Electromechanical Braking Control Method and Device Based on Wheel End Vibration Sensing

By using an electromechanical braking control method based on wheel-end vibration sensing, frequency domain features are directly extracted from wheel vibration signals to identify driving conditions and road surface types, and braking control parameters are adaptively adjusted. This solves the problem of excessive time delay in the vehicle-side indirect sensing control architecture and improves braking stability and safety.

CN123061331APending Publication Date: 2026-11-10ANHUI JIANGHUAI AUTOMOBILE GRP CORP LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611190199.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-06
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

In the existing vehicle-side indirect perception and control architecture, the operating condition recognition latency is too high, which affects emergency braking distance and driving safety.

Method used

The electromechanical braking control method based on wheel-end vibration sensing collects the original vibration signal of the wheel, performs filtering preprocessing and frequency domain analysis, extracts frequency domain feature data, uses a classification and recognition model to identify driving conditions and road surface types, and adaptively adjusts braking control parameters.

Benefits of technology

It shortens the response time of working condition recognition and braking control, adapts to different road conditions, optimizes braking performance, and significantly improves vehicle braking stability and autonomous driving safety.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123061331A_ABST
    Figure CN123061331A_ABST
Patent Text Reader

Abstract

This application discloses an electromechanical braking control method and device based on wheel-end vibration sensing. By directly relying on wheel vibration signals to directly perceive and identify road surface and driving conditions, and combining frequency domain characteristics to accurately characterize road excitation characteristics, this application overcomes the shortcomings of traditional indirect vehicle body sensing schemes, such as high latency and delayed condition feedback. Furthermore, it dynamically matches the optimal braking control parameters based on real-time road conditions. Therefore, it can solve the technical problems of existing brake-by-wire systems that rely on indirect vehicle body sensing to identify driving conditions, resulting in large response delays, untimely braking control, and excessively long emergency braking distances. This application achieves the technical effects of shortening the response delay of driving condition identification and braking control, accurately optimizing braking performance to adapt to different road conditions, effectively shortening braking distance, and significantly improving vehicle braking stability and autonomous driving safety.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This disclosure relates to the field of vehicle technology, and in particular to an electromechanical braking control method and device based on wheel-end vibration sensing. Background Technology

[0002] As the core safety actuator of the steer-by-wire chassis, the brake-by-wire system undertakes key functions such as service braking, emergency avoidance, and vehicle stability adjustment in the context of rapid iteration of new energy vehicles and advanced autonomous driving technologies. In related technologies, electromechanical braking uses electrical signals to drive the brake calipers and pistons to complete the braking operation, and relies on the coordinated operation of the vehicle's inertial measurement unit, wheel speed sensors, and suspension displacement sensors to construct an indirect perception and control architecture at the vehicle end.

[0003] In the existing indirect perception and control architecture at the vehicle end, the direct use of the vehicle inertial measurement unit, wheel speed sensor and suspension displacement sensor as perception carrier may result in excessively high delay in condition recognition, thereby affecting emergency braking distance and driving safety. Summary of the Invention

[0004] This disclosure provides an electromechanical braking control method and device based on wheel end vibration sensing.

[0005] According to a first aspect of this disclosure, an electromechanical braking control method based on wheel-end vibration sensing is provided, comprising: Based on a preset sensing unit, the original vibration signal of the wheel during the driving process is collected; The original vibration signal of the wheel is filtered, preprocessed, and analyzed in the frequency domain to extract frequency domain feature data; Based on a preset classification and recognition model, the current driving conditions and road surface type are identified according to the frequency domain feature data; Based on the identified driving conditions and road surface type, the braking control parameters of the electromechanical braking system are adaptively adjusted.

[0006] Optionally, the acquisition of raw wheel vibration signals during driving based on a preset sensing unit includes: The original vibration data of the wheel in three directions is continuously collected based on a preset high-frequency sampling rate; The original vibration data is uploaded directly, preserving the high-frequency vibration characteristics of the road surface excitation.

[0007] Optionally, the filtering preprocessing and frequency domain analysis of the original wheel vibration signal to extract frequency domain feature data includes: The original vibration signal is filtered using a bandpass filter with a preset frequency range to remove power frequency interference and device noise, while retaining the frequency signals corresponding to road excitation and braking conditions.

[0008] Optionally, the step of filtering and preprocessing the original vibration signal of the wheel and performing frequency domain analysis to extract frequency domain feature data further includes: The filtered time-domain vibration signal is subjected to a piecewise fast Fourier transform to convert the time-domain signal into a frequency-domain signal, and the frequency band energy ratio, peak frequency, amplitude range and spectral entropy are extracted as the frequency-domain feature data.

[0009] Optionally, the identification of the current driving condition and road surface type based on the frequency domain feature data according to the preset classification and recognition model includes: Based on a lightweight random forest classification model, the frequency domain feature data is matched with a preset operating condition template to identify at least one of the following driving conditions: slope-to-flat braking, high and low surface road switching, bumpy road driving, and speed bump crossing.

[0010] Optionally, the step of identifying the current driving condition and road surface type based on the frequency domain feature data according to the preset classification and recognition model further includes: The dominant frequency and maximum amplitude characteristics of the vertical vibration signal at the wheel end are extracted and compared with the built-in road surface feature database to identify at least one road surface type among asphalt road surface, gravel road surface, grass road surface and snow road surface.

[0011] Optionally, the adaptive adjustment of the braking control parameters of the electromechanical braking system based on the identified driving conditions and road surface type includes: Based on the identified driving conditions, the vehicle speed observation parameters and braking force distribution coefficient are dynamically corrected; based on the identified road surface type, the intervention thresholds and adjustment steps of the anti-lock braking system, traction control system, and electronic stability program are switched in stages.

[0012] Optionally, the method further includes: The system monitors the working status of each vibration sensing unit in real time. When any vibration sensing unit is detected to be abnormal, it automatically retrieves the vibration sensing unit data of the wheel on the same side for compensation to maintain uninterrupted operation of the system.

[0013] According to a second aspect of this disclosure, an electromechanical braking control device based on wheel-end vibration sensing is provided, comprising: The acquisition unit is used to acquire the original vibration signals of the wheels during the driving process based on the preset sensing unit; The extraction unit is used to perform filtering preprocessing and frequency domain analysis on the original vibration signal of the wheel, and extract frequency domain feature data; The identification unit is used to identify the current driving conditions and road surface type based on the frequency domain feature data according to a preset classification and identification model. The adjustment unit is used to adaptively adjust the braking control parameters of the electromechanical braking system based on the identified driving conditions and road surface type.

[0014] Optionally, the acquisition unit is further configured to: The original vibration data of the wheel in three directions is continuously collected based on a preset high-frequency sampling rate; The original vibration data is uploaded directly, preserving the high-frequency vibration characteristics of the road surface excitation.

[0015] Optionally, the extraction unit is further configured to: The original vibration signal is filtered using a bandpass filter with a preset frequency range to remove power frequency interference and device noise, while retaining the frequency signals corresponding to road excitation and braking conditions.

[0016] Optionally, the extraction unit is further configured to: The filtered time-domain vibration signal is subjected to a piecewise fast Fourier transform to convert the time-domain signal into a frequency-domain signal, and the frequency band energy ratio, peak frequency, amplitude range and spectral entropy are extracted as the frequency-domain feature data.

[0017] Optionally, the identification unit is further configured to: Based on a lightweight random forest classification model, the frequency domain feature data is matched with a preset operating condition template to identify at least one of the following driving conditions: slope-to-flat braking, high and low surface road switching, bumpy road driving, and speed bump crossing.

[0018] Optionally, the identification unit is further configured to: The dominant frequency and maximum amplitude characteristics of the vertical vibration signal at the wheel end are extracted and compared with the built-in road surface feature database to identify at least one road surface type among asphalt road surface, gravel road surface, grass road surface and snow road surface.

[0019] Optionally, the adjustment unit is further configured to: Based on the identified driving conditions, the vehicle speed observation parameters and braking force distribution coefficient are dynamically corrected; based on the identified road surface type, the intervention thresholds and adjustment steps of the anti-lock braking system, traction control system, and electronic stability program are switched in stages.

[0020] Optionally, the device further includes: The compensation unit is used to monitor the working status of each vibration sensing unit in real time. When any vibration sensing unit is detected to be abnormal, it automatically retrieves the vibration sensing unit data of the wheel on the same side for compensation to maintain the uninterrupted operation of the system.

[0021] According to a third aspect of this disclosure, an electronic device is provided, comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method described in the first aspect above.

[0022] According to a fourth aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are configured to cause the computer to perform the method described in the first aspect above.

[0023] According to a fifth aspect of this disclosure, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the method described in the first aspect above.

[0024] The electromechanical braking control method and device based on wheel-end vibration sensing disclosed herein, by directly relying on wheel vibration signals to directly perceive and identify road surface and driving conditions, and by accurately characterizing road excitation characteristics in combination with frequency domain features, overcomes the shortcomings of traditional indirect vehicle body sensing schemes, such as high latency and delayed condition feedback. Furthermore, it dynamically matches the optimal braking control parameters based on real-time road conditions. Therefore, it can solve the technical problems of existing brake-by-wire systems that rely on indirect vehicle body sensing to identify driving conditions, resulting in large response delays, untimely braking control, and excessively long emergency braking distances. It achieves the technical effects of shortening the response delay of driving condition identification and braking control, accurately optimizing braking performance to adapt to different road conditions, effectively shortening braking distance, and significantly improving vehicle braking stability and autonomous driving safety.

[0025] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent from the following description. Attached Figure Description

[0026] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein: Figure 1 A schematic flowchart of an electromechanical braking control method based on wheel end vibration sensing provided in an embodiment of this disclosure; Figure 2 A schematic diagram of an electromechanical braking control device based on wheel end vibration sensing provided in an embodiment of this disclosure; Figure 3 A schematic diagram of another electromechanical braking control device based on wheel end vibration sensing provided in an embodiment of this disclosure; Figure 4 A schematic block diagram of an example electronic device provided for embodiments of this disclosure. Detailed Implementation

[0027] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0028] The following describes an embodiment of the electromechanical braking control method and apparatus based on wheel end vibration sensing, with reference to the accompanying drawings.

[0029] Figure 1 This is a schematic flowchart of an electromechanical braking control method based on wheel end vibration sensing provided in an embodiment of the present disclosure.

[0030] like Figure 1 As shown, the method includes the following steps: Step 101: Based on the preset sensing unit, collect the original vibration signal of the wheel during the driving process; In this step, the pre-deployed sensing unit is positioned in the wheel hub area, specifically in a rigid region of the hub unaffected by driving deformation. This ensures that the acquired signals directly reflect the original vibration characteristics from road surface excitation, rather than smoothed signals filtered by the suspension's elastic structure. The sensing unit operates continuously during vehicle movement, acquiring real-time vibration data of the wheel in three orthogonal directions to obtain complete original wheel vibration signals. The original vibration signal referred to here is time-domain vibration data without noise reduction filtering or feature simplification. Its frequency band coverage is sufficient to retain the high-frequency vibration components induced by road surface excitation, thus providing sufficient and undistorted input information for subsequent condition identification and road surface recognition.

[0031] In one implementation, the sensing unit can employ a microelectromechanical system (MEMS) triaxial accelerometer to simultaneously acquire vibration data of the wheel in the radial, tangential, and axial directions at a sampling frequency of no less than tens of kilohertz. The measurement range is sufficient to accommodate vibration intensities under various conventional and complex driving conditions. A tight connection is used between the sensor and the wheel hub to ensure no relative movement between them, preventing loose contact surfaces from introducing additional noise. The acquired raw vibration signals are then transmitted to the downstream signal processing unit as the data basis for frequency domain feature extraction and operating condition identification.

[0032] Step 102: Perform filtering preprocessing and frequency domain analysis on the original vibration signal of the wheel to extract frequency domain feature data; The raw vibration signal acquired by the sensing unit is first transmitted to the signal processing unit for filtering preprocessing. Since the raw vibration signal from the wheel hub area inevitably contains power frequency interference introduced from motor operation, mechanical meshing, and inherent device noise, using it directly for feature analysis without processing would severely affect the accuracy of subsequent identification. Therefore, the preprocessing stage employs bandpass filtering, setting a reasonable frequency passband range to remove low-frequency drift components below the effective frequency band of road surface excitation and high-frequency noise exceeding the target analysis range, retaining only the effective frequency components directly related to road surface excitation and driving conditions. The filtered vibration signal then enters the frequency domain analysis stage, where a time-frequency transformation algorithm converts the time-domain signal into a frequency domain representation, thereby explicitly presenting the road texture features and driving condition information implicit in the signal in spectral form.

[0033] After frequency domain analysis, the system extracts multi-dimensional frequency domain feature data that effectively characterizes the current road surface condition and driving conditions. This feature data constitutes the core input for subsequent classification and discrimination by the recognition model. The extracted frequency domain features cover multiple dimensions, such as the peak frequency corresponding to the energy concentration region in the spectrum, the energy proportion distribution of each feature frequency band, the range statistics of vibration amplitude, and the overall complexity metric of the spectrum. Different road surface types exert significant morphological differences on the wheels in the spectrum. For example, there is a clear difference in the high-frequency energy distribution between rough and smooth roads. Different driving conditions also produce distinctive feature patterns in the frequency domain, such as the specific shift in the peak position of the spectrum when transitioning between a slope and a flat road. The combined use of multi-dimensional features enables the system to effectively distinguish road surface types and driving conditions under complex combinations, improving the reliability and anti-interference ability of the recognition results.

[0034] As one implementation method, the time-frequency transformation can be performed using the Fast Fourier Transform algorithm. The signal processing unit achieves synchronous processing of four sensor signals through a hardware-level parallel computing architecture. The filtering and transformation processing time of a single signal can be controlled within milliseconds, fully meeting the signal processing speed requirements of real-time braking control. The extracted frequency domain feature data is packaged and transmitted to the downstream decision layer via the vehicle's high-speed bus for synchronous use by the scene recognition module and the road type recognition module.

[0035] Step 103: Based on the preset classification and recognition model, identify the current driving conditions and road surface type according to the frequency domain feature data; The classification and recognition model is pre-trained based on a large amount of real-vehicle calibration data, establishing a mapping relationship between frequency domain feature vectors and various driving conditions and road surface types. After receiving frequency domain feature data, the model matches and compares it with the internally learned feature patterns, outputting the judgment result of the current driving condition and the identification result of the road surface type. Both are completed simultaneously, forming a two-dimensional perception of the vehicle's driving environment. In terms of driving condition recognition, the model classifies the current driving condition into specific scenario types such as transition braking between slopes and flat roads, abrupt switching from high-adhesion road surfaces to low-adhesion road surfaces, continuous driving on bumpy roads, or impact from speed bumps, based on the peak frequency shift pattern, the changing trend of energy distribution in the feature frequency bands, and the dynamic fluctuation characteristics of amplitude range in the frequency domain features.

[0036] Various operating conditions exhibit distinguishable cluster distributions in the frequency domain feature space, and the model classifies them based on decision boundaries. In road surface type identification, the process extracts key features such as the dominant frequency and maximum amplitude of the wheel-end vertical vibration signal, comparing and matching them with a built-in standardized road surface feature database to determine whether the current road surface is asphalt, gravel, grass, or snow-covered. Different road surfaces exhibit drastically different frequency components and energy intensities in the wheel's vertical excitation due to differences in texture and material stiffness; this physical law provides a reliable feature basis for road surface identification. The identification results are output in the form of structured data, including the current operating condition category identifier, road surface type identifier, and the confidence level information of the model's judgment. This data is then transmitted to the braking control strategy adaptation stage as the basis for braking force distribution and chassis system parameter adjustment.

[0037] The latency of the entire recognition process, from feature input to result output, is controlled to the tens of milliseconds, matching the rapid response characteristics of electromechanical braking actuators and ensuring that the braking strategy can be adjusted promptly when dynamic changes occur in operating conditions or road surfaces. As one implementation method, the classification and recognition model can be constructed using a lightweight random forest algorithm. This algorithm generates the final classification result through voting from multiple decision trees, maintaining high recognition accuracy and inference speed even under the computational constraints of the embedded controller. Both the recognition model and the road feature database are integrated within the vehicle domain controller and are permanently installed after calibration on a real vehicle.

[0038] Step 104: Based on the identified driving conditions and road surface type, adaptively adjust the braking control parameters of the electromechanical braking system.

[0039] The driving condition determination and road surface type identification results output by the recognition module are transmitted to the central control unit of the electromechanical braking system as the basis for adaptive adjustment of braking control parameters. The central control unit has pre-set sets of braking control parameters for different combinations of driving conditions and road surface types. Each parameter set defines braking force distribution rules and control logic adapted to a specific driving environment. Based on the received recognition results, the control unit retrieves the corresponding control parameters from the parameter set and dynamically corrects and switches the output behavior of the braking system, ensuring that the applied braking force always matches the actual needs of the current driving environment.

[0040] In terms of operating conditions, the adjustment of braking control parameters mainly involves the correction of the vehicle speed observation model and the dynamic adjustment of the braking force distribution coefficient. When the identification result indicates that the vehicle is in a braking scenario such as transitioning from a slope to a flat road, the control unit compensates and corrects the vehicle speed estimation model in advance to eliminate the vehicle speed observation deviation caused by changes in road slope, and adjusts the driving force output of the brake actuator accordingly to suppress sudden changes or precipitous drops in braking force at the moment of operating condition transition, maintaining the smoothness of braking deceleration. When the vehicle moves from a high-friction surface to a low-friction surface, the control unit reallocates the braking force ratio of each wheel according to the change in friction conditions to prevent wheel lock-up caused by the braking force exceeding the current road surface friction limit.

[0041] At the road surface level, the adjustment of braking control parameters focuses on the graded switching of intervention thresholds and adjustment steps of subsystems such as the chassis anti-lock braking system, traction control system, and electronic stability system. For roads with good adhesion, these subsystems can intervene at a higher slip ratio threshold, giving the braking system more room to apply braking force. For roads with poor adhesion or irregular textures, such as snow-covered roads or gravel-bumpy roads, the intervention threshold is lowered and the adjustment step is reduced, allowing the subsystems to intervene earlier and adjust more precisely, thereby reducing the interference of irregular road surface excitation on braking force stability. The parameter adjustments in these two dimensions are carried out in a coordinated manner. The control unit merges the operating condition adjustment command and the road surface adjustment command to generate the final braking execution command, which is transmitted to the brake actuators of each wheel via a high-speed bus to complete the actual output of braking force.

[0042] To ensure the system maintains basic braking control capabilities even when sensors malfunction, the control unit continuously monitors the status of all sensing signals. When an anomaly is detected in the sensing unit of a particular wheel, the system automatically activates a redundancy compensation mechanism. This mechanism reconstructs the vibration characteristics of that wheel by weighting the sensing data from adjacent wheels on the same side. Based on this reconstruction, the system continues to identify the operating conditions and road surface, ensuring the continuity of the braking control strategy is not disrupted by localized sensing interruptions. As one implementation method, the redundancy compensation mechanism can be further extended to a global reconstruction mode that integrates sensing data from all four wheels of the vehicle. By weighting and calculating multiple data streams, a more reliable compensation result is obtained, meeting the redundancy safety requirements of advanced autonomous driving systems.

[0043] In some embodiments, the acquisition of raw wheel vibration signals during driving based on a preset sensing unit includes: The original vibration data of the wheel in three directions is continuously collected based on a preset high-frequency sampling rate; The original vibration data is uploaded directly, preserving the high-frequency vibration characteristics of the road surface excitation.

[0044] The sensing unit continuously collects wheel vibration data at a fixed sampling rate far exceeding the requirements of conventional vibration monitoring. This sampling rate is set to 48,000 samples per second for each direction of vibration. This sampling frequency is sufficient to cover all effective vibration components generated at the wheel end by road surface excitation, ranging from 1 Hz to 1 kHz, including high-frequency excitation information lost by traditional vehicle-side sensors due to suspension filtering and low-pass noise reduction. During each sampling process, the sensing unit simultaneously acquires the instantaneous acceleration values ​​of the wheel along the radial, tangential, and axial directions, forming a complete three-dimensional vibration data frame. The three-axis range is set to ±16 grams, covering the vibration intensity range under various conditions, from constant speed driving on a flat road to emergency braking on a bumpy road. The sensing unit continuously outputs the above three-dimensional data frames at a fixed beat of 48 kHz. The sampling process does not rely on external trigger signals and begins from vehicle startup until the end of driving, ensuring that road surface excitation information is completely captured at any given time.

[0045] In the signal output stage, the sensing unit directly transmits the raw three-dimensional vibration data acquired in each sampling period to the downstream signal processing unit via the shortest path, without any form of low-pass filtering, mean smoothing, or downsampling processing. The engineering significance of this direct transmission mechanism is that the high-frequency transient vibrations induced at the wheel end by road surface textures such as asphalt micro-roughness, impact of gravel edges, and snow layer compaction often have concentrated energy frequencies above 100 Hz. In traditional vehicle-side sensing architectures, the suspension elastic elements have a significant amplitude attenuation effect on these high-frequency components. In addition, the multi-level filtering algorithm at the controller end will treat this frequency band signal as noise and remove it, ultimately resulting in the high-frequency road surface features being significantly smoothed out before reaching the control algorithm.

[0046] In this scheme, the sensing unit is fixed to the rigid region of the wheel hub. The vibration signal is transmitted directly from the road surface through the tire and wheel hub to the sensor sensing element. The signal path does not contain any elastic damping element, and the sensor itself does not perform any signal conditioning operation. It only completes the original conversion of physical quantity into electrical signal and then outputs it. This ensures that the road excitation characteristics of the entire frequency band from one hertz to one kilohertz are completely preserved in the signal link, providing a distortion-free input basis for subsequent frequency domain analysis.

[0047] The sensing unit is bolted to the wheel hub, and the installation location is chosen in the deformation-free base area inside the wheel hub. This area does not undergo structural deformation during wheel rotation and braking, avoiding parasitic modulation interference caused by minor deformation of the mounting surface on the sensor output signal. The sensor's sensing axis is precisely aligned with the wheel's radial, tangential, and axial directions during installation to ensure clear physical meaning of the three-axis components and reduce the computational overhead of coordinate transformation in subsequent signal processing. The acquired raw vibration data is output as a digital signal via the sensor's built-in analog-to-digital converter and transmitted to the signal processing unit via a point-to-point high-speed data link. The entire link transmission delay is controlled to the sub-millisecond level, without introducing observable signal lag.

[0048] In some embodiments, the filtering preprocessing and frequency domain analysis of the original wheel vibration signal to extract frequency domain feature data includes: The original vibration signal is filtered using a bandpass filter with a preset frequency range to remove power frequency interference and device noise, while retaining the frequency signals corresponding to road excitation and braking conditions.

[0049] The raw vibration signal is output from the sensing unit and transmitted to the signal processing unit, where it first enters the bandpass filter processing stage. The passband frequency range of this bandpass filter is preset to 1 Hz to 1 kHz. The selection of the lower and upper cutoff frequencies is based on the fact that the effective vibration energy at the wheel end caused by road excitation and braking conditions is concentrated within this frequency band, while the signal components beyond these two ends correspond to useless low-frequency drift and high-frequency electromagnetic noise, respectively. At the low-frequency end, signals below 1 Hz mainly originate from sensor zero-point temperature drift, slow changes in vehicle body posture, and low-frequency flicker noise from electronic devices. These components are unrelated to road excitation information. If retained, they would generate false low-frequency energy peaks in subsequent frequency domain transformations, interfering with the accuracy of feature extraction. Therefore, they are blocked by a high-pass filter.

[0050] At the high-frequency end, signals above 1 kHz are mainly composed of the inherent noise of the internal electronic components of the sensor, electromagnetic radiation interference from the wheel end, and parasitic high-frequency vibrations caused by microscopic defects in the bearing balls. The energy of these components is much lower than the effective signal of the road surface excitation, but if they are not suppressed, they will cause spectral leakage to the effective frequency band in the frequency domain analysis, reducing the signal-to-noise ratio of the characteristic data. Therefore, they are filtered out by a low-pass filter.

[0051] Power frequency interference is another type of noise source that bandpass filtering needs to focus on. Its frequency is concentrated at 50 Hz and its integer multiples of harmonics. It originates from periodic crosstalk injected into the sensor signal link by high-voltage components such as motor drivers and power converters in the vehicle's electrical system through electromagnetic coupling paths. The frequency of this interference falls exactly within the effective frequency band of the road excitation and cannot be avoided simply by adjusting the bandpass cutoff frequency. Therefore, the filter is equipped with notch filters targeting the three main power frequency components at 50 Hz, 100 Hz, and 150 Hz. This narrow-band blocking method accurately removes power frequency energy from the signal while preserving the effective road excitation signal on both sides of the notch frequency to the greatest extent possible, avoiding the loss of useful information in adjacent frequency bands due to excessively wide notch bandwidth.

[0052] The bandpass filter is implemented in hardware logic within the field-programmable gate array (FPGA) of the signal processing unit. Each of the four sensor signals corresponds to an independent filtering channel, and the four channels operate in parallel without blocking each other. This hardware implementation allows the filtering operation to be completed synchronously as the data flows through the FPGA, without waiting for instructions from the central processing unit. The bandpass filtering processing time for a single signal is less than one millisecond. Key parameters such as the filter order, notch center frequency, and quality factor are optimized during the system calibration phase based on the actual vehicle electromagnetic environment and sensor noise floor characteristics, and then stored in the FPGA's configuration memory. No dynamic adjustments are required during runtime, ensuring that the filtering behavior remains stable and consistent throughout the entire driving process.

[0053] The signal after bandpass filtering retains the complete frequency components of the road excitation and braking conditions in the 1 Hz to 1 kHz frequency band. Its time-domain waveform removes low-frequency drift trends, power frequency periodic spikes and high-frequency glitches, significantly improving the signal-to-noise ratio. It is then directly fed into the frequency domain transformation module for fast Fourier transform and subsequent multidimensional feature extraction.

[0054] In some embodiments, the filtering preprocessing and frequency domain analysis of the original wheel vibration signal, and the extraction of frequency domain feature data, further include: The filtered time-domain vibration signal is subjected to a piecewise fast Fourier transform to convert the time-domain signal into a frequency-domain signal, and the frequency band energy ratio, peak frequency, amplitude range and spectral entropy are extracted as the frequency-domain feature data.

[0055] The time-domain vibration signal, after bandpass filtering, is first segmented into fixed time windows before entering the frequency domain analysis. The length of each segment's time window is selected based on a balance between frequency resolution and real-time performance. For example, with 1024 sampling points per segment, the corresponding time window length at a sampling rate of 48 kHz is approximately 21.3 milliseconds. This time window length ensures sufficient frequency resolution to distinguish the frequency differences in adjacent road surface texture features while keeping the processing delay of each segment within an acceptable range. A certain overlap area is set between adjacent segments, for example, 50% overlap between consecutive segments. This overlap mechanism avoids the signal truncation effect at segment boundaries from introducing spectral leakage artifacts in the frequency domain, ensuring the continuity and consistency of the spectrum across each segment.

[0056] For each segment of the time-domain signal, the signal processing unit performs a Fast Fourier Transform (FFT) operation in a Field-Programmable Gate Array (FPGA) to convert the time-domain sampled sequence into discrete frequency-domain spectra. The transform employs a radix-2 butterfly operation structure, with the hardware logic unit performing ten levels of butterfly iterations in parallel on 1024 complex sampling points, keeping the transformation time per segment within the order of microseconds. The transform output is a set of complex sequences containing amplitude and phase information for each frequency component. Taking the magnitude of the amplitude at each frequency point yields the power spectral density distribution of that segment. The four sensor signals independently complete the above transformation, and the spectral results for each segment are sequentially written into the feature extraction buffer.

[0057] Frequency domain feature data extraction is performed segment by segment, with each frequency spectrum as a unit. The energy percentage of a frequency band is calculated by dividing the entire effective frequency band into several sub-bands, integrating and summing the power spectral density within each sub-band, and then dividing by the total energy of the entire frequency band to obtain the energy percentage of each sub-band. This feature reflects the distribution of road excitation energy in the frequency domain; there are significant differences in the energy percentage of high-frequency sub-bands between rough and smooth road surfaces. Peak frequency extraction is performed by searching for the frequency coordinate corresponding to the maximum amplitude in the power spectrum. This frequency represents the dominant excitation frequency of wheel vibration within the current segment; for example, a certain fixed frequency dominates on gravel roads, while this frequency is scattered on asphalt roads.

[0058] The amplitude range is calculated by taking the difference between the maximum and minimum amplitude values ​​in the power spectrum segment. The magnitude of the range reflects the contrast in strength of the road excitation; the range increases significantly on bumpy roads while tending to be stable on flat roads. The spectral entropy is calculated by treating the normalized energy at each frequency point of the power spectrum as a probability distribution and substituting it into the Shannon entropy formula to obtain its information entropy value. A higher entropy value indicates a more uniform distribution of spectral energy and a more disordered road excitation. For example, the spectral entropy of asphalt pavement is usually higher than that of gravel pavement with a specific dominant frequency component.

[0059] The four-dimensional feature data together constitute a complete representation of the current vibration signal segment in the frequency domain. The dimensions are complementary; a single feature may be ambiguous under specific operating conditions, while the combined use of multiple dimensions significantly improves the distinguishability between different road surface types and driving conditions. After feature extraction of each segment is completed, the processing unit writes the four-dimensional feature vector of that segment, along with the corresponding sensor number and timestamp, into the output buffer. This is then packaged by the microcontroller and transmitted via the vehicle's high-speed bus to the recognition module in the decision layer for use by the classification model. Because the field-programmable gate array (FPGA) performs segmentation, transformation, and feature extraction of the four sensor signals in parallel in a pipelined manner, the end-to-end processing delay from the input of the last sampling point to the output of all feature data does not exceed five milliseconds.

[0060] In some embodiments, identifying the current driving condition and road surface type based on the frequency domain feature data according to the preset classification and recognition model includes: Based on a lightweight random forest classification model, the frequency domain feature data is matched with a preset operating condition template to identify at least one of the following driving conditions: slope-to-flat braking, high and low surface road switching, bumpy road driving, and speed bump crossing.

[0061] Frequency domain feature data is transmitted to the scene recognition module in the decision layer via the vehicle's high-speed bus, and then sequentially fed into a lightweight random forest classification model pre-deployed within the vehicle's domain controller. This model consists of multiple decision trees, each growing independently during the training phase based on a large amount of real-vehicle calibration data. Each tree's splitting nodes select one dimension and one threshold from four features (frequency band energy percentage, peak frequency, amplitude range, and spectral entropy) as discrimination conditions, performing binary judgments layer by layer from the root node to the leaf nodes. Finally, each leaf node outputs its assigned driving condition category. During the inference phase, the four-dimensional feature vector of the current segment is simultaneously input into all decision trees, each tree independently providing a classification result. The model aggregates the outputs of all decision trees through a majority voting mechanism, and the driving condition category with the highest number of votes becomes the final result of this round of recognition. Since the inference process of a single decision tree only involves a few value comparisons and branch jumps, the computational power consumption of the embedded controller is extremely low. Even if all decision trees are executed serially, the inference time from feature input to result output can be controlled within milliseconds.

[0062] The four driving conditions identified by the model each occupy a distinguishable cluster region in the frequency domain feature space. The identification of the slope-to-flat braking condition is based on the fact that at the instant the vehicle transitions from a slope to a flat road, the contact angle and normal load between the tire and the road surface change abruptly. This results in a specific shift in the peak frequency in the spectrum, and the amplitude range also exhibits significant step fluctuations within a short period. The model can complete the condition determination within tens of milliseconds by capturing this transient frequency domain feature pattern. The identification of the high-to-low adhesion road surface switching condition is based on the fact that when the vehicle moves from a high-adhesion road surface to a low-adhesion road surface, the frictional excitation intensity between the tire and the road surface decreases sharply. The proportion of high-frequency components in the energy distribution of each frequency band is significantly reduced, the peak frequency drifts towards lower frequencies, and the spectral entropy decreases as the excitation signal becomes more regularized. These combined changes constitute the frequency domain fingerprint of the high-to-low adhesion switching condition.

[0063] The identification of the bumpy road driving condition is based on the fact that irregular road surfaces such as gravel continuously apply broadband excitation to the tires, and the main frequency of the wheel's vertical vibration stabilizes within a specific range, such as around 11 Hz. The maximum amplitude remains at a high level, the amplitude range is consistently large due to periodic impacts, and the spectral entropy value is also in a high range due to the wide energy distribution. The identification of the speed bump driving condition is based on the fact that the speed bump generates a single pulse impact on the wheel, and a narrow band of peak energy is instantaneously concentrated in the spectrum. The peak frequency and amplitude range reach extreme values ​​in a very short time and then quickly fall back. The model completes the condition identification by detecting this peak-shaped frequency domain transient response.

[0064] The calibration data used for model training comes from high-frequency vibration data collected from the wheel ends of various real vehicles under the four operating conditions mentioned above. Each operating condition includes a large number of samples under different vehicle models, road surface materials, and vehicle speeds. After frequency domain feature extraction, a labeled feature dataset is formed. During training, hyperparameters such as the maximum depth, minimum number of leaf node samples, and splitting criteria of each decision tree are optimized through cross-validation to ensure that the classification accuracy meets the preset requirements while maintaining the model's lightweight nature. After training, the model parameters are stored in the non-volatile memory of the vehicle domain controller. During runtime, the microcontroller loads the parameters into memory for inference, eliminating the need for cloud connection or external computing power. When the model outputs the current operating condition category by majority vote, this result, along with the consistency information of the votes from each decision tree, is transmitted to the braking control strategy adaptation stage. The higher the voting consistency, the stronger the decision confidence. The control unit then decides whether to execute the corresponding parameter switching action to avoid unnecessary frequent changes in braking strategy due to misjudgment.

[0065] In some embodiments, the identification of the current driving condition and road surface type based on the frequency domain feature data according to the preset classification and recognition model further includes: The dominant frequency and maximum amplitude characteristics of the vertical vibration signal at the wheel end are extracted and compared with the built-in road surface feature database to identify at least one road surface type among asphalt road surface, gravel road surface, grass road surface and snow road surface.

[0066] The road surface type identification module extracts the vertical (Z-direction) vibration components from the four sensor signals of each wheel tip. This vibration signal directly experiences the normal impact of road surface unevenness on the tire and has the closest physical causal relationship with the road surface texture structure. The extracted vertical vibration components are also segmented for frequency domain analysis. Each segment undergoes a Fast Fourier Transform (FFT) to obtain the power spectrum of the vertical vibration. The module extracts two key characteristic quantities from this spectrum: the dominant frequency, which is the frequency coordinate corresponding to the maximum amplitude in the power spectrum, reflecting the dominant excitation frequency exerted by the current road surface texture on the wheel; and the maximum amplitude, which is the amplitude value at the point corresponding to the dominant frequency in the power spectrum, reflecting the intensity level of the road surface excitation. These two characteristic quantities together constitute the two-dimensional feature vector of the current road surface vertical vibration excitation.

[0067] The built-in road surface feature database was established during the system calibration phase. Its construction process involved selecting four typical road surface types: asphalt, gravel, grass, and snow-covered surfaces. Vertical vibration signals from the wheel ends were collected at the same sampling rate under various vehicle speed conditions. After undergoing the same bandpass filtering and frequency domain transformation, the distribution range of the dominant frequency and maximum amplitude was statistically analyzed from a large number of samples for each road surface type. The mean and tolerance range of these two feature quantities for each type of road surface were then used as the baseline feature template for that type of road surface and stored in the database. The four road surface types exhibit clear separability in the two-dimensional feature space. Asphalt road surfaces have a dense and uniform surface texture, and the excitation energy applied to the wheels is mainly concentrated in the higher frequency band, resulting in a generally higher dominant frequency. However, due to the good smoothness of the road surface, the maximum amplitude of the vertical vibration is relatively small.

[0068] The large and irregularly distributed particles on the surface of gravel roads create periodic angular impacts on the tires, resulting in a distinct narrow dominant frequency component in the vertical vibration. The dominant frequency is concentrated around a mid-range frequency, such as 11 Hz, and due to the strong impact energy, the maximum amplitude remains relatively high, around 0.60 g. Grass roads, due to the elastic buffering effect of the soil and vegetation layer, have their road excitation significantly attenuated by the time it reaches the wheel end. The dominant frequency is lower and the energy is dispersed, with the maximum amplitude being the lowest among the four road types. The excitation characteristics of snow-covered roads depend on the snow layer thickness and compaction. Under shallow snow layers, the road texture can still be partially transmitted to the tire, with the dominant frequency and maximum amplitude falling between those of asphalt and grass roads. Deep snow layers further tend towards the characteristic distribution of grass roads.

[0069] The identification process employs a segment-by-segment comparison method. The road surface type identification module calculates the Euclidean distance between the two-dimensional feature vector composed of the dominant frequency and maximum amplitude of the vertical vibration of each segment and the benchmark feature templates of the four road surface types in the database. The road surface type identified by the template with the smallest distance that falls within the corresponding tolerance range is the identification result for this round. Since the comparison process only involves two numerical readings and several simple arithmetic operations, the identification time for a single segment can be controlled within ten to twenty milliseconds.

[0070] To improve the temporal consistency of road surface recognition results in continuous driving scenarios, the module also introduces a majority voting mechanism within a short-term sliding window. This mechanism statistically analyzes the recognition results for several consecutive segments, using the road surface type with the highest frequency within the window as the current output. This effectively suppresses misjudgments caused by accidental interference in a single segment and avoids frequent jumps in road surface types within a short period. The recognition results are output to the central control unit along with the driving condition recognition results. Based on this, the control unit retrieves the intervention thresholds and adjustment step parameters of the anti-lock braking system, traction control system, and electronic stability system that match the current road surface adhesion conditions from the braking parameter set, thus completing the braking strategy adaptation at the road surface level.

[0071] In some embodiments, the adaptive adjustment of the braking control parameters of the electromechanical braking system based on the identified driving conditions and road surface type includes: Based on the identified driving conditions, the vehicle speed observation parameters and braking force distribution coefficient are dynamically corrected; based on the identified road surface type, the intervention thresholds and adjustment steps of the anti-lock braking system, traction control system, and electronic stability program are switched in stages.

[0072] In terms of operating conditions, after receiving the operating condition category identifier output by the scene recognition module, the central control unit retrieves the corresponding correction amount from the subset of operating condition parameters to dynamically correct the vehicle speed observation model and the braking force distribution coefficient. The vehicle speed observation model is the core algorithm module of the braking control system for estimating the current vehicle speed. Its input depends on the integral calculation of the wheel-end acceleration signal. When the vehicle moves from a slope to a flat road, due to the sudden change in the road surface slope angle, the wheel-end acceleration signal is mixed with a step change of the gravity component. If the vehicle speed observation model does not make corresponding corrections, it will produce cumulative vehicle speed estimation errors, which will lead to a mismatch between the braking force output and the actual braking demand.

[0073] After detecting a slope-to-flat transition, the central control unit corrects the gravity compensation term in the vehicle speed observation model in real time based on the change in slope angle. This eliminates spurious acceleration components introduced by sudden slope changes, allowing the estimated vehicle speed to return to its true value. The correction process takes effect in the next control cycle after the condition is identified. The adjustment of the braking force distribution coefficient is directly related to the characteristics of the condition. During the slope-to-flat transition, the normal load on the road surface shifts from the rear axle to the front axle. The central control unit accordingly increases the front wheel braking force distribution ratio and simultaneously increases the driving force command value of the brake calipers. This compensates for the decrease in rear wheel braking force margin caused by the load transfer, suppresses the possible precipitous drop in braking force during flat road phases, and maintains a smooth transition in braking deceleration before and after the condition change.

[0074] During the transition between high and low adhesion surfaces, the moment a vehicle moves from a high-adhesion surface to a low-adhesion surface, the available road friction for each wheel drops sharply. Upon recognizing this condition, the central control unit simultaneously lowers the upper limit of braking force for all wheels within the immediate control cycle. Based on the real-time assessment of the front and rear axle adhesion conditions, it recalculates the differential braking force distribution between the left and right wheels, ensuring that the actual applied braking force is always lower than the adhesion limit that the current road surface can provide, thus preventing the wheels from locking up due to overloaded braking force.

[0075] At the road surface level, the central control unit retrieves the corresponding control parameter set from the road surface parameter subset based on the road surface type identification module. This allows for tiered switching of the intervention thresholds and adjustment steps for the three subsystems: the anti-lock braking system (ABS), traction control system (TCS), and electronic stability program (ESP). The intervention threshold defines the trigger condition boundary for each subsystem to initiate intervention, while the adjustment step determines the increment of braking force after each subsystem intervention. For asphalt pavements with good adhesion, the longitudinal and lateral friction forces provided by the pavement are both high. Therefore, the intervention thresholds for the three subsystems are set in a higher slip ratio range. For example, the intervention trigger point for the ABS is set around 15% to 20% of the slip ratio, and the adjustment step is correspondingly larger, giving the braking system ample space to apply braking force and fully utilize the road surface adhesion potential to shorten the braking distance.

[0076] For snow-covered roads with poor adhesion, the coefficient of friction between the tire and the road surface is significantly reduced. The intervention thresholds of the three subsystems are lowered to a lower slip ratio range. The intervention trigger point of the anti-lock braking system (ABS) is moved forward to around 5% to 10% of the slip ratio, and the adjustment step size is simultaneously reduced to one-half to one-third of that under asphalt road conditions. This allows the subsystems to intervene earlier and make more precise adjustments each time, avoiding over-adjustment oscillations in braking force under low adhesion conditions. On gravel roads, the rolling effect of surface particles causes intermittent fluctuations in adhesion conditions. The central control unit sets the intervention threshold in a medium slip ratio range and appropriately reduces the adjustment step size to enhance the subsystem's sensitivity to fluctuations in adhesion conditions. On grass roads, due to the elastic deformation characteristics of the soil layer, both the longitudinal and lateral stiffness of the tires decrease. The parameter configurations of the three subsystems are between those of asphalt and snow-covered roads. The ABS focuses on suppressing low-frequency slip oscillations caused by insufficient stiffness.

[0077] Under special conditions such as bumpy roads and speed bumps, irregular road surface excitations will be superimposed on the braking signal to generate periodic interference. After recognizing the above conditions, the central control unit will specifically optimize the adjustment frequency of the anti-lock braking system, reduce the adjustment frequency to weaken the coupling interference of high-frequency road vibration on the braking force control circuit, suppress the periodic false lock-up phenomenon caused by the superposition of vibration and braking, and improve the stability of the braking process and the driving comfort.

[0078] After the parameters in both dimensions are adjusted, the central control unit integrates the working condition correction and road surface parameter switching within a unified control framework to generate a comprehensive execution command that includes the target braking force value of each wheel, caliper drive current, and subsystem parameter configuration. This command is then sent to the four-wheel electromechanical brake actuators and each subsystem controller via the vehicle high-speed bus with a transmission delay of no more than two milliseconds, thus completing the adaptive adaptation of the braking strategy for this wheel.

[0079] In some embodiments, the method further includes: The system monitors the working status of each vibration sensing unit in real time. When any vibration sensing unit is detected to be abnormal, it automatically retrieves the vibration sensing unit data of the wheel on the same side for compensation to maintain uninterrupted operation of the system.

[0080] While the central control unit is running the main braking control cycle, it maintains an independent sensing unit status monitoring path in parallel. This path performs health checks on the output signals of the four vibration sensing units at fixed intervals. The diagnostic process does not occupy the computing resources of the main control loop, but is continuously executed in the microcontroller of the signal processing unit as a low-priority background task.

[0081] The health diagnosis criteria cover multiple levels: At the signal amplitude level, the microcontroller checks in real time whether the root mean square (RMS) value of each output signal is within the preset normal operating range. If the RMS value of a signal is consistently lower than the device's noise floor level or suddenly jumps outside the saturation range, it is determined that the sensor may have lost its normal sensing capability due to open circuit, short circuit, or sensitive element failure. At the signal continuity level, the microcontroller monitors whether the arrival interval of each data frame conforms to the fixed beat corresponding to the 48 kHz sampling rate. If data frame loss, abnormal jumps in arrival interval, or data frame checksum errors occur, it indicates that there is a communication interruption or data corruption in the signal link. At the signal correlation level, the microcontroller performs pairwise correlation calculations on the time-domain waveforms output by the four sensors at the same time. Under normal driving conditions, the vibration signals of two wheels on the same axle have a high correlation coefficient because the road excitation source is the same. If the correlation coefficient of a signal with the other three suddenly drops to an extremely low level and continues for several sampling cycles, it indicates that the output of that sensor has deviated from the actual physical vibration, possibly due to loose installation or sensitive shaft misalignment. The diagnostic results from the above three levels are logically synthesized to form a health status identifier for each sensing unit, which is divided into three levels: normal, warning, and fault.

[0082] When the health status indicator changes from normal to warning, the microcontroller does not immediately execute a switching action. Instead, it extends the observation window and continuously monitors for several cycles. If the signal for that path recovers to normal within the observation window, the indicator is reset to normal. If the warning status persists after the observation window ends, the indicator is upgraded to fault. The instant the indicator upgrades to fault, the microcontroller triggers a fault-tolerant compensation program, sending a sensor unit fault notification along with the sensor number information for the faulty path to the central control unit. Upon receiving the notification, the central control unit immediately removes the frequency domain characteristic data of the faulty path sensor from the current braking control parameter calculation input and instead retrieves vibration sensing unit data from the adjacent wheel on the same side as a replacement input.

[0083] Because wheels on the same side of the road travel on the same side as the faulty wheel and experience road excitation in the same direction, their vibration signal frequency domain characteristics are highly consistent and substitutable. Taking a faulty left front wheel sensor as an example, the central control unit selects the vertical vibration dominant frequency and maximum amplitude characteristics of the left rear wheel sensor as the road surface identification input for the left front wheel. The complete four-dimensional frequency domain feature vector of the left rear wheel, after being corrected by weighting coefficients, is selected as the working condition identification input for the left front wheel. The values ​​of the weighting coefficients are determined during the calibration phase based on statistical analysis of numerous differences between signals from both wheels on the same side, and are used to compensate for minor feature deviations caused by differences in wheelbase position between the front and rear wheels. After weighting correction, the compensated data is fed into a random forest classification model and a road surface feature database comparison process. The subsequent identification and control parameter adaptation logic is completely consistent with the fault-free state, and the main braking control cycle is not interrupted or degraded due to a local fault in the sensing unit.

[0084] In fault-tolerant compensation mode, the system continuously monitors the output signals of the faulty sensor. If the health index of the signal returns to the normal range during subsequent driving, the microcontroller resets the health status from faulty to normal and sends a recovery notification to the central control unit. The central control unit then switches from same-side data compensation mode back to four-way independent perception mode, and the full-function perception architecture is automatically restored. The entire fault detection, compensation switching, and recovery switching process is completed autonomously by the system without driver intervention or external command triggering. During the switching process, the execution of the braking control strategy remains continuous and uninterrupted, and there are no abrupt changes or control gaps in the output of braking force of each wheel due to the switching of perception sources. When the system is running in same-side data compensation mode, the central control unit adds a perception degradation flag to the braking control command. This flag can be read by the upper-level autonomous driving system to assess the redundancy margin of the current braking perception. If the second sensor also fails, causing both wheels on the same side to lose independent perception capability, the system will trigger a higher-level safety response strategy to remind the driver to intervene or guide the vehicle to a safe stop.

[0085] Corresponding to the aforementioned electromechanical braking control method based on wheel-end vibration sensing, this invention also proposes an electromechanical braking control device based on wheel-end vibration sensing. Since the device embodiments of this invention correspond to the aforementioned method embodiments, details not disclosed in the device embodiments can be referred to the aforementioned method embodiments, and will not be repeated here.

[0086] Figure 2 A schematic diagram of an electromechanical braking control device based on wheel-end vibration sensing provided in this disclosure is shown below. Figure 2 As shown, it includes: The acquisition unit 21 is used to acquire the original vibration signal of the wheel during driving based on the preset sensing unit; Extraction unit 22 is used to perform filtering preprocessing and frequency domain analysis on the original vibration signal of the wheel, and extract frequency domain feature data; The identification unit 23 is used to identify the current driving conditions and road surface type based on the frequency domain feature data according to the preset classification identification model. The adjustment unit 24 is used to adaptively adjust the braking control parameters of the electromechanical braking system according to the identified driving conditions and road surface type.

[0087] Furthermore, in one possible implementation of this disclosure, the acquisition unit 21 is further configured to: The original vibration data of the wheel in three directions is continuously collected based on a preset high-frequency sampling rate; The original vibration data is uploaded directly, preserving the high-frequency vibration characteristics of the road surface excitation.

[0088] Furthermore, in one possible implementation of this disclosure, the extraction unit 22 is further configured to: The original vibration signal is filtered using a bandpass filter with a preset frequency range to remove power frequency interference and device noise, while retaining the frequency signals corresponding to road excitation and braking conditions.

[0089] Furthermore, in one possible implementation of this disclosure, the extraction unit 22 is further configured to: The filtered time-domain vibration signal is subjected to a piecewise fast Fourier transform to convert the time-domain signal into a frequency-domain signal, and the frequency band energy ratio, peak frequency, amplitude range and spectral entropy are extracted as the frequency-domain feature data.

[0090] Furthermore, in one possible implementation of this disclosure embodiment, the identification unit 23 is further configured to: Based on a lightweight random forest classification model, the frequency domain feature data is matched with a preset operating condition template to identify at least one of the following driving conditions: slope-to-flat braking, high and low surface road switching, bumpy road driving, and speed bump crossing.

[0091] Furthermore, in one possible implementation of this disclosure embodiment, the identification unit 23 is further configured to: The dominant frequency and maximum amplitude characteristics of the vertical vibration signal at the wheel end are extracted and compared with the built-in road surface feature database to identify at least one road surface type among asphalt road surface, gravel road surface, grass road surface and snow road surface.

[0092] Furthermore, in one possible implementation of this disclosure, the adjustment unit 24 is further configured to: Based on the identified driving conditions, the vehicle speed observation parameters and braking force distribution coefficient are dynamically corrected; based on the identified road surface type, the intervention thresholds and adjustment steps of the anti-lock braking system, traction control system, and electronic stability program are switched in stages.

[0093] Furthermore, in one possible implementation of the embodiments of this disclosure, such as Figure 3 As shown, the device further includes: The compensation unit 25 is used to monitor the working status of each vibration sensing unit in real time. When any vibration sensing unit is detected to be abnormal, it automatically retrieves the vibration sensing unit data of the wheel on the same side for compensation to maintain the uninterrupted operation of the system.

[0094] It should be noted that the foregoing explanation of the method embodiments also applies to the apparatus of the embodiments of this disclosure, and the principle is the same. Therefore, the embodiments of this disclosure are not limited thereto.

[0095] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0096] Figure 4 A schematic block diagram of an example electronic device 400 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0097] like Figure 4 As shown, device 400 includes a computing unit 401, which can perform various appropriate actions and processes based on a computer program stored in ROM (Read-Only Memory) 402 or a computer program loaded from storage unit 408 into RAM (Random Access Memory) 403. RAM 403 may also store various programs and data required for the operation of device 400. The computing unit 401, ROM 402, and RAM 403 are interconnected via bus 404. I / O (Input / Output) interface 405 is also connected to bus 404.

[0098] Multiple components in device 400 are connected to I / O interface 405, including: input unit 406, such as keyboard, mouse, etc.; output unit 407, such as various types of monitors, speakers, etc.; storage unit 408, such as disk, optical disk, etc.; and communication unit 409, such as network card, modem, wireless transceiver, etc. Communication unit 409 allows device 400 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0099] The computing unit 401 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 401 include, but are not limited to, CPUs (Central Processing Units), GPUs (Graphics Processing Units), various special-purpose AI (Artificial Intelligence) computing chips, various computing units running machine learning model algorithms, DSPs (Digital Signal Processors), and any suitable processor, controller, microcontroller, etc. The computing unit 401 performs the various methods and processes described above, such as the electromechanical braking control method based on wheel-end vibration sensing. For example, in some embodiments, the electromechanical braking control method based on wheel-end vibration sensing can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 408. In some embodiments, part or all of the computer program can be loaded and / or installed on device 400 via ROM 402 and / or communication unit 409. When the computer program is loaded into RAM 403 and executed by computing unit 401, one or more steps of the method described above can be performed. Alternatively, in other embodiments, computing unit 401 can be configured to perform the aforementioned electromechanical braking control method based on wheel-end vibration sensing by any other suitable means (e.g., by means of firmware).

[0100] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, FPGAs (Field Programmable Gate Arrays), ASICs (Application-Specific Integrated Circuits), ASSPs (Application-Specific Standard Products), SOCs (System-on-Chips), CPLDs (Complex Programmable Logic Devices), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0101] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0102] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, RAM, ROM, EPROM (Electrically Programmable Read-Only Memory) or flash memory, optical fiber, CD-ROM (Compact Disc Read-Only Memory), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0103] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (Cathode-Ray Tube) or LCD (Liquid Crystal Display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0104] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include LANs (Local Area Networks), WANs (Wide Area Networks), the Internet, and blockchain networks.

[0105] Computer systems can include clients and servers. Clients and servers are generally geographically separated and typically interact via communication networks. The client-server relationship is established by computer programs running on the respective computers and having a client-server relationship with each other. A server can be a cloud server, also known as a cloud computing server or cloud host, a hosting product within the cloud computing service ecosystem, addressing the shortcomings of traditional physical hosts and VPS (Virtual Private Server, or VPS) services, such as high management difficulty and weak business scalability. Servers can also be servers for distributed systems or servers incorporating blockchain technology.

[0106] It's important to note that artificial intelligence (AI) is the study of enabling computers to simulate certain human thought processes and intelligent behaviors (such as learning, reasoning, thinking, and planning). It encompasses both hardware and software technologies. AI hardware technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, and big data processing. AI software technologies primarily include computer vision, speech recognition, natural language processing, machine learning / deep learning, big data processing, and knowledge graph technologies.

[0107] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0108] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. An electromechanical braking control method based on wheel-end vibration sensing, characterized in that, include: Based on a preset sensing unit, the original vibration signal of the wheel during the driving process is collected; The original vibration signal of the wheel is filtered, preprocessed, and analyzed in the frequency domain to extract frequency domain feature data; Based on a preset classification and recognition model, the current driving conditions and road surface type are identified according to the frequency domain feature data; Based on the identified driving conditions and road surface type, the braking control parameters of the electromechanical braking system are adaptively adjusted.

2. The method according to claim 1, characterized in that, The acquisition of raw wheel vibration signals during driving, based on a preset sensing unit, includes: The original vibration data of the wheel in three directions is continuously collected based on a preset high-frequency sampling rate; The original vibration data is uploaded directly, preserving the high-frequency vibration characteristics of the road surface excitation.

3. The method according to claim 1, characterized in that, The filtering preprocessing and frequency domain analysis of the original vibration signal of the wheel, and the extraction of frequency domain feature data, include: The original vibration signal is filtered using a bandpass filter with a preset frequency range to remove power frequency interference and device noise, while retaining the frequency signals corresponding to road excitation and braking conditions.

4. The method according to claim 1, characterized in that, The step of filtering and preprocessing the original vibration signal of the wheel and performing frequency domain analysis to extract frequency domain feature data further includes: The filtered time-domain vibration signal is subjected to a piecewise fast Fourier transform to convert the time-domain signal into a frequency-domain signal, and the frequency band energy ratio, peak frequency, amplitude range and spectral entropy are extracted as the frequency-domain feature data.

5. The method according to claim 1, characterized in that, The classification and recognition model based on the preset criteria, which identifies the current driving conditions and road surface type according to the frequency domain feature data, includes: Based on a lightweight random forest classification model, the frequency domain feature data is matched with a preset operating condition template to identify at least one of the following driving conditions: slope-to-flat braking, high and low surface road switching, bumpy road driving, and speed bump crossing.

6. The method according to claim 5, characterized in that, The method for identifying the current driving condition and road surface type based on the preset classification and recognition model and the frequency domain feature data also includes: The dominant frequency and maximum amplitude characteristics of the vertical vibration signal at the wheel end are extracted and compared with the built-in road surface feature database to identify at least one road surface type among asphalt road surface, gravel road surface, grass road surface and snow road surface.

7. The method according to claim 1, characterized in that, The adaptive adjustment of the braking control parameters of the electromechanical braking system based on the identified driving conditions and road surface type includes: Based on the identified driving conditions, the vehicle speed observation parameters and braking force distribution coefficient are dynamically corrected; based on the identified road surface type, the intervention thresholds and adjustment steps of the anti-lock braking system, traction control system, and electronic stability program are switched in stages.

8. The method according to claim 1, characterized in that, The method further includes: The system monitors the working status of each vibration sensing unit in real time. When any vibration sensing unit is detected to be abnormal, it automatically retrieves the vibration sensing unit data of the wheel on the same side for compensation to maintain uninterrupted operation of the system.

9. An electromechanical braking control device based on wheel-end vibration sensing, characterized in that, include: The acquisition unit is used to acquire the original vibration signals of the wheels during the driving process based on the preset sensing unit; The extraction unit is used to perform filtering preprocessing and frequency domain analysis on the original vibration signal of the wheel, and extract frequency domain feature data; The identification unit is used to identify the current driving conditions and road surface type based on the frequency domain feature data according to a preset classification and identification model. The adjustment unit is used to adaptively adjust the braking control parameters of the electromechanical braking system based on the identified driving conditions and road surface type.

10. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-8.