Vehicle control methods and devices, vehicles, storage media and software products

Abstract

This application provides a vehicle control method and device, a vehicle, a storage medium, and a program product. The control method includes: determining signal anomaly characteristics of the network signal based on communication parameters of the vehicle's network signal collected in the current driving scenario, wherein the current driving scenario is determined based on synchronously collected vehicle driving data; identifying the signal anomaly characteristics using an anomaly rule set corresponding to the current driving scenario in a preset rule base to obtain a first identification result; identifying the signal anomaly characteristics in the current driving scenario using a preset identification model to obtain a second identification result; determining the anomaly level of the network signal based on the first and second identification results; and controlling the vehicle to execute a preset response strategy based on the anomaly level of the network signal. This solution identifies signal anomaly characteristics in specific driving scenarios through a rule base and identification model, significantly reducing the possibility of false alarms and missed alarms.

Description

Technical Field

[0001] This application relates to, but is not limited to, the field of vehicle technology, and in particular to a vehicle control method and apparatus, a vehicle, a storage medium, and a program product. Background Technology

[0002] With the widespread adoption of intelligent connected vehicles, 4G / 5G networks have become the foundation for core functions such as in-vehicle navigation, remote control, OTA (Over-the-Air) upgrades, V2X (Vehicle to Everything) vehicle-to-everything (V2X) interaction, and in-vehicle entertainment. The stability of the 4G / 5G signal on the vehicle side directly affects vehicle driving safety and user experience. However, related technologies for monitoring 4G / 5G signals suffer from problems such as false alarms, missed alarms, and low efficiency. Summary of the Invention

[0003] One objective of this application is to provide a vehicle control method to solve the problems of false alarms, missed alarms, and low efficiency in the monitoring of 4G / 5G signals in related technologies; a second objective is to provide a vehicle control device; a third objective is to provide a vehicle; a fourth objective is to provide a computer-readable storage medium; and a fifth objective is to provide a computer program product.

[0004] The technical solution of this application embodiment is implemented as follows: This application provides a vehicle control method, the control method including: Based on the communication parameters of the vehicle's network signal collected in the current driving scenario, the signal anomaly characteristics of the network signal are determined, wherein the current driving scenario is determined based on the vehicle's driving data collected synchronously. Using the abnormal rule set corresponding to the current driving scenario in the preset rule base, the abnormal features of the signal are identified to obtain a first identification result. The rule base includes at least two abnormal rule sets corresponding to driving scenarios. Using a preset recognition model, the abnormal signal features in the current driving scenario are identified to obtain a second recognition result; Based on the first identification result and the second identification result, the anomaly level of the network signal is determined; Based on the anomaly level of the network signal, the vehicle is controlled to execute a preset response strategy.

[0005] In this embodiment, firstly, signal anomaly characteristics of the current driving scenario are determined based on simultaneously collected driving data and network signal communication parameters, achieving a high correlation between signal anomaly characteristics and the actual driving scenario, thus improving the accuracy and scenario adaptability of signal anomaly feature extraction. Secondly, signal anomaly characteristics are identified using a rule base and a recognition model respectively, yielding a first recognition result and a second recognition result. This combines the determinism of rule matching with the generalization ability of model recognition, taking into account both known and unknown anomaly recognition, reducing the possibility of false alarms and missed alarms caused by a single method, and greatly improving the comprehensiveness and accuracy of anomaly recognition. Finally, the anomaly level of the network signal is determined based on the two recognition results, and the vehicle is controlled according to the anomaly level, realizing refined classification and differentiated control of network signal anomalies, improving the intelligence and safety of the vehicle in dealing with network signal anomalies. In addition, since the technical solution of this application can be quickly integrated into existing vehicles without large-scale modification of the original vehicle hardware, deployment is convenient and compatibility is strong.

[0006] In some implementations, determining the anomaly level of the network signal based on the first identification result and the second identification result includes: determining a target identification result based on the first identification result and the second identification result; and determining the anomaly level of the network signal based on the target identification result.

[0007] In this embodiment of the application, by first determining a unified target identification result based on two identification results, and then determining the anomaly level based on the target identification result, a hierarchical conflict arbitration logic is realized to be compatible with known, unknown, complex and other anomalies. This reduces the possibility of directly using two potentially conflicting identification results for classification, making the anomaly level determination process more reasonable and accurate.

[0008] In some implementations, determining the target identification result based on the first identification result and the second identification result includes: if the first identification result and the second identification result are compatible, using either the first identification result or the second identification result as the target identification result; if the first identification result and the second identification result are not compatible, determining the target identification result based on the anomaly level corresponding to the first identification result and the anomaly level corresponding to the second identification result.

[0009] In the embodiments of this application, when the two identification results match, one of them is directly used as the target identification result, ensuring processing efficiency; when the two identification results do not match, a comprehensive judgment is made based on their respective anomaly levels to make full use of the advantages of rule matching and intelligent identification, thereby obtaining a more reliable target identification result, improving the accuracy and robustness of anomaly judgment, and thus ensuring the reliability and security of vehicle-side 4G / 5G signal early warning.

[0010] In some implementations, the target identification result includes target anomaly type, target severity, and target duration. The target anomaly type includes one of the following: signal quality type, transmission type, and security type. Determining the anomaly level of the network signal based on the target identification result includes: acquiring a target grading strategy adapted to the target anomaly type, wherein the target grading strategy characterizes the relationship between each severity, each duration, and each anomaly level under the target anomaly type; and determining an anomaly level that matches both the target severity and the target duration according to the target grading strategy.

[0011] In this application, by refining the target identification results into three dimensions—anomaly type, severity, and duration—and adopting different grading strategies for different anomaly types to determine the anomaly level, a multi-dimensional and refined grading of network signal anomalies is achieved. This makes the determination of anomaly levels more scientific and in line with actual driving scenarios, avoiding misjudgments caused by a one-size-fits-all approach. Consequently, it can provide a more reliable and refined decision-making basis for subsequent differentiated early warning and emergency response.

[0012] In some implementations, determining the signal anomaly characteristics of the network signal based on the communication parameters of the vehicle's network signal collected in the current driving scenario includes: preprocessing the communication parameters to obtain target communication parameters; using principal component analysis dimensionality reduction algorithm to extract features from the target communication parameters to obtain a first feature characterizing the overall signal quality trend; constructing anomaly features from the target communication parameters to obtain a second feature; and fusing the first feature and the second feature to obtain the signal anomaly characteristics.

[0013] In this embodiment, communication parameters are preprocessed to filter out noise and invalid data in the original signal, thereby improving data quality. Principal component analysis is used to extract a first feature representing the overall signal quality trend and reduce the data dimension. At the same time, a second feature representing specific anomalies is constructed. Finally, the two features are fused, which preserves the macro trend of the signal and highlights the micro anomaly details, making the extracted signal anomaly features more comprehensive and effective. This improves the accuracy of anomaly identification and reduces false alarms and false negatives.

[0014] In some implementations, the response strategy includes at least one of the following: maintaining the vehicle's network function, switching the vehicle's network frequency band, issuing an alarm, or disconnecting some of the vehicle's applications from the network; the step of controlling the vehicle to execute a preset response strategy based on the anomaly level of the network signal includes: maintaining the vehicle's network function when the anomaly level of the network signal is at a first level; switching the vehicle's network frequency band and issuing an alarm according to a preset first display method when the anomaly level of the network signal is at a second level; and disconnecting some of the vehicle's applications from the network and issuing an alarm according to a preset second display method when the anomaly level of the network signal is at a third level.

[0015] In this application embodiment, based on different levels of network signal anomalies, differentiated control strategies are implemented, ranging from maintaining network functionality to switching frequency bands and issuing alarms, and then disconnecting some application networks and issuing alarms. This achieves gradient and refined emergency response to network signal anomalies, which can not only effectively avoid the risk of vehicle function failure caused by signal anomalies, but also improve the user experience while ensuring the safety of the vehicle's core functions.

[0016] In some embodiments, the control method further includes: uploading the communication parameters, the first identification result, and the second identification result to a server, wherein the server is used to update the rule base and the identification model based on the communication parameters, the first identification result, and the second identification result.

[0017] In this embodiment, by uploading data such as communication parameters and two recognition results to the server, a data foundation is provided for the server to update the rule base and recognition model, realizing collaborative iteration and closed-loop optimization between the vehicle and the cloud, which can continuously improve the accuracy of anomaly recognition, thereby effectively reducing false alarms and false negatives.

[0018] This application provides a vehicle control device, characterized in that the control device includes: The first determining module is used to determine the signal anomaly characteristics of the network signal based on the communication parameters of the network signal of the vehicle collected in the current driving scenario, wherein the current driving scenario is determined based on the driving data of the vehicle collected synchronously. The first identification module is used to identify the abnormal features of the signal by using the abnormal rule set corresponding to the current driving scenario in the preset rule base, and to obtain a first identification result. The rule base includes at least two abnormal rule sets corresponding to driving scenarios. The second identification module is used to identify the abnormal signal features in the current driving scenario using a preset identification model, and obtain a second identification result; The second determining module is used to determine the anomaly level of the network signal based on the first identification result and the second identification result; The control module is used to control the vehicle to execute a preset response strategy based on the anomaly level of the network signal.

[0019] This application also provides a vehicle, including a processor and a memory, wherein the memory stores a computer program or instructions that, when executed by the processor, implement any of the methods described above.

[0020] This application also provides a computer-readable storage medium storing a computer program or instructions that, when executed by a processor, implement any of the methods described above.

[0021] This application also provides a computer program product, which includes a computer program or instructions that, when executed by a processor, implement the method described above.

[0022] The beneficial effects of this application are: (1) The signal anomaly features are identified by rule base and recognition model respectively, and the first recognition result and the second recognition result are obtained. This combines the determinism of rule matching and the generalization ability of model recognition, takes into account both known and unknown anomaly recognition, reduces the possibility of false alarms and missed alarms caused by a single method, and greatly improves the comprehensiveness and accuracy of anomaly recognition.

[0023] (2) Based on the communication parameters of the driving data and network signals collected simultaneously, the signal anomaly characteristics of the current driving scenario are determined, so as to achieve a high correlation between the signal anomaly characteristics and the actual driving scenario, thereby improving the accuracy of signal anomaly characteristic extraction and scenario adaptability.

[0024] (3) The communication parameters are preprocessed to filter out noise and invalid data in the original signal, improve the data quality, and the principal component analysis dimensionality reduction algorithm is used to extract the first feature representing the overall signal quality trend and reduce the data dimension. At the same time, the second feature representing the specific anomaly is constructed. Finally, the two features are fused, which not only preserves the macro trend of the signal, but also highlights the micro anomaly details, making the extracted signal anomaly features more comprehensive and effective.

[0025] (4) By first determining a unified target identification result based on two identification results, and then determining the anomaly level based on the target identification result, the hierarchical conflict arbitration logic is realized to be compatible with known, unknown, complex and other anomalies, reducing the possibility of directly using two potentially conflicting identification results for classification, making the anomaly level determination process more reasonable and accurate.

[0026] (5) When the two identification results match, one of them is directly used as the target identification result, which ensures processing efficiency; when the two identification results do not match, a comprehensive judgment is made based on their respective anomaly levels, so as to make full use of the advantages of rule matching and intelligent identification, thereby obtaining a more reliable target identification result, improving the accuracy and robustness of anomaly judgment, and thus ensuring the reliability and security of vehicle-side 4G / 5G signal early warning.

[0027] (6) The target identification results are refined into three dimensions: anomaly type, severity and duration. Different grading strategies are adopted for different anomaly types to determine the anomaly level. This achieves multi-dimensional and refined grading of network signal anomalies, making the anomaly level judgment more scientific and in line with actual driving scenarios. It avoids misjudgment caused by a one-size-fits-all approach, and can provide a more reliable and refined decision-making basis for subsequent differentiated early warning and emergency response.

[0028] (7) Based on different levels of network signal anomalies, differentiated control strategies are implemented, ranging from maintaining network functions to switching frequency bands and issuing alarms, and then disconnecting some application networks and issuing alarms. This achieves gradient and refined emergency response to network signal anomalies, which can not only effectively avoid the risk of vehicle function failure caused by signal anomalies, but also improve user experience while ensuring the safety of the vehicle's core functions.

[0029] (8) Upload communication parameters, two recognition results and other data to the server, providing a data foundation for the server to update the rule base and recognition model, realizing collaborative iteration and closed-loop optimization between the vehicle and the cloud, which can continuously improve the accuracy of anomaly recognition, thereby effectively reducing false alarms and false negatives.

[0030] (9) The technical solution of this application can be quickly integrated into existing vehicles without large-scale modification of the original vehicle hardware, making it easy to deploy and highly compatible. Attached Figure Description

[0031] Figure 1 A schematic diagram of the implementation process of a vehicle control method provided in this application embodiment. Figure 1 ; Figure 2 This is a schematic diagram of the composition structure of a signal warning system provided in an embodiment of this application; Figure 3 A schematic diagram of the implementation process of a vehicle control method provided in this application embodiment. Figure 2 ; Figure 4 This is a schematic diagram of the composition structure of a vehicle control device provided in an embodiment of this application; Figure 5 This is a schematic diagram of the hardware structure of a vehicle provided in an embodiment of this application.

[0032] It should be noted that the terms "first" and "second" mentioned above are only used to distinguish between different options and do not represent the degree of superiority or inferiority of the options or their priority in the implementation process. Detailed Implementation

[0033] The embodiments of this application will be described below with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be understood that the preferred embodiments are only for illustrating this application and are not intended to limit the scope of protection of this application.

[0034] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this application. Therefore, the drawings only show the components related to this application and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0035] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0036] In the following description, the terms "first, second, third" are used merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first, second, third" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0037] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0038] The technical solutions in the embodiments of this application will now be clearly and completely described with reference to the accompanying drawings.

[0039] Figure 1 This is a schematic diagram illustrating the implementation flow of a vehicle control method provided in an embodiment of this application, as shown below. Figure 1 As shown, the control method includes steps S11 to S15, wherein: Step S11: Based on the communication parameters of the vehicle's network signal collected in the current driving scenario, determine the signal anomaly characteristics of the network signal. The current driving scenario is determined based on the synchronously collected vehicle driving data.

[0040] Here, network signal refers to the radio signal transmitted by a vehicle when interacting with an external network via an onboard communication module (such as a 4G / 5G module). The quality of the network signal directly affects the stability and security of functions such as in-vehicle navigation, remote control, OTA upgrades, and V2X vehicle-to-everything (V2X) communication. In some implementations, the network signal may include, but is not limited to, 4G and 5G.

[0041] It is understood that this application collects multi-dimensional communication parameters of network signals. These communication parameters may include, but are not limited to, at least two of the following: signal strength (e.g., RSRP (Reference Signal Receiving Power)), signal quality (e.g., SINR (Signal to Interference plus Noise Ratio)), network latency (e.g., RTT (Round-Trip Time)), bit error rate (BER), network frequency band, base station identifier, handover frequency and handover success rate, data transmission rate (uplink / downlink), and signal hijacking characteristics (e.g., abnormal base station access, data packet tampering identifier). RSRP measures the strength of the current cellular signal; a higher value indicates a stronger signal. SINR measures the purity of the useful signal relative to the interfering signal; a higher value indicates better signal quality and faster network speed. RTT refers to the total time it takes for the vehicle terminal to send a data packet to the network side and receive a return acknowledgment; it is a key indicator for measuring the real-time performance of vehicle-to-everything (V2X) networks. BER is the proportion of erroneous bits to the total number of transmitted bits, used to measure the probability of transmission errors; a lower value indicates higher transmission reliability. Network frequency bands primarily refer to the designated transmission frequency ranges for vehicle-mounted communication signals. A base station identifier (ID) is a unique identification code of the mobile communication base station currently accessed by the vehicle. The network side can use this ID to accurately locate the cell where the vehicle is located, assisting in signal scheduling, neighbor cell handover, and other operations. Handover frequency refers to the total number of times the vehicle terminal switches from one base station cell to other adjacent cells during vehicle operation. Handover success rate refers to the percentage of successful access after the vehicle initiates a cell handover request. Downlink rate refers to the speed at which data is transmitted from the network to the vehicle terminal. Uplink rate refers to the speed at which data is uploaded from the vehicle terminal to the network. Abnormal base station access occurs when a vehicle, while driving normally, connects to an unfamiliar, non-operator-compliant base station without a reasonable cause; this is often a deceptive tampering attempt initiated by a fake base station. Data packet tampering identifiers indicate unauthorized tampering of vehicle location, driving status, and other data transmitted by the vehicle.

[0042] The frequency at which communication parameters are collected can be any suitable frequency. The collection frequency can refer to the number of discrete samples extracted and recorded from a continuously changing network signal per unit time. In practice, it is dynamically adjusted based on the vehicle's driving status.

[0043] In some implementations, the sampling frequency can be a preset fixed sampling frequency or a dynamically changing sampling frequency.

[0044] In some implementations, the data acquisition frequency is determined based on the vehicle's driving state. Driving state refers to the vehicle's current motion state. This driving state can include, but is not limited to, high-speed driving, low-speed driving, and stationary driving. In some implementations, the driving state can be determined based on vehicle speed; for example, a speed greater than 80 km / h can be considered high-speed driving, a speed between 10-80 km / h can be considered low-speed driving, and a speed less than 10 km / h can be considered stationary. In practice, different driving states have different requirements for the network signal acquisition frequency. Thus, dynamically adjusting the data acquisition frequency according to the driving state fully considers the low-power consumption requirements of the vehicle, achieving a balance between real-time monitoring and low power consumption, thereby significantly improving monitoring efficiency.

[0045] In some implementations, the relationship between each driving state and each acquisition frequency can be pre-established. Based on this correspondence, the acquisition frequency adapted to that driving state can be obtained. In practice, when the vehicle is traveling at high speed, the network environment changes drastically due to rapid changes in vehicle position, requiring more frequent signal acquisition to capture transient anomalies. Therefore, the acquisition frequency is shortened, for example, set to 100ms. When the vehicle is traveling at low speed or stationary, the network environment is relatively stable, and the acquisition frequency can be appropriately extended to reduce system power consumption, for example, set to 500ms.

[0046] In some implementations, the data collection frequency can be the same for the same driving conditions, or it can be further adjusted according to vehicle speed. For example, when a vehicle is traveling at 120 km / h on a highway, a network signal can be collected every 100 ms; when the vehicle is stationary in a parking lot, a network signal can be collected every 500 ms. In practice, this collection frequency can be positively correlated with vehicle speed; that is, the faster the vehicle speed, the higher the collection frequency can be. It is understandable that the vehicle's data processing performance must be considered in conjunction with vehicle speed.

[0047] Driving data refers to data that reflects the current driving status of a vehicle and its environment. This driving data may include, but is not limited to, vehicle speed, geographical location (such as GPS coordinates), and road segment type (such as highway, urban road, tunnel, remote road segment, etc.). For example, when a vehicle is traveling at 120 km / h on a highway, the driving data collected may include the vehicle speed of 120 km / h, geographical location coordinates (such as 120.5°E, 30.2°N), and road segment type: highway.

[0048] Driving scenarios represent the driving environment and state of a vehicle. These scenarios can include, but are not limited to, highway driving, tunnel travel, remote road sections, urban areas, suburban mountain areas, stationary conditions in underground parking garages, idling in open-air parking spaces, low-speed traffic congestion, and elevated bridges. It is understandable that different driving scenarios have different impacts on network signal stability; for example, tunnel scenarios are prone to signal interruptions or attenuation, while highway scenarios tend to result in frequent signal switching.

[0049] The driving scenario can be determined by comprehensively analyzing synchronously collected vehicle driving data (such as vehicle speed, geographical location, and road segment type). In some implementations, a correspondence between each driving data point and each driving scenario can be pre-established. Based on this correspondence, the current driving scenario adapted to the driving data can be obtained. For example, when the vehicle speed in the driving data is 120 km / h, the geographical location shows the vehicle is in a tunnel, and the road segment type is a highway tunnel, the current driving scenario can be determined as a highway tunnel passage scenario. In some implementations, the driving data can be input into a scene recognition model to obtain the driving scenario. This scene recognition model can be any suitable neural network model capable of performing this function.

[0050] Signal anomaly features refer to multi-dimensional quantitative information extracted from acquired network signals that characterizes abnormal signal states. Examples include the magnitude and duration of sudden changes in signal strength, persistent exceedances of network latency, and fluctuations in bit error rate. These signal anomaly features are primarily used for subsequent anomaly identification.

[0051] The abnormal signal characteristics may include, but are not limited to, timing characteristics, scene-related characteristics, and abnormal behavior characteristics. Timing characteristics include the magnitude and duration of sudden changes in signal strength, delay, and bit error rate. Scene-related characteristics include the deviation of signal parameter thresholds in scenarios such as tunnels, highways, and remote road sections. Abnormal behavior characteristics include abnormal base station access, frequent handover failures, persistently exceeding the bit error rate limit, and sudden drops in data rate.

[0052] For example, in a high-speed tunnel scenario, signal anomaly characteristics may include: the signal strength RSRP suddenly drops from -90dBm to -125dBm and lasts for 6 seconds (timing characteristic); the signal strength value is lower than the preset threshold of -110dBm for the tunnel scenario (scenario-related characteristic); and it is accompanied by two base station handover failures (abnormal behavior characteristic).

[0053] In this way, the scenario-based and multi-dimensional feature extraction of network signals is realized, providing a high-quality data foundation for subsequent accurate anomaly identification. At the same time, by dynamically adjusting the acquisition frequency and lightweight processing, the requirements for real-time monitoring and low power consumption are taken into account.

[0054] Step S12: Using the abnormal rule set corresponding to the current driving scenario in the preset rule base, identify the abnormal signal features of the current driving scenario to obtain the first identification result. The rule base includes at least two abnormal rule sets corresponding to driving scenarios.

[0055] Here, the rule base can be a set of judgment rules based on expert experience or historical data, used to determine the types of known signal anomalies. The rule sets in the rule base are usually associated with specific driving scenarios and customized for the characteristics of 4G / 5G signals on the vehicle side.

[0056] For example, the rule base can be pre-set with multiple sets of rules for abnormal 4G / 5G signals on the vehicle side under various driving scenarios. For instance, in high-speed scenarios, if the signal strength RSRP is not greater than -120dBm and lasts for more than 5 seconds, the rule base determines it as a weak signal anomaly; if the network latency RTT is not less than 500ms and lasts for more than 3 seconds, the rule base determines it as a latency anomaly; if the bit error rate BER is not less than 10... -4 If the error persists for more than 2 seconds, the rule base will determine it as an abnormality due to excessive bit error rate.

[0057] The first identification result refers to the judgment result obtained after matching the signal anomaly features with the rule base. The first identification result may include, but is not limited to, whether it is an anomaly, the type of anomaly, etc. For example, when the signal anomaly features show that RSRP is not greater than -125dBm, lasts for 6 seconds, and the current driving scenario is a high-speed tunnel passage scenario, the high-speed scenario weak signal anomaly rule in the rule base is triggered, then the first identification result is weak signal anomaly (known anomaly). If the signal anomaly features do not match any rule, then the first identification result is no anomaly.

[0058] During implementation, signal anomaly features (such as temporal features and scene-related features) can be input into a preset rule base. The rule base automatically matches the corresponding set of anomaly rules based on the current driving scenario (such as highway, tunnel, or remote road section), and compares each signal anomaly feature to see if it meets the threshold, duration, and other conditions in the anomaly rules to obtain the first identification result. For example, in a highway tunnel scenario, the rule base first retrieves the set of anomaly rules for judging signal anomalies in tunnel scenarios. This set of anomaly rules stipulates that an RSRP of no more than -115dBm and lasting for more than 4 seconds is an anomaly. If the RSRP in the signal anomaly features is -125dBm and lasts for 6 seconds, then the condition is met, and the rule base outputs the first identification result as weak signal anomaly (tunnel scenario).

[0059] In this way, the rule base can quickly identify known and typical signal anomaly types, providing a basis for subsequent fusion judgment. At the same time, since the rule base can be deployed locally on the vehicle, the response speed is fast, ensuring real-time performance.

[0060] Step S13: Using a preset recognition model, identify the abnormal signal features in the current driving scenario to obtain a second recognition result.

[0061] Here, the recognition model can be a computational model trained based on machine learning or deep learning algorithms to identify signal anomalies. This model can learn complex patterns between normal and anomalous signals, thereby identifying unknown anomalies not defined in the rule base, such as novel signal hijacking or abnormal base station interference. The recognition model can be any suitable neural network model capable of performing this function. For example, the recognition model can employ a lightweight neural network model (such as one optimized based on TensorFlow Lite), trained only with normal 4G / 5G signal data, capable of identifying unknown anomalies and outputting anomaly confidence levels (0~1). In some implementations, this recognition model can be deployed in an in-vehicle controller, exhibiting low model inference latency and adapting to the low-power requirements of automotive embedded systems.

[0062] The second identification result refers to the judgment result obtained after reasoning about the abnormal features of the signal through the identification model. The second identification result may include, but is not limited to, the anomaly confidence level and possible anomaly types. For example, when the abnormal features of the signal are input into the identification model, the model outputs an anomaly confidence level of 0.85 and labels the anomaly type as unknown signal interference. If the anomaly confidence level output by the model is lower than a preset threshold (such as 0.5), then the network signal can be considered to be without anomalies.

[0063] In implementation, the identification model can perform inference calculations on the input signal anomaly features and output an anomaly confidence value and anomaly type between 0 and 1. For example, for a set of signal anomaly features including sudden changes in signal strength, persistent delay exceeding the limit, and fluctuations in bit error rate, the identification model outputs an anomaly confidence value of 0.92 after inference and identifies that the anomaly pattern is highly similar to the base station anomaly interference pattern in the training data. Therefore, the second identification result is an anomaly confidence value of 0.92 and a type of base station anomaly interference.

[0064] In this way, through the intelligent analysis of the identification model, unknown anomalies that cannot be covered by the rule base can be captured, significantly improving the comprehensiveness and accuracy of anomaly identification and effectively reducing the false negative rate and false positive rate.

[0065] Step S14: Based on the first identification result and the second identification result, determine the anomaly level of the network signal.

[0066] Here, the anomaly level is a classification of signal anomalies based on the severity of the anomaly, the extent of its impact on vehicle functions, and the potential safety risks. This anomaly level may include, but is not limited to, Level 1, Level 2, and Level 3. Level 1 (i.e., minor anomaly) can be characterized by slightly low signal strength (e.g., RSRP between -110 and -120 dBm) and a slightly excessive bit error rate (BER less than 10). -6 First-level anomalies (short delays, such as RTT between 300 and 500 ms) have no impact on core functions. Second-level anomalies (general abnormalities) can include low signal strength (such as RSRP between -120 and -130 dBm) and excessive bit error rate (such as 10 dBm). -6 ≤BER<10 -4 Level 2 (high latency, such as RTT between 500 and 1000 ms) and at least two failed handover attempts may affect in-vehicle entertainment and navigation accuracy. Level 3 (serious anomaly) may include signal interruption, signal hijacking, or a severely excessive bit error rate (such as BER not less than 10). -4 Level 3 anomalies include extremely high latency (e.g., RTT not less than 1000ms), at least 3 failed handover attempts, and impact core security functions such as remote control, OTA upgrades, and V2X interaction. In implementation, those skilled in the art can set more or fewer anomaly levels according to actual needs; this application's embodiments do not impose such limitations.

[0067] The method for determining the anomaly level can be any suitable method.

[0068] In some implementations, a correspondence between each first identification result, each second identification result, and each anomaly level can be established in advance. Based on this correspondence, an anomaly level that matches both the first and second identification results can be obtained.

[0069] In some implementations, the anomaly level can be determined by weighted fusion of the first identification result and the second identification result.

[0070] In some implementations, the target identification result can be determined first based on the first identification result and the second identification result, and then the anomaly level can be determined based on the target identification result.

[0071] For example, when rule matching triggers an anomaly (the first identification result is an anomaly) or the confidence level of the intelligent model (i.e., the second identification result) is not less than the confidence threshold (e.g., 0.8), it is determined to be a signal anomaly. The final anomaly level is then determined based on the anomaly type, network signal parameters (e.g., RSRP value, RTT value, BER value, duration, number of handover failures, etc.), and the current driving scenario. For example, if the signal anomaly characteristics are RSRP = -125dBm for 6 seconds, RTT = 600ms for 4 seconds, and a bit error rate of 5 × 10⁻⁶... -5 If the current driving scenario is a high-speed tunnel, then according to the definition of anomaly level, the signal anomaly level belongs to the second level (general anomaly). If both the first and second identification results show no anomaly, then the anomaly level is determined to be no anomaly.

[0072] In this way, by combining the speed of rule matching and the comprehensiveness of intelligent models, the severity of anomalies can be accurately quantified, providing a reliable basis for subsequent differentiated vehicle control and avoiding false alarms or missed alarms caused by a single judgment mode.

[0073] Step S15: Based on the anomaly level of the network signal, control the vehicle to execute a preset response strategy.

[0074] Here, the response strategy refers to implementing differentiated control measures (such as early warning prompts and emergency response strategies) based on the determined anomaly level to ensure vehicle driving safety and the stability of in-vehicle network services. It is understandable that different anomaly levels may require different response strategies. In practice, warnings can be issued through methods such as the in-vehicle central control screen, instrument panel, and voice prompts.

[0075] The response strategy can be determined in any suitable way. In some implementations, a correspondence between each anomaly level and each response strategy can be established in advance, and a response strategy adapted to the anomaly level can be obtained based on this correspondence. In some implementations, the anomaly level can be input into a preset control model to obtain the response strategy. The control model can be any suitable neural network model capable of implementing this function.

[0076] In some implementations, the response strategy may include, but is not limited to, at least one of the following: maintaining the vehicle's network functionality, switching the vehicle's network frequency band, issuing alarms, disconnecting some of the vehicle's applications from the network, logging, and providing emergency alerts.

[0077] For example, for Level 1 (minor anomaly), the response strategy could include maintaining the vehicle's network functionality and logging. This could involve keeping background logs running without pop-up notifications and continuing to monitor signal changes.

[0078] For example, for Level 2 (general anomaly), the response strategy may include switching the vehicle's network frequency band, issuing alarms, and recording data. This could involve a light-colored notification box popping up on the central control screen (e.g., indicating weak 4G / 5G signal and optimizing connection), a gentle voice prompt (e.g., indicating signal anomaly and potential impact on navigation accuracy, without interrupting driving), simultaneous recording of abnormal data, and automatic switching between 4G / 5G frequency bands to attempt reconnecting to the optimal base station.

[0079] For example, for Level 3 (Severe Anomaly), the response strategy may include disconnecting certain vehicle applications from the network, alarms, and recording functions. This could involve displaying a red warning box on the central control screen, issuing an emergency voice prompt (such as "Severe network signal anomaly, please drive with caution"), providing auxiliary prompts on the instrument panel, simultaneously uploading abnormal data to the cloud, suspending non-essential in-vehicle networking functions (such as entertainment functions), and prioritizing core safety functions (such as V2X emergency communication).

[0080] During implementation, if the signal remains abnormal for a period of time not less than a preset time threshold (e.g., 10 seconds), an emergency vehicle alert will be triggered (e.g., reminding the driver to switch to the backup communication mode).

[0081] In this way, by implementing hierarchical control based on the level of anomaly, we can achieve differentiated early warning methods (from silent recording to emergency alerts) and precise emergency response (from automatic base station switching to suspending non-essential functions). While avoiding interference with driving, we can effectively deal with signal anomalies of different severity and ensure the normal operation of the vehicle's core safety functions.

[0082] In this embodiment, firstly, signal anomaly characteristics of the current driving scenario are determined based on simultaneously collected driving data and network signal communication parameters, achieving a high correlation between signal anomaly characteristics and the actual driving scenario, thus improving the accuracy and scenario adaptability of signal anomaly feature extraction. Secondly, signal anomaly characteristics are identified using a rule base and a recognition model respectively, yielding a first recognition result and a second recognition result. This combines the determinism of rule matching with the generalization ability of model recognition, taking into account both known and unknown anomalies, improving the comprehensiveness and accuracy of anomaly identification, and significantly reducing the possibility of false alarms and missed alarms. Finally, the anomaly level of the network signal is determined based on the two recognition results, and the vehicle is controlled according to the anomaly level, realizing refined classification and differentiated control of network signal anomalies, improving the intelligence and safety of the vehicle in dealing with network signal anomalies. In addition, since the technical solution of this application can be quickly integrated into existing vehicles without large-scale modification of the original vehicle hardware, deployment is convenient and compatibility is strong.

[0083] In some embodiments, step S11 includes steps S111 to S114, wherein: Step S111: Preprocess the communication parameters to obtain the target communication parameters.

[0084] Here, communication parameters are the fundamental data source for determining vehicle communication status and network environment. Preprocessing can involve pre-operating on the raw acquired communication parameters to improve data quality and provide reliable input for subsequent feature extraction and anomaly identification. Pre-operations may include, but are not limited to, filtering invalid data (such as missing values ​​caused by acquisition interruptions, extreme values ​​of abnormal fluctuations), data denoising (eliminating random noise caused by environmental interference), missing data imputation, and data normalization (mapping parameters of different dimensions to a uniform scale, such as the 0-1 range). For example, if the RSRP value acquired at a certain moment is -200dBm, which is significantly outside the normal range (usually -140dBm to -80dBm), then the RSRP value of -200dBm can be determined as invalid data and discarded.

[0085] Target communication parameters refer to preprocessed data that is of higher quality, purer, and in a uniform format. It can be understood that target communication parameters retain the core characteristics of the original signal while eliminating the influence of noise and outliers, providing an accurate and reliable data foundation for subsequent feature extraction and anomaly identification, thereby improving the reliability of the entire signal warning system.

[0086] Step S112: Using principal component analysis dimensionality reduction algorithm, feature extraction is performed on the target communication parameters to obtain the first feature characterizing the overall signal quality trend.

[0087] Here, Principal Component Analysis (PCA) dimensionality reduction algorithm is a statistical method used to map high-dimensional data to a low-dimensional space while preserving the main features of the data, thereby reducing data complexity and computational resource consumption.

[0088] Feature extraction refers to the process of extracting key information that reflects the essential characteristics or changing patterns of a signal from preprocessed target communication parameters using specific algorithms or methods. In this embodiment, the purpose of feature extraction is to transform high-dimensional, complex network signal data into low-dimensional, more representative feature vectors to facilitate subsequent anomaly identification.

[0089] The first feature refers to one or more principal components obtained after feature extraction of the target communication parameters using the PCA dimensionality reduction algorithm. The first feature characterizes the trend of overall signal quality changes over time. For example, the first feature can be a comprehensive signal quality index, integrating information from multiple parameters such as signal strength, delay, and bit error rate, and can intuitively reflect whether the signal quality is improving or deteriorating. The overall signal quality trend refers to the direction or pattern of change in the overall network signal quality (such as stability and reliability) over a period of time; for example, a continuous decline in signal quality or increased fluctuations in signal quality.

[0090] In implementation, since the target communication parameters contain multiple dimensions (such as signal strength, delay, and bit error rate), resulting in high data dimensionality, the PCA dimensionality reduction algorithm uses linear transformation to identify the directions with the largest variance in the data (i.e., principal components) and projects the original data onto these principal components, thereby achieving dimensionality reduction and feature extraction. For example, for data containing four dimensions—RSRP, SINR, RTT, and BER—the PCA dimensionality reduction algorithm can reduce the dimensionality to two principal components, which can explain most of the variation information in the original data. Therefore, using the PCA dimensionality reduction algorithm can reduce data dimensionality, decrease computational load, and reduce processing latency. It not only effectively adapts to the low power consumption and real-time requirements of automotive embedded systems but also extracts key features reflecting the overall trend of signal changes, thus providing a more macroscopic and stable basis for subsequent anomaly identification.

[0091] Step S113: Construct abnormal features for the target communication parameters to obtain the second feature.

[0092] Here, anomaly feature construction refers to the process of constructing features that can directly reflect or indicate abnormal signal states based on the target communication parameters and through specific rules or algorithms. Anomaly feature construction focuses more on capturing sudden changes, deviations, or behaviors that do not conform to normal patterns in the signal. For example, anomaly feature construction can construct features such as the amplitude of signal strength changes within a short period of time (e.g., within 1 second), the duration of network latency exceeding a threshold, and the frequency of base station handover failures.

[0093] The second feature refers to a series of features constructed through anomaly features, directly pointing to specific, potential abnormal behaviors or states. For example, the second feature may include specific values ​​such as the magnitude of signal strength fluctuations (e.g., -15dBm / s), the duration of delay exceeding limits (e.g., 3 seconds), and the number of handover failures (e.g., 2 times). The second feature provides a direct, quantitative basis for subsequent anomaly identification. In some implementations, the second feature may include, but is not limited to, time-domain fluctuation features, event statistical features, and vehicle-to-network coupling features. Time-domain fluctuation features extract indicators such as variance, extreme value difference, and Z-score coefficient of variation within a sliding window to capture local mutations and oscillation characteristics of the signal, and are sensitive to time-domain abnormal fluctuations. Event statistical features statistically analyze event-dimensional indicators such as the number of abnormal points within a statistical window, the frequency of signal exceeding limits, and the duration of events, quantifying the intensity and distribution patterns of abnormal events. Vehicle-to-network coupling features combine indicators such as traffic matching degree, spatiotemporal synchronization deviation, and coupling parameter linkage deviation in vehicle-to-network interaction scenarios to characterize the abnormal coupling state during vehicle-to-network interaction.

[0094] In this way, by constructing abnormal features of target communication parameters, abnormal behaviors or states in network signals can be captured in a targeted manner. This can compensate for local and sudden anomalies that may be ignored by relying solely on overall trend features, thereby improving the comprehensiveness and sensitivity of anomaly identification.

[0095] It is understandable that steps S112 and S113 can be executed simultaneously or separately.

[0096] Step S114: Fuse the first feature and the second feature to obtain the signal abnormality feature.

[0097] Here, fusion refers to combining the first and second features according to a certain strategy or algorithm to form a comprehensive and more informative feature vector, i.e., signal anomaly features. Fusion methods can be simple concatenation (connecting the feature vectors of the first and second features), weighted summation (assigning different weights based on the importance of the features), or more complex machine learning models. For example, a user can concatenate a first feature representing the overall signal quality trend (such as a comprehensive quality index) with a second feature representing specific abnormal behavior (such as the magnitude of signal strength fluctuations or the duration of delay exceeding limits) to form a comprehensive feature vector containing both macroscopic trends and microscopic anomaly information.

[0098] Signal anomaly features refer to the comprehensive features ultimately used to determine whether a network signal is abnormal and the type of anomaly, serving as standard input data for subsequent dual-mode anomaly identification. These signal anomaly features integrate the overall signal quality trend and specific abnormal behavior information, enabling a more comprehensive and accurate reflection of the signal state in the current driving scenario. For example, a signal anomaly feature can be a multi-dimensional vector containing information such as a declining signal quality trend and persistent delay exceeding limits, thus determining that the current signal is in a state of declining signal quality and abnormal delay.

[0099] In this embodiment, communication parameters are preprocessed to filter out noise and invalid data in the original signal, thereby improving data quality. Principal component analysis is used to extract a first feature representing the overall signal quality trend and reduce the data dimension. At the same time, a second feature representing specific anomalies is constructed. Finally, the two features are fused, which preserves the macro trend of the signal and highlights the micro anomaly details, making the extracted signal anomaly features more comprehensive and effective. This improves the accuracy of anomaly identification and reduces false alarms and false negatives.

[0100] In some embodiments, step S12 includes steps S121 and S122, wherein: Step S121: Based on the first recognition result and the second recognition result, determine the target recognition result.

[0101] Here, the first identification result is the judgment result output by the rule base after matching the signal anomaly features in the current driving scenario according to preset rules, regarding whether there is a known type of anomaly (such as weak signal, interruption, excessive latency, handover anomaly, or excessive bit error rate). The form of the first identification result can be a specific anomaly type label, a Boolean value (yes / no anomaly), or a data packet containing the anomaly type and confidence level, etc. For example, in a high-speed scenario, if the RTT is not less than 500ms and lasts for 3 seconds, the rule base determines it as a latency anomaly.

[0102] The second identification result is an anomaly confidence score (a value between 0 and 1) output by the intelligent model after analyzing the signal anomaly characteristics of the current network signal. This score indicates the probability that the current signal has unknown anomalies. It can identify unknown anomalies not defined in the rule base (such as new types of signal hijacking, abnormal interference from base stations, etc.).

[0103] The target identification result is a final judgment on the current network signal state made by combining the first and second identification results. The target identification result integrates the deterministic judgment of known anomalies by the rule matching layer and the probabilistic assessment of unknown anomalies by the intelligent identification layer, aiming to provide a more comprehensive and accurate description of the signal state.

[0104] The target identification result can be a first identification result, a second identification result, etc. For example, when the rule base determines that there is no anomaly, but the confidence level output by the intelligent model is 0.9, the target identification result may be identified as a suspected unknown anomaly; when the rule base determines that there is a weak signal anomaly, and the confidence level output by the intelligent model is 0.3, the target identification result is identified as a known anomaly: weak signal. In this way, the advantages of fast and deterministic rule matching are utilized to handle known problems, while the generalization ability of the intelligent model is used to discover potential new threats, thereby improving the comprehensiveness and accuracy of anomaly identification.

[0105] Step S122: Based on the target recognition results, determine the anomaly level of the network signal.

[0106] Here, the anomaly level is a quantitative classification of the severity of network signal anomalies. The anomaly level is used to trigger differentiated early warning and emergency response strategies (i.e., the aforementioned response strategies).

[0107] The target identification result provides the specific type and characteristics of the signal anomaly (e.g., weak signal + high delay), while the anomaly level is a risk level label assigned after a comprehensive assessment based on the specific type and severity of the signal anomaly (e.g., the specific value of RSRP, the specific value of RTT, and the duration). The anomaly level can be determined using any suitable method. In some implementations, a correspondence between each identification result and each anomaly level can be pre-established; based on this correspondence, an anomaly level suitable for the target identification result can be obtained. In some implementations, the target identification result can be input into a level recognition model to obtain the anomaly level; this level recognition model can be any suitable neural network model capable of performing this function.

[0108] For example, if the target identification result is a known anomaly: weak signal (RSRP=-125dBm, lasting 6s), the target identification result will be mapped to a level 2 warning (general anomaly) according to the preset classification standard.

[0109] For example, if the target identification result is a known anomaly: signal interruption (RSRP=-140dBm, lasting 3s), the target identification result will be mapped to a level three warning (serious anomaly).

[0110] In this embodiment of the application, by first determining a unified target identification result based on two identification results, and then determining the anomaly level based on the target identification result, a hierarchical conflict arbitration logic is realized to be compatible with known, unknown, complex and other anomalies. This reduces the possibility of directly using two potentially conflicting identification results for classification, making the anomaly level determination process more reasonable and accurate.

[0111] In some implementations, step S121 includes: if the first identification result and the second identification result are compatible, using the first identification result or the second identification result as the target identification result; if the first identification result and the second identification result are not compatible, determining the target identification result based on the anomaly level corresponding to the first identification result and the anomaly level corresponding to the second identification result.

[0112] Here, "fit" refers to the absence of fundamental conflict between the first and second identification results in their judgment conclusions, meaning they can corroborate or be compatible with each other. For example, if the rule base determines the signal as weakly abnormal, and the intelligent identification model outputs an anomaly confidence level of 0.9 (highly abnormal), then both the first and second identification results indicate the presence of an anomaly, and thus the result is considered a fit. Conversely, if the rule base determines the signal as normal, and the intelligent identification model outputs an anomaly confidence level of 0.2 (lowly abnormal), then both the first and second identification results indicate no anomaly or a slight anomaly, and this is also considered a fit. The criteria for fit can be pre-defined. For instance, when the result from the rule base matches the abnormal / normal state indicated by the confidence threshold (e.g., 0.8) of the intelligent identification model, they are considered to be a fit.

[0113] The target identification result is the final judgment adopted to trigger subsequent tiered early warnings and emergency responses. The target identification result integrates the results of rule matching and intelligent recognition, ensuring the accuracy and reliability of the judgment. This mechanism ensures that when the first and second identification results are consistent, the final conclusion can be determined quickly and accurately, avoiding erroneous warnings caused by misjudgments based on a single pattern.

[0114] For example, when a vehicle is traveling at high speed, the rule base detects an RSRP of -125dBm for 6 seconds, triggering a weak signal anomaly rule and outputting a first identification result of weak signal anomaly. Simultaneously, the intelligent recognition model analyzes the signal characteristics and outputs a second identification result of an anomaly confidence level of 0.9. Therefore, it can be determined that both the first and second identification results point to an anomaly, and these two results are compatible; thus, the weak signal anomaly can be directly used as the target identification result.

[0115] Mismatch refers to a fundamental conflict between the first and second identification results in their judgments, making them unable to corroborate each other. For example, the rule base determines the first identification result to be normal, but the intelligent identification model outputs a second identification result with an anomaly confidence level of 0.95 (high anomaly). In other words, the first and second identification results have completely opposite conclusions, thus the first and second identification results are considered mismatched. Another example is that the rule base determines the first identification result to be a weak signal anomaly, while the intelligent identification model outputs a second identification result with an anomaly confidence level of 0.3 (low anomaly). The first and second identification results point to different anomaly states, which also constitutes mismatch.

[0116] The first and second identification results each correspond to an anomaly level. For example, if the rule base determines the signal is weak, the anomaly level corresponding to the first identification result can be level two; if the intelligent identification model outputs an anomaly confidence level of 0.95, the anomaly level corresponding to the second identification result can be level three.

[0117] When the first identification result and the second identification result do not match, the first identification result or the second identification result cannot be simply adopted. Instead, a comprehensive judgment should be made based on the anomaly level of each of the first and second identification results to provide a more robust and secure final judgment.

[0118] The determination of the target identification result can be done in any suitable way. In some implementations, a strategy of selecting the highest level can be adopted, that is, selecting the result with the higher anomaly level as the target identification result to ensure that no response to potentially high-risk anomalies is missed. In some implementations, a weighted average strategy can be adopted, assigning different weights to the identification results of the rule base and the intelligent identification model, and then selecting the target identification result based on the weighted identification result. In this way, the advantages of rule matching and intelligent identification can be fully utilized, making rapid decisions when the conclusions are consistent, and weighing risks through level comparison when the conclusions conflict, avoiding warning errors or omissions caused by misjudgment due to a single pattern or conflict between the first and second identification results.

[0119] In the embodiments of this application, when the two identification results match, one of them is directly used as the target identification result, ensuring processing efficiency; when the two identification results do not match, a comprehensive judgment is made based on their respective anomaly levels to make full use of the advantages of rule matching and intelligent identification, thereby obtaining a more reliable target identification result, improving the accuracy and robustness of anomaly judgment, and thus ensuring the reliability and security of vehicle-side 4G / 5G signal early warning.

[0120] In some implementations, the target identification result includes the target anomaly type, target severity, and target duration. The target anomaly type includes one of the following: signal quality, transmission, and security. Step S122 includes steps S1221 and S1222, wherein: Step S1221: Obtain the target grading strategy that is compatible with the target anomaly type. The target grading strategy represents the relationship between each severity, each duration and each anomaly level under the target anomaly type.

[0121] Here, the target recognition result is a comprehensive judgment information output after analyzing the network signal, which is used to fully describe the specific situation of the current network signal anomaly.

[0122] The target anomaly type is used to define the fundamental nature of the current network signal anomaly; it is a classification label for the essential attributes of the network signal anomaly. The target anomaly type is used to distinguish network signal anomalies with different causes and manifestations. This target anomaly type can include, but is not limited to, signal quality anomalies, transmission anomalies, and security anomalies. Signal quality anomalies mainly refer to network signal anomalies caused by the degradation of physical layer parameters such as signal strength and signal-to-noise ratio, such as weak signal, high interference, significant signal jitter, cell ping-pong handover / handover failure, abnormal RF module power, and air interface RF and network access failures such as network search and registration failure. Transmission anomalies mainly refer to performance problems in the data transmission process, such as excessive latency, excessive bit error rate, sudden drop in data rate, high packet loss, insufficient throughput bandwidth, link congestion, and heartbeat timeout of long-term vehicle-to-cloud connections. Security anomalies refer to network signal anomalies involving network attacks or illegal access, such as signal hijacking, abnormal / fake base station access, data packet tampering, malicious neighbor cell deception, abnormal signaling tampering, data packet attacks, encryption and authentication failure, and external signal suppression and interference. By clearly identifying the type of target anomaly, the system can select the most appropriate handling strategy.

[0123] Target severity is a metric used to quantify the depth of impact of network signal anomalies on network performance and user experience. In some implementations, it can typically be determined through preset thresholds or model scoring. For example, a signal strength RSRP below -120 dBm can be classified as moderately severe, and below -130 dBm as severe. Target severity provides a key basis for tiered early warning systems.

[0124] The target duration measures the duration of network signal anomalies, primarily distinguishing between transient network signal fluctuations and persistent network signal failures. For example, a delay spike lasting 3 seconds might just be transient interference, while a signal interruption lasting more than 10 seconds constitutes a severe, persistent failure. The target duration is a key parameter for determining whether a network signal anomaly requires emergency response.

[0125] During implementation, the target anomaly type, target severity, and target duration together constitute a three-dimensional network signal anomaly profile, providing a data foundation for subsequent accurate classification and differentiated handling.

[0126] A target grading strategy is a set of pre-defined grading rules or models for specific target anomaly types. It maps multi-dimensional network signal anomaly parameters (severity, duration) to a unified anomaly level. Understandably, different anomaly types, even with the same target severity and duration, may correspond to different anomaly levels; that is, different anomaly types can correspond to different grading strategies. For example, for signal quality-related network signal anomalies, an RSRP below -120dBm lasting for 5 seconds might be classified as a Level 2 anomaly. Conversely, for security-related network signal anomalies, once signal hijacking characteristics are detected, even if the duration of the hijacking characteristic is extremely short, it might be directly classified as a Level 3 anomaly. This differentiated grading standard significantly improves the accuracy and scenario adaptability of target anomaly level determination, avoiding misjudgments caused by a one-size-fits-all approach to target anomaly level grading. This, in turn, provides a more reliable and refined decision-making basis for subsequent differentiated target anomaly level early warning and emergency response.

[0127] Step S1222: Determine the anomaly level that matches both the severity and duration of the target, according to the target classification strategy.

[0128] Here, the anomaly level is the final assessment of the severity of the current network signal anomaly. The anomaly level is used to trigger different levels of early warning and emergency response measures. Understandably, the higher the anomaly level, the greater the impact of the network signal anomaly on vehicle functionality and safety, and the more urgent the required response.

[0129] During implementation, the severity and duration of the target are used as inputs to determine its anomaly level from the target classification strategy. For example, the classification strategy for transmission-related target anomalies specifies: low severity and short duration are classified as Level 1 Minor Warning; medium severity or low severity with long duration is classified as Level 2 General Warning; high severity or medium severity with long duration is directly classified as Level 3 Severe Warning. Therefore, when the severity is severe and the duration is not less than 5 seconds, it is classified as Level 3 anomaly. This achieves automated and standardized conversion from multi-dimensional anomaly parameters to a single-level label, ensuring the objectivity and consistency of the classification results.

[0130] In this application, by refining the target identification results into three dimensions—anomaly type, severity, and duration—and adopting different grading strategies for different anomaly types to determine the anomaly level, a multi-dimensional and refined grading of network signal anomalies is achieved. This makes the determination of anomaly levels more scientific and in line with actual driving scenarios, avoiding misjudgments caused by a one-size-fits-all approach. Consequently, it can provide a more reliable and refined decision-making basis for subsequent differentiated early warning and emergency response.

[0131] In some implementations, the response strategy includes at least one of the following: maintaining the vehicle's network functionality, switching the vehicle's network frequency band, issuing an alarm, or disconnecting some applications on the vehicle from the network; step S15 includes steps S151 to S153, wherein: Step S151: Maintain the vehicle's network function when the network signal abnormality level is Level 1.

[0132] Here, the network signal anomaly level is determined by a comprehensive evaluation of multi-dimensional communication parameters (such as RSRP, SINR, RTT, BER, etc.) of the vehicle-side 4G / 5G network signal. The network signal anomaly level is used to differentiate the degree of impact of signal anomalies on vehicle functions, allowing for differentiated response strategies. In this embodiment, a Level 1 anomaly will not affect the vehicle's core safety functions.

[0133] The network functions of a vehicle refer to all online service capabilities that a vehicle can achieve by relying on 4G / 5G networks, which may include, but are not limited to, in-vehicle navigation, remote control, OTA upgrades, V2X vehicle-to-everything (V2X) interaction, and in-vehicle entertainment.

[0134] Under Level 1 anomalies, due to their minor severity, they do not affect the vehicle's core safety functions. Therefore, the vehicle can maintain normal operation of all network functions, ensuring the stable operation of core safety features. During implementation, logs can be recorded in the background to continuously monitor signal changes without any notification, ensuring the user remains unaware. This avoids unnecessary interference to vehicle users during minor signal anomalies, thus maintaining a good user experience.

[0135] Step S152: When the network signal abnormality level is the second level, switch the vehicle's network frequency band and issue an alarm according to the preset first display method.

[0136] Here, the second level is a general anomaly. A level two anomaly may affect non-core functions such as in-vehicle entertainment and navigation accuracy, but it has not yet endangered driving safety.

[0137] Network frequency bands refer to the specific frequency ranges used by 4G / 5G communication, such as Band 1 (2100MHz), Band 3 (1800MHz), and Band 5 (850MHz) commonly used in 4G networks, and n78 (3500MHz) and n41 (2515MHz) commonly used in 5G networks. Switching a vehicle's network frequency band means that the vehicle automatically switches from the currently used network frequency band to another available network frequency band to find a connection with better signal quality. For example, when a vehicle encounters signal interference while traveling at high speed, it can automatically switch from a high-frequency band (such as 5G n78) to a low-frequency band (such as 4G Band 5), taking advantage of the strong penetration and wide coverage of low-frequency signals to improve communication quality.

[0138] An alarm is a notification sent to a vehicle user about an abnormal signal. The purpose of an alarm is to remind the vehicle user that there is an anomaly in the current network status and / or to inform the vehicle user that optimization measures are being taken.

[0139] The first display method refers to the specific alarm presentation format triggered under the second-level anomaly, which does not interrupt the ongoing driving operation. This first display method can be any suitable display method. For example, a light-colored prompt box may pop up on the central control display screen, displaying information such as weak 4G / 5G signal or connection optimization. In some implementations, a gentle voice prompt may also be issued, such as indicating signal abnormality or potential impact on navigation accuracy.

[0140] When implemented, network connectivity can be proactively optimized and vehicle users can be notified of alarm information when signal anomalies affect certain functions. This can improve the stability of the vehicle's network connectivity and reduce the impact on the user experience.

[0141] Step S153: When the network signal abnormality level is level three, disconnect the network of some applications of the vehicle and issue an alarm according to the preset second display mode.

[0142] Here, the third level is a serious anomaly. A third-level anomaly will directly affect core security functions such as remote control, OTA upgrades, and V2X interaction, and requires immediate emergency measures.

[0143] Vehicle applications refer to various in-vehicle applications installed on the vehicle. Based on their network dependence and security requirements, these applications can be categorized into core and non-core applications. Core applications include V2X emergency communication, remote control, and security-related OTA upgrades. Non-core applications include in-vehicle entertainment (such as online music and video) and information browsing. In the event of a Level 3 anomaly, the vehicle system will disconnect non-core applications from the network, prioritizing network resources for core applications to avoid security risks caused by network signal anomalies and maximize driving safety. For example, when signal hijacking is detected, all non-essential entertainment applications will be immediately suspended from network access, retaining only the V2X emergency communication channel.

[0144] The second display method refers to a stronger alarm triggered under third-level anomalies. Understandably, the intensity and urgency of the second display method are significantly higher than the first, aiming to draw the vehicle user's attention. This second display method can be any suitable display method. For example, a red warning box may pop up on the central control display screen, displaying an emergency warning such as "Severe network signal anomaly, please drive with caution." In some implementations, a concise and clear emergency voice prompt may also be issued, such as "Signal anomaly, non-essential functions have been suspended." In some implementations, relevant warning icons may also be displayed on the instrument panel.

[0145] In this application embodiment, based on different levels of network signal anomalies, differentiated control strategies are implemented, ranging from maintaining network functionality to switching frequency bands and issuing alarms, and then disconnecting some application networks and issuing alarms. This achieves gradient and refined emergency response to network signal anomalies, which can not only effectively avoid the risk of vehicle function failure caused by signal anomalies, but also improve the user experience while ensuring the safety of the vehicle's core functions.

[0146] In some implementations, the control method further includes uploading communication parameters, a first identification result, and a second identification result to a server, which is used to update the rule base and the identification model based on the communication parameters, the first identification result, and the second identification result.

[0147] Here, the multi-dimensional communication parameters of the network signal can comprehensively reflect the current network environment quality and communication status of the vehicle. The collection of communication parameters is the basic data source for subsequent anomaly identification and early warning.

[0148] The first identification result is the rapid and efficient identification of known types of signal anomalies based on a pre-defined vehicle-side scenario-based anomaly judgment rule base. The rule base is a set of pre-defined rules stored on the vehicle side used to determine whether the vehicle-side 4G / 5G signal is abnormal.

[0149] The second identification result is the identification of undefined / novel signal anomalies (such as signal hijacking or base station interference) or complex signal anomalies identified by the intelligent identification model in the rule base. The identification model refers to a lightweight neural network model deployed on the vehicle. The parameters and performance of the identification model directly affect its ability to identify unknown anomalies.

[0150] The server can be a computing and storage platform deployed in the cloud, responsible for receiving, storing, and analyzing data uploaded from the vehicle. The server possesses powerful computing capabilities and storage space, enabling in-depth mining and statistical analysis of massive amounts of data from a large number of vehicles, providing data support for vehicle operation and maintenance and network optimization. In implementation, the server and vehicle interact via a vehicle-cloud collaborative communication unit (such as one based on the MQTT protocol), serving as the core hub for achieving dynamic iterative optimization.

[0151] During implementation, after receiving the communication parameters, first identification result, and second identification result uploaded by the vehicle, the server will conduct in-depth analysis to optimize the rule base and / or identification model, so as to continuously improve the accuracy and adaptability of the warning, thereby effectively reducing the false alarm and false alarm rates, better coping with the complex and ever-changing real road environment, and ensuring the long-term stability and security of vehicle networking services.

[0152] For example, the server can analyze abnormal patterns that occur frequently in a specific road segment (such as a tunnel) but are not covered by the rule base, thereby optimizing and updating the thresholds in the rule base or adding new rules.

[0153] For example, the server can also use the accumulated abnormal data as new training samples to retrain or fine-tune the recognition model, optimize the model parameters, and improve the recognition model's accuracy in recognizing unknown anomalies.

[0154] During implementation, the optimized rule base and / or recognition model will be delivered to the vehicle via OTA to complete dynamic iteration.

[0155] In this embodiment, by uploading data such as communication parameters and two recognition results to the server, a data foundation is provided for the server to update the rule base and recognition model, realizing collaborative iteration and closed-loop optimization between the vehicle and the cloud, which can continuously improve the accuracy of anomaly recognition, thereby effectively reducing false alarms and false negatives.

[0156] The technical solution of this application is described in detail below.

[0157] With the widespread adoption of intelligent connected vehicles, 4G / 5G networks have become the foundation for core functions such as in-vehicle navigation, remote control, OTA upgrades, V2X vehicle-to-everything (V2X) interaction, and in-vehicle entertainment. The stability of the 4G / 5G signal on the vehicle side directly affects driving safety and user experience. Currently, the transmission of 4G / 5G signals on the vehicle side is susceptible to factors such as complex road conditions, environmental interference, network coverage blind spots, signal hijacking, and base station failures, resulting in problems such as weak signals, interruptions, excessive latency, excessive bit error rates, and abnormal handover.

[0158] Existing 4G / 5G signal-related technologies for vehicles have the following shortcomings: 1) It can only provide simple feedback on signal strength and cannot monitor and identify deeper anomalies such as signal delay, bit error rate, handover anomalies, and signal hijacking, resulting in a single warning dimension; 2) There is a lack of targeted anomaly detection criteria. The system relies heavily on general network monitoring algorithms and does not take into account the signal characteristics of vehicle driving scenarios (such as high-speed movement, tunnel passage, and remote road sections), resulting in a high rate of false alarms and false negatives. 3) The early warning method is simple, relying solely on in-vehicle system prompts, without developing differentiated emergency response strategies based on the level of abnormality, thus failing to address the risk of in-vehicle function failure caused by signal anomalies; 4) The retention, analysis and iterative optimization of abnormal signal data have not been implemented, making it impossible to improve the accuracy of early warnings based on historical data, and also unable to provide data support for vehicle operation and maintenance; 5) The low power consumption requirements of the vehicle were not considered, and the complex monitoring algorithm consumed too many vehicle embedded resources, affecting the real-time performance of signal monitoring.

[0159] Therefore, there is an urgent need for a vehicle-mounted 4G / 5G network signal early warning system that is adapted to vehicle scenarios, focuses on 4G / 5G signals, provides comprehensive monitoring, accurate early warning, timely response, and low power consumption, in order to address the shortcomings of related technologies.

[0160] This application aims to overcome the shortcomings of existing technologies and solve the technical problems of single-dimensional vehicle-side 4G / 5G signal monitoring, low accuracy of anomaly identification, imperfect early warning and emergency response, poor low-power adaptability, and lack of iterative optimization mechanism. It provides a vehicle-side 4G / 5G network signal early warning system and method that can monitor multi-dimensional parameters of 4G / 5G signals in real time, accurately identify anomalies, provide graded early warning and link emergency response, and is low-power and dynamically optimizable.

[0161] To achieve the above objectives, this application provides a vehicle-mounted 4G / 5G network signal warning system, such as... Figure 2 As shown, the early warning system includes the following core modules, which work together to achieve full-process monitoring and early warning of 4G / 5G signals: 1. Signal acquisition module 21 (or signal acquisition unit) As a system input terminal, it is integrated into the vehicle-mounted T-BOX (Telematics BOX, remote information processing terminal) and directly connects to the vehicle's 4G / 5G module. It uses a bypass acquisition method to avoid interfering with the normal transmission of the original vehicle's 4G / 5G signal. It collects multi-dimensional core parameters of the vehicle's 4G / 5G signal in real time, and the acquisition period can be dynamically adjusted according to the vehicle's driving status (shortened to, for example, 100ms at high speeds, and adjusted to, for example, 500ms at low speeds / stationary conditions). The acquired parameters include: 1) Signal strength (RSRP), signal quality (SINR); 2) Network latency (round-trip time, RTT) and bit error rate (BER); 3) Network frequency band, base station ID, handover frequency, and handover success rate; 4) Data transmission rate (uplink / downlink); 5) Signal hijacking characteristics (abnormal base station access, data packet tampering identifiers).

[0162] Simultaneously, it connects to the vehicle's CAN bus to collect vehicle driving assistance data (such as vehicle speed, geographical location, and road segment type) for scenario-based judgment of abnormal signals.

[0163] 2. Signal preprocessing and feature extraction module 22 The collected 4G / 5G signal parameters and vehicle driving assistance data are preprocessed to filter out invalid data (such as missing values ​​caused by data collection interruption and extreme values ​​of abnormal fluctuations), and the data accuracy is improved through normalization and noise reduction. Extracting signal anomaly features, including: 1) Signal timing characteristics: amplitude and duration of sudden changes in signal strength, delay, and bit error rate; 2) Scene-related features: deviation of signal parameter thresholds in scenarios such as tunnels, highways, and remote road sections; 3) Abnormal behavior characteristics: abnormal base station access, frequent handover failures, continuous exceedance of the bit error rate, sudden drop in data rate, etc.

[0164] A lightweight PCA dimensionality reduction algorithm is adopted to reduce the data dimensionality, adapt to the low power consumption requirements of automotive embedded systems, and ensure that the processing latency is no more than 50ms.

[0165] Output high-dimensional feature vectors to the next stage to ensure data quality and reduce subsequent computational load.

[0166] 3. Anomaly Detection Engine Module 23 It includes a rule matching layer and a lightweight intelligent model. By adopting a dual-mode recognition approach of "rule matching + intelligent model," it can identify both known and unknown anomalies while controlling power consumption. 1) Rule matching layer: Pre-set vehicle-side 4G / 5G signal anomaly judgment rule base. The rules are customized based on vehicle-side scenarios (such as RSRP not greater than -120dBm and lasting for more than 5 seconds in high-speed scenarios are judged as abnormal, RTT not less than 500ms and lasting for 3 seconds are judged as latency anomaly). It covers known scenarios such as weak signal, interruption, excessive latency, handover anomaly, and excessive bit error rate, so as to quickly identify known anomalies. 2) Intelligent recognition layer: Deploy a lightweight neural network model (optimized based on TensorFlow Lite, model size less than 300KB), trained only with normal 4G / 5G signal data, which can identify unknown anomalies (such as new types of signal hijacking, abnormal interference from base stations, etc.), and output anomaly confidence (0-1), so as to achieve accurate identification of unknown attacks or complex anomalies; 3) Fusion judgment: Combining the dual-mode recognition results, when the rule matching triggers an anomaly or the confidence of the intelligent model is ≥0.8, it is judged as a signal anomaly, and the anomaly type (signal quality type, transmission type, security type) is marked.

[0167] 4. Tiered Early Warning and Emergency Response Module 24 Based on the type, severity, and duration of the anomaly, the warnings are divided into three levels, each with a corresponding differentiated warning method and emergency response strategy to avoid interfering with driving while ensuring the safety of vehicle functions. 1) Level 1 Warning (Minor Anomaly): Signal strength is slightly low (e.g., RSRP is between -110 and 120 dBm), and the bit error rate is slightly exceeded (e.g., BER is less than 10). -6 A brief delay (such as RTT of 300~500ms) has no impact on core functions; Warning method: The vehicle's infotainment system records logs in the background without popping up a notification, thus not affecting user experience; Emergency response: Maintain signal acquisition and continuously monitor signal changes.

[0168] 2) Level 2 Warning (General Abnormality): Low signal strength (e.g., RSRP between -120 and 130 dBm), excessive bit error rate (10 dBm). -6 ≤BER<10 -4 High latency (e.g., RTT between 500 and 1000 ms) and no more than two failed switching attempts may affect in-vehicle entertainment and navigation accuracy. Warning method: A light-colored warning box pops up on the vehicle's central control screen, and a gentle voice prompt (without interrupting driving) is given, while abnormal data is recorded simultaneously; Emergency response: Automatically switch 4G / 5G frequency bands and attempt to reconnect to the optimal base station; 3) Level 3 Warning (Severe Anomaly): Signal interruption, signal hijacking, or severely excessive bit error rate (BER not less than 10). -4Extremely high latency (RTT not less than 1000ms), no less than 3 switching failures, affecting core security functions such as remote control, OTA upgrades, and V2X interaction; Warning methods: A red warning box pops up in the vehicle's central control screen, an emergency voice prompt (concise and clear), and auxiliary prompts are displayed on the instrument panel, while abnormal data is simultaneously uploaded to the cloud; Emergency Response: Suspend non-essential vehicle networking functions (such as entertainment functions), prioritize core safety functions (such as V2X emergency communication), and if the signal remains abnormal for no less than 10 seconds, trigger the vehicle emergency prompt (such as reminding the driver to switch to backup communication mode).

[0169] 5. Data storage and vehicle-cloud collaborative iteration module 25 The vehicle has a built-in small storage unit that stores signal acquisition data, anomaly identification results, and early warning records from the past few months for local traceability. It is responsible for data interaction between the vehicle and the cloud (corresponding to the aforementioned server). At the same time, it uploads abnormal data (including signal parameters, vehicle location, and driving status) to the cloud platform through the vehicle's T-BOX. The cloud platform performs in-depth analysis of the abnormal data, optimizes the abnormal judgment rule base and intelligent model parameters, and sends the optimized rules and model parameters to the vehicle through OTA upgrades to achieve dynamic iteration of warning accuracy. All abnormal data from the vehicle are stored in the cloud at the same time, providing data support for vehicle operation and maintenance and network optimization.

[0170] 6. Low-power control module 26 Based on the vehicle's driving status (driving / stationary) and network function usage, the operating mode and power consumption of each unit are dynamically adjusted to ensure low-power operation of the system in the in-vehicle embedded environment. When the vehicle is stationary and no network function is used, the signal acquisition frequency is reduced and some computing units of the intelligent model are turned off, reducing power consumption to no more than 50mW; when the vehicle is driving or the core network function is enabled, the normal monitoring mode is restored to ensure a balance between real-time monitoring and low power consumption.

[0171] This application also provides a method for early warning of 4G / 5G network signals on a vehicle, such as... Figure 3 As shown, it includes the following steps: Step S1: Signal acquisition; The signal acquisition module collects multi-dimensional parameters of the vehicle's 4G / 5G signal and vehicle driving assistance data in real time, and dynamically adjusts the acquisition cycle.

[0172] Step S2: Preprocessing and Feature Extraction; The collected data is denoised, normalized, and dimensionality reduced to extract abnormal signal features.

[0173] Step S3: Anomaly identification; By using a dual-mode collaboration of rule matching layer and intelligent recognition layer, it determines whether the signal is abnormal and labels the type of abnormality and the confidence level.

[0174] Step S4: Tiered early warning and emergency response; Based on the level of abnormality, the corresponding early warning method is triggered, and differentiated emergency response strategies are implemented.

[0175] Step S5: Data storage and upload; Abnormal data is retained on the vehicle and simultaneously uploaded to the cloud platform.

[0176] Step S6: Iterative optimization; The system analyzes abnormal data in the cloud, optimizes the rule base and intelligent model, and delivers the results to the vehicle via OTA to complete dynamic iteration.

[0177] This system features a simple hardware deployment, requiring no modification to the original vehicle's core structure, specifically including: (1) Core control unit: The vehicle-grade low-power MCU (such as STM32H7 series) is integrated into the vehicle T-BOX and is responsible for the coordinated control, data processing and instruction issuance of each module; (2) Signal acquisition unit: It is directly connected to the vehicle-side 4G / 5G module (such as Qualcomm SDX55) and uses SPI (Serial Peripheral Interface) interface for communication to realize bypass acquisition of signal parameters without interfering with the original vehicle signal transmission. (3) Storage unit: Use vehicle-grade Flash storage (e.g., capacity not less than 16GB) to retain signal data and warning records; (4) Warning output unit: connected to the vehicle central control screen, instrument panel and voice module to realize different levels of warning prompts; (5) Communication unit: Relying on the original vehicle 4G / 5G module, it realizes data interaction between the vehicle and the cloud and supports OTA upgrade.

[0178] The software in this application adopts a layered design to adapt to automotive embedded systems, and the specific implementation is as follows: (1) Acquisition layer: Based on AT (Attention, standard) commands and SPI communication protocol, it realizes real-time acquisition of 4G / 5G signal parameters, dynamically adjusts the acquisition cycle, and filters invalid data; (2) Preprocessing layer: Lightweight scripts such as Python are used to perform data normalization, noise reduction, and PCA dimensionality reduction, with a processing latency of no more than 50ms; (3) Anomaly identification layer: The rule base is written in C language and has 12 preset rules for 4G / 5G signal anomalies on the vehicle side; the intelligent model is optimized using TensorFlow Lite and deployed on the MCU, with a model inference latency of no more than 30ms. (4) Early warning and emergency response layer: Communicates with the vehicle central control system and voice module to realize graded early warning prompts; Writes emergency response logic to automatically trigger frequency band switching, function pause and other operations; (5) Vehicle-Cloud Collaboration Layer: Based on the MQTT protocol, it realizes the uploading of abnormal data from the vehicle and the reception of instructions from the cloud, supports OTA upgrade logic, and ensures the dynamic updating of the rule base and model.

[0179] For example, while the vehicle is traveling at high speed, the signal acquisition module collects 4G / 5G signal parameters in real time and finds that RSRP = -125dBm (for 6 seconds), RTT = 600ms (for 4 seconds), and bit error rate = 5 × 10⁻⁶. -5 Meanwhile, the vehicle was in a high-speed tunnel section; The signal preprocessing and feature extraction module extracts temporal features (signal strength abrupt changes, continuous delay exceeding the standard) and scene-related features (threshold deviation in tunnel scenes), and inputs them into the anomaly recognition engine module after dimensionality reduction by PCA. The rule matching layer of the anomaly detection engine module triggers a level 2 warning rule, and the intelligent model outputs an anomaly confidence score of 0.85, which is fused and judged as a level 2 anomaly (general anomaly). The graded early warning and emergency response module triggers a light-colored prompt box in the central control system (displaying "4G / 5G signal is weak, optimizing connection"), and gently prompts with a voice "Signal abnormality, navigation accuracy may be affected", automatically switching 4G / 5G frequency bands and attempting to connect to a backup base station in the tunnel; Meanwhile, the vehicle stores the abnormal data (including signal parameters, vehicle location, and speed) and uploads it to the cloud platform. The cloud platform analyzes the signal anomaly patterns in the tunnel section, optimizes the anomaly judgment threshold in this scenario, and subsequently sends it to the vehicle via OTA upgrade to improve the accuracy of the warning in this scenario.

[0180] Compared with related technologies, this application has the following advantages: (1) Focusing on the core of 4G / 5G signals, with comprehensive monitoring dimensions: Specifically designed for vehicle-side 4G / 5G signals, covering multiple parameters such as signal strength, quality, latency, bit error rate, handover status, and security risks, solving the problem of single monitoring in existing technologies; (2) Accurate anomaly identification with low false alarm and false negative rates: Customized judgment rules are combined with vehicle driving scenarios and lightweight intelligent models are used to identify both known and unknown anomalies, which improves the accuracy of identification while adapting to vehicle scenarios. (3) Improved graded early warning and emergency response: graded according to the severity of the anomaly, the early warning method does not interfere with driving, and the emergency response strategy is highly targeted, which can effectively avoid the risk of vehicle function failure caused by signal anomalies; (4) Good low power consumption adaptability: Through dynamic power consumption control and lightweight algorithm optimization, it adapts to the limited resources of automotive embedded systems, and balances real-time monitoring with low power consumption. (5) Dynamically iteratively optimized: Through vehicle-cloud collaboration, the rule base and intelligent model can be continuously upgraded, the accuracy of early warning can be continuously improved with the usage scenario, and at the same time, data support can be provided for vehicle operation and maintenance; (6) Easy to deploy and highly compatible: It is integrated into the vehicle T-BOX and is compatible with existing vehicle 4G / 5G modules and central control systems. It does not require large-scale modification of the original vehicle hardware and is easy to promote and apply.

[0181] Based on the above embodiments, this application also provides a vehicle control device, such as... Figure 4 As shown, the control device 40 includes: The first determining module 41 is used to determine the signal anomaly characteristics of the network signal based on the communication parameters of the network signal of the vehicle collected in the current driving scenario. The current driving scenario is determined based on the driving data of the vehicle collected synchronously. The first identification module 42 is used to identify signal abnormality features by using the abnormal rule set corresponding to the current driving scenario in the preset rule base to obtain the first identification result. The rule base includes at least two abnormal rule sets corresponding to driving scenarios. The second identification module 43 is used to identify the abnormal signal features in the current driving scenario using a preset identification model, and obtain a second identification result; The second determining module 44 is used to determine the anomaly level of the network signal based on the first identification result and the second identification result; The control module 45 is used to control the vehicle to execute a preset response strategy based on the anomaly level of the network signal.

[0182] In some implementations, the second determining module 44 is used to determine the target identification result based on the first identification result and the second identification result; and to determine the anomaly level of the network signal based on the target identification result.

[0183] In some implementations, the second determining module 44 is used to take the first identification result or the second identification result as the target identification result if the first identification result and the second identification result are compatible; and to determine the target identification result based on the anomaly level corresponding to the first identification result and the anomaly level corresponding to the second identification result if the first identification result and the second identification result are not compatible.

[0184] In some implementations, the target identification result includes target anomaly type, target severity, and target duration. The target anomaly type includes one of the following: signal quality type, transmission type, and security type. The second determination module 44 is used to obtain a target classification strategy that matches the target anomaly type. The target classification strategy represents the relationship between each severity, each duration, and each anomaly level under the target anomaly type. According to the target classification strategy, an anomaly level that matches both the target severity and the target duration is determined.

[0185] In some implementations, the first determining module 41 is used to preprocess the communication parameters to obtain target communication parameters; use principal component analysis dimensionality reduction algorithm to extract features from the target communication parameters to obtain a first feature characterizing the overall signal quality trend; construct abnormal features from the target communication parameters to obtain a second feature; and fuse the first feature and the second feature to obtain signal abnormal features.

[0186] In some implementations, the response strategy includes at least one of the following: maintaining the vehicle's network function, switching the vehicle's network frequency band, issuing an alarm, or disconnecting some of the vehicle's applications from the network; the control module 45 is configured to maintain the vehicle's network function when the network signal anomaly level is at the first level; switch the vehicle's network frequency band and issue an alarm according to a preset first display method when the network signal anomaly level is at the second level; and disconnect some of the vehicle's applications from the network and issue an alarm according to a preset second display method when the network signal anomaly level is at the third level.

[0187] In some embodiments, the control method further includes a sending module for uploading communication parameters, a first identification result, and a second identification result to a server, which updates the rule base and identification model based on the communication parameters, the first identification result, and the second identification result.

[0188] The descriptions of the above device embodiments are similar to those of the above method embodiments, and have similar beneficial effects. For technical details not disclosed in the device embodiments of this application, please refer to the descriptions of the method embodiments of this application for understanding.

[0189] It should be noted that, in the embodiments of this application, if the above methods are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, or the parts that contribute to related technologies, can be embodied in the form of software products. These software products are stored in a storage medium and include several instructions to cause an electronic device (which may be a personal computer, server, vehicle, or network device, etc.) to execute all or part of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), magnetic disks, or optical disks. Thus, the embodiments of this application are not limited to any specific hardware and software combination.

[0190] This application also provides a vehicle including a memory and a processor, the memory storing a computer program that can run on the processor, and the processor executing the computer program to implement any of the methods described above.

[0191] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method. The computer-readable storage medium can be transient or non-transient.

[0192] This application also provides a computer program product, which includes a computer program or instructions that, when executed by a processor, implement some or all of the steps in any of the above-described methods. The computer program product can be implemented specifically through hardware, software, or a combination thereof. In one optional embodiment, the computer program product is specifically embodied in a computer storage medium; in another optional embodiment, the computer program product is specifically embodied in a software product, such as a software development kit (SDK), etc.

[0193] It should be noted that, Figure 5 This is a schematic diagram of the hardware entity of a vehicle provided in an embodiment of this application, such as... Figure 5 As shown, the hardware entity of the vehicle 500 includes: a processor 501, a communication interface 502, and a memory 503, wherein: The processor 501 typically controls the overall operation of the vehicle 500.

[0194] Communication interface 502 enables the vehicle to communicate with other terminals or servers via a network.

[0195] The memory 503 is configured to store instructions and applications executable by the processor 501, and can also cache data to be processed or already processed by the processor 501 and various modules in the vehicle 500 (e.g., image data, audio data, voice communication data, and video communication data). It can be implemented using flash memory or random access memory (RAM). Data transfer between the processor 501, the communication interface 502, and the memory 503 can be performed via bus 504.

[0196] It should be noted that the descriptions of the storage media and product embodiments above are similar to those of the method embodiments above, and have similar beneficial effects. For technical details not disclosed in the storage media and product embodiments of this application, please refer to the descriptions of the method embodiments of this application for understanding.

[0197] The above embodiments are merely preferred embodiments provided to fully illustrate this application, and the scope of protection of this application is not limited thereto. Equivalent substitutions or modifications made by those skilled in the art based on this application are all within the scope of protection of this application.

Claims

1. A method for controlling a vehicle, characterized in that, The control method includes: Based on the communication parameters of the vehicle's network signal collected in the current driving scenario, the signal anomaly characteristics of the network signal are determined, wherein the current driving scenario is determined based on the vehicle's driving data collected synchronously. Using the abnormal rule set corresponding to the current driving scenario in the preset rule base, the abnormal features of the signal are identified to obtain a first identification result. The rule base includes at least two abnormal rule sets corresponding to driving scenarios. Using a preset recognition model, the abnormal signal features in the current driving scenario are identified to obtain a second recognition result; Based on the first identification result and the second identification result, the anomaly level of the network signal is determined; Based on the anomaly level of the network signal, the vehicle is controlled to execute a preset response strategy.

2. The control method according to claim 1, characterized in that, The step of determining the anomaly level of the network signal based on the first identification result and the second identification result includes: Based on the first identification result and the second identification result, the target identification result is determined; Based on the target identification results, the anomaly level of the network signal is determined.

3. The control method according to claim 2, characterized in that, The step of determining the target identification result based on the first identification result and the second identification result includes: If the first identification result and the second identification result are compatible, the first identification result or the second identification result shall be used as the target identification result; If the first identification result and the second identification result do not match, the target identification result is determined based on the anomaly level corresponding to the first identification result and the anomaly level corresponding to the second identification result.

4. The control method according to claim 2, characterized in that, The target identification results include target anomaly type, target severity, and target duration. The target anomaly type includes one of the following: signal quality, transmission, and security. Determining the anomaly level of the network signal based on the target identification result includes: Obtain a target grading strategy that is compatible with the target anomaly type, wherein the target grading strategy characterizes the relationship between each severity, each duration and each anomaly level under the target anomaly type; According to the target classification strategy, an anomaly level is determined that matches both the severity and duration of the target.

5. The control method according to claim 1, characterized in that, The determination of signal anomaly characteristics of the network signal based on communication parameters of the vehicle's network signal collected in the current driving scenario includes: The communication parameters are preprocessed to obtain the target communication parameters; Using principal component analysis dimensionality reduction algorithm, feature extraction is performed on the target communication parameters to obtain the first feature characterizing the overall signal quality trend; Anomaly features are constructed from the target communication parameters to obtain a second feature; The first feature and the second feature are fused to obtain the signal anomaly feature.

6. The control method according to any one of claims 1 to 5, characterized in that, The response strategy includes at least one of the following: maintaining the vehicle's network function, switching the vehicle's network frequency band, issuing an alarm, or disconnecting some of the vehicle's applications from the network; The method of controlling the vehicle to execute a preset response strategy based on the anomaly level of the network signal includes: If the network signal anomaly level is Level 1, maintain the vehicle's network functionality; If the network signal anomaly level is level two, switch the vehicle's network frequency band and issue an alarm according to the preset first display method; If the network signal anomaly level is level three, disconnect the network of some applications of the vehicle and issue an alarm according to the preset second display mode.

7. The control method according to claim 6, characterized in that, The control method further includes: The communication parameters, the first identification result, and the second identification result are uploaded to the server, and the server is used to update the rule base and the identification model based on the communication parameters, the first identification result, and the second identification result.

8. A vehicle control device, characterized in that, The control device includes: The first determining module is used to determine the signal anomaly characteristics of the network signal based on the communication parameters of the network signal of the vehicle collected in the current driving scenario, wherein the current driving scenario is determined based on the driving data of the vehicle collected synchronously. The first identification module is used to identify the abnormal features of the signal by using the abnormal rule set corresponding to the current driving scenario in the preset rule base, and to obtain a first identification result. The rule base includes at least two abnormal rule sets corresponding to driving scenarios. The second identification module is used to identify the abnormal signal features in the current driving scenario using a preset identification model, and obtain a second identification result; The second determining module is used to determine the anomaly level of the network signal based on the first identification result and the second identification result; The control module is used to control the vehicle to execute a preset response strategy based on the anomaly level of the network signal.

9. A vehicle, characterized in that, The method includes a memory and a processor, wherein the memory stores a computer program or instructions, which, when executed by the processor, implement the method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program or instructions, which, when executed by a processor, implement the method described in any one of claims 1-7.

11. A computer program product, characterized in that, The computer program product includes a computer program or instructions, which, when executed by a processor, implement the method described in any one of claims 1-7.

Images