Automatic tire positioning system for automobile tire pressure monitoring

Abstract

The invention discloses an automatic tire positioning system for automobile tire pressure monitoring, and relates to the technical field of tire pressure monitoring and positioning. Comprising a storage building module, a static processing module, a dynamic processing module, a positioning decision module and a dynamic calibration module, wherein the storage building module is used for collecting historical data to build a historical database and dividing independent storage areas; the static processing module is used for independently extracting effective historical static data in N days from the storage area, and executing static feature vector construction after preprocessing; the technical key points are that accurate positioning of an automobile in most scenes can be realized, the defects of easy positioning deviation and low applicability during dynamic use caused by single static signal analysis in the prior art can be effectively solved, a mode of combining static characteristics and dynamic characteristics is adopted in the scheme, and further subdivision is performed in combination with a working condition environment, so that the positioning accuracy is improved. The dynamic scene positioning accuracy can be greatly improved, the applicability is wide, and the application prospect is good.

Description

Technical Field

[0001] This invention relates to the field of tire pressure monitoring and positioning technology, specifically to an automatic tire positioning system for automobile tire pressure monitoring. Background Technology

[0002] The function of a tire pressure monitoring system is to monitor the air pressure inside the tires in real time, ensuring that the air pressure is within the appropriate range. This prevents tire blowouts caused by excessively high air pressure or increased tire wear and fuel consumption caused by excessively low air pressure. The working principle of a tire pressure monitoring system is similar to measuring body temperature with a household thermometer. A small sensor is installed inside the tire to collect tire pressure data in real time and transmit it to the vehicle's dashboard. When the tire pressure is low, a yellow warning light will illuminate on the dashboard; if the air pressure is too high or there is a rapid leak, a red warning light will illuminate. This avoids the uncertainty of drivers subjectively judging tire pressure by tapping the tires. Compared to judging tire pressure by manually squeezing the tires, the tire pressure monitoring system provides more accurate results.

[0003] Many car owners often find themselves in a predicament when they encounter a tire pressure warning: although it is clear that the tires are low on air, checking each of the four tires one by one is really troublesome. In this case, the tire positioning system in the tire pressure monitoring system will directly display the specific tire positions, such as the front left and rear right, in red on the dashboard when the warning occurs. Car owners do not need to squat down to measure the tire pressure one by one, which saves time and is accurate and efficient.

[0004] The principle of a tire alignment system is similar to assigning a number to each tire. The tire pressure sensors in the four tires each have a unique code and are bound to the vehicle system at the factory. When the system receives sensor data, it will match the data to a fixed position according to the code.

[0005] Existing tire alignment systems rely on a single static signal for tire positioning, which can achieve tire positioning when the tire is stationary. However, in complex dynamic operating scenarios, the tire positioning accuracy is low. Moreover, they use a fixed scheme, and the sensors will age over time, resulting in a decrease in positioning accuracy over time. They also have insufficient anti-interference performance, and the vehicle sensor number needs to be manually bound after each tire change, requiring professional personnel to operate, thus limiting their applicability.

[0006] In summary, existing automatic tire positioning systems do not meet market requirements. Therefore, we propose an automatic tire positioning system based on tire pressure monitoring for automobiles. Summary of the Invention

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] An automatic tire positioning system for automobile tire pressure monitoring includes a storage module, a static processing module, a dynamic processing module, a positioning decision module, and a dynamic calibration module.

[0009] Storage setup module: Collects historical data to build a historical database and allocates independent storage areas;

[0010] Static processing module: Extracts valid historical static data from the storage area within N days, performs preprocessing, constructs static feature vectors, builds clusters using clustering algorithms, calculates the similarity between the clusters and the static feature vectors, and sets the initial location;

[0011] Dynamic processing module: Extracts valid historical dynamic data within N days from the storage area, performs preprocessing, extracts dynamic feature vectors of operating conditions, performs static feature vector processing and fusion with dynamic feature vectors to obtain historical fused feature vectors;

[0012] Location decision module: acquires real-time data, extracts and updates cluster centers when the cumulative mileage reaches the basic trigger condition, obtains new cluster centers, calculates weighted Euclidean distance based on real-time data and cluster centers and converts it into location similarity, and executes location determination strategy;

[0013] Dynamic calibration module: Real-time monitoring of location similarity, determining whether the abnormal triggering conditions are met. If not, it is determined to be normal and no action is taken; otherwise, it is determined to be abnormal and an abnormal handling strategy is executed.

[0014] Preferably, the independent storage area is group H, which is established based on the identification of the four tire pressure sensors installed on the car tires. The storage area includes a static reference cluster, a dynamic optimization cluster, historical iteration records, and an anomaly storage repository. Both the static reference cluster and the dynamic optimization cluster include the mean of the received signal strength, the variance of the received signal strength, the signal attenuation coefficient, and the generation time.

[0015] Preferably, extracting valid historical static data within N days from the storage area means extracting historical static data with a data integrity rate higher than 96% within the most recent N days. Static data acquisition is executed after receiving a trigger condition, which is that the ignition switch is in the on state. After receiving the trigger condition, a noise judgment strategy is executed to detect the intensity of environmental interference and calculate the noise value. If the noise value is less than a set threshold, historical static data extraction is executed; otherwise, execution is temporarily suspended until the noise value is less than the set threshold.

[0016] Preferably, the step of constructing static feature vectors after preprocessing is as follows:

[0017] Extract valid historical static data for N days;

[0018] The Laida criterion was used to process the extracted historical static data and remove outliers.

[0019] Count the number of historical static data after removing outliers. If it exceeds M groups, no action is taken; otherwise, perform supplementary data collection to ensure that the number of historical static data after removing outliers exceeds M groups.

[0020] Perform standardization processing on historical static data to transform it into standardized values;

[0021] The mean received signal strength, variance of received signal strength, and signal attenuation coefficient in the standardized historical static data are used as static feature vectors.

[0022] Preferably, clusters are constructed using a clustering algorithm, and the similarity between the clusters and the static feature vectors is calculated. The initial localization steps are set as follows:

[0023] We use random sampling to select initial points from the static feature vectors of H storage regions, which will be used as H cluster centers.

[0024] The K-means clustering algorithm is used for iterative calculation until the cluster center offset between two adjacent iterations meets the convergence threshold, and the clustering is completed, resulting in H independent clusters.

[0025] The clusters are associated with the tire positions. The average similarity of the static feature vectors in each cluster is calculated separately and compared with a preset similarity threshold. If the average similarity exceeds the preset similarity threshold, the clustering is deemed valid; otherwise, the clustering is deemed invalid and the clusters are reconstructed.

[0026] Extract the most recent historical static data or real-time static data and convert it into H groups of static feature vectors for localization;

[0027] Extract the centroids of H clusters to obtain H cluster centers;

[0028] The distance between the H groups of static feature vectors used for localization and the H cluster centers is calculated using Euclidean distance, and the distance values ​​are converted into similarity.

[0029] If the highest similarity among the H similarities is greater than the standard similarity value, then the tire position corresponding to the cluster with the highest similarity is taken as the initial positioning result, thus realizing the association between the vehicle tire pressure sensor identifier and the tire position;

[0030] Conversely, if the initial localization fails, the cluster is reconstructed.

[0031] Preferably, the effective historical dynamic data includes historical real-time received signal strength values, historical driving state data, and historical dynamic operating condition identification results. The historical driving state data includes steering angle, vehicle speed, braking signal, and accelerator pedal opening. The historical dynamic operating condition identification results include steering, acceleration, and constant speed. When the historical dynamic operating condition is steering, the dynamic feature vector is the average received signal strength of the inner wheel minus the average received signal strength of the outer wheel and the standard deviation of the sequence of the difference between the instantaneous received signal strength of the inner wheel and the instantaneous received signal strength of the outer wheel. When the historical dynamic operating condition is acceleration, the dynamic feature vector is the variance of the received signal strength of the front wheel minus the variance of the received signal strength of the rear wheel and the received signal strength stability data. When the historical dynamic operating condition is constant speed, the dynamic feature vector is the difference between the short-term average received signal strength and the average received signal strength in the cluster center and the fluctuation amplitude of the received signal.

[0032] Preferably, the steps of performing static feature vector processing and fusing with dynamic feature vectors are as follows:

[0033] Extract the mean received signal strength from the cluster center data, calculate the absolute difference between the short-time mean of the dynamic received signal strength and the mean received signal strength, and obtain the mean received signal deviation.

[0034] The installation angle attenuation coefficient deviation is calculated based on the difference between the attenuation coefficient corresponding to the installation angle of the tire pressure monitoring sensor and the factory calibration reference value.

[0035] The mean deviation of the received signal and the deviation of the installation angle attenuation coefficient are used to construct a new static feature vector.

[0036] The weights of the dynamic feature vector and the new static feature vector are determined based on the dynamic operating conditions.

[0037] The dynamic feature vectors at the same tire position are associated and aligned with the new static feature vectors to form a set of feature dimensions;

[0038] The feature values ​​of each dimension in the feature dimension set are weighted and assembled into a historical fusion feature vector.

[0039] Preferably, when the cumulative mileage reaches the basic trigger condition, the cluster centers are extracted and updated:

[0040] Real-time statistics of vehicle mileage are obtained to calculate the cumulative mileage. The cumulative mileage is compared with the basic triggering conditions. If the cumulative mileage reaches the basic triggering conditions, the cluster center iterative update operation is performed.

[0041] Acquire real-time dynamic data and group it according to tire position, then analyze temporary cluster centers;

[0042] A weighted fusion is performed on the cluster centers and temporary cluster centers to obtain new cluster centers.

[0043] Preferably, when executing the positioning determination strategy, the weighted Euclidean distance is calculated based on real-time data and the cluster center and converted into positioning similarity. First, real-time data is acquired, the real-time fused feature vector is analyzed, and the weighted Euclidean distance between the real-time fused feature vector and the new cluster center is calculated. The weighted Euclidean distance is converted into positioning similarity using a preset maximum distance threshold combined with a similarity formula. The positioning similarity values ​​of H tire positions are sorted in descending order, and the tire position corresponding to the maximum positioning similarity is taken as the candidate positioning result. If the maximum positioning similarity is greater than the preset standard determination threshold, the candidate positioning result is considered valid; otherwise, the positioning determination strategy is re-executed. If in the results of L consecutive executions of the positioning determination strategy, there is a single tire position whose maximum positioning similarity is less than or equal to the preset standard determination threshold, or multiple tire positions have a maximum positioning similarity less than or equal to the preset standard determination threshold, an exception handling strategy is executed.

[0044] Preferably, the abnormal triggering conditions also include detecting a new tire pressure monitoring sensor identifier. The steps of executing the abnormal handling strategy are as follows: based on real-time data analysis, the real-time fusion feature vectors of H tire positions and the positioning similarity of each new cluster center are used to form a similarity matrix. Anomalies are identified in the similarity matrix, and a calibration strategy is executed based on the anomaly identification results.

[0045] This invention provides an automatic tire alignment system for automobile tire pressure monitoring, which has the following advantages:

[0046] This invention employs a combination of static and dynamic data analysis to determine tire positioning, enabling accurate positioning in most automotive scenarios. It effectively addresses the shortcomings of existing methods that rely on single static signal analysis, which often result in positioning errors and low applicability during dynamic use. This solution combines static and dynamic features, further subdividing the operating environment to significantly improve positioning accuracy in dynamic scenarios. It is widely applicable and delivers excellent results.

[0047] This invention uses a similarity matrix to achieve anomaly identification and executes a calibration strategy based on the anomaly identification results. It can effectively solve the problem that existing technologies require full data re-collection and model training when anomalies occur, resulting in long calibration time and high resource consumption. This solution can achieve accurate anomaly identification, perform data collection and positioning on abnormal tires, while the position of normal tires remains undisturbed. This can greatly reduce the amount of data processing, significantly reduce calibration time, and has good performance.

[0048] This invention combines cumulative mileage-triggered iteration with historical weight linear decay technology to update the cluster center. This effectively avoids the problem of decreased positioning accuracy due to sensor aging and environmental changes. Each time the cumulative mileage is large enough, an iteration is performed, thereby dynamically optimizing the cluster center and ensuring positioning accuracy over long-term use. It has good performance and promising application prospects. Attached Figure Description

[0049] Figure 1 This is a flowchart of an automatic tire positioning system for automobile tire pressure monitoring according to the present invention. Detailed Implementation

[0050] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] This invention is primarily aimed at four-wheeled vehicles, and it can automatically identify and position tires under dynamic conditions. It does not require driver operation and is easy to use.

[0052] Example:

[0053] Please see Figure 1 This embodiment provides an automatic tire positioning system for automobile tire pressure monitoring, including a storage module, a static processing module, a dynamic processing module, a positioning decision module, a dynamic calibration module, and a result output module.

[0054] Storage setup module: Collects historical data to build a historical database and allocates independent storage areas;

[0055] The independent storage area is group H, corresponding to H tire positions, and is set up based on the identification of the four tire pressure sensors installed on the car tires. The independent storage area includes a static reference cluster, a dynamic optimization cluster, historical iteration records, and an anomaly storage repository. Both the static reference cluster and the dynamic optimization cluster contain the mean received signal strength, the variance of received signal strength, the signal attenuation coefficient, and the generation time.

[0056] The number of independent storage areas is equal to the number of tires. A typical car has four tires, so there are four sets of independent storage areas.

[0057] The historical database contains data such as the installation angle of the tire pressure monitoring sensor during installation, the steering angle, vehicle speed, braking, and accelerator pedal readings obtained by connecting a diagnostic tool through the vehicle's OBD interface, and the corresponding data is extracted and stored in a separate storage area.

[0058] The historical database uses the lightweight SQLite database and establishes data tables for sensor identification, installation location, installation time, communication status, and more records, facilitating subsequent retrieval and analysis.

[0059] The historical database uses scheduled automatic backups to ensure that no data is lost.

[0060] Static processing module: Extracts valid historical static data from the storage area within N days, performs preprocessing, constructs static feature vectors, builds clusters using clustering algorithms, calculates the similarity between the clusters and the static feature vectors, and sets the initial location;

[0061] Extracting valid historical static data within N days from the storage area means extracting historical static data with a data integrity rate of over 96% within the most recent N days.

[0062] When the system receives a trigger condition, it will perform static data acquisition. The trigger condition is that the ignition switch is in the on state, and the static data is the data collected when the vehicle speed is 0km / h. After receiving the trigger condition, the system will execute a noise judgment strategy to detect the intensity of environmental interference and calculate the noise value. If the noise value is less than the set threshold, the system will extract historical static data. Otherwise, the system will trigger a pause. The threshold is usually set to 88dBm. When the noise value is greater than or equal to 88dBm, the surface environmental interference is relatively serious. The system will continue to extract historical static data until the noise value is less than the set threshold.

[0063] The steps for constructing static feature vectors after preprocessing are as follows:

[0064] Extract N days of valid historical static data, or use a fixed number of records, or combine the two methods, such as collecting the 8 most similar sets of valid historical static data within 10 days;

[0065] The extracted historical static data was processed using the Raida criterion to remove outliers. The Raida criterion is a commonly used method for removing outliers, so it will not be described in detail here.

[0066] Count the number of historical static data after removing outliers. If it exceeds M groups, no processing is performed; otherwise, perform supplementary data collection to make the number of historical static data after removing outliers exceed M groups. M is generally set to 1000. The larger M is, the higher the accuracy, but the larger the corresponding data processing volume.

[0067] Perform standardization processing on historical static data to transform it into standardized values;

[0068] When performing standardization, the minimum-maximum standardization method is used. After standardization, verification can be performed by calculating the coefficient of variation of the received signal mean, which is the ratio of the standardized received signal strength to the received signal mean. The ratio is then compared with a preset value. If the coefficient of variation of the received signal mean is greater than the preset value, the data is considered invalid. The preset value is generally 0.05.

[0069] The mean received signal strength, variance of received signal strength, and signal attenuation coefficient in the standardized historical static data are used as static feature vectors, that is, the static feature vectors are three-dimensional feature vectors.

[0070] Clusters are constructed using a clustering algorithm, and the similarity between the clusters and the static feature vectors is calculated. The initial localization steps are set as follows:

[0071] We use random sampling to select initial points from the static feature vectors of H storage regions, which will be used as H cluster centers.

[0072] The K-means clustering algorithm is used for iterative calculation until the cluster center offset between two adjacent iterations meets the convergence threshold, and the clustering is completed, resulting in H independent clusters.

[0073] The process continues until the convergence threshold is met: the offset of the cluster centers in two consecutive iterations is less than 0.001, or the number of iterations reaches the set upper limit of 60.

[0074] Based on the static feature vector range of four tires of the same model vehicle, clustering is performed and associated with tire position. That is, it is determined which static feature vector range each of the four tires falls into, and thus associated with the corresponding cluster. Then, the verification is performed by calculating the average similarity of the static feature vectors within each cluster and comparing the average similarity with a preset similarity threshold. If the average similarity exceeds the preset similarity threshold, the clustering is deemed valid; otherwise, the clustering is deemed invalid and the clusters are reconstructed.

[0075] Generally, the average similarity is set to 92%. This step involves selecting the static feature vector of the tire, calculating the similarity of all static feature vectors within its corresponding cluster, and then taking the average value to obtain the average similarity. It is then determined whether the average similarity exceeds the preset similarity threshold. If the average similarity exceeds the preset similarity threshold, the clustering is valid. The cluster is then reconstructed by replacing more than 30% of the historical static data, re-analyzing and processing it, and constructing the cluster.

[0076] The above method establishes the association between clusters and tire locations, thus defining the cluster corresponding to each tire.

[0077] Extract the most recent historical static data or real-time static data and convert it into H groups of static feature vectors for localization;

[0078] The most recent historical static data is the set of static data that is closest to the current time;

[0079] Extract the centroids of H clusters to obtain H cluster centers;

[0080] The distance between the H groups of static feature vectors used for localization and the H cluster centers is calculated using Euclidean distance. The distance values ​​are then converted into similarity scores. The smaller the distance value, the higher the similarity score. The similarity score is calculated as: Similarity = 1 - Distance value / Preset maximum distance threshold. The preset maximum distance threshold is based on the maximum possible distance calibrated for vehicles of the same model.

[0081] If the highest similarity among H similarities is greater than the standard similarity value (which is typically set to 88%), then the tire position corresponding to the cluster with the highest similarity is taken as the initial positioning result, thus associating the tire pressure sensor identifier with the tire position and achieving the positioning of the sensor identifier with the tire position.

[0082] Conversely, if the initial positioning fails, the clusters are reconstructed. Similar to the previous process, reconstructing the clusters involves replacing more than 30% of the historical static data, re-analyzing and processing it, and then reconstructing the clusters.

[0083] Dynamic processing module: Extracts valid historical dynamic data within N days from the storage area, performs preprocessing, extracts dynamic feature vectors of operating conditions, performs static feature vector processing and fusion with dynamic feature vectors to obtain historical fused feature vectors;

[0084] Valid historical dynamic data refers to data collected when the vehicle speed is greater than 0.

[0085] Valid historical dynamic data includes historical real-time received signal strength values, historical driving state data, and historical dynamic condition identification results. Historical driving state data includes steering angle, vehicle speed, braking signal, and accelerator pedal opening. Historical dynamic condition identification results include steering, acceleration, and constant speed. Historical real-time received signal strength values ​​are preprocessed using a moving average filter to remove noise. Braking signals include braking and no braking; 0 is used for braking, and 1 for no braking. To determine if steering is present, the absolute value of the steering angle is checked to see if it exceeds 10°. When the historical dynamic condition is steering, the dynamic feature vector is the average received signal strength of the inner wheel minus the average received signal strength of the outer wheel, plus the standard deviation of the sequence of instantaneous differences between the instantaneous received signal strengths of the inner and outer wheels. For example, first, the received signal strength of the inner wheel is calculated separately within the sampling data of a single steering condition duration. The average signal strength of the inner wheel and the average signal strength of the outer wheel are collected, and then the difference between the two can be calculated. If the data collection is frequent, for example, 10 times per second, and the turning time is long, such as 20 seconds, in order to reduce the amount of data, the standard deviation of the difference sequence between the instantaneous signal strength of the inner wheel and the instantaneous signal strength of the outer wheel can be changed to the standard deviation of the difference sequence between the short-term mean of the signal strength of the inner wheel and the short-term mean of the signal strength of the outer wheel. For example, the average signal strength of the inner wheel and the signal strength of the outer wheel are collected and calculated every 2 seconds to obtain the short-term mean of the signal strength of the inner wheel and the short-term mean of the signal strength of the outer wheel. This yields 10 sets of short-term mean difference data, forming a short-term mean difference sequence. The standard deviation of these 10 sets of data can then be calculated, which can greatly reduce the amount of data processing. The inner wheel is the wheel closest to the turning center when turning.

[0086] In the historical dynamic operating condition of acceleration, the dynamic feature vector consists of the variance of the front wheel received signal strength minus the variance of the rear wheel received signal strength, and the stability data of the received signal strength. To reduce the computational load, when analyzing the dynamic feature vector in the historical dynamic operating condition of acceleration, the calculation of the variance of the front wheel received signal strength minus the variance of the rear wheel received signal strength involves statistically analyzing all front wheel and rear wheel received signal strength data during a single complete acceleration cycle, calculating the variances of the front and rear wheel received signal strength, and then calculating the difference between the two sets of variances. When calculating the stability data of the received signal strength, the stability coefficients of each of the four tires are analyzed separately, using a fixed time window for the analysis of the stability coefficients of the four tires. The data was divided into 5 segments with 10-second intervals over 50 seconds. The mean received signal strength of each segment was calculated, resulting in a mean sequence of 5 mean data. The rate of change was calculated using the linear fitting slope method. The time of each segment was plotted on the x-axis and the mean of each segment was plotted on the y-axis. A linear regression equation was obtained by fitting the data using the least squares method. The absolute value of the slope is the tire's stability coefficient. The smaller the tire's stability coefficient, the more stable the signal received by the tire during acceleration, and vice versa. To facilitate subsequent calculations, the calculated stability coefficient was standardized to the range of 0-1. During standardization, the stability coefficient was divided by the maximum coefficient threshold, which is the maximum fluctuation slope based on the acceleration conditions calibrated on the actual vehicle.

[0087] When the historical dynamic operating condition is constant speed, the dynamic feature vector consists of the absolute difference between the short-time mean of the received signal strength and the mean of the received signal strength in the cluster center, and the fluctuation amplitude of the received signal. When calculating the difference between the short-time mean of the received signal strength and the mean of the received signal strength in the cluster center, each of the four tires is analyzed separately. The short-time mean of the received signal strength is the mean of the received signal strength within one minute under constant speed conditions, and the absolute difference is the absolute value of the difference. When analyzing the fluctuation amplitude of the received signal, each of the four tires is analyzed separately. The received signal strength of the target tire within one minute is statistically analyzed, the maximum and minimum received signal strengths are identified, and the difference between the maximum and minimum received signal strengths is calculated, which is the fluctuation amplitude of the received signal.

[0088] Since the deceleration condition is similar to the acceleration condition, no deceleration condition is distinguished.

[0089] The steps for performing static feature vector processing and fusing with dynamic feature vectors are as follows:

[0090] Extract the mean received signal strength from the cluster center data, calculate the absolute difference between the short-time mean of the dynamic received signal strength and the mean received signal strength, and obtain the mean received signal deviation.

[0091] The installation angle attenuation coefficient deviation is calculated based on the difference between the attenuation coefficient corresponding to the installation angle of the tire pressure monitoring sensor and the factory calibration reference value.

[0092] The mean deviation of the received signal and the deviation of the installation angle attenuation coefficient are used to construct a new static feature vector.

[0093] The weights of the dynamic feature vector and the new static feature vector are determined based on the dynamic operating conditions. When the historical dynamic operating condition is constant speed, the weights of the dynamic feature vector and the new static feature vector are 0.7 and 0.3, respectively. When the historical dynamic operating condition is acceleration, the weights of the dynamic feature vector and the new static feature vector are 0.75 and 0.25, respectively. When the historical dynamic operating condition is turning, the weights of the dynamic feature vector and the new static feature vector are 0.8 and 0.2, respectively. The main reason for the complexity is that the situation is more complicated during turning and acceleration.

[0094] The dynamic feature vectors at the same tire position are associated and aligned with the new static feature vectors to form a set of feature dimensions, which determines the working condition and thus retrieves the corresponding weights.

[0095] The feature values ​​of each dimension in the feature dimension set are weighted and assembled into a historical fusion feature vector.

[0096] Then, for each working condition, the weights of the mean deviation of the two sets of received signals and the deviation of the installation angle attenuation coefficient are further set, and the weights of the two sets of dynamic feature vectors corresponding to each working condition are set. The weighted values ​​are calculated to obtain the corresponding weighted values, that is, four sets of weighted values ​​are obtained. No summation operation is performed, and they are arranged in a preset order to obtain the fused feature vector.

[0097] This invention employs a combination of static and dynamic data analysis to determine tire positioning, enabling accurate positioning in most automotive scenarios. It effectively addresses the shortcomings of existing methods that rely on single static signal analysis, which often result in positioning errors and low applicability during dynamic use. This solution combines static and dynamic features, further subdividing the operating environment to significantly improve positioning accuracy in dynamic scenarios. It is widely applicable and delivers excellent results.

[0098] Location decision module: acquires real-time data, extracts and updates cluster centers when the cumulative mileage reaches the basic trigger condition, obtains new cluster centers, calculates weighted Euclidean distance based on real-time data and cluster centers and converts it into location similarity, and executes location determination strategy;

[0099] When the accumulated mileage reaches the basic trigger condition, the cluster centers are extracted and updated:

[0100] The vehicle's mileage is counted in real time to obtain the cumulative mileage. The cumulative mileage is compared with the basic trigger condition. If the cumulative mileage reaches the basic trigger condition, the cluster center iterative update operation is performed. For example, if the trigger condition is 10 kilometers, the cluster center iterative update operation is performed every 10 kilometers.

[0101] When performing iterative update operations on cluster centers, the fused feature vector data within a 5-kilometer radius is extracted from the historical database. The current cluster center is used as the initial benchmark, and the weights are adjusted according to the number of historical iterations. For example, in the first iteration, the weight of the historical center is set to 0.5; in the second and subsequent iterations, the weight of the historical center is linearly decayed according to the formula: historical center weight = 0.5 - 0.05 × number of iterations.

[0102] Real-time dynamic data is acquired and grouped according to tire position, i.e., divided into four groups. Temporary cluster centers are analyzed. The specific steps are the same as the previous clustering steps, except that the features used at this time are different, such as the mean deviation of received signal strength.

[0103] A weighted fusion is performed on the cluster centers and temporary cluster centers to obtain new cluster centers.

[0104] The formula for weighted fusion is: New cluster center = Cluster center × Historical center weight + Temporary cluster center × (1 - Historical center weight).

[0105] For example, during the first iteration, the new cluster center = cluster center × 0.5 + temporary cluster center × (1 - 0.5), and during the second iteration, the new cluster center = cluster center × (0.5 - 0.05 × 2) + temporary cluster center × [1 - (0.5 - 0.05 × 2)].

[0106] The same verification process is performed after the update. The average similarity of the static feature vectors within each cluster is calculated separately, and the average similarity is compared with a preset similarity threshold. If the average similarity exceeds the preset similarity threshold, the update is valid.

[0107] Based on real-time data and the cluster centers, a weighted Euclidean distance is calculated and converted into a location similarity. When executing the location determination strategy, real-time data is first acquired, and the real-time fused feature vector is analyzed. The weighted Euclidean distance between the real-time fused feature vector and the new cluster centers is calculated. A preset maximum distance threshold is used in conjunction with a similarity formula to convert the weighted Euclidean distance into a location similarity. The location similarity values ​​of H tire positions are sorted in descending order, and the tire position corresponding to the maximum location similarity is taken as the candidate location result. If the maximum location similarity is greater than the preset standard determination threshold, the candidate location result is considered valid; otherwise, the location determination strategy is re-executed. This part is the same as the previous one. If in the results of L consecutive executions of the location determination strategy, there is a single tire position whose maximum location similarity is less than or equal to the preset standard determination threshold, or multiple tire positions have a maximum location similarity less than or equal to the preset standard determination threshold, then an exception handling strategy is executed.

[0108] An abnormal trigger condition is when, in the results of L consecutive executions of the positioning determination strategy, the maximum value of the positioning similarity of a single tire position is less than or equal to the preset threshold, or the maximum value of the positioning similarity of multiple tire positions is less than or equal to the preset threshold.

[0109] This invention combines cumulative mileage-triggered iteration with historical weight linear decay technology to update the cluster center. This effectively avoids the problem of decreased positioning accuracy due to sensor aging and environmental changes. Each time the cumulative mileage is large enough, an iteration is performed, thereby dynamically optimizing the cluster center and ensuring positioning accuracy over long-term use. It has good performance and promising application prospects.

[0110] Dynamic calibration module: Real-time monitoring of location similarity, determining whether the abnormal triggering conditions are met. If not, it is determined to be normal and no action is taken; otherwise, it is determined to be abnormal and an abnormal handling strategy is executed.

[0111] Abnormal triggering conditions also include the detection of a new tire pressure monitoring sensor identifier.

[0112] Typically, in the results of L consecutive executions of the positioning determination strategy, there are instances where the maximum positioning similarity of a single tire position is less than or equal to the preset threshold and a new tire pressure monitoring sensor identifier is detected simultaneously, indicating that a single tire has been replaced. However, in actual use, there are also cases where the tire positions are changed. For such cases, further identification is required through anomaly handling strategies.

[0113] The abnormal triggering conditions also include the detection of new tire pressure monitoring sensor identifiers. The steps for implementing the abnormal handling strategy are as follows: based on real-time data analysis, the feature vectors of H tire positions are fused in real time and the similarity between the location of each new cluster center is calculated to form a similarity matrix. Anomalies are identified in the similarity matrix, and a calibration strategy is executed based on the anomaly identification results.

[0114] When performing anomaly identification on the similarity matrix, it is determined whether there are any differences between the current sensor identification and positioning results and the historical sensor identification and positioning results. If such differences exist, it indicates that the tires have been swapped.

[0115] For tires that have been swapped or partially replaced, the calibration strategy is to lock the abnormal locations, collect data specifically, retrain only the clusters corresponding to the abnormal locations, update the cluster centers by inputting the specially collected data, and keep the clusters at the other three locations unchanged, thereby reducing the consumption of computing resources.

[0116] If the situation is determined to be a complete tire replacement or swap, the full re-clustering process is initiated to reconstruct the fused feature vectors of the four tire positions, retrain the clustering model, and generate four new cluster centers.

[0117] This invention uses a similarity matrix to achieve anomaly identification and executes a calibration strategy based on the anomaly identification results. It can effectively solve the problem that existing technologies require full data re-collection and model training when anomalies occur, resulting in long calibration time and high resource consumption. This solution can achieve accurate anomaly identification, perform data collection and positioning on abnormal tires, while the position of normal tires remains undisturbed. This can greatly reduce the amount of data processing, significantly reduce calibration time, and has good performance.

[0118] Results output module: Displays the results of positioning determination strategy, anomaly handling strategy and calibration strategy, making it convenient for relevant staff and drivers to view and for subsequent analysis.

[0119] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0120] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0121] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A tire automatic positioning system for automotive tire pressure monitoring, characterized by, Comprise: Storage building module: collect historical data to build a historical database, and divide independent storage areas; Static processing module: extract N-day effective historical static data from the storage area alone, perform static feature vector construction after preprocessing, build clustering clusters through clustering algorithm, perform clustering cluster and static feature vector similarity calculation, and set initial positioning; Dynamic processing module: extract N-day effective historical dynamic data from the storage area, perform working condition dynamic feature vector extraction after preprocessing, perform static feature vector processing and fusion with dynamic feature vector, and obtain historical fusion feature vector; Positioning decision module: obtain real-time data, extract clustering cluster center and update when the cumulative driving mileage reaches the basic trigger condition, obtain new clustering cluster center, calculate weighted Euclidean distance based on real-time data and clustering cluster center and convert it into positioning similarity, and perform positioning judgment strategy; Dynamic calibration module: real-time monitor positioning similarity, judge whether the abnormal trigger condition is met, if not, determine as normal and do not process, otherwise, determine as abnormal and perform abnormal processing strategy.

2. The automatic positioning system for tire of automobile tire pressure monitoring according to claim 1, characterized in that: The independent storage area is H group, which is set up according to four automobile tire pressure sensors installed on the automobile tire, and the storage area contains static reference cluster, dynamic optimization cluster, historical iteration record and exception storage library. The static reference cluster and the dynamic optimization cluster both contain received signal strength average, received signal strength variance, signal attenuation coefficient and generation time.

3. The automatic positioning system for tire of automobile tire pressure monitoring according to claim 1, characterized in that: Extracting N-day effective historical static data from the storage area alone is to extract historical static data with data integrity higher than 96% in the last N days. After receiving the trigger condition, static data acquisition is performed. The trigger condition is that the ignition switch is in the on state. After receiving the trigger condition, noise judgment strategy is performed. The environmental interference intensity is detected and the noise value is calculated. If the noise value is less than the set threshold, historical static data extraction is performed. Otherwise, temporary execution is triggered until the noise value is less than the set threshold.

4. The automatic positioning system for tire of automobile tire pressure monitoring according to claim 3, characterized in that: The steps of performing static feature vector construction after preprocessing are: Extract N-day effective historical static data; Use the Laplace criterion to process the extracted historical static data and eliminate outliers; Statistical the number of historical static data after eliminating outliers. If it exceeds M groups, do not process; otherwise, perform data supplement collection to make the historical static data after eliminating outliers exceed M groups; Perform standardization processing on the historical static data to convert the historical static data into standardized values; Use the received signal strength average, received signal strength variance and signal attenuation coefficient in the standardized historical static data as the static feature vector.

5. The automatic positioning system for tire pressure monitoring of a vehicle tire according to claim 1, wherein: The steps of building clustering clusters through clustering algorithm and performing clustering cluster and static feature vector similarity calculation to set initial positioning are: Use random sampling method to select initial points from the static feature vectors of H storage areas as H clustering centers; Use K-means clustering algorithm for iterative calculation until the clustering center offset of adjacent two iterations meets the convergence threshold after clustering is completed, and H independent clustering clusters are obtained; The clustering cluster is associated with the tire position, the average similarity of the static feature vector in each clustering cluster is calculated separately, and the average similarity is compared with the preset similarity threshold. If the average similarity exceeds the preset similarity threshold, it is determined that the clustering is effective, otherwise, it is determined that the clustering is invalid, and the clustering cluster is reconstructed; The recent historical static data or real-time static data is extracted and converted into H sets of static feature vectors for positioning; The centroids of the H clustering clusters are extracted to obtain H clustering cluster centers; The Euclidean distance is used to calculate the distance value of the H sets of static feature vectors for positioning and the H clustering cluster centers, and the distance value is converted into similarity; If the highest similarity in the H similarities is greater than the standard similarity value, the clustering cluster corresponding to the highest similarity is taken as the initial positioning result, and the tire position is associated with the tire position. Otherwise, it is determined that the initial positioning fails, and the clustering cluster is reconstructed.

6. The automatic positioning system for tire of automobile tire pressure monitoring according to claim 5, characterized in that: The effective historical dynamic data includes historical real-time received signal strength values, historical driving state data, and historical dynamic working condition recognition results. The historical driving state data includes steering angle, vehicle speed, brake signal, and accelerator pedal opening. The historical dynamic working condition recognition results include steering, acceleration, and constant speed. When the historical dynamic working condition is steering, the dynamic feature vector is the average of the inside wheel received signal strength and the average of the outside received signal strength, and the standard deviation of the inside wheel received signal instantaneous strength and the outside received signal instantaneous strength difference sequence. When the historical dynamic working condition is acceleration, the dynamic feature vector is the front wheel received signal strength variance and the rear wheel received signal strength variance and the received signal strength stability data. When the historical dynamic working condition is constant speed, the dynamic feature vector is the difference between the received signal strength short-time average and the received signal strength average in the clustering cluster center and the received signal fluctuation amplitude.

7. The automatic positioning system for tire of automobile tire pressure monitoring according to claim 6, characterized in that: The steps of performing static feature vector processing and fusion with dynamic feature vector are: Extract the received signal strength average in the clustering cluster center data, calculate the absolute difference between the dynamic received signal strength short-time average and the received signal strength average, and obtain the received signal average deviation. Calculate the installation angle attenuation coefficient deviation value based on the difference between the installation angle attenuation coefficient corresponding to the tire pressure monitoring sensor and the factory calibration reference value, and obtain the installation angle attenuation coefficient deviation. The received signal average deviation and the installation angle attenuation coefficient deviation form a new static feature vector. Determine the weight of the dynamic feature vector and the new static feature vector according to the dynamic working condition. Align the dynamic feature vector and the new static feature vector of the same tire position to form a feature dimension set. Perform weighted operation on the feature value of each dimension in the feature dimension set to assemble a historical fusion feature vector.

8. The automatic positioning system for tire of automobile tire pressure monitoring according to claim 7, characterized in that: When the cumulative driving mileage reaches the basic trigger condition, the clustering cluster center is extracted and updated: Real-time statistics of vehicle driving mileage are obtained to obtain the cumulative driving mileage, and the cumulative driving mileage is compared with the basic trigger condition. If the cumulative driving mileage reaches the basic trigger condition, the clustering cluster center iteration update operation is performed. Real-time dynamic data is obtained and grouped according to the tire position, and the temporary clustering center is analyzed. The clustering cluster center and the temporary clustering center are weighted and fused to obtain a new clustering cluster center.

9. The automatic positioning system for tire of automobile tire pressure monitoring according to claim 1, characterized in that: In the positioning judgment strategy, the real-time data is obtained first, the real-time fusion feature vector is analyzed, the weighted Euclidean distance between the real-time fusion feature vector and the new cluster center is calculated, the weighted Euclidean distance is converted into positioning similarity by using a preset maximum distance threshold and a similarity formula, the positioning similarity values of the H tire positions are sorted in descending order, the tire position corresponding to the maximum positioning similarity value is taken as the selected positioning result, if the maximum positioning similarity value is greater than a preset threshold, the selected positioning result is determined to be valid, otherwise, the positioning judgment strategy is re-executed, if the maximum positioning similarity value of a single tire position is less than or equal to the preset threshold or the maximum positioning similarity values of multiple tire positions are less than or equal to the preset threshold in the results of the positioning judgment strategy executed for L times in succession, an abnormality processing strategy is executed.

10. The automatic positioning system for tire of automobile tire pressure monitoring according to claim 9, characterized in that: The abnormality triggering condition also includes monitoring a new tire pressure monitoring sensor identifier, and the steps of the abnormality processing strategy include: analyzing the positioning similarity between the real-time fusion feature vectors of the H tire positions and each new cluster center based on the real-time data, constructing a similarity matrix, performing abnormality identification on the similarity matrix, and executing a calibration strategy based on the abnormality identification result.

Images