An EMB pre-judgment braking control system and method based on intelligent tire multi-source global perception

Abstract

The application discloses an EMB pre-judgment type brake control system and method based on intelligent tire multi-source global perception, relates to the brake control technical field, and obtains vibration signals, deformation signals, load signals, friction coefficient signals and tire pressure and temperature signals collected by four-wheel intelligent tires. The vibration signals and the deformation signals are subjected to multi-stage filtering processing, the load signals, the friction coefficient signals and the tire pressure and temperature signals are subjected to consistency verification, and screened multi-source heterogeneous original data are obtained. The application moves the perception to the tire ground interface, realizes road identification and working condition pre-judgment in advance, changes the brake from response to pre-adjustment, shortens the time delay, and improves the brake stability and control continuity. Meanwhile, the brake parameters are adjusted through dynamic coupling of the tire state and the road condition, and in combination with multi-source verification and a layered fault-tolerant mechanism, the system is kept stable in operation under the data abnormal condition, the structural complexity is reduced, and the engineering adaptation capability is improved.

Description

Technical Field

[0001] This application relates to the field of braking control technology, specifically to an EMB predictive braking control system and method based on intelligent tire multi-source global perception. Background Technology

[0002] Current new energy vehicle chassis technology is developing towards drive-by-wire and advanced autonomous driving. Electromechanical brakes (EMB), due to their use of pure electric signals to drive braking, offer advantages such as fast response, high control precision, simplified structure, and the absence of hydraulic fluid, gradually becoming an important development direction for next-generation braking systems. In existing technologies, EMB systems typically rely on vehicle-side sensors as the primary sensing source, including inertial measurement units (IMUs), wheel speed sensors, and suspension displacement sensors. By collecting signals such as changes in vehicle attitude, wheel speed differences, and suspension motion, and after filtering and feature extraction, they indirectly estimate road adhesion and vehicle operating conditions. Upon detecting wheel slippage or vehicle vibration, they trigger control strategies such as ABS, TCS, or ESP, combining preset calibration parameters to complete the distribution and adjustment of braking force. This type of solution belongs to a technical architecture of "indirect perception at the vehicle side + passive response control," and its core characteristic lies in relying on low-frequency time-domain signals for operating condition identification and making control decisions through lookup tables or empirical models.

[0003] However, the aforementioned existing technical solutions have significant shortcomings in practical applications: First, because the sensing signal needs to be transmitted through elastic structures such as tires and suspension, the high-frequency road excitation information is severely attenuated or even lost. In addition, the overall perception has a large time delay, which is difficult to match the millisecond-level fast response capability of the EMB system, thus forming a performance mismatch between perception and execution. Secondly, existing methods mainly rely on single vibration or wheel speed characteristics for identification, lacking direct measurement of the tire-road contact state, and cannot accurately distinguish complex working conditions with low adhesion such as wet and slippery roads and icy and snowy roads, thus limiting the identification accuracy and stability. Secondly, the control strategies rely heavily on fixed calibration parameters and fail to fully consider the dynamic impact of factors such as tire wear, tire pressure changes, and temperature on braking performance, resulting in insufficient system adaptability. In addition, existing systems generally lack multi-source data redundancy and fault tolerance mechanisms. Once a key sensor fails, it will directly affect braking safety and make it difficult to meet the functional safety requirements of high-level autonomous driving. At the same time, the need for extensive calibration for different operating conditions results in a long overall development cycle and high cost, which limits its further promotion and application.

[0004] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this application is to provide an EMB predictive braking control system and method based on intelligent tire multi-source global perception, so as to solve the problems in the background art mentioned above.

[0006] To achieve the above objectives, this application provides the following technical solution: an EMB predictive braking control method based on intelligent tire multi-source full-domain perception, comprising the following steps: The vibration signal, deformation signal, load signal, friction coefficient signal, and tire pressure and temperature signal collected by the four-wheel intelligent tire are acquired. The vibration signal and deformation signal are subjected to multi-level filtering processing, and the load signal, friction coefficient signal, and tire pressure and temperature signal are subjected to consistency verification to obtain the filtered multi-source heterogeneous raw data. Based on multi-source heterogeneous raw data, frequency domain analysis and time domain abrupt change analysis are performed on the vibration signal. Steady-state characteristic parameters are extracted from the load signal, friction coefficient signal, and tire pressure and temperature signal. Multidimensional characteristic data containing frequency domain features, time domain features, load features, friction features, and temperature and pressure features are output. For multidimensional feature data, inputting it into the road surface recognition model yields the road surface type determination result, inputting it into the working condition prediction model yields the working condition change trend within the future time window, and outputs the corresponding prediction information; Based on the road surface type determination results and the working condition change trend, read the tire wear rate, tire pressure deviation and tire temperature parameters, substitute them into the tire, road surface and braking coupling relationship, and calculate the four-wheel braking basic threshold, braking force distribution parameters and braking adjustment parameters. Predictive braking control is performed based on the four-wheel braking base threshold and braking force distribution parameters. Preload control is performed during the stable operating condition phase, and dynamic adjustment control is performed during the changing operating condition phase to adjust the output of each wheel brake actuator. Fault determination is carried out based on multi-source heterogeneous raw data and multi-dimensional feature data. Abnormal data is removed, data on the same side of the wheel is compensated, data on the diagonal wheel is reconstructed, and degraded braking control is performed under abnormal conditions of multi-source data, and tire status warning information is output.

[0007] Multi-level filtering processing includes a combination of three methods: bandpass filtering, adaptive filtering, and frequency band separation processing. By partitioning and extracting the spectral energy distribution of the vibration signal, the frequency band signal corresponding to the road excitation is separated from the tire structure vibration signal. Combined with the synchronous change characteristics of the deformation signal, the filtering results are dynamically corrected, thereby constructing a high-fidelity vibration information set with road sensitivity to improve the effectiveness of subsequent feature extraction.

[0008] The multidimensional feature data contains at least eleven fused feature vectors, including vibration signal frequency domain energy distribution parameters, wavelet decomposition abrupt change coefficients, load change rate parameters, friction coefficient fluctuation amplitude parameters, and tire pressure and tire temperature coupling state parameters. Through the synergistic mapping relationship between multidimensional features, the differences in road surface types can be finely characterized, thereby improving the ability to distinguish complex road surfaces.

[0009] The road surface recognition model adopts a classification mechanism based on feature space partitioning, which maps multi-dimensional feature data to a preset set of road surface categories, including dry road surface, wet and slippery road surface, icy and snowy road surface and loose particle road surface. By dynamically updating the feature distribution boundary, it achieves adaptive recognition of different road surface conditions, thereby enhancing the stability of the recognition results.

[0010] The working condition prediction model constructs a time series input from multi-dimensional feature data, extracts the feature change trend within a continuous time window, and uses the slope and fluctuation pattern of the sequence change to predict future changes in road surface adhesion. It outputs working condition trend information within a time span of 50 to 100 milliseconds, thereby realizing the ability to adjust braking control in advance.

[0011] The tire-road-braking coupling relationship includes tire wear rate correction factors, tire pressure deviation correction factors, and temperature influence correction factors. By assigning weights to different correction factors, the braking threshold and braking force distribution parameters can be dynamically adjusted, thereby ensuring the consistency of braking behavior under different tire conditions.

[0012] Preload control outputs the initial braking torque in advance, putting the braking actuator in a ready-to-respond state. Dynamic adjustment control corrects the change in braking torque output in real time, keeping the braking output curve continuously changing, thereby avoiding vehicle instability caused by sudden changes.

[0013] The abnormal data removal process identifies abnormal vibration amplitude, sudden changes in friction coefficient, and load deviation by setting multidimensional threshold rules, and filters out abnormal data by combining time continuity judgment methods, thereby ensuring the consistency and reliability of the data involved in the calculation.

[0014] Diagonal wheel data reconstruction establishes a coupled mapping relationship between the four wheels of a vehicle, utilizes the vibration characteristics and load distribution information of the diagonal wheels to infer and calculate the road surface state of the target wheel, and makes corrections by combining historical data trends, thereby achieving the ability to recover the state under conditions of missing multi-source data.

[0015] An EMB predictive braking control system based on intelligent tire multi-source full-domain perception includes a multi-source signal acquisition and processing module, a multi-dimensional feature extraction and analysis module, a road surface identification and working condition prediction module, a braking parameter coupling calculation module, a predictive braking control execution module, and a multi-level fault-tolerant safety management module. The multi-source signal acquisition and processing module acquires vibration signals, deformation signals, load signals, friction coefficient signals, and tire pressure and temperature signals collected by the four-wheel intelligent tires. It performs multi-level filtering on the vibration and deformation signals and performs consistency verification on the load, friction coefficient, and tire pressure and temperature signals to obtain the filtered multi-source heterogeneous raw data. The multidimensional feature extraction and analysis module performs frequency domain analysis and time domain abrupt change analysis on vibration signals based on multi-source heterogeneous raw data. It extracts steady-state feature parameters from load signals, friction coefficient signals, and tire pressure and temperature signals, and outputs multidimensional feature data containing frequency domain features, time domain features, load features, friction features, and temperature and pressure features. The road surface recognition and working condition prediction module takes multi-dimensional feature data, inputs it into the road surface recognition model to obtain the road surface type determination result, inputs it into the working condition prediction model to obtain the working condition change trend within the future time window, and outputs the corresponding prediction information. The braking parameter coupling calculation module reads tire wear rate, tire pressure offset and tire temperature parameters based on the road surface type determination result and the working condition change trend, substitutes the tire, road surface and braking coupling relationship, and calculates the four-wheel braking basic threshold, braking force distribution parameters and braking adjustment parameters. The predictive braking control execution module performs predictive braking control according to the four-wheel braking basic threshold and braking force distribution parameters. It performs pre-load control in the stable working condition stage and dynamic adjustment control in the changing working condition stage, adjusting the output of each wheel brake actuator. The multi-level fault-tolerant safety management module performs fault determination based on multi-source heterogeneous raw data and multi-dimensional feature data, performs elimination processing on abnormal data, performs compensation processing on data of wheels on the same side, performs reconstruction processing on data of wheels diagonally opposite, performs degraded braking control under the condition of multi-source data anomalies, and outputs tire status warning information.

[0016] The technical effects and advantages provided by this application in the above technical solution are as follows: This application moves the perception level forward to the tire contact interface, directly acquiring the raw interaction information between the tire and the road surface. This allows for road condition identification and operational trend judgment before the vehicle body responds. This approach transforms braking control from a traditional response mode to a pre-adjustment mode, enabling the control strategy to be matched before execution, effectively shortening the time interval between perception and execution. Under complex road conditions, braking force output maintains higher consistency with adhesion conditions, reducing unstable fluctuations during braking. This improves the vehicle's driving stability and controllability in emergency braking scenarios, while also making the braking response more consistent, thus enhancing the overall vehicle control quality.

[0017] This application constructs a dynamic control mechanism that integrates tire condition and road surface conditions, enabling braking parameters to adaptively adjust according to tire wear, tire pressure changes, and temperature, thereby avoiding the adaptation limitations of fixed parameters. Based on this, through multi-source data cross-validation and a hierarchical fault-tolerant strategy, the system can maintain stable operation and control continuity even when some data is abnormal or sensor information is missing. Simultaneously, by uniformly utilizing multi-source information, the reliance on distributed sensing components is reduced, simplifying the chassis sensing structure, which helps reduce system implementation complexity and improve overall consistency. This allows the control system to meet safety requirements while possessing better engineering adaptability and scalability. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings.

[0019] Figure 1 This is a flowchart of the predictive EMB braking method of this application.

[0020] Figure 2 This is a schematic diagram of the system modules of this application. Detailed Implementation

[0021] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0022] like Figure 1 As shown, this application provides an EMB predictive braking control method based on intelligent tire multi-source global perception, including the following steps: The vibration signal, deformation signal, load signal, friction coefficient signal, and tire pressure and temperature signal collected by the four-wheel intelligent tire are acquired. The vibration signal and deformation signal are subjected to multi-level filtering processing, and the load signal, friction coefficient signal, and tire pressure and temperature signal are subjected to consistency verification to obtain the filtered multi-source heterogeneous raw data. Based on multi-source heterogeneous raw data, frequency domain analysis and time domain abrupt change analysis are performed on the vibration signal. Steady-state characteristic parameters are extracted from the load signal, friction coefficient signal, and tire pressure and temperature signal. Multidimensional characteristic data containing frequency domain features, time domain features, load features, friction features, and temperature and pressure features are output. For multidimensional feature data, inputting it into the road surface recognition model yields the road surface type determination result, inputting it into the working condition prediction model yields the working condition change trend within the future time window, and outputs the corresponding prediction information; Based on the road surface type determination results and the working condition change trend, read the tire wear rate, tire pressure deviation and tire temperature parameters, substitute them into the tire, road surface and braking coupling relationship, and calculate the four-wheel braking basic threshold, braking force distribution parameters and braking adjustment parameters. Predictive braking control is performed based on the four-wheel braking base threshold and braking force distribution parameters. Preload control is performed during the stable operating condition phase, and dynamic adjustment control is performed during the changing operating condition phase to adjust the output of each wheel brake actuator. Fault determination is carried out based on multi-source heterogeneous raw data and multi-dimensional feature data. Abnormal data is removed, data on the same side of the wheel is compensated, data on the diagonal wheel is reconstructed, and degraded braking control is performed under abnormal conditions of multi-source data, and tire status warning information is output.

[0023] Implementation Method 1: In this implementation method, the application scenario is set as a vehicle traveling at high speed in a straight line, such as on a highway or urban expressway, with a speed exceeding 80 km / h. Under this condition, the vehicle's braking system requires extremely high response speed and stability. Once it enters a low-traction road surface area, it is highly susceptible to safety risks such as increased braking distance, wheel lock-up, or vehicle skidding. Therefore, this implementation method focuses on demonstrating the application process and technical advantages of a predictive braking control method based on intelligent tire multi-source full-domain perception in low-traction road surface conditions.

[0024] During the data acquisition phase, multiple types of sensing units integrated within the four-wheeled intelligent tire operate synchronously, including high-frequency acceleration sensors, flexible strain sensors, pressure sensors, and temperature detection units. These units continuously collect dynamic information about the tire's contact with the ground at a high sampling frequency. Vibration signals reflect the road surface's microstructure and excitation characteristics; deformation signals characterize the tire's stress state and ground contact pattern changes; load signals reflect the tire's vertical force distribution; friction coefficient signals characterize the tire's adhesion to the ground; and tire pressure and temperature signals describe the tire's internal working state. This multi-source data constitutes a complete tire-ground interactive information system.

[0025] In the signal preprocessing stage, a multi-level combined filtering strategy is employed to process vibration and deformation signals. This includes methods such as bandpass filtering, low-pass filtering, and adaptive filtering to separate signals from different frequency bands, extracting effective high-frequency road surface excitation information and low-frequency structural response information. During this process, interference signals from tire structural vibration and vehicle suspension system transmission are filtered out, preserving the original road surface excitation characteristics. Simultaneously, consistency checks are performed on load signals, friction coefficient signals, and tire pressure and temperature data. A multi-parameter cross-comparison mechanism identifies anomalous data, such as sudden changes, drift values, or distorted signals, and these anomalous data are removed or marked, providing a reliable data foundation for subsequent processing.

[0026] In the feature extraction stage, for vibration signals, the Fast Fourier Transform (FFT) method is used for frequency domain analysis to extract the energy distribution characteristics of the vibration signals in different frequency bands, reflecting the road surface roughness and material properties. Simultaneously, wavelet transform is used for temporal abrupt change analysis of the vibration signals to capture road surface abrupt change points and transient impact characteristics, thereby identifying complex road surface features such as water films, ice layers, or gravel. For load signals, friction coefficient signals, and tire pressure and temperature signals, key feature parameters, such as average values, rates of change, and fluctuation amplitudes, are extracted through statistical analysis and steady-state modeling. Finally, the above frequency domain features, time domain features, load features, friction features, and temperature and pressure features are fused to form multi-dimensional feature data, which has high information density and strong expressive power.

[0027] In the identification and prediction phase, multi-dimensional feature data is input into the road surface identification model and the working condition prediction model. The road surface identification model classifies the current road surface type based on multi-dimensional features, including various typical working conditions such as dry asphalt, wet asphalt, waterlogged roads, icy and snowy roads, and gravel roads. The working condition prediction model analyzes the changing trends within a certain time window based on time series features. By analyzing changes in vibration spectrum, friction coefficient change rate, and load change trends, it predicts possible sudden changes in the adhesion coefficient within the next 50 to 100 milliseconds. Through this identification and prediction process, information about the road surface condition ahead can be obtained before the vehicle enters a low-adhesion area.

[0028] During the parameter calculation phase, based on the road surface identification results and the trend of operating conditions, and combined with tire wear rate, tire pressure deviation, and tire temperature status, a coupling relationship model between tire, road surface, and braking is constructed. This model comprehensively considers the influence of tire condition and road surface adhesion conditions on braking performance, and dynamically calculates the basic braking threshold, braking force distribution ratio, and braking adjustment parameters for each wheel. For example, under wet and slippery road conditions, the model will automatically reduce the maximum braking force threshold and increase the braking adjustment frequency to avoid wheel lock-up; under conditions of high tire temperature or abnormal tire pressure, the model will further adjust the braking parameters to adapt to changes in tire performance.

[0029] During the control execution phase, based on the calculated braking parameters, a pre-loading operation of the braking force is performed in advance. That is, before the vehicle enters the low-traction area, an appropriate initial braking force is applied to prepare the braking system. Simultaneously, after the vehicle actually enters the low-traction surface, the braking output is dynamically adjusted based on real-time feedback. By adjusting the output torque of the brake actuators, precise control of the braking force on each wheel is achieved. This process avoids the lag problem inherent in traditional passive response methods, making the braking process smoother and the response faster, thereby effectively preventing tire slippage and sudden changes in braking force.

[0030] During the safety and fault-tolerant phase, the system continuously monitors multi-source data. If a single sensor signal anomaly is detected, such as an interruption in vibration signal or an abnormal jump in friction coefficient, it is processed through a data redundancy mechanism. By weighted compensation of data from wheels on the same side, data continuity can be maintained even in the event of local sensor failure. Further state reconstruction using data from diagonally opposite wheels can restore the road surface feature information of the target wheel. In extreme cases, when all multi-source data are abnormal, the system enters a safety degradation mode, employing a conservative braking strategy to ensure driving safety. Simultaneously, it outputs tire condition warnings, such as abnormal tire pressure, overheating, or excessive wear, to prompt the vehicle system or driver to take appropriate measures.

[0031] Through the entire process described above, this implementation method achieves complete closed-loop control from data acquisition, feature extraction, state recognition, trend prediction to control execution and safety assurance. Compared with traditional braking control methods based on indirect vehicle body perception, this method moves the perception layer forward to the tire contact interface, significantly shortening the perception delay, improving road surface recognition accuracy, and enabling proactive adjustment of braking control. Under low-adhesion road surface conditions, this method can significantly reduce braking distance, improve vehicle driving stability, and enhance the reliability and safety of the system in complex environments.

[0032] Implementation Method Two: In this implementation method, the application scenario is set as the braking phase of a vehicle traveling on a continuous slope. Typical scenarios include long downhill sections, mountain roads, or elevated bridge approach areas. This type of condition exhibits significant longitudinal load variation characteristics. During downhill driving, the load distribution on the front and rear wheels dynamically changes due to the gravitational component. As the vehicle gradually transitions from a downhill to a flat road, the vertical load on the tires will significantly increase or reverse its trend. Traditional braking systems in this type of condition often rely on fixed calibration parameters for control, making it difficult to adapt to load changes in a timely manner. This can easily lead to sudden changes in braking force or delayed braking response at the moment of transition from a slope to a flat road, resulting in decreased braking comfort or even safety hazards. Therefore, this implementation method focuses on using an intelligent tire multi-source sensing and prediction mechanism to achieve early identification of downhill-to-flat road conditions and adaptive adjustment of braking parameters.

[0033] During the data acquisition phase, multiple sensing units inside the four-wheeled intelligent tire operate continuously, acquiring vibration signals, deformation signals, load signals, friction coefficient signals, and tire pressure and temperature information in real time. Specifically, the load signal, acquired through embedded strain sensors or piezoelectric elements, reflects the changing trend of tire contact pressure; the vibration signal, collected by a high-frequency accelerometer, describes the road excitation and tire response characteristics; the friction coefficient signal is estimated through tire contact characteristics and used to characterize adhesion; and the tire pressure and temperature signals reflect changes in the tire's internal state. These multi-source signals are continuously output through high-frequency sampling, providing fundamental data for subsequent analysis.

[0034] In the preprocessing stage, a multi-level filtering strategy is employed to separate vibration and deformation signals across different frequency bands. This includes low-pass filtering for extracting structural response features, high-pass filtering for extracting road surface excitation information, and adaptive filtering for suppressing random noise. During this process, analysis of the signal spectrum distribution distinguishes between vibration components caused by the road surface and those generated by the tire's own structure. Simultaneously, consistency checks are performed on load signals, friction coefficient signals, and tire pressure and temperature information. Multi-parameter cross-comparison identifies abnormal data, such as sensor drift, abrupt changes, or sampling distortion, and these abnormal data are either removed or flagged to ensure the reliability of the input data.

[0035] During feature extraction, vibration signals are analyzed in the frequency domain using Fast Fourier Transform (FFT) to obtain energy distribution characteristics across different frequency bands, reflecting road surface roughness and tire contact characteristics. Simultaneously, wavelet transform is employed to perform temporal abrupt change analysis on the vibration signals, capturing transient impact changes to identify dynamic responses caused by changes in road structure or slope. Load signals are analyzed through time series analysis to extract trend parameters, including the load mean, rate of change, and fluctuation range; these parameters are crucial for determining slope changes. Friction coefficient signals are used to supplement adhesion information. Tire pressure and temperature signals are analyzed using steady-state analysis to extract current tire operating state characteristics. Through these processes, frequency domain features, time domain features, load trend features, friction features, and temperature / pressure features are fused to form a high-dimensional feature dataset.

[0036] In the condition identification and trend prediction stage, high-dimensional feature data is input into the condition prediction model. This model analyzes the changing trends over a future period based on time-series data. By continuously tracking the load change trend, when it detects a continuous downward trend in the tire vertical load with a gradually slowing rate of change, coupled with an increase in low-frequency components in the vibration spectrum, it can be determined that the vehicle is on a downhill slope. When the load change trend reaches an inflection point and gradually increases, the model determines that it is about to enter a level road phase. By identifying this trend in advance, predictive information can be output before the vehicle has fully entered a level road, thus providing advance warning for subsequent braking control.

[0037] During the parameter calculation phase, based on the predicted results and current tire condition information, including tire wear rate, tire pressure deviation, and tire temperature, braking parameters are dynamically calculated using a tire-road-brake coupling model. This coupling model considers the interaction between the tire and the road surface, while also incorporating tire health factors, allowing braking parameters to be adjusted in real time according to actual operating conditions. For example, when high tire wear is detected, the model will appropriately lower the braking limit to avoid the risk of lock-up due to decreased adhesion; when tire pressure is too low or tire temperature is too high, the model will adjust the braking adjustment step size to improve control accuracy. During the transition from a slope to a flat surface, as the load increases, the model will gradually increase the braking force output to ensure braking continuity.

[0038] During the control execution phase, the brake actuator is pre-adjusted based on the calculated braking baseline threshold and braking force distribution parameters. During downhill sections, the rate of increase in braking force is appropriately limited to avoid braking instability caused by load changes. As the vehicle approaches level ground, the braking output capacity is gradually increased based on predictions, enabling a smooth transition to level road conditions. Upon actually entering level ground, the braking output is dynamically adjusted through real-time feedback to prevent sudden increases or decreases in braking force, thereby improving braking comfort and stability.

[0039] Regarding tire condition adaptive control, when tire wear rate is detected to exceed a preset threshold, such as 20%, the system will automatically adjust the braking control strategy. By reducing the maximum braking force threshold and increasing the adjustment step size, the system improves its adaptability to aging tires. Simultaneously, in cases of abnormal tire pressure or excessively high tire temperature, the system corrects the coupling relationship model parameters to make braking control more conservative, thereby reducing risk.

[0040] Regarding safety and fault tolerance mechanisms, the system continuously monitors multi-source data and identifies anomalies through cross-validation of multi-source data. When a single sensor signal is abnormal, it is handled through data elimination and replacement strategies; when data from the same side wheel is abnormal, it is compensated using data from adjacent wheels; when multiple data points are abnormal, it is reconstructed using data from diagonally opposite wheels to restore the target wheel's status information. In extreme cases, when data reliability cannot be guaranteed, the system enters a degraded operation mode, employing a conservative braking strategy to ensure vehicle safety. Simultaneously, it outputs tire status warning information, including excessive wear, abnormal tire pressure, and abnormal temperature, providing a basis for subsequent maintenance.

[0041] Through the complete process described above, this implementation method achieves forward-looking identification and adaptive braking control under slope-to-flat conditions, effectively solving problems such as discontinuous braking force, response lag, and insufficient adaptability in traditional control methods. Under complex slope conditions and varying tire conditions, this method can significantly improve braking stability and comfort, while enhancing the system's environmental adaptability and safety assurance capabilities, providing a reliable braking control foundation for advanced autonomous driving scenarios.

[0042] Implementation Method 3: This implementation method focuses on addressing the reliability issues of multi-source perception data in complex environments and the extremely high safety requirements of braking systems in advanced autonomous driving scenarios, constructing a multi-source fault-tolerant and full-domain collaborative control mechanism. This implementation method not only covers typical fault scenarios such as single sensor failure and multi-sensor anomalies, but also extends to the cross-system collaborative control level, realizing information sharing and dynamic coordination between braking, suspension, and steering, thereby constructing an intelligent braking control system with high robustness, high safety, and high scalability.

[0043] During system operation, the four-wheel intelligent tires continuously output vibration signals, deformation signals, load signals, friction coefficient information, and tire pressure and temperature data. These multi-source heterogeneous data, after filtering and feature extraction, form a multi-dimensional feature data set. Simultaneously, the system monitors the raw and feature data in real time, evaluating data quality through multi-dimensional anomaly detection rules. These rules include amplitude threshold judgment, rate of change anomaly detection, temporal continuity detection, and multi-source data consistency comparison. For example, if a vibration signal experiences a sudden interruption or an abnormal spike in amplitude, it can be determined that the signal acquisition is faulty; if the friction coefficient signal and load change trends show significant inconsistency, it can be determined that the data is distorted.

[0044] In the first-level fault-tolerance phase, when a single sensor signal anomaly is detected, the system initiates a self-checking mechanism to eliminate abnormal data. This mechanism cross-validates data from multiple sensor channels within the same wheel, such as by analyzing the correlation between vibration and deformation signals, to identify the source of the anomaly and eliminate faulty data. Simultaneously, it utilizes historical data for short-term compensation, thus preventing a single anomaly from affecting the overall judgment. During this phase, the system maintains complete sensing capabilities, and the braking control strategy remains largely unchanged.

[0045] In the secondary fault-tolerance stage, when data is missing from the same-side wheels or multiple sensor signals are simultaneously abnormal, the system performs weighted compensation using data from the same-side tires. For example, if some data from the left front wheel fails, compensation can be made using the corresponding signal from the left rear wheel. By considering vehicle dynamics and load distribution, the compensation data is weighted and corrected to reflect the true state of the current wheels. During this process, a vehicle lateral stability model is introduced to constrain the compensation results, ensuring the rationality of the compensation data.

[0046] In the third-level fault-tolerant stage, when multiple wheels exhibit data anomalies or same-side compensation fails to meet accuracy requirements, the system further reconstructs the state using data from the diagonal wheels. By establishing a coupling relationship model between the four wheels and utilizing the dynamic symmetry of the diagonal wheels, the system infers the road surface and tire states of the target wheels. For example, a mapping relationship is established between the left front wheel and the right rear wheel. By analyzing the vibration spectrum characteristics and load variation trends of the right rear wheel, the road surface excitation characteristics of the left front wheel are calculated, thereby restoring key control parameters. This stage can maintain the continuity and stability of braking control under multi-point fault conditions.

[0047] In the Level 4 fault-tolerance stage, when the anomalies in multi-source data worsen, such as severe distortion or interruption of data from multiple wheel sensors, the system enters a degraded control mode. In this mode, the braking control strategy switches from predictive dynamic control to a conservative control strategy, reducing the upper limit of braking force output and increasing control redundancy to ensure the vehicle retains basic braking capability even in extreme conditions. Simultaneously, it outputs warning information such as tire wear, abnormal tire temperature, and abnormal tire pressure to provide a basis for subsequent maintenance. The control strategy in degraded mode prioritizes safety over optimal performance.

[0048] Building upon the aforementioned multi-level fault-tolerant mechanism, this implementation further extends to the full-domain collaborative control level. Supported by multi-source fusion feature data and operational condition prediction results, road surface condition information and future trends are output to other chassis control units, including the suspension control unit and steering control unit. Through an information sharing mechanism, coordinated adjustment between braking, suspension, and steering is achieved. For example, when a decrease in the road surface adhesion coefficient is detected ahead, the braking system adjusts its braking force distribution strategy in advance, the suspension system simultaneously adjusts its damping parameters to improve tire contact with the road, and the steering system appropriately reduces its steering sensitivity, thereby improving overall vehicle stability.

[0049] In terms of hardware expansion, this implementation supports the replacement of various sensing solutions. For example, traditional MEMS sensors can be replaced with high-frequency piezoelectric sensing films to improve signal sampling frequency and shock resistance; the sensor module installation method can be adjusted according to the tire structure, including embedded, sidewall embedded, or bearing integrated designs, to adapt to the needs of different vehicle platforms. At the communication level, traditional CAN communication can be upgraded to in-vehicle Ethernet communication to improve data transmission bandwidth and real-time performance, thereby supporting higher frequency data acquisition and more complex algorithm processing.

[0050] At the algorithmic level, this implementation also possesses excellent scalability. Existing recognition methods based on feature engineering and lightweight models can be upgraded to deep learning models, such as convolutional neural networks or temporal network structures, to improve the recognition accuracy of complex road surfaces. Simultaneously, long-term prediction models can be introduced to achieve condition prediction over longer time windows, thereby further enhancing predictive control capabilities. Under extreme conditions, nonlinear signal analysis methods, such as the Hilbert-Huang transform, can be employed to finely analyze non-stationary signals, improving the system's adaptability to complex environments.

[0051] Furthermore, in terms of functional expansion, multi-source data analysis can further enhance fault diagnosis capabilities, such as identifying potential problems like brake caliper wear, suspension loosening, or wheel bearing abnormalities, thereby achieving comprehensive vehicle status monitoring. Simultaneously, by analyzing driver operating habits, braking control strategies can be personalized, ensuring the system balances safety with a superior driving experience.

[0052] Through the aforementioned multi-source fault-tolerant mechanism and global collaborative control strategy, this implementation constructs a highly robust and scalable intelligent braking control system. Under complex operating conditions and multiple fault conditions, the system can maintain stable operation and provide reliable control output, while also possessing good scalability to adapt to the future development needs of autonomous driving technology, providing a solid safety guarantee for high-level autonomous vehicles.

[0053] like Figure 2As shown, this application provides an EMB predictive braking control system based on intelligent tire multi-source full-domain perception, including a multi-source signal acquisition and processing module, a multi-dimensional feature extraction and analysis module, a road surface identification and working condition prediction module, a braking parameter coupling calculation module, a predictive braking control execution module, and a multi-level fault-tolerant safety management module. The multi-source signal acquisition and processing module acquires vibration signals, deformation signals, load signals, friction coefficient signals, and tire pressure and temperature signals collected by the four-wheel intelligent tires. It performs multi-level filtering on the vibration and deformation signals and performs consistency verification on the load, friction coefficient, and tire pressure and temperature signals to obtain the filtered multi-source heterogeneous raw data. The multidimensional feature extraction and analysis module performs frequency domain analysis and time domain abrupt change analysis on vibration signals based on multi-source heterogeneous raw data. It extracts steady-state feature parameters from load signals, friction coefficient signals, and tire pressure and temperature signals, and outputs multidimensional feature data containing frequency domain features, time domain features, load features, friction features, and temperature and pressure features. The road surface recognition and working condition prediction module takes multi-dimensional feature data, inputs it into the road surface recognition model to obtain the road surface type determination result, inputs it into the working condition prediction model to obtain the working condition change trend within the future time window, and outputs the corresponding prediction information. The braking parameter coupling calculation module reads tire wear rate, tire pressure offset and tire temperature parameters based on the road surface type determination result and the working condition change trend, substitutes the tire, road surface and braking coupling relationship, and calculates the four-wheel braking basic threshold, braking force distribution parameters and braking adjustment parameters. The predictive braking control execution module performs predictive braking control according to the four-wheel braking basic threshold and braking force distribution parameters. It performs pre-load control in the stable working condition stage and dynamic adjustment control in the changing working condition stage, adjusting the output of each wheel brake actuator. The multi-level fault-tolerant safety management module performs fault determination based on multi-source heterogeneous raw data and multi-dimensional feature data, performs elimination processing on abnormal data, performs compensation processing on data of wheels on the same side, performs reconstruction processing on data of wheels diagonally opposite, performs degraded braking control under the condition of multi-source data anomalies, and outputs tire status warning information.

[0054] The present invention provides an EMB predictive braking control method based on intelligent tire multi-source full-domain perception, which is implemented by the above-mentioned EMB predictive braking control system based on intelligent tire multi-source full-domain perception. For details of the specific method and process of the EMB predictive braking control system based on intelligent tire multi-source full-domain perception, please refer to the above-mentioned embodiment of the EMB predictive braking control method based on intelligent tire multi-source full-domain perception, which will not be repeated here.

[0055] The foregoing has only described certain exemplary embodiments of this application by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of this application. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of this application.

Claims

1. A predictive braking control method based on intelligent tire multi-source global perception (EMB), characterized in that, Includes the following steps: The vibration signal, deformation signal, load signal, friction coefficient signal, and tire pressure and temperature signal collected by the four-wheel intelligent tire are acquired. The vibration signal and deformation signal are subjected to multi-level filtering processing, and the load signal, friction coefficient signal, and tire pressure and temperature signal are subjected to consistency verification to obtain the filtered multi-source heterogeneous raw data. Based on multi-source heterogeneous raw data, frequency domain analysis and time domain abrupt change analysis are performed on the vibration signal. Steady-state characteristic parameters are extracted from the load signal, friction coefficient signal, and tire pressure and temperature signal. Multidimensional characteristic data containing frequency domain features, time domain features, load features, friction features, and temperature and pressure features are output. For multidimensional feature data, inputting it into the road surface recognition model yields the road surface type determination result, inputting it into the working condition prediction model yields the working condition change trend within the future time window, and outputs the corresponding prediction information; Based on the road surface type determination results and the working condition change trend, read the tire wear rate, tire pressure deviation and tire temperature parameters, substitute them into the tire, road surface and braking coupling relationship, and calculate the four-wheel braking basic threshold, braking force distribution parameters and braking adjustment parameters. Predictive braking control is performed based on the four-wheel braking base threshold and braking force distribution parameters. Preload control is performed during the stable operating condition phase, and dynamic adjustment control is performed during the changing operating condition phase to adjust the output of each wheel brake actuator. Fault determination is carried out based on multi-source heterogeneous raw data and multi-dimensional feature data. Abnormal data is removed, data on the same side of the wheel is compensated, data on the diagonal wheel is reconstructed, and degraded braking control is performed under abnormal conditions of multi-source data, and tire status warning information is output.

2. The EMB predictive braking control method based on intelligent tire multi-source global perception as described in claim 1, characterized in that, Multi-level filtering processing includes three combined methods: bandpass filtering, adaptive filtering, and frequency band separation processing. By partitioning and extracting the spectral energy distribution of the vibration signal, the frequency band signal corresponding to the road excitation is separated from the tire structure vibration signal. Combined with the synchronous change characteristics of the deformation signal, the filtering result is dynamically corrected, thereby constructing a high-fidelity vibration information set with road sensitivity.

3. The EMB predictive braking control method based on intelligent tire multi-source global perception as described in claim 1, characterized in that, The multidimensional feature data contains at least eleven fused feature vectors, including vibration signal frequency domain energy distribution parameters, wavelet decomposition abrupt change coefficients, load change rate parameters, friction coefficient fluctuation amplitude parameters, and tire pressure and tire temperature coupling state parameters.

4. The EMB predictive braking control method based on intelligent tire multi-source global perception as described in claim 1, characterized in that, The road surface recognition model adopts a classification mechanism based on feature space partitioning, which maps multi-dimensional feature data to a preset set of road surface categories, including dry road surface, wet and slippery road surface, icy and snowy road surface and loose particle road surface. By dynamically updating the feature distribution boundary, it achieves adaptive recognition of different road surface conditions.

5. The EMB predictive braking control method based on intelligent tire multi-source global perception according to claim 1, characterized in that, The working condition prediction model constructs a time series input from multi-dimensional feature data, extracts the feature change trend within a continuous time window, and uses the slope and fluctuation pattern of the sequence change to predict future changes in road surface adhesion, outputting working condition trend information within a time span of 50 to 100 milliseconds.

6. The EMB predictive braking control method based on intelligent tire multi-source global perception as described in claim 1, characterized in that, The tire-road-braking coupling relationship includes tire wear rate correction factors, tire pressure deviation correction factors, and temperature influence correction factors. By assigning weights to different correction factors, the braking threshold and braking force distribution parameters can be dynamically adjusted.

7. The EMB predictive braking control method based on intelligent tire multi-source global perception according to claim 6, characterized in that, Preload control prepares the braking actuator for response by outputting the initial braking torque in advance, while dynamic adjustment control maintains the continuous variation characteristic of the braking force output curve by correcting the change in braking torque output in real time.

8. The EMB predictive braking control method based on intelligent tire multi-source global perception according to claim 1, characterized in that, The abnormal data removal process identifies abnormal vibration amplitude, sudden changes in friction coefficient, and load deviation by setting multidimensional threshold rules, and then filters out abnormal data by combining time continuity judgment methods.

9. The EMB predictive braking control method based on intelligent tire multi-source global perception according to claim 1, characterized in that, Diagonal wheel data reconstruction establishes a coupled mapping relationship between the four wheels of a vehicle, utilizes the vibration characteristics and load distribution information of the diagonal wheels to infer and calculate the road surface condition of the target wheel, and makes corrections by combining historical data trends.

10. A predictive braking control system based on intelligent tire multi-source full-domain perception, used to implement the predictive braking control method based on intelligent tire multi-source full-domain perception as described in any one of claims 1-9, characterized in that, It includes a multi-source signal acquisition and processing module, a multi-dimensional feature extraction and analysis module, a road surface identification and condition prediction module, a braking parameter coupling calculation module, a predictive braking control execution module, and a multi-level fault-tolerant safety management module; The multi-source signal acquisition and processing module acquires vibration signals, deformation signals, load signals, friction coefficient signals, and tire pressure and temperature signals collected by the four-wheel intelligent tires. It performs multi-level filtering on the vibration and deformation signals and performs consistency verification on the load, friction coefficient, and tire pressure and temperature signals to obtain the filtered multi-source heterogeneous raw data. The multidimensional feature extraction and analysis module performs frequency domain analysis and time domain abrupt change analysis on vibration signals based on multi-source heterogeneous raw data. It extracts steady-state feature parameters from load signals, friction coefficient signals, and tire pressure and temperature signals, and outputs multidimensional feature data containing frequency domain features, time domain features, load features, friction features, and temperature and pressure features. The road surface recognition and working condition prediction module takes multi-dimensional feature data, inputs it into the road surface recognition model to obtain the road surface type determination result, inputs it into the working condition prediction model to obtain the working condition change trend within the future time window, and outputs the corresponding prediction information. The braking parameter coupling calculation module reads tire wear rate, tire pressure offset and tire temperature parameters based on the road surface type determination result and the working condition change trend, substitutes the tire, road surface and braking coupling relationship, and calculates the four-wheel braking basic threshold, braking force distribution parameters and braking adjustment parameters. The predictive braking control execution module performs predictive braking control according to the four-wheel braking basic threshold and braking force distribution parameters. It performs pre-load control in the stable working condition stage and dynamic adjustment control in the changing working condition stage, adjusting the output of each wheel brake actuator. The multi-level fault-tolerant safety management module performs fault determination based on multi-source heterogeneous raw data and multi-dimensional feature data, performs elimination processing on abnormal data, performs compensation processing on data of wheels on the same side, performs reconstruction processing on data of wheels diagonally opposite, performs degraded braking control under the condition of multi-source data anomalies, and outputs tire status warning information.

Images