis acceleration and angular velocity collected by the IMU, it calculates the dynamic load transfer during vehicle acceleration, braking, and steering. Height sensor subgroup: Kalman filtering (KF) is used to assist in dynamic load estimation, and the suspension deformation is inferred from the change in vehicle height to correct the amplitude of the dynamic load.

[0056] 2) Main filter fuses local results. The main filter receives the outputs of the three local filters (static load estimation, dynamic load estimation, and correction), constructs the fusion state equation, and introduces the sensor confidence weight factor into the covariance matrix update formula. The specific formula is: ; in, is the updated covariance matrix of the main filter, is the covariance matrix of the local filter; represents the sensor confidence weight factor, Corresponding to three subgroups respectively.

[0057] Confidence weight factor Calculation: Based on the consistency test of sensor data (such as residual analysis method) and environmental adaptability (such as IMU on bumpy roads Lower, pressure sensor on flat road Increase), dynamic adjustment and .

[0058] Based on the above design, the separation of local filtering from main filtering can reduce the impact of individual sensor failures on the overall results and improve the system's fault tolerance. The confidence weighting factor makes the fusion process more "intelligent," namely, it prioritizes trust in data from highly reliable sensors, reduces errors caused by noise or interference (such as the reduced weight of a pressure sensor due to temperature), and improves fusion accuracy. At the same time, it achieves effective separation and fusion of static and dynamic loads, solving the problem of a single sensor's difficulty in accurately capturing both static (such as load during parking) and dynamic (such as load transfer during rapid acceleration) characteristics.

[0059] Step S5-2: verify the estimated value based on the triple-module redundancy architecture, and determine whether to trigger the vehicle load fault recovery process based on the verification result.

[0060] The principles of the triple-module redundant architecture are as follows: 1) Triple-module redundant design: The primary execution channel performs the primary control function; the verification channel independently runs the same algorithm; and the arbitration channel determines the final output. 2) Safety monitoring mechanism: Failures are detected by comparing the outputs of the primary and backup channels, using a "2 out of 2" voting logic.

[0061] The key components of the triple-module redundant architecture are: 1) a deviation detection unit that calculates the absolute difference between the primary and backup channels in real time; 2) a safety monitoring mechanism that uses threshold judgment to determine whether to initiate a fault response, activates the safety mode, and simultaneously records the fault code and actual deviation value; 3) an implicit arbitration logic that trusts the primary channel output by default and switches to a safe state in the event of a fault.

[0062] Furthermore, the triple-module redundant architecture includes a main channel, a monitoring channel, and a protection channel; the main channel is used to execute vehicle load estimation; the monitoring channel adopts a 2-degree-of-freedom simplified suspension model to run a 7-degree-of-freedom vehicle dynamics model; the protection channel is used to preset the load change rate threshold and implement the rule checker logic. When the result deviation of the three channels exceeds the set threshold (load change rate threshold), the fault recovery process is triggered.

[0063] The fault recovery process includes: switching to the backup model when the sensor fails, and starting the memory double buffering strategy when the computing resources are abnormal. Figure 4 As shown, in the fault monitoring stage, the output value of the main algorithm is first temporarily stored in the buffer, and the monitoring module simultaneously requests the verification value from the simplified model, and compares the main output with the simplified model verification value; in the fault judgment stage, when a deviation is found, the fault type is analyzed (value out of bounds, logical anomaly, etc.), and the specific fault mode is identified through type matching; in the recovery execution stage, the safety module triggers three-level recovery, including: 1) soft reset: clear the abnormal state of the algorithm; 2) data rollback: restore to the last valid state; 3) mode switching, down to safe operation mode. In the recovery confirmation stage, continuous monitoring for 3-5 cycles is carried out to confirm that normal recovery has been achieved, and a reply completion signal is sent through the CAN bus. The entire process can be completed within 10ms, and an instantaneous fault recovery rate of more than 99% is achieved through designs such as output buffering, dual verification mechanism, and hierarchical recovery. Specifically: 1) Implementation of a triple-module redundant architecture. The triple-module redundant architecture calculates vehicle load estimates through three independent paths, including: Path 1: Analytical calculation based on the 7-DOF model and physical formulas; Path 2: Data-driven prediction based on micro LSTM networks; Path 3: Fusion result based on the federated Kalman filter (i.e., the output of step S3 above).

[0064] Verification process: Set the deviation threshold to 5% of the load estimate and calculate the pairwise deviation of the three path results: 、 、 ; “Pairwise deviation” refers to the deviation of the estimated results between path 1 and path 2, the deviation of the estimated results between path 1 and path 3, and the deviation of the estimated results between path 2 and path 3; where, 、 、 The weight factors corresponding to the three groups of pairwise deviations are used to quantify the influence of different deviations on the overall verification results. The weight values ​​are based on the inherent accuracy, dynamic response characteristics and reliability of each sensor subgroup under the current working conditions: for example, if the pressure sensor has higher accuracy under low-speed static conditions, the weight corresponding to the deviation it participates in (such as path 1 and path 2) is It can be set larger to highlight the impact of the deviation on the verification results; under high-speed dynamic conditions, the reliability of the IMU subgroup is improved, and the weight of the corresponding deviation (such as ) Adaptable adjustment. In this embodiment, if all deviations are less than or equal to the deviation threshold, the estimated value is determined to be valid and the fusion result is output; if any deviation is greater than the deviation threshold, the vehicle load fault recovery process is triggered.

[0065] 2) The fault recovery process includes: Level 1 recovery: Enable backup sensors (such as redundant pressure sensors) to recollect data and repeat the three-path calculation. Secondary recovery: If the primary recovery fails, switch to a simplified estimation model (such as dual sensor fusion based only on pressure sensor and IMU) to ensure basic load estimation function; Warning prompt: Send a fault code to the vehicle controller and prompt the driver through the instrument panel that "the load estimation system is degraded" and recommend that the driver go for inspection.

[0066] Based on the above design, the triple-module redundant architecture can significantly reduce the risk of miscalculation caused by single algorithm or sensor failure through multi-path verification, thereby improving system reliability. At the same time, the hierarchical fault recovery process ensures that when some components fail, the system can still maintain basic functions, avoiding the impact of complete failure on vehicle safety (such as overload detection, suspension control and other functions that rely on load data).

[0067] The vehicle load estimation method based on a parallel hybrid architecture provided by the present invention also includes an online self-calibration mechanism, namely: a) The calibration process is triggered every time the accumulated driving distance is 50,000 kilometers, such as Figure 3As shown in the figure, during the initialization phase, the ECU sends a zero-position calibration command and confirms that the sensor is fault-free and the communication is normal; then, the hydraulic press applies a 20% benchmark load (to establish an initial reference point); during the data acquisition phase, the sensor starts uploading real-time data, and the hydraulic press gradually loads according to 5 gradients, and records data after each load level remains stable; during the parameter calibration phase, the system automatically identifies feature points, calculates calibration coefficients, and writes them into the ECU non-volatile memory; during the verification phase, the hydraulic press unloads to zero position, and the ECU verifies the validity of the calibration parameters.

[0068] b) On a level road with zero speed, compare the current load estimate with the baseline value; c) Update the spring stiffness parameter matrix using the least squares method .

[0069] Example 2 This embodiment discloses a vehicle load estimation system based on a parallel hybrid architecture.

[0070] A vehicle load estimation system based on a parallel hybrid architecture, comprising: The data acquisition module is configured to: acquire vehicle composite data in real time based on multimodal sensors; A preprocessing module is configured to: perform a preprocessing operation on the obtained vehicle composite data; a parallel hybrid architecture module configured to input preprocessed vehicle composite data into the parallel hybrid architecture to predict a load motion state and a load compensation value; wherein the parallel hybrid architecture includes a 7-degree-of-freedom vehicle dynamics model with a temperature compensation term and a data-driven model using a lightweight LSTM network; a dynamic load separation module configured to: perform dynamic load separation on the predicted value of the parallel hybrid architecture using an improved EEMD algorithm, and separate static load and dynamic load by adaptive white noise amplitude adjustment; The vehicle load estimation module is configured to: fuse the separation results of static load and dynamic load based on the federated Kalman filter to obtain an estimated value of the vehicle load; verify the obtained estimated value based on the triple-module redundant architecture, and determine whether to trigger the vehicle load fault recovery process based on the verification result. Example 3 The purpose of this embodiment is to provide a computer-readable storage medium.

[0071] A computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the steps of a vehicle load estimation method based on a parallel hybrid architecture as described in the first embodiment of the present disclosure.

[0072] Example 4 The purpose of this embodiment is to provide an electronic device.

[0073] An electronic device includes a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps of a vehicle load estimation method based on a parallel hybrid architecture as described in the first embodiment of the present disclosure are implemented.

[0074] The steps involved in the apparatuses of Examples 2, 3, and 4 above correspond to those of Method Example 1. For detailed implementations, please refer to the relevant description of Example 1. The term "computer-readable storage medium" should be understood to mean a single medium or multiple media containing one or more instruction sets; it should also be understood to include any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and causing the processor to perform any method of the present invention.

[0075] Those skilled in the art will appreciate that the modules or steps of the present invention described above can be implemented using a general-purpose computer device. Alternatively, they can be implemented using program code executable by a computing device, which can then be stored in a storage device and executed by the computing device. Alternatively, they can be fabricated into separate integrated circuit modules, or multiple modules or steps can be fabricated into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0076] Although the above describes the specific embodiments of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without any creative work are still within the scope of protection of the present invention.

Claims

1. A vehicle load estimation method based on a parallel hybrid architecture, characterized in that: include: Acquire vehicle composite data in real time based on multimodal sensors; performing preprocessing operations on the obtained vehicle composite data; Inputting the pre-processed vehicle composite data into a parallel hybrid architecture to predict load motion and load compensation values; wherein the parallel hybrid architecture includes a 7-degree-of-freedom vehicle dynamics model with temperature compensation and a data-driven model using a lightweight LSTM network; An improved EEMD algorithm is used to perform dynamic load separation on the predicted value of the parallel hybrid architecture, and static load and dynamic load are separated by adaptive white noise amplitude adjustment; The separation results of static load and dynamic load are fused based on the federated Kalman filter to obtain an estimated value of the vehicle load; the obtained estimated value is verified based on the triple-module redundant architecture, and the triggering of the vehicle load fault recovery process is determined based on the verification result.

2. The vehicle load estimation method based on a parallel hybrid architecture according to claim 1, characterized in that: An improved EEMD algorithm is used to perform dynamic load separation on the predicted values ​​of the parallel hybrid architecture, and the static load and dynamic load are separated by adaptive white noise amplitude adjustment. The method includes: first, injecting adaptive amplitude white noise into the original signal; then, extracting the intrinsic mode function through Hilbert transform, and using the kurtosis index to screen out the effective IMF components to reconstruct the static load to achieve dynamic load separation.

3. The vehicle load estimation method based on a parallel hybrid architecture according to claim 1, characterized in that: Preprocessing operations are performed on the vehicle composite data, including: performing sliding mean filtering, temperature compensation, wavelet noise reduction and time alignment operations on the vehicle composite data in sequence.

4. The vehicle load estimation method based on a parallel hybrid architecture according to claim 1, characterized in that: The 7-DOF vehicle dynamics model includes: vertical, pitch, and roll motions of the sprung mass, vertical motions of four unsprung masses, and tire stiffness and damping characteristic equations; wherein the model parameters of the 7-DOF vehicle dynamics model are updated through a real-time parameter identification algorithm.

5. The vehicle load estimation method based on a parallel hybrid architecture according to claim 1, characterized in that: The structure of the data-driven model using a lightweight LSTM network is: an input layer with 10 nodes, a single hidden layer with 8 nodes, and an output layer with 1 node. The weights are solidified into DSRAM after INT8 quantization. Among them, the 10 nodes of the input layer contain pressure gradient, altitude change rate, and frequency domain features.

6. The vehicle load estimation method based on a parallel hybrid architecture according to claim 1, characterized in that: The separation results of static load and dynamic load are fused based on the federated Kalman filter. First, local filters are used to process the pressure sensor subgroup, IMU subgroup, and altitude sensor subgroup respectively. Then, the main filter is used to fuse the local results, and the sensor confidence weight factor is introduced into the covariance matrix update formula of the main filter.

7. The vehicle load estimation method based on a parallel hybrid architecture according to claim 1, characterized in that: The triple-module redundant architecture includes a main channel, a monitoring channel, and a protection channel. The main channel is used to estimate vehicle loads, the monitoring channel is used to run a 7-DOF vehicle dynamics model, and the protection channel is used to implement rule checker logic. When the deviation of the results of the three channels exceeds a set threshold, a fault recovery process is triggered.

8. A vehicle load estimation system based on a parallel hybrid architecture, characterized in that: include: The data acquisition module is configured to: acquire vehicle composite data in real time based on multimodal sensors; A preprocessing module is configured to: perform a preprocessing operation on the obtained vehicle composite data; a parallel hybrid architecture module configured to input preprocessed vehicle composite data into the parallel hybrid architecture to predict a load motion state and a load compensation value; wherein the parallel hybrid architecture includes a 7-degree-of-freedom vehicle dynamics model with a temperature compensation term and a data-driven model using a lightweight LSTM network; a dynamic load separation module configured to: perform dynamic load separation on the predicted value of the parallel hybrid architecture using an improved EEMD algorithm, and separate static load and dynamic load by adaptive white noise amplitude adjustment; The vehicle load estimation module is configured to: fuse the separation results of static load and dynamic load based on the federated Kalman filter to obtain an estimated value of the vehicle load; verify the obtained estimated value based on the triple-module redundant architecture, and determine whether to trigger the vehicle load fault recovery process based on the verification result.

9. A computer-readable storage medium having a program stored thereon, characterized in that: When the program is executed by a processor, the steps of the vehicle load estimation method based on a parallel hybrid architecture as described in any one of claims 1 to 7 are implemented.

10. An electronic device comprising a memory, a processor, and a program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the vehicle load estimation method based on a parallel hybrid architecture as described in any one of claims 1 to 7 are implemented.

Images