An ETC transaction failure judgment method based on inertial navigation and an ETC transaction system

Abstract

The application provides an ETC transaction failure judgment method and system based on inertial navigation, an inertial navigation module is introduced into a vehicle-mounted label to record continuous motion state data, geographical environment matching is performed in combination with positioning data to identify the scene type of a road section where a transaction fails, transaction process data, equipment working conditions, the inertial navigation data, the positioning data and the scene type where the transaction is located are aligned to form a failure log, a preset failure space-time correlation analysis logic framework is used for failure troubleshooting or a pre-trained failure analysis model is introduced to identify the transaction failure reason. The application establishes a complete cause-and-effect chain to analyze the ETC transaction failure reason for a driving route, a transaction failure road section, motion state parameters and an ETC transaction failure phenotype, and is suitable for failure detection and cause investigation in a multi-scene complex transaction environment.

Description

Technical Field

[0001] This invention relates to the field of ETC transaction technology, and in particular to an ETC transaction fault judgment method and ETC transaction system based on inertial navigation. Background Technology

[0002] With the widespread application of ETC (Electronic Toll Collection) systems in complex traffic scenarios, the problem of transaction failures caused by various reasons during vehicle dynamic driving is becoming increasingly prominent. Traditional on-board tags record data in limited formats, only recording fragmented location and transaction data in dynamic scenarios, failing to fully reconstruct the vehicle's movement trajectory during a transaction. Lacking multi-dimensional motion parameters, the impact of geographical environment and motion state on ETC transactions cannot be incorporated. Furthermore, the trajectory recorded by traditional on-board tags has a weak correlation with faults, only acquiring local data such as acceleration, and cannot establish a complete causal chain of "driving route - change in motion state - transaction failure," for example, it cannot determine whether continuous bumps on a long downhill section are the cause of the fault. Faced with increasingly complex causes of transaction failures, solutions without motion perception capabilities are significantly insufficient for fault analysis. Therefore, a new ETC transaction fault judgment and identification scheme is urgently needed, aiming to achieve full-dimensional perception of motion state, provide a dynamic evidence chain in spatiotemporal coordinates for transaction faults, and realize the analysis and identification of ETC transaction faults in complex scenarios. Summary of the Invention

[0003] In view of this, embodiments of the present invention provide an ETC transaction fault judgment method and an ETC transaction system based on inertial navigation, so as to eliminate or improve one or more defects in the prior art and solve the problem that the prior art cannot combine continuous motion state data to identify and judge the cause of ETC transaction failure.

[0004] One aspect of the present invention provides a method for determining ETC transaction faults based on inertial navigation, the method comprising the following steps: The inertial navigation module collects inertial navigation data at a set sampling frequency, including three-dimensional acceleration, angular velocity, speed, heading angle, and pitch angle; the positioning module collects and updates positioning data at the set sampling frequency, and performs geographic environment matching based on the positioning data to determine the scene type of the road segment. In response to ETC transaction failure information, the transaction process data, equipment status, inertial navigation data, positioning data, and scene type within a predetermined time range prior to the ETC transaction failure information are aligned according to timestamps and stored in the fault log; In response to an access request from a roadside unit, the fault log is sent to the roadside unit via a preset communication path. The roadside unit then uses the fault data in the fault log to perform fault investigation based on a preset fault spatiotemporal correlation analysis logic framework to determine the cause of transaction failure, or inputs the fault data into a pre-trained fault analysis model and outputs the cause of transaction failure. The fault spatiotemporal correlation analysis logic framework is obtained by analyzing the primary correlation between the inertial navigation data, the scene type of the road segment, and the equipment operating condition; analyzing the secondary correlation between the equipment operating condition and the ETC transaction failure phenotype; and logically fusing the primary and secondary correlations. The fault analysis model is obtained by learning the mapping relationship between fault data and transaction failure causes in historical fault logs. The ETC transaction failure phenotype is an abnormal transaction state presented based on the transaction process data. In some embodiments, the scene types include: tunnel sections, mountain curved road sections, toll station sections, parking lot sections, urban elevated roads and interchange sections, long straight highway sections, and service area and ramp connection sections. The equipment operating conditions include vehicle-mounted tag operating conditions and roadside unit operating conditions. The vehicle-mounted tag operating conditions include: vehicle-mounted tag communication performance parameters, power supply status parameters, operating environment status parameters, and installation and operation status parameters. The roadside unit operating conditions include roadside unit communication performance parameters and system operation status parameters.

[0005] In some embodiments, the reasons for transaction failure are obtained by troubleshooting based on a preset fault spatiotemporal correlation analysis logic framework, including: The inertial navigation data is compared with a predefined range of motion state parameters to determine the abnormal motion state of the vehicle tag. Based on the scenario type at the time of the fault, the standard parameter range of the equipment under standard conditions is corrected; the corrected standard parameter range is compared with the equipment operating conditions to determine the abnormal operating condition parameters. Based on the ETC transaction failure phenotype, the abnormal motion state, and the abnormal operating condition parameters, the cause of the transaction failure is obtained by troubleshooting according to the secondary correlation recorded in the fault spatiotemporal correlation analysis logic framework; wherein, the ETC transaction failure phenotype includes: abnormal signal state and abnormal equipment operating state.

[0006] In some embodiments, the fault analysis model pre-training step includes: Obtain a training sample set containing multiple samples. Each sample records one historical fault data. The historical fault data includes the transaction process data, the equipment condition, the inertial navigation data, the positioning data, and the scene type. The reason for the transaction failure is added as a label. An initial neural network is obtained, taking the transaction process data, the equipment operating condition, the inertial navigation data, the positioning data, and the scene type as inputs, and outputting a predicted value for the cause of the transaction failure; the initial neural network adopts a convolutional neural network, a fully connected neural network, or a recurrent neural network; the parameters of the initial neural network are updated by constructing a loss based on the deviation between the predicted value and the label, and the fault analysis model is obtained.

[0007] In some embodiments, the method further includes: Gaussian noise is added to the training sample set to simulate signal interference for data enhancement, and SMOTE oversampling is used to synthesize data for samples in which the proportion of one or more transaction failure reasons is less than a set value. The initial neural network adopts the form of transfer learning, which is pre-trained using sample data generated in the highway scenario and then transferred to the training sample set for training.

[0008] In some embodiments, the fault analysis model updates its parameters using the newly generated fault data at a set period.

[0009] In some embodiments, the method further includes: when the reason for the transaction failure is indicated as an abnormal device working state, initiating a self-check and self-correction of the abnormal device, and feeding back the correction result; The fault log is sent to the roadside unit via a preset communication path, including via a Bluetooth module, a Wi-Fi module, or a Zigbee communication module.

[0010] On the other hand, the present invention also provides an ETC transaction system based on inertial navigation, the system comprising: an on-board tag and a roadside unit; The vehicle-mounted tag includes an MCU controller and an inertial navigation module, a positioning module, a Bluetooth communication module, an radio frequency module, an embedded security control module, an anti-tamper circuit, and a power supply module connected to the MCU controller; the vehicle-mounted tag executes the above-described ETC transaction fault judgment method based on inertial navigation. The roadside unit includes an industrial control computer and a transaction antenna connected to the industrial control computer. The roadside unit performs fault investigation based on a preset fault spatiotemporal correlation analysis logic framework to obtain the reason for transaction failure, or inputs the fault data into a pre-trained fault analysis model and outputs the reason for transaction failure.

[0011] On the other hand, the present invention also provides a computer-readable storage medium having a computer program / instructions stored thereon, which, when executed by a processor, implement the steps of the above-described method.

[0012] On the other hand, the present invention also provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the above-described method.

[0013] The ETC transaction fault judgment method and ETC transaction system based on inertial navigation described in this invention introduces an inertial navigation module into the vehicle tag to record continuous motion state data. This module connects to a mobile phone via Bluetooth to update location information, eliminating accumulated errors in the inertial navigation system. Combined with location data, it performs geographic environment matching to identify the scenario type of the road segment where the transaction failed. A fault log is formed by aligning transaction process data, equipment status, the inertial navigation data, the location data, and the scenario type. A preset fault spatiotemporal correlation analysis logic framework is used for fault investigation, or a pre-trained fault analysis model is introduced to identify the cause of transaction failure. This invention establishes a complete causal chain for the driving route, the road segment where the transaction failed, motion state parameters, and the ETC transaction failure phenotype. It analyzes the causes of ETC transaction failures under the condition of full-dimensional perception of motion state, adapting to fault detection and cause investigation in complex transaction environments across multiple scenarios. It provides objective evidence from the motion trajectory dimension for warranty tag failures and user complaint scenarios, improving the accuracy and fairness of ETC operation services.

[0014] Additional advantages, objects, and features of the invention will be set forth in part in the description which follows, and will also become apparent in part to those skilled in the art upon studying the description, or may be learned by practice of the invention. The objects and other advantages of the invention can be realized and obtained by means of the structures specifically pointed out in the description and drawings.

[0015] Those skilled in the art will understand that the objectives and advantages achievable with the present invention are not limited to those specifically described above, and that the above and other objectives achievable with the present invention will become clearer from the following detailed description. Attached Figure Description

[0016] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, are not intended to limit the scope of the invention. In the drawings: Figure 1 This is a flowchart illustrating an embodiment of the ETC transaction fault judgment method based on inertial navigation according to an embodiment of the present invention.

[0017] Figure 2 This is a schematic diagram of the structure of the vehicle-mounted tag in an ETC transaction system based on inertial navigation, according to another embodiment of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the embodiments and accompanying drawings. Here, the illustrative embodiments and descriptions of this invention are used to explain the invention, but are not intended to limit the invention.

[0019] It should also be noted that, in order to avoid obscuring the invention with unnecessary details, only the structures and / or processing steps closely related to the solution according to the invention are shown in the accompanying drawings, while other details that are not closely related to the invention are omitted.

[0020] It should be emphasized that the term "including / comprises" as used herein refers to the presence of a feature, element, step, or component, but does not exclude the presence or addition of one or more other features, elements, steps, or components.

[0021] It should also be noted that, unless otherwise specified, the term "connection" in this article can refer not only to a direct connection, but also to an indirect connection involving an intermediary.

[0022] In the following description, embodiments of the invention will be illustrated with reference to the accompanying drawings. In the drawings, the same reference numerals represent the same or similar parts, or the same or similar steps.

[0023] This application aims to address the difficulty in locating transaction failures caused by traditional ETC tags in complex traffic environments. Existing ETC tags lack vehicle motion and environmental perception capabilities, failing to fully reconstruct the vehicle's trajectory and geographical information during a transaction. Due to fragmented dynamic scenes, missing multi-dimensional motion parameters, and weak correlation between trajectory and fault, it is impossible to determine and identify the cause of electronic tag transaction failures. For example, in mountainous areas with curves, tunnels, and multi-vehicle sections, it is difficult to determine whether the transaction failure is caused by dynamic driving states such as rapid acceleration or sharp turns, or by geographical factors. This invention embeds an inertial navigation chip within the ETC tag to collect vehicle dynamic parameters in real time. Combined with an optimized log storage structure, it achieves synchronous recording of motion status and environmental parameters during the transaction process, accurately locating transaction anomalies, improving the ability to analyze the causes of electronic tag faults, and thus proposing targeted improvement measures to enhance product quality.

[0024] This invention provides a novel vehicle-mounted ETC electronic tag, which, in addition to a DSRC communication module, a power supply module, a processor module, and an IC card reader / writer module, also includes an inertial navigation module and a Bluetooth communication module. The inertial navigation module collects the electronic tag's position and records its navigation position parameters; the Bluetooth communication module is bound to a mobile phone, receives the phone's location information, updates the electronic tag's position, and eliminates the accumulated errors of the inertial navigation module. The electronic tag can also have its own built-in satellite navigation positioning module, eliminating the need to receive location information from a mobile phone, and automatically updating its position and sending the updated position information to the inertial navigation module.

[0025] Specifically, this invention provides a method for judging ETC transaction faults based on inertial navigation, such as... Figure 1 As shown, the method includes the following steps S101~S103: Step S101: The inertial navigation module collects inertial navigation data at a set sampling frequency. This data includes three-dimensional acceleration, angular velocity, speed, heading angle, and pitch angle. The positioning module collects and updates positioning data at the same set sampling frequency. Based on this positioning data, geographic environment matching is performed to determine the scene type of the current road segment. Specifically, the system can communicate with the vehicle owner's mobile phone via Bluetooth at a set frequency to receive positioning data, update location information, and eliminate accumulated errors from the inertial navigation module.

[0026] Step S102: In response to the ETC transaction failure information, align the transaction process data, equipment status, inertial navigation data, positioning data and scene type within the time range set before the ETC transaction failure information according to the timestamp, and store them in the fault log.

[0027] Step S103: In response to the access request from the roadside unit, the fault log is sent to the roadside unit through a preset communication path. The roadside unit can then use the fault data in the fault log to troubleshoot the transaction failure based on a preset fault spatiotemporal correlation analysis logic framework, or input the fault data into a pre-trained fault analysis model and output the transaction failure reason. The fault spatiotemporal correlation analysis logic framework is obtained by analyzing the primary correlation between inertial navigation data, the scene type of the road segment, and the equipment operating condition, analyzing the secondary correlation between the equipment operating condition and the ETC transaction failure phenotype, and logically fusing the primary and secondary correlations. The fault analysis model is obtained by learning the mapping relationship between fault data and transaction failure reasons in historical fault logs. The ETC transaction failure phenotype is an abnormal transaction state presented based on the transaction process data.

[0028] Steps S101 to S103 can be performed based on the vehicle-mounted tag. In step S101, the vehicle-mounted tag can continuously collect data according to a fixed sampling period, or it can respond to the wake-up command (BST) from the roadside unit to collect relevant data. The roadside unit sends a wake-up command to the vehicle-mounted tag, activating it from its dormant state and preparing it for data collection. This command simultaneously activates the data collection functions of the inertial navigation module and the positioning module.

[0029] To incorporate complete motion state data, the vehicle tag in this application includes an inertial navigation module for collecting inertial navigation data. The inertial navigation module collects inertial navigation data at a set sampling frequency, including three-dimensional acceleration (acceleration in the x, y, and z directions), which can be used to analyze the vehicle's acceleration, deceleration, and roll; angular velocity (the speed at which the vehicle rotates around three axes), which reflects the vehicle's turning and swaying; travel speed; heading angle (the angle between the vehicle's direction of travel and a reference direction in the geographic coordinate system, such as the angle with true north); and pitch angle (the angle at which the front of the vehicle tilts up and down, such as when climbing or descending a slope).

[0030] The positioning module collects positioning data, such as latitude and longitude coordinates, at a set sampling frequency. Then, it matches this positioning data with a pre-stored road database to determine the scene type of the road segment where the vehicle is located, such as whether it is in a tunnel, a mountain bend, a bridge, or a flat urban road, or other special road segments or scenes.

[0031] In some embodiments, the scenario types include: tunnel sections, mountainous curved road sections, toll station sections, parking lot sections, urban elevated roads and interchange sections, long straight highway sections, and service area and ramp connection sections. The scenario types of different road sections have different logics and identification standards for fault detection, which can be used to guide the establishment of a logical framework for fault spatiotemporal correlation analysis and guide fault analysis.

[0032] For example, tunnel sections, being semi-enclosed road environments, experience multiple interferences to the signal transmission of the ETC system due to their unique physical structure. These include: 1) Signal shielding effect: The metal structure and reinforced concrete within the tunnel create a "Faraday cage" effect, severely attenuating the 5.8GHz microwave signal of the ETC system. Actual measurement data shows that in tunnels longer than 1km, the microwave signal strength decreases by an average of 62%, leading to communication interruptions between the vehicle tag and the roadside unit. When a vehicle attempts to transact in the middle of the tunnel, the success rate is less than 35%. 2) Multipath reflection interference: The regular surface of the tunnel walls causes multiple reflections of electromagnetic waves, resulting in signal delay extension. 3) Position drift: GPS signals are completely shielded within the tunnel. The ETC system relies on inertial navigation to calculate its position, and the cumulative error increases with the tunnel length. In tunnels longer than 3km, position drift can reach ±200 meters, leading to incorrect path marker matching. 4) Light change interference: The alternating light and dark conditions at tunnel entrances and exits affect the license plate recognition system. During testing in the Zhongnanshan Tunnel in Shaanxi, the license plate recognition error rate reached 28% under strong backlight conditions, leading to a "license plate mismatch" alarm in ETC transactions.

[0033] For example, in mountainous curved sections, the alignment characteristics of mountain highways pose unique challenges to ETC transactions. The main failure modes can be categorized into three types: 1) Curvature radius-related failures: When the curve radius is less than 800m, vehicles exhibit significant deflection when passing the ETC gantry. Test data shows that when the angle between the vehicle's longitudinal axis and the gantry normal exceeds 15°, the read / write success rate drops to 68%; at 30°, it further decreases to 41%. 2) Slope-related failures: Steep slopes (longitudinal slope > 5%) cause changes in the vehicle's pitch angle. When the pitch angle > 10°, the effective communication distance between the OBU device on the windshield and the gantry antenna is shortened by 28%. On continuous downhill sections, the braking of heavy-duty trucks creates a water mist barrier, further attenuating signal strength. 3) Obstruction effects: Mountain curves commonly encounter metal obstacles such as rock walls and crash barriers.

[0034] It is evident that the differences in the characteristics of objective environmental factors under different road scenarios can lead to some unique problems in the ETC transaction process. Based on this, a fault spatiotemporal correlation analysis logical framework can be proposed for different types of road segments by summarizing and generalizing, connecting the driving route, the road segment where the transaction failed, the motion state parameters, and the ETC transaction failure phenotype to form a complete causal chain.

[0035] In step S102, when an ETC transaction fails, the on-board unit (OIN) responds to the ETC transaction failure information, triggering corresponding data processing and storage operations. Multiple data points within a pre-defined time range are integrated to form a fault log, which can be exemplarily set at 5 seconds. Transaction process data may include communication data related to the ETC transaction, transaction request and response information, etc. Equipment operating conditions include OIN operating conditions and roadside unit operating conditions. OIN operating conditions include: OIN communication performance parameters, power supply status parameters, operating environment status parameters, and installation and operation status parameters; roadside unit operating conditions include roadside unit communication performance parameters and system operation status parameters. Inertial navigation data includes motion parameters such as three-dimensional acceleration and angular velocity collected in step S101. Positioning data includes the vehicle's latitude and longitude location information. Scene type data includes the scene type of the road segment. These data are aligned according to timestamps to ensure consistency and correlation in the time dimension, and then the aligned data is stored in the fault log for subsequent analysis.

[0036] In step S103, after the roadside unit confirms the transaction failure, it can send an access request to the vehicle tag, requesting the upload of fault data for a set period of time before the fault. Specifically, the data is encrypted with AES-256 and uses "vehicle VIN code + timestamp" as a dynamic key to prevent tampering or leakage.

[0037] This application introduces two diagnostic paths for transaction failure reasons: a rule-driven path that quickly analyzes common faults based on a preset logical framework, and a model-driven path that uses AI models to handle complex faults.

[0038] Among them, the rule-driven path introduces a rule engine to build a logical framework for spatiotemporal correlation analysis of faults, and establishes a causal chain between the environment and motion state, equipment operating conditions and ETC transaction failure phenotypes.

[0039] For example, if the on-board tag has a pitch angle greater than 12°, the angle between the antenna and the RSU signal will be greater than 22°, causing signal attenuation. The phenotype of ETC transaction failure is transaction link interruption.

[0040] In some embodiments, the cause of transaction failure is determined by troubleshooting based on a preset fault spatiotemporal correlation analysis logic framework, including steps S201~S203: Step S201: Compare the inertial navigation data with the predefined motion state parameter range to determine the abnormal motion state of the vehicle tag.

[0041] Step S202: Correct the standard parameter range of the equipment under standard conditions according to the scenario type when the fault occurs; compare the corrected standard parameter range with the equipment operating conditions to determine the abnormal operating condition parameters.

[0042] Step S203: Based on the ETC transaction failure phenotype, abnormal motion state, and abnormal operating condition parameters, and in accordance with the secondary correlation recorded in the fault spatiotemporal correlation analysis logical framework, conduct fault investigation to obtain the cause of transaction failure; among which, the ETC transaction failure phenotype includes: abnormal signal state and abnormal equipment operating state.

[0043] In step S201, by comparing and analyzing the motion state presented by the inertial navigation data in the fault data according to the preset motion state parameter range, predefined abnormal motion states can be identified. For example, continuous turning, continuous shaking, continuous downhill, etc. can be identified.

[0044] In step S202, the standard parameter range for the operating status of the equipment is adjusted and corrected according to the characteristics of different road segment scenarios to prevent excessive anomaly detection. For example, on mountain curves, the signal reception strength and transmission power of the vehicle tag will fluctuate reasonably due to obstructions; on urban flat roads with complex signals, the interference sources are complex, and the bit error rate will reasonably increase. Therefore, the standard parameter range for the operating conditions is corrected to adapt to reasonable changes in operating conditions under different scenarios.

[0045] In step S203, the fault spatiotemporal correlation analysis logical framework establishes a causal chain of motion state, equipment condition and transaction failure phenotype. By comparing the abnormal objects identified in steps S201 and S202 with the fault spatiotemporal correlation analysis logical framework, the cause of ETC transaction failure can be obtained.

[0046] In some embodiments, the fault analysis model pre-training step includes steps S301-S302: Step S301: Obtain a training sample set containing multiple samples. Each sample records one historical fault data. The historical fault data includes the transaction process data, the equipment operating condition, the inertial navigation data, the positioning data, and the scene type. The reason for the transaction failure is added as a label. Step S302: Obtain an initial neural network, taking the transaction process data, the equipment condition, the inertial navigation data, the positioning data, and the scene type as inputs, and outputting a predicted value for the reason for the transaction failure; the initial neural network adopts a convolutional neural network, a fully connected neural network, or a recurrent neural network; the parameters of the initial neural network are updated by constructing a loss based on the deviation between the predicted value and the label, to obtain the fault analysis model.

[0047] A large sample of fault data was collected from the historical records of the ETC system. An expert team or existing fault analysis reports were organized to conduct a detailed analysis of each historical fault data point, determine the specific reasons for the transaction failure, and add these reasons as tags to the sample.

[0048] For example, reasons for transaction failure may include: 1) Signal transmission problems, including insufficient signal strength, signal interference such as multipath interference, electromagnetic interference, and excessive signal transmission distance. 2) Equipment failure, such as OBU hardware failure (e.g., damaged antenna, chip failure), depleted battery, and RSU failure. 3) Vehicle movement, such as the vehicle speed exceeding the ETC system's design speed range, rapid acceleration, rapid deceleration, or severe bumps, leading to transaction failure. For example, a vehicle passing through an ETC lane at a speed exceeding 160 km / h exceeds the speed range that the system can accurately identify and process.

[0049] Before adding labels, the collected data is preprocessed to ensure data quality and consistency. This includes data cleaning to remove erroneous or unreasonable data, such as acceleration values ​​that are significantly beyond the physical range; data standardization to convert data of different dimensions to a uniform scale for easier model processing; and data imputation to fill in missing data appropriately, such as by interpolating data from adjacent time points.

[0050] Choosing an initial neural network structure, a Convolutional Neural Network (CNN) can be used, suitable for processing spatially correlated data. In ETC fault analysis, transaction process data and inertial navigation data can be considered to have temporal correlation. For example, arranging vehicle acceleration data over a period of time into a one-dimensional sequence, a CNN can extract local features, such as short-term trends in acceleration, through convolutional kernels. These trends may be related to abnormal motion states such as rapid acceleration and sudden braking of the vehicle, thus affecting ETC transactions. Alternatively, a Fully Connected Neural Network (FNN) can be chosen, suitable for processing simple data with no obvious correlation between features. If all features, such as transaction process data, equipment operating conditions, and inertial navigation data, are flattened into one-dimensional vectors and then input into a fully connected neural network, the network can learn the complex nonlinear relationships between the features. For example, a fully connected layer can detect that the risk of transaction failure increases when the OBU battery is low and the vehicle is traveling at high speed. Furthermore, Recurrent Neural Networks (RNNs) and their variants, such as LSTM and GRU, can be chosen, specifically for processing sequential data and capable of capturing time-series features in the data. In ETC scenarios, the vehicle's motion state and transaction process form a time-varying sequence. For example, RNNs can store sequential data such as speed and acceleration changes over a period before a vehicle enters the ETC lane, combining this information with current data to better understand the impact of the vehicle's trajectory and state changes on the transaction. LSTMs, for instance, can selectively retain or forget information in the sequence through their gating structure, making them highly effective for analyzing the dynamic behavior of vehicles during transactions.

[0051] During training, information such as transaction process data, equipment operating conditions, inertial navigation data, positioning data, and scene type are integrated to construct an input feature vector.

[0052] The initial neural network's weights and biases are randomly initialized, or predefined initialization methods (such as He initialization or Xavier initialization) are used to set them to facilitate model training. The constructed input feature vector is fed into the initial neural network, which then performs forward propagation to obtain predicted values ​​for the reasons for transaction failure. The predicted value is a probability distribution vector representing the probability that the sample belongs to each possible category of transaction failure reason; or it is a specific class label. The predicted value is compared with the sample label to calculate the loss function. The loss function measures the difference between the predicted value and the true value; commonly used loss functions include cross-entropy loss and mean squared error loss. For example, in multi-class classification problems, cross-entropy loss can calculate the difference between the predicted probability distribution and the true class distribution. Based on the calculated loss value, the gradient of each parameter in the network with respect to the loss is calculated using the backpropagation algorithm. Then, optimization algorithms are used to update the network parameters, such as stochastic gradient descent and the Adam optimizer. The purpose of the update is to reduce the value of the loss function, making the model's predictions closer to the true labels. For example, the Adam optimizer can efficiently adjust network parameters and accelerate model convergence by combining momentum and adaptive learning rate.

[0053] The process of forward propagation, loss calculation, backpropagation, and parameter update is repeated multiple times across the entire training sample set, i.e., multiple training epochs. Within each epoch, batch gradient descent, stochastic gradient descent, or mini-batch gradient descent can be used for training. For example, mini-batch gradient descent uses a small subset of samples each time to calculate the gradient and update the parameters, thus improving training efficiency while ensuring model convergence.

[0054] In some embodiments, the method further includes steps S301 and S302: Step S301: Add Gaussian noise to the training sample set to simulate signal interference for data augmentation, and use SMOTE oversampling to synthesize data for samples where the proportion of one or more transaction failure reasons is less than a set value.

[0055] Step S302: The initial neural network adopts the form of transfer learning, which is pre-trained using sample data generated in the highway scenario and then transferred to the training sample set for training.

[0056] In step S301, Gaussian noise is a type of noise whose probability density function follows a normal distribution. Signal interference is a common problem in ETC (Electronic Toll Collection) transactions, such as electromagnetic interference from other electronic devices and multipath effects. By adding Gaussian noise, such as received signal strength and signal quality indicators, to the signal-related data in the training sample set, we can simulate signal fluctuations and distortions caused by these interference factors in real-world environments. This makes the trained model more robust and better able to cope with various complex signal environments in practical applications.

[0057] In the training sample set, there may be instances where the number of samples for certain reasons for transaction failure is too small. For example, transaction failures caused by "device hardware failure" may only account for about 1% of the total samples, while "signal transmission problems" may account for 50%. This data imbalance can cause the model to perform well in predicting the majority class during training, but poorly in predicting the minority class, because the model tends to predict the majority class to achieve higher overall accuracy. SMOTE (Synthetic Minority Over-sampling Technique) is a commonly used oversampling technique to address the data imbalance problem. For each minority class sample, the algorithm calculates the feature space distance between it and its nearest neighbor sample, and then randomly generates a synthetic sample on the line connecting the sample and its nearest neighbor. In the ETC failure analysis scenario, for minority class samples where the proportion of transaction failure reasons is less than a set value, such as "OBU software failure" accounting for only 1% of the samples, the k nearest neighbors of each "OBU software failure" sample are first determined, with k typically set to around 5. Then, for each sample, a nearest neighbor is randomly selected, and the difference vector between the two is calculated. A new synthetic sample is then generated along this difference vector in a certain proportion. In this way, the number of minority class samples can be increased to a relatively reasonable proportion, thereby balancing the training dataset and improving the model's ability to identify the reasons for minority class transaction failures.

[0058] In step S302, transfer learning is a machine learning method that applies knowledge and model parameters learned in the source domain to the target domain. In this step, the source domain is ETC transaction data in highway scenarios, and the target domain is a training sample set containing various complex scenarios. By pre-training the initial neural network on the source domain data, the network can learn general feature representations related to ETC transactions, such as basic signal transmission patterns and general characteristics of equipment operating conditions. These features also have certain applicability in the target domain. Using transfer learning can reduce training time and computational resource consumption on the target domain training sample set. Because the network has already learned some basic and general features during the source domain pre-training stage, these features do not need to be learned from scratch, thereby accelerating the model's convergence speed. In addition, transfer learning can also improve model performance, especially when the number of training samples in the target domain is limited. The pre-training knowledge in the source domain can provide a useful supplement, preventing problems such as overfitting during model training in the target domain.

[0059] In some embodiments, the fault analysis model updates its parameters using newly generated fault data at set intervals.

[0060] In some embodiments, the method further includes: when the reason for transaction failure is indicated as abnormal device operating status, initiating self-check and self-correction of the abnormal device, and feeding back the correction result.

[0061] The fault log is sent to the roadside unit via a preset communication path (DSRC communication), or after authentication, it is sent to a dedicated receiving device via a wireless communication interface (including Bluetooth, Wi-Fi, or Zigbee, etc.). The dedicated receiving device can be handheld or a fixed roadside device.

[0062] On the other hand, the present invention also provides an ETC transaction system based on inertial navigation, the system comprising: an on-board tag and a roadside unit.

[0063] like Figure 2 As shown, the vehicle tag includes an MCU controller and an inertial navigation module, a Bluetooth module, a positioning module, a radio frequency module, an embedded security control module, an anti-tamper circuit, and a power module connected to the MCU controller; the vehicle tag performs the ETC transaction fault judgment method based on inertial navigation described in steps S101 to S103 above.

[0064] The roadside unit includes an industrial control computer and a transaction antenna connected to the industrial control computer. The roadside unit performs fault investigation based on a preset fault spatiotemporal correlation analysis logic framework to obtain the reason for transaction failure, or inputs the fault data into a pre-trained fault analysis model and outputs the reason for transaction failure.

[0065] Furthermore, the vehicle-mounted tag also includes a buzzer and LED indicator lights. Embedded safety control module. On the other hand, the present invention also provides a computer-readable storage medium having a computer program / instructions stored thereon, which, when executed by a processor, implement the steps of the above-described method.

[0066] On the other hand, the present invention also provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the above-described method.

[0067] The present invention will now be described in conjunction with specific embodiments: The purpose of this invention is to provide a method and system for logging and determining problems using ETC tags integrated with inertial navigation. By integrating an inertial navigation chip into the tag, it collects real-time vehicle motion parameters in all dimensions, including 3D acceleration, angular velocity, latitude and longitude, heading angle, and pitch angle. It receives location data from the vehicle owner's mobile phone, updates location information, and establishes a 3D spatiotemporal log of "transaction time - geographical location - motion state," enabling spatiotemporal tracing of faults. Based on the fusion analysis of inertial navigation data and transaction logs, it accurately identifies transaction anomalies caused by vehicle dynamics (such as sudden braking or continuous lane changes) or geographical environment (such as tunnels or mountainous areas). This provides objective evidence of motion trajectory dimensions for tag malfunctions during the warranty period and user complaint scenarios, accurately analyzes the reasons for electronic tag transaction failures, and improves the accuracy and fairness of ETC operation services.

[0068] In this solution, the vehicle-mounted tag integrates an inertial navigation chip, which can output motion state parameters, including three-dimensional acceleration, angular velocity, speed, heading angle, and pitch angle. An integrated positioning chip acquires location parameters, such as latitude and longitude. In response to transaction failure information during ETC transactions, the vehicle-mounted tag records the driving trajectory points for the previous 5 seconds, one point per second, including latitude, longitude, and motion state parameters.

[0069] The log generation process includes the following steps: After the vehicle tag is activated, the inertial navigation chip collects motion and position data at 100ms intervals. Scene and trajectory association: After determining the lane / gantry scene through BeaconID, the inertial navigation data is aligned with the timestamps of transaction process data (such as DSRC communication status and IC card information) to generate a spatiotemporal log.

[0070] Failure log generation rules: If a transaction fails, the latitude, longitude, heading angle, acceleration, and other data at the time of failure are stored in EF02, and the trajectory for the previous 5 seconds is recorded; if a transaction succeeds but triggers an abnormal movement (such as speed > 100km / h, sudden change in heading angle > 30°), it is stored in the "Successful Transaction with Abnormal Movement" sub-record of EF02. Every hour, the electronic tag's Bluetooth module communicates with the vehicle owner's mobile phone to update the current location information and eliminate the accumulated error of the electronic tag's inertial navigation module.

[0071] For some electronic tag models with navigation and positioning functions, they can update their location information periodically without needing to receive location information from mobile phones. The collection and updating of location information are carried out independently, do not participate in the transaction process, and do not affect the information security of the existing ETC system.

[0072] Log storage optimization includes the following: Transaction success rate file (EF07) extension: Add a motion status statistics field to the EF07 file under the DF01 directory.

[0073] High-speed transaction percentage: The percentage of transactions recorded at speeds ≥ 80 km / h.

[0074] Statistical analysis of the percentage of transactions involving sharp turns: Record the percentage of transactions where the rate of change in heading angle is ≥15° / s.

[0075] Statistical tunnel transaction percentage: The percentage of transactions based on latitude and longitude matching with the tunnel geographic database.

[0076] Example 1: For scenarios where transactions fail on mountain slopes, ETC transactions may fail when a vehicle is climbing a mountain highway.

[0077] The log records show that during the uphill phase, the pitch angle was greater than 12° for 15 seconds, reaching a maximum of 18°; uphill resistance caused the longitudinal acceleration to fluctuate between 0.8 and 1.2 m / s²; the speed at the time of failure was 45 km / h, within the medium-speed trading range. File EF02 records a pitch angle of 17.5° at the time of failure, which gradually increased from 10° to 17.5° in the preceding 5 seconds; the longitudinal acceleration was 1.1 m / s² in the second before failure, with a velocity change rate of... .

[0078] Problem identification: The business system matched the pitch angle data with the road database and confirmed that the location was a continuous uphill section with a 5% slope. Based on the threshold of the optimal communication angle of the roadside unit antenna, it was determined that the excessive pitch angle of the vehicle body caused the antenna and the vehicle tag signal to be too large, with the measured angle reaching 22°, which caused signal attenuation and interruption of the transaction link.

[0079] Example 2: A scenario where a transaction fails due to rapid acceleration on a highway: the transaction fails when the vehicle accelerates rapidly to enter the ETC lane.

[0080] The log records show that in the 3 seconds prior to the failure, the longitudinal acceleration suddenly increased from 0.5 m / s² to 3.8 m / s², indicating rapid acceleration; the speed increased from 30 km / h to 65 km / h within 2 seconds, with a rate of change of 17.5%. The peak three-dimensional acceleration along the x-axis is 3.8 m / s², and along the y-axis it is 0.5 m / s², indicating slight body roll. At the moment the EF02 file failed to record, the longitudinal acceleration was 3.5 m / s², and the speed was 65 km / h. The trajectory for the first 5 seconds shows the vehicle merging from the service area ramp onto the main road, with a heading angle change of 12°.

[0081] Problem identified: Triggering a rapid acceleration caused the trading area to shrink, resulting in trading failure.

[0082] The parameters such as the information recording time interval, sampling time, and update frequency in this embodiment can be reset according to the performance of the electronic tag and the collection requirements.

[0083] In summary, the ETC transaction fault judgment method and ETC transaction system based on inertial navigation described in this invention introduces an inertial navigation module into the vehicle tag to record continuous motion state data. Combined with positioning data, it performs geographic environment matching to identify the scenario type of the road segment where the transaction failed. It aligns transaction process data, equipment status, the inertial navigation data, the positioning data, and the scenario type to form a fault log. A preset fault spatiotemporal correlation analysis logic framework is used for fault investigation, or a pre-trained fault analysis model is introduced to identify the cause of transaction failure. This invention establishes a complete causal chain for the driving route, the road segment where the transaction failed, motion state parameters, and the ETC transaction failure phenotype. It analyzes the cause of ETC transaction failure under the condition of full-dimensional perception of motion state, adapting to fault detection and cause investigation in complex transaction environments across multiple scenarios. It provides objective evidence of motion trajectory dimensions for warranty period tag failures and user complaint scenarios, improving the accuracy and fairness of ETC operation services.

[0084] Those skilled in the art will understand that the exemplary components, systems, and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software, or a combination of both. Whether implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention. When implemented in hardware, it can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this invention are programs or code segments used to perform the desired tasks. The programs or code segments can be stored in a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried in a carrier wave.

[0085] It should be clarified that the present invention is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of the present invention.

[0086] In this invention, features described and / or illustrated for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with or in place of features of other embodiments.

[0087] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, various modifications and variations of the embodiments of the present invention are possible. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for judging ETC transaction faults based on inertial navigation, characterized in that, The method includes the following steps: The inertial navigation module collects inertial navigation data at a set sampling frequency, including three-dimensional acceleration, angular velocity, speed, heading angle, and pitch angle; the positioning module collects and updates positioning data at the set sampling frequency, and performs geographic environment matching based on the positioning data to determine the scene type of the road segment. In response to ETC transaction failure information, the transaction process data, equipment status, inertial navigation data, positioning data, and scene type within a predetermined time range prior to the ETC transaction failure information are aligned according to timestamps and stored in the fault log; In response to an access request from a roadside unit, the fault log is sent to the roadside unit via a preset communication path. The roadside unit then uses the fault data in the fault log to troubleshoot and determine the cause of the transaction failure based on a preset fault spatiotemporal correlation analysis logic framework, or it inputs the fault data into a pre-trained fault analysis model and outputs the cause of the transaction failure. The fault spatiotemporal correlation analysis logic framework is obtained by analyzing the primary correlation between the inertial navigation data, the scene type of the road segment, and the equipment operating condition; analyzing the secondary correlation between the equipment operating condition and the ETC transaction failure phenotype; and logically fusing the primary and secondary correlations. The fault analysis model is obtained by learning the mapping relationship between fault data and transaction failure causes in historical fault logs. The ETC transaction failure phenotype is an abnormal transaction state presented based on the transaction process data. The process of troubleshooting transaction failures based on a pre-defined spatiotemporal correlation analysis framework includes: comparing the inertial navigation data with a predefined range of motion state parameters to determine the abnormal motion state of the vehicle tag; correcting the standard parameter range of the equipment under standard conditions according to the scenario type at the time of the fault; comparing the corrected standard parameter range with the equipment operating conditions to determine abnormal operating condition parameters; and, based on the ETC transaction failure phenotype, the abnormal motion state, and the abnormal operating condition parameters, troubleshooting to obtain the transaction failure reasons according to the secondary correlation relationships recorded in the spatiotemporal correlation analysis framework. The ETC transaction failure phenotypes include: abnormal signal state and abnormal equipment operating state. The pre-training steps of the fault analysis model include: acquiring a training sample set containing multiple samples, each sample recording one historical fault data point, the historical fault data including the transaction process data, the equipment operating condition, the inertial navigation data, the positioning data, and the scene type, and adding the transaction failure reason as a label; acquiring an initial neural network, taking the transaction process data, the equipment operating condition, the inertial navigation data, the positioning data, and the scene type as input, and outputting a predicted value for the transaction failure reason; the initial neural network adopts a convolutional neural network, a fully connected neural network, or a recurrent neural network; using the deviation between the predicted value and the label to construct a loss and update the parameters of the initial neural network to obtain the fault analysis model.

2. The ETC transaction fault judgment method based on inertial navigation according to claim 1, characterized in that, The scenario types include: tunnel sections, mountain winding road sections, toll station sections, parking lot sections, urban elevated roads and interchange sections, long straight highway sections, and service area and ramp connection sections; The equipment operating conditions include vehicle-mounted tag operating conditions and roadside unit operating conditions. The vehicle-mounted tag operating conditions include: vehicle-mounted tag communication performance parameters, power supply status parameters, operating environment status parameters, and installation and operation status parameters. The roadside unit operating conditions include roadside unit communication performance parameters and system operation status parameters.

3. The ETC transaction fault judgment method based on inertial navigation according to claim 1, characterized in that, The method further includes: Gaussian noise is added to the training sample set to simulate signal interference for data enhancement, and SMOTE oversampling is used to synthesize data for samples in which the proportion of one or more transaction failure reasons is less than a set value. The initial neural network adopts the form of transfer learning, which is pre-trained using sample data generated in the highway scenario and then transferred to the training sample set for training.

4. The ETC transaction fault judgment method based on inertial navigation according to claim 3, characterized in that, The fault analysis model updates its parameters using the newly generated fault data at a set periodic rate.

5. The ETC transaction fault judgment method based on inertial navigation according to claim 1, characterized in that, The method further includes: When the reason for the transaction failure is indicated as abnormal device operating status, a self-check and self-correction of the abnormal device is initiated, and the correction results are fed back. The fault log is sent to the roadside unit via a preset communication path, including via a Bluetooth module, a Wi-Fi module, or a Zigbee communication module.

6. An ETC transaction system based on inertial navigation, characterized in that, The system includes: vehicle-mounted tags and roadside units; The vehicle tag includes an MCU controller and an inertial navigation module, a positioning module, an radio frequency module, an embedded security control module, an anti-tamper circuit, and a power supply module connected to the MCU controller; the vehicle tag performs the ETC transaction fault judgment method based on inertial navigation as described in any one of claims 1 to 5; The roadside unit includes an industrial control computer and a transaction antenna connected to the industrial control computer. The roadside unit performs fault investigation based on a preset fault spatiotemporal correlation analysis logic framework to obtain the reason for transaction failure, or inputs the fault data into a pre-trained fault analysis model and outputs the reason for transaction failure.

7. A computer-readable storage medium having a computer program / instructions stored thereon, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method as described in any one of claims 1 to 5.

8. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1 to 5.

Images