Vehicle parameter self-adaptive control method and device and vehicle

Abstract

The invention relates to the technical field of vehicle active safety control, in particular to a vehicle parameter self-adaptive control method and device and a vehicle. The method comprises the steps that pressure distribution data, temperature gradient data and deformation acceleration data of tires are obtained; according to the pressure distribution data, the temperature gradient data and the deformation acceleration data, the health degree index of the tire is calculated, and the vehicle speed and the longitudinal acceleration of the vehicle are obtained; and adjusting at least one of a trigger threshold, a steering compensation angle and a torque limit of the vehicle body electronic stabilization system according to at least one of the health index, the vehicle speed and the longitudinal acceleration, and controlling the vehicle based on the at least one of the trigger threshold, the steering compensation angle and the torque limit. Therefore, the problems that in the related technology, a vehicle monitoring system can only sense the basic state of the tire part, and once the tire state is abnormal, potential safety hazards are likely to happen to the vehicle are solved.

Description

Technical Field

[0001] This application relates to the field of vehicle active safety control technology, and in particular to a vehicle parameter adaptive control method, device and vehicle. Background Technology

[0002] In the field of vehicle driving safety control, electronic stability control, steering, and torque control are key components in ensuring driving safety, and their control parameters are typically set using fixed calibration methods. In related technologies, vehicle monitoring systems can only sense the basic condition of the tires; if the tire condition becomes abnormal, the vehicle is prone to safety hazards. Summary of the Invention

[0003] This application provides a vehicle parameter adaptive control method, device, and vehicle to solve the problems in related technologies where vehicle monitoring systems can only detect basic tire abnormalities, leading to potential safety hazards once the tire condition becomes abnormal.

[0004] The first aspect of this application provides a vehicle parameter adaptive control method, comprising the following steps: acquiring tire pressure distribution data, temperature gradient data, and deformation acceleration data; calculating a tire health index based on the pressure distribution data, temperature gradient data, and deformation acceleration data, and acquiring the vehicle speed and longitudinal acceleration; adjusting at least one of a trigger threshold, steering compensation angle, and torque limit of the vehicle electronic stability system based on at least one of the health index, vehicle speed, and longitudinal acceleration, and controlling the vehicle based on at least one of the trigger threshold, steering compensation angle, and torque limit.

[0005] Based on the aforementioned technical means, this application embodiment can first calculate a health index that reflects the true health status of the tire by acquiring tire pressure distribution data, temperature gradient data, and deformation acceleration data, combined with the vehicle's real-time speed and longitudinal acceleration. Then, based on this index and vehicle dynamic parameters, at least one of the trigger threshold, steering compensation angle, and torque limit of the vehicle's electronic stability system is dynamically adjusted, thereby achieving precise control of the vehicle. This not only allows the control strategy to be adapted in a timely manner according to the actual health status of the tire (such as abnormal local stress or potential damage), but also avoids the risk of loss of control caused by parameters that cannot be dynamically adjusted with the operating conditions. It allows the control logic to conform to real-time driving conditions, reducing the problems of over-control or under-control, while enhancing the coordination between the tire and the vehicle, indirectly extending tire life and improving driving safety.

[0006] Optionally, a sensor array is installed inside the vehicle's tires to collect pressure distribution data, temperature gradient data, and deformation acceleration data.

[0007] Based on the aforementioned technical means, this embodiment of the application can collect tire pressure distribution data, temperature gradient data, and deformation acceleration data by setting a sensor array inside the vehicle tire. This overcomes the limitation of single-point sensors in related technologies, which can only acquire isolated local data. Through the spatially distributed layout of the array, it can reflect the state differences of different parts of the tire (such as the pressure difference between the tire shoulder and the tire crown, and the temperature changes between the inner and outer sides), avoiding the omission of local anomalies (such as localized wear and stress concentration). Furthermore, it provides comprehensive and accurate foundational support for subsequent tire health index calculations through multi-dimensional data. Simultaneously, cross-validation of multi-sensor data reduces the impact of single sensor failures or errors on the perception results, further improving data reliability.

[0008] Optionally, the sensor array includes an accelerometer, a thermopile sensor, and a piezoelectric sensor, with the piezoelectric sensor generating power from tire deformation to power the sensor array.

[0009] Based on the aforementioned technical means, this embodiment of the application specifically configures the sensor array as a combination of an accelerometer, a thermopile sensor, and a piezoelectric sensor. The accelerometer is responsible for collecting tire deformation acceleration data, the thermopile sensor focuses on acquiring temperature gradient data, and the piezoelectric sensor undertakes the task of collecting pressure distribution data. Simultaneously, the piezoelectric sensor can generate electricity using the natural deformation of the tire during driving, directly providing power to the entire sensor array. Through the clearly defined division of labor among multiple types of sensors, the accuracy and specificity of multi-dimensional tire state data acquisition are ensured. Furthermore, the piezoelectric sensor achieves integrated sensing and power supply, eliminating the dependence on batteries in related technologies and avoiding the problems of short battery life and high replacement and maintenance costs. At the same time, the continuous deformation of the tire ensures the stability of the power supply, effectively preventing monitoring failures due to power interruptions and providing support for the long-term stable operation of the sensor array.

[0010] Optionally, the tire health index is calculated based on pressure distribution data, temperature gradient data, and deformation acceleration data, including: generating an input vector based on the pressure distribution data, temperature gradient data, and deformation acceleration data; inputting the input vector into a pre-built prediction model; and the prediction model outputting the tire health index. The prediction model includes a long short-term memory network and a random forest classification network. The long short-term memory network performs temporal encoding on the input vector, and the random forest classification network calculates the health index based on the encoding results.

[0011] Based on the aforementioned technical means, in calculating the tire health index, this embodiment first generates a corresponding input vector based on the tire pressure distribution data, temperature gradient data, and deformation acceleration data. This input vector is then input into a pre-constructed prediction model. This prediction model consists of a Long Short-Term Memory (LSTM) network and a Random Forest classification network. The LTM network encodes the temporal features in the input vector, capturing the dynamic changes in tire condition over time. The Random Forest classification network calculates the tire health index based on the encoded feature results. Leveraging the excellent processing capabilities of the LTM network for temporal data, the dynamic correlation information of tire condition is accurately extracted. Simultaneously, the high robustness and computational accuracy of the Random Forest classification network ensure the accuracy of the health index. The synergy of these two approaches avoids the limitations of a single model processing complex data and efficiently outputs reliable health results, providing a precise basis for the subsequent dynamic adjustment of vehicle control parameters, while also meeting the data processing efficiency requirements of in-vehicle scenarios.

[0012] Optionally, an input vector is generated based on pressure distribution data, temperature gradient data, and deformation acceleration data, including: calculating the tire's pressure equalization coefficient based on the pressure distribution data; extracting the temperature gradient data to calculate the tire's temperature risk factor; extracting the deformation acceleration data to calculate the tire's deformation loss index; and generating the input vector based on the pressure equalization coefficient, temperature risk factor, and deformation loss index.

[0013] Based on the aforementioned technical means, this application embodiment, when generating the input vector required for the prediction model, does not directly use the original pressure distribution, temperature gradient, and deformation acceleration data. Instead, it first performs feature quantization on the three types of data: calculating the pressure equalization coefficient reflecting tire pressure uniformity based on the pressure distribution data; extracting the temperature risk factor characterizing tire overheating risk from the temperature gradient data; and obtaining the deformation loss index reflecting tire structural wear through the deformation acceleration data. These three quantified feature parameters are then integrated into the input vector. Transforming complex original multidimensional data into highly targeted key features avoids interference from redundant information in the original data on the model, while allowing the core states of pressure, temperature, and deformation dimensions to be accurately presented through quantified indicators. Simultaneously, the standardized feature vector can better adapt to the subsequent prediction model composed of a long short-term memory network and a random forest, providing high-quality input for time-series coding and health calculation, ultimately improving the accuracy and reliability of tire health index calculation.

[0014] Optionally, the calculation formula for Long Short-Term Memory (LSTM) networks is as follows:

[0015] in, This represents the encoding result at time t. Represents the input vector. This represents the encoding result at time t-1. This represents the weight matrix from the hidden state to the gated unit. This represents the bias vector. Represents the hyperbolic tangent activation function; The formula for calculating a random forest classification network is:

[0016] in, This indicates the health index. Let represent the prediction function of the i-th random forest decision tree, with the final hidden state as the input. The output is the prediction of tire health by the random forest decision tree. Indicates the number of decision trees.

[0017] Optionally, the vehicle parameter adaptive control method further includes: obtaining the actual yaw rate error of the vehicle; and updating the weights of the long short-term memory network of the prediction model based on the actual yaw rate error.

[0018] Based on the aforementioned technical means, this application embodiment, in addition to calculating the health index based on multi-dimensional tire data, introduces the actual yaw rate error of the vehicle (a key parameter reflecting the vehicle's dynamic stability). This error is used as the basis for adjustment, and the weights of the long short-term memory network in the prediction model are optimized in real time, forming a closed-loop optimization mechanism of data acquisition, model calculation, error feedback, and weight update. During the adaptive control of vehicle parameters, the weights of the long short-term memory network are dynamically updated by acquiring this actual yaw rate error, constructing a closed-loop logic of "data perception, model calculation, error feedback, and parameter optimization." The yaw rate error is used to calibrate the long short-term memory network's encoding capability of the tire's temporal characteristics in real time. This avoids the accuracy decay of the model due to long-term operating condition differences such as tire aging and road condition changes, ensuring that the health index always accurately matches the actual state of the tire and vehicle. This allows subsequent adjustments to vehicle control parameters based on the health index to better reflect real driving dynamics, further reducing safety risks caused by insufficient model accuracy and improving vehicle safety assurance capabilities.

[0019] Optionally, before calculating the tire health index based on pressure distribution data, temperature gradient data, and deformation acceleration data, the process includes: aligning the pressure distribution data, temperature gradient data, and deformation acceleration data for data preprocessing, wherein the preprocessing method includes at least one of time alignment processing, pressure wave delay compensation processing, and temperature diffusion rate correction processing.

[0020] Based on the aforementioned technical means, before calculating tire health indicators based on pressure distribution data, temperature gradient data, and deformation acceleration data, this embodiment of the application performs targeted preprocessing on these three types of multi-source sensor data. Specific preprocessing methods include at least one of time alignment processing, pressure wave delay compensation processing, and temperature diffusion rate correction processing. This accurately solves the problem of spatiotemporal misalignment and measurement deviation caused by differences in acquisition mechanisms and physical characteristics (such as pressure signal propagation speed and temperature conduction rate) of multi-source data, ensuring that the preprocessed pressure distribution, temperature gradient, and deformation acceleration data are synchronous in time and accurate in value, forming high-quality model input data. High-quality input data not only reduces the interference of original data errors on subsequent health indicator calculations, helping the model more accurately capture tire state characteristics, thereby improving the accuracy of tire health indicator calculations, but also lays a solid data foundation for subsequent adaptive adjustment of vehicle control parameters based on health indicators.

[0021] A second aspect of this application provides a vehicle parameter adaptive control device, comprising: an acquisition module for acquiring tire pressure distribution data, temperature gradient data, and deformation acceleration data; a calculation module for calculating a tire health index based on the pressure distribution data, temperature gradient data, and deformation acceleration data, and acquiring the vehicle speed and longitudinal acceleration; and a control module for adjusting at least one of a trigger threshold, steering compensation angle, and torque limit of the vehicle electronic stability system based on at least one of the health index, vehicle speed, and longitudinal acceleration, and controlling the vehicle based on at least one of the trigger threshold, steering compensation angle, and torque limit.

[0022] Optionally, the acquisition module is further configured to collect pressure distribution data, temperature gradient data, and deformation acceleration data using a sensor array located within the vehicle's tires.

[0023] Optionally, the sensor array includes an accelerometer, a thermopile sensor, and a piezoelectric sensor, with the piezoelectric sensor generating power from tire deformation to power the sensor array.

[0024] Optionally, the calculation module is further used to calculate the tire health index based on pressure distribution data, temperature gradient data, and deformation acceleration data, including: generating an input vector based on the pressure distribution data, temperature gradient data, and deformation acceleration data, inputting the input vector into a pre-built prediction model, and the prediction model outputting the tire health index. The prediction model includes a long short-term memory network and a random forest classification network. The long short-term memory network performs temporal encoding on the input vector, and the random forest classification network calculates the health index based on the encoding result.

[0025] Optionally, the calculation module is further configured to generate an input vector based on pressure distribution data, temperature gradient data, and deformation acceleration data, including: calculating the tire's pressure equalization coefficient based on the pressure distribution data; extracting the temperature gradient data to calculate the tire's temperature risk factor; extracting the deformation acceleration data to calculate the tire's deformation loss index; and generating the input vector based on the pressure equalization coefficient, temperature risk factor, and deformation loss index.

[0026] Optionally, the calculation formula for Long Short-Term Memory (LSTM) networks is as follows:

[0027] in, This represents the encoding result at time t. Represents the input vector. This represents the encoding result at time t-1. express, express, express; The formula for calculating a random forest classification network is:

[0028] in, Indicators of health status express, This represents the encoding result at time t. express.

[0029] Optionally, it also includes an update module for obtaining the actual yaw rate error of the vehicle and updating the long short-term memory network weights of the prediction model based on the actual yaw rate error.

[0030] Optionally, it also includes a preprocessing module for aligning the pressure distribution data, temperature gradient data, and deformation acceleration data for data preprocessing before calculating the tire health index based on the pressure distribution data, temperature gradient data, and deformation acceleration data. The preprocessing method includes at least one of time alignment processing, pressure wave delay compensation processing, and temperature diffusion rate correction processing.

[0031] A third aspect of this application provides a vehicle, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the program to implement the vehicle parameter adaptive control method as described in the above embodiments.

[0032] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0033] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart of a vehicle parameter adaptive control method provided according to an embodiment of this application; Figure 2 This is a flowchart of a vehicle parameter adaptive control method according to an embodiment of this application; Figure 3 This is a schematic diagram of the hardware system structure provided according to an embodiment of this application; Figure 4 This is a block diagram of a vehicle parameter adaptive control device according to an embodiment of this application; Figure 5 This is a structural schematic diagram of a vehicle provided according to an embodiment of this application. Detailed Implementation

[0034] The embodiments of this application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0035] In related technologies, tire monitoring systems include a sensing layer, a processing layer, a control layer, and a protection layer, but they have the following shortcomings: (1) The sensing layer has a single monitoring dimension and lacks spatial information. Related technologies are based on TPMS (Tire Pressure Monitoring System), which collects basic parameters such as tire pressure and temperature through single-point sensors. It can only identify surface anomalies such as air leakage and overheating, and cannot capture deep health conditions such as internal stress distribution, local deformation (such as tire shoulder delamination and cord breakage) or material fatigue. Indirect TPMS relies on wheel speed difference to judge tire pressure changes, which is prone to misjudgment in situations where multiple tires on the same axle are simultaneously deflated, high-speed driving (>100km / h), or complex road conditions; and it lacks spatial distribution information of pressure and temperature, cannot distinguish the differences in the condition of different areas of the tire, and has a very high risk of missing local damage, making it difficult to meet the needs of refined monitoring.

[0036] (2) Processing Layer: Poor real-time data processing and weak feature fusion capabilities. Relying on physical models (such as finite element simulation) or cloud computing, a digital tire model needs to be built in advance. Data transmission and analysis have minute-level delays, which cannot meet the requirements of vehicle real-time control for second-level and millisecond-level responses. Signal processing methods (such as FFT (Fast Fourier Transform)) can only extract simple time-domain and frequency-domain features. They are not capable of processing high-dimensional nonlinear data (such as cord breakage harmonics in vibration spectrum and the spatiotemporal correlation between pressure and temperature), making it difficult to accurately extract damage features under complex working conditions. Moreover, simple threshold rules (such as triggering a warning when tire pressure drops by 10%) cannot distinguish between slow leaks and normal fluctuations caused by ambient temperature, resulting in a high false alarm rate.

[0037] (3) Control Layer: Early warning and control are disconnected, and parameters cannot be dynamically adapted. All tire monitoring technologies adopt independent early warning modes, only outputting abnormal information to the driver through pop-ups and beeps, without linkage with the vehicle's active control system (such as ESP (Electronic Stability Program) and EPS (Electric Power Steering)). Even if some high-end models support emergency warnings before tire blowout, they cannot automatically adjust control parameters such as steering compensation angle, torque limit, and ESP trigger threshold, relying entirely on manual driver response; and at high speeds (such as 120km / h), the reaction time left for the driver after a tire blowout is less than 3 seconds, and the lag in manual intervention can easily lead to safety accidents, forming an execution gap of warning without intervention. Vehicle stability control strategies (such as ABS (Anti-lock Braking System) intervention threshold and ESP trigger logic) are generally designed based on fixed calibration values, without considering the impact of real-time tire health status (such as wear and aging) or dynamic road conditions (such as reduced grip on wet roads). For example, severely worn tires have a significantly reduced coefficient of adhesion on wet and slippery surfaces, but the ESP threshold still uses the parameters of new tires, which may lead to excessive intervention or insufficient response during braking; moreover, static parameters cannot adapt to the dynamic changes of slow tire pressure and tire temperature rise during driving, and the control strategy deviates from the actual needs, making it difficult to ensure driving safety under all working conditions.

[0038] (4) Protection layer: Limited energy supply and insufficient reliability in harsh environments. Sensors generally rely on lithium thionyl chloride batteries for power supply, with a lifespan of only 2 - 5 years and high replacement costs (including hardware replacement and recalibration), which severely restricts the deployment of high-density sensing arrays. If tire pressure distribution monitoring is to be achieved (such as 9-point sampling), the battery replacement frequency and maintenance costs of multiple sensors will increase significantly. Although attempts have been made to use new energy supply methods such as piezoelectric power generation and solar energy, due to the limited internal space of the tire and the energy conversion efficiency, the generated power is generally low (for example, the average output is only 1 mW when driving at 105 km / h), which can only meet the low sampling rate data transmission of the basic TPMS and cannot support the acquisition of millisecond-level pressure distribution and deformation acceleration data. The inside of the tire is in an environment with a temperature above 120°C and a centrifugal force above 10000g for a long time, and the reliability of MEMS (Micro-Electro-Mechanical Systems) sensors drops significantly. The mean time between failures can only meet the basic tire pressure and temperature monitoring, and it is prone to failure under complex working conditions such as off-road and heavy load. The continuous change of the sensor antenna direction caused by tire rotation leads to wireless transmission signal attenuation and is easily interfered by devices such as in-vehicle Bluetooth and Wi-Fi, resulting in a high data loss rate. In addition, the deformation and extrusion of the tire rubber may damage the sensor package, further reducing the overall reliability of the system and making it difficult to meet the requirements of long-term stable monitoring.

[0039] In summary, in the related technologies, the sensing layer misses the detection of deep states due to single dimension and lack of spatial information; the processing layer cannot provide accurate decision-making basis due to poor real-time performance and weak fusion ability; the control layer makes the warning and intervention disjointed due to the lack of closed-loop and static parameters; the protection layer restricts the scale application due to limited energy supply and insufficient reliability. Ultimately, the related technologies can only achieve basic abnormal alarms and cannot meet the vehicle's demand for real-time health assessment - dynamic control adaptation.

[0040] Next, the vehicle parameter adaptive control method, device, and vehicle of the embodiments of the present application will be described with reference to the accompanying drawings. Aiming at the problems in the related technologies mentioned in the above background technology that the vehicle monitoring system can only sense basic tire abnormalities and cannot real-time link and control the vehicle to avoid safety risks. The present application provides a tire health monitoring system based on multi-source sensing fusion, which installs a piezoelectric-thermoelectric composite sensor array in the tire to collect pressure distribution, temperature gradient, and dynamic deformation data in real time; uses a LSTM-RF (Long Short-Term Memory-Random Forest) hybrid algorithm to construct a THI (Tire Health Index) model; combines vehicle dynamics parameters to establish a closed-loop control strategy to dynamically adjust system parameters such as ESP / ABS. The system includes a self-powered module that uses tire deformation to generate electricity for lifelong maintenance-free. Verified by the bench test, it can reduce the flat tire accident rate by 83% and extend the tire life by 15%.

[0041] Specifically, Figure 1 This is a flowchart of a vehicle parameter adaptive control method provided in an embodiment of this application.

[0042] like Figure 1 As shown, the vehicle parameter adaptive control method includes the following steps: In step S101, tire pressure distribution data, temperature gradient data, and deformation acceleration data are acquired.

[0043] It is understood that the pressure distribution data in the embodiments of this application can reflect the stress differences in different areas of the tire (such as the tire shoulder and the center of the tread), the temperature gradient data can capture the uneven heat distribution in local areas of the tire (such as the inner liner and the tread), and the deformation acceleration data can sense the material deformation trend under the dynamic working conditions of the tire (such as cord vibration and tire fatigue). The three together constitute the data basis for reflecting the abnormality of the tire surface and the deep health status.

[0044] Therefore, this technology overcomes the limitations of tire monitoring systems that rely solely on basic data such as tire pressure and temperature at a single point. It solves the problem of missing deeper conditions due to the single data dimension in related technologies. Compared to TPMS, which can only identify surface anomalies such as leaks and overheating, the combination of these three types of data can cover deeper risks such as localized stress concentration (e.g., precursors to tire shoulder delamination), regional overheating (e.g., high temperatures on the tire sidewall near the brake drum), and abnormal dynamic deformation (e.g., changes in the deformation spectrum caused by cord breakage), avoiding the omission of hidden dangers due to incomplete data. Secondly, it provides a data foundation for subsequent calculations of tire health indicators. Additionally, the length... The hybrid model of the periodic memory network needs to accurately assess the health status based on the spatiotemporal correlation characteristics of multi-dimensional data (such as the synchronicity of abnormal pressure distribution and local temperature rise). Single data cannot support refined analysis, and obtaining multiple types of data ensures the accuracy of health index calculation. Finally, when dynamically adjusting parameters such as ESP trigger threshold and EPS steering compensation angle, it is necessary to base it on the actual condition of the tire. The three types of data can fully reflect the actual performance of the tire (such as grip and structural strength), avoiding the disconnect between the control strategy and the actual condition of the tire due to data loss, and ensuring the rationality and safety of control adjustment.

[0045] In this embodiment of the application, a sensor array is installed inside the tire of the vehicle to collect pressure distribution data, temperature gradient data, and deformation acceleration data.

[0046] It is understood that the embodiments of this application can collect tire pressure distribution data, temperature gradient data, and deformation acceleration data by setting a sensor array inside the vehicle tire. This overcomes the limitation of single-point sensors in related technologies that can only acquire local and isolated data. With the help of the spatially distributed layout of the array, it can reflect the state differences of different parts of the tire (such as the pressure difference between the tire shoulder and the tire crown, and the temperature changes between the inner and outer sides), avoiding the omission of local anomalies (such as local wear and stress concentration). It can also provide comprehensive and accurate basic support for the subsequent calculation of tire health indicators through multi-dimensional data. At the same time, cross-validation of multi-sensor data can reduce the impact of single sensor failure or error on the perception results, further improving data reliability.

[0047] In this embodiment, the sensor array includes an accelerometer, a thermopile sensor, and a piezoelectric sensor. The piezoelectric sensor generates electricity using tire deformation to power the sensor array.

[0048] It is understood that the sensor array in this application embodiment is specifically configured as a combination of an accelerometer, a thermopile sensor and a piezoelectric sensor. The accelerometer is responsible for collecting tire deformation acceleration data, the thermopile sensor focuses on acquiring temperature gradient data, and the piezoelectric sensor undertakes the task of acquiring pressure distribution data. At the same time, the piezoelectric sensor can also generate electricity by utilizing the natural deformation of the tire during driving, directly providing power to the entire sensor array.

[0049] Therefore, this embodiment of the application ensures the accuracy and specificity of multi-dimensional tire status data acquisition through a variety of sensors with clearly defined functions. It also enables the piezoelectric sensor to achieve integrated sensing and power supply, eliminating the reliance on batteries for internal tire sensors in related technologies and avoiding the problems of short battery life and high replacement and maintenance costs. Simultaneously, the continuous deformation of the tire ensures the stability of the power supply, effectively preventing monitoring failures due to power interruptions and providing support for the long-term stable operation of the sensor array.

[0050] Specifically, in this embodiment of the application, data such as pressure distribution, temperature gradient, and deformation acceleration can be simultaneously collected by a piezoelectric array, thermopile, and MEMS accelerometer embedded in the tire.

[0051] In step S102, the tire health index is calculated based on pressure distribution data, temperature gradient data, and deformation acceleration data, and the vehicle speed and longitudinal acceleration are obtained.

[0052] It is understood that the embodiments of this application can calculate a unified tire health index (THI) through the model, transforming the complex tire condition into a quantifiable and comparable value; on the other hand, by simultaneously acquiring vehicle speed and longitudinal acceleration, the subsequent control strategy can not only consider whether the tire itself is healthy, but also combine the actual working conditions of whether the vehicle is currently driving fast or accelerating, thus avoiding deviation from the real driving scenario.

[0053] Compared to related technologies that can only issue alarms for a single parameter (such as "low tire pressure" or "high temperature"), this application can intuitively distinguish tire health levels through THI quantification. Related technologies rely on single data points, which are prone to misjudgment. For example, high tire temperature could be due to normal high-speed driving or abnormal overheating caused by damage. THI, on the other hand, combines pressure distribution (whether there is localized pressure concentration) and deformation acceleration (whether there is abnormal vibration) for comprehensive calculation, which can accurately distinguish tire conditions and avoid false alarms from single-parameter monitoring. Obtaining vehicle speed and longitudinal acceleration is because the impact of tire health status on driving safety is related to operating conditions. By obtaining these two parameters, it is possible to adapt to different driving scenarios and achieve adaptive vehicle control.

[0054] In this embodiment of the application, before calculating the health index of the tire based on the pressure distribution data, temperature gradient data, and deformation acceleration data, the method includes: aligning the pressure distribution data, temperature gradient data, and deformation acceleration data for data preprocessing, wherein the preprocessing method includes at least one of time alignment processing, pressure wave delay compensation processing, and temperature diffusion rate correction processing.

[0055] It is understood that, before calculating the tire health index based on pressure distribution data, temperature gradient data, and deformation acceleration data, the embodiments of this application will first perform targeted preprocessing on these three types of multi-source sensor data. The specific preprocessing methods include at least one of time alignment processing, pressure wave delay compensation processing, and temperature diffusion rate correction processing.

[0056] Therefore, the embodiments of this application accurately solve the problems of spatiotemporal misalignment and measurement deviation caused by different acquisition mechanisms and physical characteristics of multi-source data. This ensures that the preprocessed pressure distribution, temperature gradient, and deformation acceleration data are synchronized in time and accurate in numerical values, forming high-quality model input data. High-quality input data not only reduces the interference of original data errors on subsequent health index calculations, but also helps the model more accurately capture tire state characteristics, thereby improving the accuracy of tire health index calculations. Furthermore, it lays a solid data foundation for subsequent adaptive adjustment of vehicle control parameters based on health indices.

[0057] Specifically, the spatiotemporal alignment preprocessing is used to compensate for differences in the physical positions of the sensors: pressure wave delay compensation: Temperature diffusion rate correction: The control conditions and parameters for spatiotemporal alignment preprocessing are shown in Table 1.

[0058] Table 1

[0059] The spatiotemporal alignment preprocessing operation compensates for pressure signal delay and corrects temperature signal deviation, resolving the issue of data asynchrony and inaccuracy caused by multiple sensors installed in different locations. This is because the sensors are mounted in different positions on the tire (e.g., sensor...). On the tire shoulder, sensor On the tire tread, the distance between the two is Pressure signal from Transmitted to It takes a little time, and this time difference is... , This refers to the propagation speed of pressure waves within the rubber, which can be fixed at 240±10 m / s (this is an inherent property of rubber materials). Inaccurate values ​​(e.g., deviating too much from 240m / s) will lead to Calculation errors can prevent the compensation of pressure data time differences caused by sensor position variations, ultimately affecting the effectiveness of spatiotemporal alignment preprocessing (multi-sensor data asynchrony); in addition, local temperature changes in the tire (such as sidewall warming near the brakes) will slowly spread to the surrounding area, but the rate of temperature change will decrease during the diffusion process, for example, the originally measured temperature gradient ( (This refers to the temperature difference per unit distance), and when it is transmitted to a distant sensor, the actual effective gradient will become smaller. It is the thermal diffusion attenuation coefficient, which is fixed at 0.05 (this value is calibrated based on the thermal conductivity characteristics of rubber).

[0060] For example, if the two sensors are 0.05 meters apart, then After determining the time difference, the pressure data from one of the sensors is advanced or delayed by 0.2 milliseconds to align the pressure data from the two sensors in time. This is the maximum temperature gradient threshold, used to determine if there is a risk of localized overheating in the tire. After correction by the temperature diffusion rate, it represents the actual temperature gradient of the tire. (The corrected temperature gradient) if it exceeds (15℃ / cm) indicates an abnormal, rapid change in localized tire temperature, which may be a precursor to damage such as accelerated rubber aging or decreased heat resistance of the cords. In this case, the system will trigger an emergency mode (such as actively reducing torque and prompting the driver to slow down to prevent localized overheating that could lead to a tire blowout or structural failure). This indicates that the tire temperature distribution is uniform, with no risk of localized overheating, and the pre-processed temperature data can be used for subsequent health assessments. This application does not specify the exact values ​​for the parameter control conditions in Table 1.

[0061] In this embodiment, the tire health index is calculated based on pressure distribution data, temperature gradient data, and deformation acceleration data. This includes: generating an input vector based on the pressure distribution data, temperature gradient data, and deformation acceleration data; inputting the input vector into a pre-built prediction model; and outputting the tire health index by the prediction model. The prediction model includes a long short-term memory network and a random forest classification network. The long short-term memory network performs temporal encoding on the input vector, and the random forest classification network calculates the health index based on the encoding result.

[0062] It is understood that, in calculating the tire health index, the embodiments of this application first generate a corresponding input vector based on the tire pressure distribution data, temperature gradient data, and deformation acceleration data, and then input the input vector into a pre-built prediction model. The prediction model consists of a long short-term memory network and a random forest classification network. The long short-term memory network is responsible for encoding the temporal features in the input vector to capture the dynamic changes in tire condition over time, while the random forest classification network calculates the tire health index based on the encoded feature results.

[0063] Therefore, the embodiments of this application can leverage the excellent processing capabilities of long short-term memory networks for time-series data to accurately extract dynamic correlation information of tire status, while relying on the high robustness and computational accuracy of random forest classification networks to ensure the accuracy of health indicators. The synergy of the two avoids the limitations of a single model when processing complex data, and can efficiently output reliable health results, providing an accurate basis for the dynamic adjustment of subsequent vehicle control parameters, while also adapting to the data processing efficiency requirements of in-vehicle scenarios.

[0064] For example, embodiments of this application can calculate the THI health index using a hybrid model of Long Short-Term Memory Network and Random Forest. Under the real-time requirements of in-vehicle scenarios (completing within 50ms), the input is a 15-dimensional feature vector (multi-source data), and the output is a quantized THI∈[0,1]. At the same time, parameter constraints ensure computational accuracy and reliability.

[0065] In this embodiment of the application, the input vector is generated based on pressure distribution data, temperature gradient data, and deformation acceleration data, including: calculating the tire's pressure equalization coefficient based on the pressure distribution data; extracting the temperature gradient data to calculate the tire's temperature risk factor; extracting the deformation acceleration data to calculate the tire's deformation loss index; and generating the input vector based on the pressure equalization coefficient, temperature risk factor, and deformation loss index.

[0066] It is understood that, in generating the input vector required for the prediction model, the embodiments of this application do not directly use the original pressure distribution, temperature gradient and deformation acceleration data, but first perform feature quantization on the three types of data respectively: calculate the pressure balance coefficient reflecting the uniformity of tire pressure based on the pressure distribution data, extract the temperature risk factor representing the risk of tire overheating from the temperature gradient data, obtain the deformation loss index reflecting the tire structural loss through the deformation acceleration data, and then integrate the three quantized feature parameters into the input vector.

[0067] Therefore, the embodiments of this application can transform complex raw multidimensional data into highly targeted key features, which not only avoids the interference of redundant information in the raw data on the model, but also allows state data such as pressure, temperature, and deformation dimension to be accurately presented through quantitative indicators. At the same time, the standardized feature vectors can better adapt to the prediction model composed of the subsequent long short-term memory network and random forest, providing high-quality input for time-series coding and health calculation, and ultimately improving the calculation accuracy and reliability of tire health indicators.

[0068] Specifically, the feature extraction formulas for the pressure equilibrium coefficient, temperature risk factor, and deformation damage index are as follows:

[0069] In this embodiment of the application, the calculation formula for the Long Short-Term Memory network is as follows:

[0070] in, This represents the encoding result at time t. Represents the input vector. This represents the encoding result at time t-1. The weight matrix represents the weights from the hidden state to the gated unit, controlling the influence of historical information. This represents the bias vector, used to adjust the linear transformation result of the neural network at each layer. This represents the hyperbolic tangent activation function, which compresses the input values ​​to the interval [-1, 1], enabling the neural network to better express nonlinear relationships and prevent gradient divergence. The formula for calculating a random forest classification network is:

[0071] in, This indicates the health index. Let represent the prediction function of the i-th random forest decision tree, with the input being the final hidden state output by the LSTM model. The output is the independent prediction of tire health by this decision tree. Indicates the number of decision trees.

[0072] It should be noted that the hidden state in LSTM ( , This is a time-series information integration unit, whose function is to integrate historical hidden states. With current input ,through Weight adjustment Bias correction and After nonlinear mapping, the current time-time feature representation is output; where It is a local feature at a single time step. ( The final step of the sequence (the global feature summary across all time series) provides input for the random forest. In the random forest formula... This refers to 100 ensemble decision trees, which are analyzed based on... Calculate the mean of the predicted values ​​and output them. (This is a regression task) to reduce single-tree error and balance prediction accuracy with onboard real-time performance.

[0073] Specifically, since a single model cannot simultaneously meet the vehicle requirements of capturing time-series trends, fast computation, and high accuracy, LSTM is good at processing time-series data (such as tire pressure fluctuations over time and temperature changes), but it is not robust enough for the final classification / regression of multi-dimensional features and is prone to overfitting. RF is good at robust computation of multiple features (strong noise resistance), but it cannot capture the temporal correlation of data.

[0074] Therefore, this application uses LSTM to first encode the time dynamic information (such as pressure changes over 10 consecutive time moments) in the 15-dimensional input vector into fixed-dimensional features; RF then uses this encoding result to output THI through integrated calculation of multiple trees, which solves the problem of capturing time-series trends, ensures computational accuracy and anti-interference capability, and the 50ms computation time adapts to the requirements of real-time vehicle control (vehicle control requires millisecond-level response). The control conditions and parameters for calculating the THI health index are shown in Table 2.

[0075] Table 2

[0076] The number of hidden units in the LSTM determines the... The 128-dimensionality of the algorithm is sufficient to accommodate the temporal features of a 15-dimensional input vector (avoiding information loss) without slowing down computation due to excessive dimensionality (ensuring completion within 50ms), representing a balance between feature extraction capability and real-time performance. Random Forest is an algorithm that integrates multiple decision trees, including the number of RF trees. Different vehicle models have different computing capabilities for their onboard chips. By adjusting the number of trees (e.g., 80 trees for weaker computing power, 100 trees for stronger computing power), a flexible balance between computation time and accuracy can be achieved, ensuring that different vehicle models can meet the 50ms real-time requirement. Confidence score represents the model's confidence in its own output (e.g., a confidence score of 0.9 indicates that the model is 90% certain that the THI is accurate). If the confidence score is <0.85, it indicates that the data may be abnormal (e.g., sensor malfunction), triggering redundancy checks to avoid adjusting vehicle control parameters with incorrect THI, which could lead to safety risks.

[0077] In this embodiment of the application, the vehicle parameter adaptive control method further includes: obtaining the actual yaw rate error of the vehicle; and updating the long short-term memory network weights of the prediction model based on the actual yaw rate error.

[0078] It is understood that, based on the calculation of health indicators using multi-dimensional tire data, this application's embodiments introduce the actual yaw rate error (a key parameter reflecting vehicle dynamic stability) and use this error as an adjustment basis to optimize the weights of the long short-term memory network in the prediction model in real time, forming a closed-loop optimization mechanism of data acquisition, model calculation, error feedback, and weight update. During the vehicle parameter adaptive control process, the weights of the long short-term memory network are dynamically updated by acquiring this actual yaw rate error, constructing a closed-loop logic of "data perception, model calculation, error feedback, and parameter optimization," and leveraging the yaw rate error to calibrate the long short-term memory network's encoding capability of the tire state temporal characteristics in real time.

[0079] Therefore, the embodiments of this application can avoid the accuracy decay of the model due to long-term operating condition differences such as tire aging and road condition changes, and ensure that the health index always accurately matches the actual state of the tire and the vehicle. This allows the subsequent adjustment of vehicle control parameters based on the health index to better fit the real driving dynamics, further reducing the safety risks caused by insufficient model accuracy and improving the vehicle's safety assurance capabilities.

[0080] Specifically, the formula for online correction of the THI model is:

[0081] Among them, based on the actual yaw rate error Backpropagation updates the LSTM weights.

[0082] In step S103, at least one of the trigger threshold, steering compensation angle, and torque limit of the vehicle electronic stability system is adjusted based on at least one of the health index, vehicle speed, and longitudinal acceleration, and the vehicle is controlled based on at least one of the trigger threshold, steering compensation angle, and torque limit.

[0083] It is understood that the embodiments of this application differ from the related technologies that rely on fixed preset parameters. Instead, based on at least one of the health index (THI), vehicle speed, and longitudinal acceleration obtained in step S102, the key control parameters such as the trigger threshold of the vehicle electronic stability system, steering compensation angle, and torque limit are flexibly adjusted, and the adjusted parameters are directly used to control the vehicle, so that the control strategy is adapted to the actual tire performance and the current driving scenario.

[0084] In related technologies, parameters such as the ESP trigger threshold and steering compensation angle of a vehicle are calibrated based on new tires or ideal conditions. Once the tires wear out or age, the actual grip and structural strength will weaken, but the control parameters remain unchanged, which may lead to two risks: either the intervention is too late (such as the ESP being triggered too late, and the vehicle has already skidded), or the intervention is insufficient (such as insufficient steering compensation, resulting in handling deviation).

[0085] In this application, if the health index is low, the ESP trigger threshold will be actively lowered (so that the ESP can intervene earlier) and the steering compensation angle will be increased (to counteract the deviation caused by tire deformation), ensuring that the control force matches the actual capacity of the tire. In addition, by combining vehicle speed and longitudinal acceleration, parameters can be dynamically adjusted: during high-speed rapid acceleration, not only will the ESP threshold be tightened, but the torque limit will also be reduced (reducing power output and tire load), avoiding unilateral control that leads to insufficient safety redundancy or a deterioration in the driving experience.

[0086] Therefore, the embodiments of this application automatically adjust and execute control parameters based on monitoring data (such as ESP active braking to correct the trajectory of one side wheel), realizing proactive intervention and early resolution of tire abnormal risks. This solves the problem in related technologies where tire condition monitoring systems are limited to abnormal warning functions and do not form a linkage with vehicle active control systems (such as ESP, EPS), requiring manual intervention and response delays that lead to vehicle loss of control risks. This significantly improves the driving stability and active safety of vehicles under abnormal tire conditions.

[0087] Specifically, based on the tire health index (THI) and vehicle speed... Longitudinal acceleration The control parameters adapted to the current state are generated through a quantization model, specifically including: ESP threshold, formula is The logic is that the lower the THI (the worse the tire health) and the higher the vehicle speed (the higher the risk level). The smaller the value, the lower the threshold for the ESP system to trigger intervention, enabling a safety redundancy design that allows for earlier intervention when the system is in poor health or operating at high speeds. Steering compensation gain, the formula is: ,in Its logic is based on the target turning angle Compared with actual steering angle The deviation is taken as input, and steering compensation is achieved through PD control; the lower the THI (the more significant the tire deformation). , The smaller the value, the less likely it is to cause handling fluctuations due to overcompensation, thus ensuring the smoothness of steering corrections.

[0088] The rules for limiting power torque are as follows: The logic is to limit the engine output torque according to the THI rating. The lower the health rating, the stricter the torque limit, so as to avoid the tires from being damaged or slipping due to overload (such as rapid acceleration), thereby matching the actual load-bearing capacity of the tires.

[0089] After the parameters are generated, implementation is achieved through actuator-based coordinated control. Specifically, the generated control parameters are converted into execution commands via the vehicle's CAN bus and sent to the control module to enable coordinated action across multiple systems: the ESP hydraulic unit, according to... Threshold adjustment of braking force distribution ratio to each wheel (e.g., enhanced braking of one side of wheel to correct sideslip); EPS motor based on Calculate the steering compensation angle and output additional steering torque to counteract tire deformation-induced pull-off; EMS follows... Limit engine output torque to reduce tire load. To improve control accuracy, a real-time feedback mechanism is introduced: collect the vehicle's actual driving state (such as yaw rate and tire slip ratio), compare it with the preset target value (ideal state calculated based on THI and operating conditions) to obtain the deviation value; based on the deviation, correct the feature weights of the THI model (such as increasing the weight of temperature gradient data under wet road surface) to optimize the health assessment accuracy, and then dynamically adjust the control parameters to form an adaptive loop of assessment-control-correction.

[0090] Finally, continuous operation of the entire chain is ensured through autonomous energy supply: Energy harvesting: Utilizing the deformation of the tire during driving (such as tire tread compression upon contact with the ground) to drive piezoelectric materials to generate electricity (piezoelectric effect formula: piezoelectric coefficient, For stress, (Area); Energy storage: Electrical energy is stored in supercapacitors (to meet the needs of high-frequency charging and discharging); Power supply objects: Powering the sensor array and edge computing nodes (local THI model calculations), ensuring the stability of the self-powered module. The control conditions and parameters for the dynamic mapping of vehicle parameters are shown in Table 3, which are supplementary constraints to the above quantitative model.

[0091] Table 3

[0092] in, The dynamic range is to avoid Too low a pressure (e.g., <0.3g) can lead to excessive ESP intervention (frequent braking, affecting driving smoothness), and also avoids... Excessive amounts (e.g., >0.8g) lead to intervention being too late (missing the opportunity for correction); It is a proportional coefficient that controls the vehicle's steering sensitivity, determining the ratio of the driver's steering wheel angle to the vehicle's actual steering angle: when =1.0 (Normal condition, THI≥0.5, good tire health): Steering sensitivity is normal; when the driver turns the steering wheel 10°, the vehicle actually turns 10°, with timely response, consistent with normal driving habits; when THI<0.5 (poor tire health, such as uneven tread wear or tire deformation): When the torque limit is reduced to 0.5~0.9 (e.g., 0.5), steering sensitivity is downgraded; if the driver turns the steering wheel 10°, the vehicle only actually turns 5°, and the response becomes sluggish. The torque limiter activates when THI < 0.4 (≤50% output): this clearly defines the triggering conditions for strict torque limiting. It only activates when tire health enters a high-risk range, preventing torque limiting from affecting the driving experience when tire health is still acceptable.

[0093] It should be noted that the steering compensation gain ( ) and steering gain ( All of these aim to address steering problems caused by poor tire health. When THI is low, tire deformation (such as shoulder delamination) can cause steering drift, or uneven grip can lead to abnormal steering response. and By addressing both deviation correction and response limitation, we can jointly ensure safe steering. Correcting existing deviations is achieved through PD control (proportional-derivative control). The input is the deviation between the target steering angle and the actual steering angle, and the output is the steering angle / torque that needs to be supplemented. It's to prevent the deviation from becoming too large when the tire health is too low. It will automatically descend. At this time, when the driver turns the steering wheel (e.g., to turn 10°), the vehicle will only turn 5°. Even if the tire grip is insufficient, it will not cause oversteering (vehicle fishtailing) due to excessive steering, thus reducing the possibility of deviation from the source.

[0094] The vehicle parameter adaptive control method proposed in this application acquires tire pressure distribution data, temperature gradient data, and deformation acceleration data. Combined with real-time vehicle speed and longitudinal acceleration, a health index reflecting the true health state of the tire is first calculated. Then, based on this index and vehicle dynamic parameters, at least one of the trigger threshold, steering compensation angle, and torque limit of the vehicle's electronic stability system is dynamically adjusted to achieve precise vehicle control. This method overcomes the limitations of related technologies that rely on fixed parameter calibration for vehicle control. It can adapt the control strategy promptly according to the actual health condition of the tire (such as abnormal local stress or potential damage), avoiding the risk of loss of control due to parameters not being dynamically adjusted with operating conditions. It also allows the control logic to conform to real-time driving conditions, reducing over-control or under-control issues, while enhancing the coordination between the tire and the vehicle, indirectly extending tire lifespan.

[0095] The process of the vehicle parameter adaptive control method will be described below through a specific embodiment, such as... Figure 2 As shown, the specific steps are as follows: In step one, tire multi-source data acquisition is performed by Figure 3 Once the sensing layer at the tire end is completed, a piezoelectric sensing array (piezoelectric ceramic array) (PZT-5H, 9-point grid, 0.1kPa resolution) captures the tread stress distribution, a thermopile sensor (TS318-1B03, 90° field of view, ±0.5℃ accuracy) monitors the shoulder temperature gradient, and a MEMS accelerometer (ADXL357, ±40g range, 500Hz bandwidth) analyzes the deformation frequency domain characteristics. This multi-dimensional data provides a foundation for subsequent health assessments.

[0096] In step two, the self-powered module relies on... Figure 3 The self-powered system (pressure generator module and supercapacitor bank) is implemented through a tire-end power management module connecting the two. The pressure generator module (piezoelectric generator: The system generates electricity using tire deformation (with a power generation efficiency of >85% at 120℃). After being regulated by the power management module, the electricity is used to directly power the sensing layer and edge preprocessing unit. At the same time, the electricity is stored in a supercapacitor bank (Maxwell120F, cycle life >1 million times) to provide continuous and stable energy support for the system.

[0097] In step three, the spatiotemporal alignment preprocessing is performed by... Figure 3 The edge preprocessing unit at the tire end performs signal conditioning circuitry in conjunction with a 24-bit ADC to perform spatiotemporal alignment and noise reduction on the multi-source data (pressure data sampling rate 1kHz, temperature data sampling rate 100Hz) collected by the sensing layer, ensuring that data of different types and time series can be uniformly input into the subsequent calculation stage.

[0098] In step four, the THI health index is calculated by... Figure 3 Once the core unit of the central control domain is completed, multi-source data is transmitted to the central control domain via the CANFD bus. It is first integrated by the multi-source data fusion module, and then input into the THI computing engine equipped with an ARM Cortex-M7+NPU accelerator (4 GOPS computing power, <2W power consumption). The THI value, which reflects the health status of the tire, is obtained through real-time calculation using the LSTM-RF hybrid model.

[0099] In step five, the vehicle parameter dynamic mapping is performed by... Figure 3 The dynamic parameter mapper in the central control domain is based on the THI value and the vehicle's real-time operating conditions (such as vehicle speed and acceleration). The dynamic parameter mapper generates control commands such as ESP trigger threshold, steering compensation angle, and torque limit ratio that are adapted to the current tire condition by switching online and performing matrix operations on 20 built-in control curves.

[0100] In step six, the actuator coordinated control is performed by... Figure 3 The vehicle's execution layer is completed by multiple actuators. The ESP hydraulic unit adjusts the wheel braking force distribution according to the trigger threshold (response delay <10ms), the EPS brushless motor drives the steering rack to correct deviation according to the steering compensation angle, and the EMS torque controller adjusts the throttle output according to the torque limit ratio. The multi-system collaboration ensures the vehicle's driving stability.

[0101] In step seven, closed-loop feedback optimization is achieved. Figure 3 The system implements wheel speed feedback, steering angle feedback, and other links to collect the actual driving status of the vehicle (such as yaw rate, tire slip ratio, and actual steering angle), and sends it back to the multi-source data fusion module and THI computing engine in the central control domain. This process optimizes the model feature weights (such as increasing the weight of temperature gradient data under wet and slippery road conditions), and then dynamically adjusts the control parameters to form an adaptive cycle of evaluation-control-correction.

[0102] In summary, the embodiments of this application have at least the following beneficial effects: (1) Safety Performance Dimension. Proactive Accident and Risk Prevention: Reduced tire blowout accident rate, with early warnings for localized damage such as tire shoulder delamination and cord breakage; reduced risk of loss of control on slippery roads; in the event of a sudden tire blowout, dynamic control measures such as steering compensation and torque limiting provide drivers with time to recover. From Passive Warning to Active Control: Upgrading tires to active safety control units, dynamically adjusting vehicle parameters such as ESP and steering based on their health status, proactively mitigating tire abnormality risks, and filling the gap in the "warning without intervention" approach.

[0103] (2) Economic Benefits. Tire Life Cycle and Cost Optimization: Extended tire life (wear leveling and on-demand maintenance); self-powered design using piezoelectric power generation and supercapacitors, achieving lifetime maintenance-free operation and eliminating the need for battery replacement. Energy Efficiency Improvement: Improved fuel efficiency, reducing vehicle operating energy consumption; the self-powered system supports high-frequency sampling while requiring no additional maintenance costs, reducing long-term operating costs.

[0104] (3) Technological Innovation Dimension. Multi-dimensional Perception and High-Precision Assessment: Sensor Fusion: Through a 9-point piezoelectric array, a 4-point thermopile, and a MEMS accelerometer, pressure distribution, temperature gradient, and deformation harmonics (5-10th order) are captured in real time, providing comprehensive perception of tire structure and thermal state. Algorithm Model: Utilizing an LSTM-RF hybrid model, vehicle-mounted deployment can output THI (accuracy > 92%) within 50ms; a remaining life prediction model is established (accuracy > 92%), achieving precise management throughout the entire lifecycle. Response Speed: THI response speed < 200ms, 40 times faster than related technologies. Energy and Deployment Breakthrough: Employing high-temperature resistant piezoelectric materials ( It features power generation efficiency >85% at 120℃ and supercapacitor energy storage, supports 500Hz high-frequency sampling, solves the energy bottleneck of high-density sensing, and achieves lifelong maintenance-free system operation.

[0105] (4) Industry and Ecosystem Dimensions. Autonomous Driving and Intelligent Transportation: Provide mechanical redundancy for autonomous driving, supplement the safety dimension through real-time tire health perception; support V2X sharing of road pothole data (accuracy ±0.3m), contributing to the construction of an intelligent transportation ecosystem. Environmental Protection and Resource Sustainability: Reduce rubber waste, reduce rubber resource waste by extending tire life and maintaining as needed, and promote the green development of the automotive industry.

[0106] (5) System architecture dimension. Achieve full-link closed loop of perception → real-time decision-making → dynamic control: For the first time, integrate multi-source sensor fusion (pressure / temperature / deformation), LSTM-RF real-time evaluation, and dynamic mapping of vehicle parameters (coupled with vehicle speed / acceleration), upgrading the tire from a passive monitoring component to an active safety control unit, solving core pain points such as control delay, open-loop strategy, and energy dependence.

[0107] Next, the vehicle parameter adaptive control device proposed according to the embodiments of this application is described with reference to the accompanying drawings.

[0108] Figure 4 This is a block diagram of a vehicle parameter adaptive control device according to an embodiment of this application.

[0109] like Figure 4 As shown, the vehicle parameter adaptive control device 40 includes: an acquisition module 401, a calculation module 402, and a control module 403.

[0110] The acquisition module 401 is used to acquire tire pressure distribution data, temperature gradient data, and deformation acceleration data; the calculation module 402 is used to calculate the tire health index and acquire the vehicle speed and longitudinal acceleration based on the pressure distribution data, temperature gradient data, and deformation acceleration data; the control module 403 is used to adjust at least one of the trigger threshold, steering compensation angle, and torque limit of the vehicle electronic stability system based on at least one of the health index, vehicle speed, and longitudinal acceleration, and to control the vehicle based on at least one of the trigger threshold, steering compensation angle, and torque limit.

[0111] In this embodiment of the application, the acquisition module 401 is further used to collect pressure distribution data, temperature gradient data and deformation acceleration data using a sensor array installed inside the tires of the vehicle.

[0112] In this embodiment, the sensor array includes an accelerometer, a thermopile sensor, and a piezoelectric sensor. The piezoelectric sensor generates electricity using tire deformation to power the sensor array.

[0113] In this embodiment, the calculation module 402 is further configured to calculate the tire health index based on pressure distribution data, temperature gradient data, and deformation acceleration data, including: generating an input vector based on the pressure distribution data, temperature gradient data, and deformation acceleration data; inputting the input vector into a pre-built prediction model; and outputting the tire health index by the prediction model. The prediction model includes a long short-term memory network and a random forest classification network. The long short-term memory network performs temporal encoding on the input vector, and the random forest classification network calculates the health index based on the encoding result.

[0114] In this embodiment, the calculation module 402 is further configured to generate an input vector based on pressure distribution data, temperature gradient data, and deformation acceleration data, including: calculating the tire's pressure equalization coefficient based on the pressure distribution data; extracting the temperature gradient data to calculate the tire's temperature risk factor; extracting the deformation acceleration data to calculate the tire's deformation loss index; and generating the input vector based on the pressure equalization coefficient, temperature risk factor, and deformation loss index.

[0115] In this embodiment of the application, the calculation formula for the Long Short-Term Memory network is as follows:

[0116] in, This represents the encoding result at time t. Represents the input vector. This represents the encoding result at time t-1. The weight matrix represents the weights from the hidden state to the gated unit, controlling the influence of historical information. This represents the bias vector, used to adjust the linear transformation result of the neural network at each layer. This represents the hyperbolic tangent activation function, which compresses the input values ​​to the interval [-1, 1], enabling the neural network to better express nonlinear relationships and prevent gradient divergence. The formula for calculating a random forest classification network is:

[0117] in, This indicates the health index. Let represent the prediction function of the i-th random forest decision tree, with the input being the final hidden state output by the LSTM model. The output is the independent prediction of tire health by this decision tree. Indicates the number of decision trees.

[0118] In this embodiment of the application, the apparatus 40 further includes an update module.

[0119] The update module is used to obtain the actual yaw rate error of the vehicle and update the weights of the long short-term memory network of the prediction model based on the actual yaw rate error.

[0120] In this embodiment of the application, the apparatus 40 further includes a preprocessing module.

[0121] The preprocessing module is used to perform data preprocessing on the pressure distribution data, temperature gradient data, and deformation acceleration data before calculating the tire health index based on the pressure distribution data, temperature gradient data, and deformation acceleration data. The preprocessing method includes at least one of time alignment processing, pressure wave delay compensation processing, and temperature diffusion rate correction processing.

[0122] It should be noted that the foregoing explanation of the vehicle parameter adaptive control method embodiment also applies to the vehicle parameter adaptive control device of this embodiment, and will not be repeated here.

[0123] The vehicle parameter adaptive control device proposed in this application acquires tire pressure distribution data, temperature gradient data, and deformation acceleration data. Combined with real-time vehicle speed and longitudinal acceleration, it first calculates a health index reflecting the true health state of the tire. Then, based on this index and vehicle dynamic parameters, it dynamically adjusts at least one of the trigger threshold, steering compensation angle, and torque limit of the vehicle's electronic stability system, thereby achieving precise vehicle control. This overcomes the limitations of related technologies that rely on fixed parameter calibration for vehicle control. It can adapt the control strategy promptly according to the actual health condition of the tire (such as abnormal local stress or potential damage), avoiding the risk of loss of control due to parameters not being dynamically adjusted with operating conditions. It also allows the control logic to conform to real-time driving conditions, reducing over-control or under-control issues, while enhancing the coordination between the tire and the vehicle, indirectly extending tire lifespan.

[0124] Figure 5 A schematic diagram of the structure of a vehicle provided in an embodiment of this application. The vehicle may include: The memory 501, the processor 502, and the computer program stored on the memory 501 and capable of running on the processor 502.

[0125] When the processor 502 executes the program, it implements the vehicle parameter adaptive control method provided in the above embodiments.

[0126] Furthermore, the vehicle also includes: Communication interface 503 is used for communication between memory 501 and processor 502.

[0127] The memory 501 is used to store computer programs that can run on the processor 502.

[0128] The memory 501 may include high-speed RAM (Random Access Memory) memory, and may also include non-volatile memory, such as at least one disk storage.

[0129] If the memory 501, processor 502, and communication interface 503 are implemented independently, then the communication interface 503, memory 501, and processor 502 can be interconnected via a bus to complete communication between them. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 5The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0130] Optionally, in a specific implementation, if the memory 501, processor 502, and communication interface 503 are integrated on a single chip, then the memory 501, processor 502, and communication interface 503 can communicate with each other through an internal interface.

[0131] The processor 502 may be a CPU (Central Processing Unit), an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of this application.

[0132] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0133] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0134] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0135] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any of the following techniques known in the art, or a combination thereof: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (FPGAs), field-programmable gate arrays (FPGAs), etc.

[0136] Those skilled in the art will understand that all or part of the steps of the methods implementing the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0137] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A vehicle parameter adaptive control method, characterized in that, Includes the following steps: Acquire tire pressure distribution data, temperature gradient data, and deformation acceleration data; Based on the pressure distribution data, the temperature gradient data, and the deformation acceleration data, the health index of the tire is calculated, and the vehicle speed and longitudinal acceleration are obtained. Based on at least one of the health index, the vehicle speed, and the longitudinal acceleration, at least one of the trigger threshold, steering compensation angle, and torque limit of the vehicle electronic stability system is adjusted, and the vehicle is controlled based on at least one of the trigger threshold, steering compensation angle, and torque limit.

2. The vehicle parameter adaptive control method according to claim 1, characterized in that, The vehicle's tires are equipped with a sensor array, which is used to collect pressure distribution data, temperature gradient data, and deformation acceleration data.

3. The vehicle parameter adaptive control method according to claim 2, characterized in that, The sensor array includes an accelerometer, a thermopile sensor, and a piezoelectric sensor. The piezoelectric sensor generates electricity using tire deformation to power the sensor array.

4. The vehicle parameter adaptive control method according to claim 1, characterized in that, The step of calculating the tire's health index based on the pressure distribution data, the temperature gradient data, and the deformation acceleration data includes: An input vector is generated based on the pressure distribution data, the temperature gradient data, and the deformation acceleration data. The input vector is then input into a pre-built prediction model, which outputs the health index of the tire. The prediction model includes a long short-term memory network and a random forest classification network. The long short-term memory network performs temporal encoding on the input vector, and the random forest classification network calculates the health index based on the encoding results.

5. The vehicle parameter adaptive control method according to claim 4, characterized in that, The step of generating an input vector based on the pressure distribution data, the temperature gradient data, and the deformation acceleration data includes: Calculate the pressure equalization coefficient of the tire based on the pressure distribution data; Extract the temperature gradient data to calculate the temperature risk factor of the tire; The deformation acceleration data is extracted to calculate the deformation loss index of the tire; The input vector is generated based on the pressure balance coefficient, the temperature risk factor, and the deformation loss index.

6. The vehicle parameter adaptive control method according to claim 4, characterized in that, The calculation formula for the Long Short-Term Memory network is as follows: in, This represents the encoding result at time t. Represents the input vector. This represents the encoding result at time t-1. Indicates hidden state The weight matrix to the gated unit, This represents the bias vector. Represents the hyperbolic tangent activation function; The calculation formula for the random forest classification network is as follows: in, This indicates the health index. Let represent the prediction function of the i-th random forest decision tree, with the final hidden state as the input. The output is the prediction of tire health by the random forest decision tree. Indicates the number of decision trees.

7. The vehicle parameter adaptive control method according to claim 4, characterized in that, Also includes: Obtain the actual yaw rate error of the vehicle; The weights of the long short-term memory network in the prediction model are updated based on the actual yaw rate error.

8. The vehicle parameter adaptive control method according to claim 1 or 4, characterized in that, Before calculating the tire's health index based on the pressure distribution data, the temperature gradient data, and the deformation acceleration data, the process includes: Align the pressure distribution data, the temperature gradient data, and the deformation acceleration data and perform data preprocessing, wherein the preprocessing method includes at least one of time alignment processing, pressure wave delay compensation processing, and temperature diffusion rate correction processing.

9. A vehicle parameter adaptive control device, characterized in that, include: The acquisition module is used to acquire tire pressure distribution data, temperature gradient data, and deformation acceleration data; The calculation module is used to calculate the health index of the tire based on the pressure distribution data, the temperature gradient data and the deformation acceleration data, and to obtain the vehicle speed and longitudinal acceleration. The control module is configured to adjust at least one of the trigger threshold, steering compensation angle, and torque limit of the vehicle electronic stability system based on at least one of the health index, the vehicle speed, and the longitudinal acceleration, and to control the vehicle based on at least one of the trigger threshold, the steering compensation angle, and the torque limit.

10. A vehicle, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the program to implement the vehicle parameter adaptive control method according to any one of claims 1-8.

Images