Vehicle control method, control device, electronic device, and vehicle

Abstract

Embodiments of the present application provide a vehicle control method, a control device, an electronic device and a vehicle. The method comprises: after the vehicle is powered off, obtaining an actual behavior feature of a target node in the vehicle; determining a hibernation deviation degree according to the actual behavior feature and a behavior fingerprint library of the target node, the behavior fingerprint library comprising a hibernation behavior feature of the target node in a standby period in a normal state; and when the hibernation deviation degree reaches a preset risk threshold, performing pre-disposal on the target node to enable the target node to successfully enter a hibernation state before an anomaly occurs. Thus, the method realizes active prediction and management of node anomalies by collecting actual behavior features of nodes in real time, judges risks in advance and intervenes in advance before node anomalies occur, significantly reduces the probability of anomalies, and solves the technical problem in related technologies that a relay power-off control strategy is executed after an abnormal node is monitored, resulting in continuous consumption of battery power during the node anomaly period and possible interruption of user functions.

Description

Technical Field

[0001] This application relates to the field of vehicle power management technology, and in particular to a vehicle control method, control device, electronic device, and vehicle. Background Technology

[0002] In the field of vehicle power management, with the development of vehicle intelligence and connectivity, the number of onboard ECUs (Electronic Control Units) has surged, continuously increasing the complexity of power management. Ideally, all ECUs should enter a sleep state after the vehicle is powered off to save power. However, due to software design flaws, occasional hardware failures, or abnormal network management logic, some ECUs may remain awake, preventing the vehicle network from sleeping and continuously consuming battery power, ultimately leading to vehicle battery depletion and inability to start.

[0003] In related technologies, a passive response power management solution is adopted: after detecting that a node has been abnormally woken up, measures such as resetting and powering off are taken. Although this type of solution can locate and handle abnormal nodes, the battery power consumed during the node's abnormality is irreversible, and the handling process may cause node function interruption, affecting user experience. Summary of the Invention

[0004] This application provides a vehicle control method, control device, electronic device, and vehicle, aiming to improve the problem in related technologies where passive response power management schemes continuously consume battery power during node anomalies, which may cause node function interruption and affect user experience during the processing.

[0005] The first aspect of this application proposes a vehicle control method, which includes: after the vehicle is powered off, acquiring the actual behavioral characteristics of a target node in the vehicle during the current standby cycle, the actual behavioral characteristics including at least the non-sleep duration of the target node from the moment of power-off; determining a sleep deviation degree based on the actual behavioral characteristics and a behavioral fingerprint database of the target node, the behavioral fingerprint database including the sleep behavior characteristics of the target node in a normal state during the standby cycle; and pre-processing the target node when the sleep deviation degree reaches a preset risk threshold, so that the target node can successfully enter a sleep state before an anomaly occurs.

[0006] According to the vehicle control method of this application embodiment, after the vehicle is powered off, the actual behavioral characteristics of the target node in the vehicle during the current standby cycle are obtained. The actual behavioral characteristics include at least the duration of the target node's non-sleep time from the moment of power-off. The sleep deviation degree is determined in real time based on the actual behavioral characteristics and the behavioral fingerprint database of the target node. When the sleep deviation degree reaches a preset risk threshold, the target node is pre-processed so that the target node can successfully enter the sleep state before being identified as an abnormal wake-up node. This achieves proactive prediction and management of node abnormalities, predicts risks and intervenes in advance before node abnormalities occur, significantly reduces the probability of abnormality, and enables the target node to successfully enter the sleep state. This solves the technical problem in related technologies where passive response power management schemes continuously consume battery power and user functions may be interrupted during node abnormalities.

[0007] In some embodiments of this application, the number of feature types of actual behavioral features is multiple. Determining the dormancy deviation degree based on the actual behavioral features and the behavioral fingerprint database of the target node includes: identifying the behavioral deviation of the corresponding actual behavioral feature based on the actual behavioral feature and the corresponding dormancy behavioral feature in the behavioral fingerprint database; and determining the dormancy deviation degree based on the behavioral deviation of multiple actual behavioral features.

[0008] By employing the above technical solutions and using multi-dimensional behavioral characteristics to evaluate the dormancy deviation of target nodes, the accuracy of risk assessment for dormancy anomalies can be improved.

[0009] In some embodiments of this application, the actual behavioral characteristics also include network management message characteristics and environmental association characteristics of the target node from the moment of power-off. The sleep deviation degree is determined based on the behavioral deviation of multiple actual behavioral characteristics, including: determining the feature weights corresponding to multiple actual behavioral characteristics; and performing a weighted calculation on the behavioral deviation of multiple actual behavioral characteristics based on the feature weights corresponding to multiple actual behavioral characteristics to obtain the sleep deviation.

[0010] By employing the above technical solution, which uses a weighted fusion of behavioral deviations from multiple actual behavioral characteristics to determine the dormancy deviation of the target node, the computational resource requirements are reduced and computational efficiency is improved while ensuring computational accuracy.

[0011] In some embodiments of this application, the vehicle control method further includes: acquiring at least one historical behavior sequence data of a target node, wherein the historical behavior sequence data characterizes the actual behavior characteristics of the target node during the process of successfully entering a hibernation state within a historical standby period; determining current behavior sequence data based on the actual behavior characteristics of the target node, wherein the current behavior sequence data characterizes the actual behavior characteristics of the target node during the process of successfully entering a hibernation state within the current standby period; predicting the anomalous probability of the target node based on the current behavior sequence data and at least one historical behavior sequence data; and adjusting a preset risk threshold and / or feature weights based on the anomalous probability.

[0012] By combining the above technical solutions with the vehicle's historical behavior sequence data from past cycles and the current behavior sequence data, the power-down behavior trend of the target node is analyzed, and the probability of the target node's abnormality is predicted. This allows for the adjustment of the preset risk threshold and / or feature weights. For example, when the probability of an abnormality in the target node increases, the preset risk threshold is lowered, so that the pre-treatment strategy matches the actual state of the vehicle, thus ensuring the control effect.

[0013] In some embodiments of this application, pre-processing is performed on the target node when the sleep deviation reaches a preset risk threshold, including: generating and sending a diagnostic request to the target node when the sleep deviation reaches the preset risk threshold, wherein the target node reads self-diagnostic data based on the diagnostic request and generates a self-diagnostic feedback signal based on the self-diagnostic data; if no self-diagnostic feedback signal is received within a preset time, or if a self-diagnostic feedback signal is received within a preset time and it is determined based on the self-diagnostic feedback signal that the target node has a non-sleep record and / or an abnormal wake-up source, controlling the target node to reset and restart; if a self-diagnostic feedback signal is received within a preset time and it is determined based on the self-diagnostic feedback signal that the target node does not have a non-sleep record and an abnormal wake-up source, continuously monitoring the sleep deviation, and controlling the target node to reset and restart when the sleep deviation reaches a preset reset threshold, wherein the preset reset threshold is greater than the preset risk threshold.

[0014] Through the above technical solution, risk prediction and graded handling of target nodes are carried out based on the sleep deviation degree obtained by real-time monitoring. Before the node reaches the abnormal judgment standard, the abnormal risk is predicted in advance according to the preset threshold, and graded intervention measures are taken to achieve predictive power management.

[0015] In some embodiments of this application, after the target node is reset and restarted, the vehicle control method further includes: acquiring the actual behavior characteristics of the target node after reset within a preset reset window; determining the reset sleep deviation degree based on the actual behavior characteristics after reset and the behavior fingerprint database; and controlling the target node to power off for protection when the reset sleep deviation degree is greater than or equal to a preset risk threshold.

[0016] Through the above technical solution, after the target node completes the restart after the pre-treatment measures are implemented, the master control node enters the verification window period to continuously monitor the network behavior of the target node. When the node is found to be abnormal, power-off protection measures are implemented for the target node to prevent situations such as battery depletion caused by abnormal hibernation.

[0017] In some embodiments of this application, the vehicle control method further includes: updating the sleep behavior characteristics based on the actual behavior characteristics after reset when the reset sleep deviation is less than a preset risk threshold.

[0018] The above technical solution enables the execution of pre-treatment measures and, after determining that the target node has returned to normal, to update the behavior fingerprint database of the node by combining the actual behavior characteristics continuously acquired after the reset, making the behavior fingerprint database more closely approximate the vehicle's actual performance.

[0019] A second aspect of this application provides a vehicle control device, comprising: an acquisition module, configured to acquire, after the vehicle is powered off, the actual behavioral characteristics of a target node in the vehicle during the current standby cycle, the actual behavioral characteristics including at least the duration of the target node's non-sleep state from the moment of power-off; a determination module, configured to determine a sleep deviation degree based on the actual behavioral characteristics and a behavioral fingerprint database of the target node, the behavioral fingerprint database including the sleep behavioral characteristics of the target node in a normal state during the standby cycle; and a control module, configured to pre-process the target node when the sleep deviation degree reaches a preset risk threshold, so that the target node successfully enters a sleep state before an anomaly occurs.

[0020] A third aspect of this application provides an electronic device including a processor and a memory, wherein the memory is used to store a computer program; and the processor is used to execute the program stored in the memory to implement the vehicle control method proposed in the first aspect.

[0021] The fourth aspect of this application provides a vehicle that includes the aforementioned electronic equipment. Attached Figure Description

[0022] Figure 1 This is a flowchart of a vehicle control method provided in an embodiment of this application; Figure 2 This is a schematic diagram of the system architecture of a vehicle provided in an embodiment of this application; Figure 3 This is a flowchart of a vehicle control method provided in a specific embodiment of this application; Figure 4 This is a connection diagram of a vehicle control device provided in an embodiment of this application; Figure 5 This is a connection diagram of an electronic device provided in an embodiment of this application; Figure 6 This is a block diagram of a vehicle provided in one embodiment of this application. Detailed Implementation

[0023] To make the technical problems, technical solutions, and beneficial effects solved by this application clearer, the following detailed description is provided in conjunction with embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0024] In the field of vehicle power management, with the development of vehicle intelligence and connectivity, the number of onboard ECUs has surged, and the complexity of power management has continuously increased. Ideally, after the vehicle is powered off, all ECUs should enter a sleep state to save power. However, due to software design flaws, occasional hardware failures, or abnormal network management logic, some ECUs may remain awake, preventing the vehicle network from sleeping, continuously consuming battery power, and ultimately causing the vehicle to run out of power and fail to start.

[0025] In related technologies, methods for preventing vehicle battery drain can be divided into three categories: (1) Relay-level power-off scheme: In the vehicle, the battery connects to multiple ECUs and loads via relays. After the vehicle is powered off, the duration of network wake-up is recorded. When the duration reaches a preset time, a network anomaly is detected, and the relays are disconnected to prevent the vehicle from running out of power. However, this scheme is simple and crude, using a one-size-fits-all power-off control method, which will cause all related functions to fail.

[0026] (2) Direct power-off solution for network monitoring: After the vehicle leaves the station, non-dormant nodes in the vehicle network are identified by monitoring network management messages. Once a node is found, its power is directly cut off. Although this solution can locate abnormal nodes, it is a reactive approach, and intervention is only taken after the node has been abnormally consuming power for a period of time.

[0027] (3) Cloud-assisted analysis solution: The vehicle network status is uploaded to the cloud, and the cloud analyzes the data and issues commands. This solution relies on network connectivity and is also a reactive response.

[0028] Therefore, the above technical solutions are all passive response management, meaning the system only detects and handles issues after a node has already malfunctioned and started consuming power. This management approach cannot prevent malfunctions from occurring, and the power consumption during a node malfunction is irreversible. Furthermore, this solution fails to utilize historical node behavior patterns to predict future malfunctions, lacking proactive prevention capabilities. Therefore, while the above solutions achieve the identification and handling of abnormal nodes, they have the following shortcomings: (1) Passive response, no prevention: All processing measures are only triggered after the node has been abnormally woken up and started consuming power. Power consumption during the node abnormality period cannot be recovered, and user functions may have been affected.

[0029] (2) Lack of historical behavior learning: The historical dormant behavior patterns of nodes are not utilized, and each standby cycle is judged independently, making it impossible to identify abnormal precursors or risk trends of nodes.

[0030] (3) Single decision-making basis: Decisions are made solely based on the current abnormal state (such as the duration of network management messages) without taking into account the node’s historical performance, environmental factors, vehicle status and other multi-dimensional information for comprehensive risk assessment.

[0031] (4) Unable to achieve self-optimization: The results of abnormal handling (such as successful reset or successful power failure) are not effectively used to optimize subsequent decision parameters, and the system cannot learn from historical experience and improve itself.

[0032] To address at least one of the aforementioned technical problems, this application proposes a vehicle control method. This method monitors the actual behavioral characteristics of a target node in real time after the vehicle is powered off. By combining this with pre-defined dormant behavior characteristics in the target node's behavioral fingerprint database, the dormant deviation of the target node is determined. When the dormant deviation reaches a preset risk threshold, pre-processing is performed on the target node, such as restarting or power-off, to ensure the target node successfully enters a dormant state before an anomaly occurs. This achieves proactive prediction and management of node anomalies, anticipating risks and intervening in advance before anomalies occur, significantly reducing the probability of anomalies and enabling the target node to successfully enter a dormant state. This solves the technical problem in related technologies where passive response power management schemes continuously consume battery power and may interrupt user functions during node anomalies.

[0033] The vehicle control method of this application will be described in detail below with reference to the accompanying drawings.

[0034] Combination Figure 1 As shown in the figure, a vehicle control method according to an embodiment of this application includes the following steps: S1. After the vehicle is powered off, obtain the actual behavior characteristics of the target node in the vehicle during the current standby cycle. The actual behavior characteristics include at least the duration of the target node's non-sleep time from the moment of power-off.

[0035] Specifically, upon receiving a vehicle power-down command, and after the vehicle is powered down, the actual behavioral characteristics of each target node in the vehicle during the current standby cycle are monitored in real time. The target node is the electronic control unit (ECU) on the vehicle, and the actual behavioral characteristics are those used to characterize the ECU's behavior during the process of entering a sleep state after the vehicle is powered down. These characteristics could include sleep duration curves, NM message characteristics, etc., to assess whether the target node has successfully entered sleep mode.

[0036] The sleep duration curve refers to the time distribution of the target node entering sleep mode after the vehicle is powered off. For example, the probability of a node entering sleep mode within 30 seconds after power-off is 95%, and the probability is 5% between 30 and 60 seconds. Network management message characteristics refer to the frequency and content pattern of NM (Network Management) messages sent by the target node during standby. For example, when the ECU is in normal operation / ready sleep mode, it will periodically send NM messages; after the ECU enters sleep mode, it will stop sending NM messages.

[0037] S2, determine the hibernation deviation based on the actual behavioral characteristics and the behavioral fingerprint database of the target node. The behavioral fingerprint database includes the hibernation behavioral characteristics of the target node in the standby period when it is in a normal state.

[0038] Specifically, the dormant behavior characteristics of the target nodes can be pre-defined according to the actual behavior characteristic types. That is, a behavioral fingerprint database is established for each target node (i.e., each electronic control unit (ECU) on the vehicle), recording its dormant behavior patterns under normal conditions as the dormant behavior characteristics of the target nodes. The establishment of the behavioral fingerprint database can adopt a phased, gradual approach, including the following stages: Phase 1: Factory Calibration. During the vehicle development phase, bench testing or real-vehicle calibration is used to obtain the sleep behavior characteristics of each ECU of the vehicle model under normal conditions (such as the distribution of non-sleep duration, NM message frequency, etc.), which are then used as an initial fingerprint database and pre-installed into the vehicle software.

[0039] Phase Two: Real-Vehicle Adaptive Learning. After vehicle delivery, the system continuously records the actual sleep behavior of each node during each normal power-down standby cycle. After several normal cycles (e.g., 10 cycles), the system dynamically corrects the fingerprint database using the actual data from the vehicle, making it approximate the true behavioral characteristics of the vehicle's nodes.

[0040] Phase 3: Cloud Aggregation and Optimization. This phase involves aggregating massive amounts of normal behavior data from vehicles of the same model, statistically analyzing this data to obtain a standard fingerprint database for that model, and then distributing updates via vehicle-to-cloud communication to continuously improve the accuracy of the fingerprint database.

[0041] In practical applications, standard sleep characteristics, i.e., sleep behavior characteristics, of the target node are retrieved from the behavioral fingerprint database. The real-time collected actual behavior characteristics are compared with the standard sleep characteristics to calculate the sleep deviation of the target node. For example, when the actual behavior characteristic is a sleep duration curve, the current of the target node after receiving a power-down command is monitored in real time to plot the actual sleep duration curve of the node. The actual sleep duration curve is then compared with the standard sleep duration curve retrieved from the behavioral fingerprint database. The sleep deviation is evaluated based on the curve similarity. For example, if the curve similarity is greater than 99%, the sleep deviation is considered to be 0%.

[0042] S3, when the hibernation deviation reaches the preset risk threshold, pre-process the target node to ensure that the target node successfully enters hibernation state before hibernation anomaly occurs.

[0043] Specifically, the preset risk threshold is used to assess the risk level of the target node's hibernation anomaly. It can be set according to the actual situation. For example, the preset risk threshold can be set to 30%. When the hibernation deviation of the target node reaches 30%, it is considered that the target node has a hibernation anomaly and is at medium risk. At this time, the target node is preprocessed, such as controlling the target node to reset and restart, so as to re-execute the power-down hibernation process, so that the target node can complete hibernation before the hibernation anomaly occurs.

[0044] Compared to related technologies, this embodiment upgrades from passive response power management to active predictive management, predicting risks and intervening in advance before node hibernation anomalies occur, significantly reducing the occurrence rate of hibernation anomalies and reducing battery energy waste.

[0045] In some embodiments of this application, the number of feature types of actual behavioral features is multiple. The dormancy deviation degree is determined based on the actual behavioral features and the behavioral fingerprint database of the target node, including: identifying the behavioral deviation of the corresponding actual behavioral feature based on the actual behavioral features and the corresponding dormancy behavioral features in the behavioral fingerprint database; and determining the dormancy deviation degree based on the behavioral deviation of multiple actual behavioral features.

[0046] Specifically, when there are multiple actual behavioral characteristics, the vehicle behavior fingerprint database sets corresponding dormant behavior characteristics for each characteristic type. In practical applications, each actual behavioral characteristic collected is compared with the corresponding dormant behavior characteristics to calculate the behavior deviation of each actual behavioral characteristic. Then, the behavior deviations of multiple actual behavioral characteristics are fused to calculate the dormant deviation degree of the target node, which is used to predict the risk of the node experiencing dormant anomalies.

[0047] This embodiment uses multi-dimensional behavioral characteristics to evaluate the dormancy deviation of the target node, thereby improving the accuracy of risk assessment for dormancy anomalies.

[0048] In some embodiments of this application, multiple actual behavioral characteristics include non-sleep duration curves, network management message characteristics, and environmental association characteristics.

[0049] Specifically, the non-sleep duration curve refers to the time distribution of the target node entering sleep mode after the vehicle is powered off. For example, the probability of a node entering sleep mode within 30 seconds after power-off is 95%, and the probability is 5% between 30 and 60 seconds. Network management message characteristics refer to the frequency and content pattern of NM (Network Management) messages sent by the target node during standby. Environmental correlation characteristics refer to the sleep behavior of the target node under different ambient temperatures (e.g., sleep time may be prolonged in low-temperature environments).

[0050] After the vehicle is powered off and enters standby mode, the master control node monitors the network behavior of each target node in real time, including: whether the node sends NM messages; the frequency and content of the NM messages sent by the node; and the "non-sleep duration" of the node in the current standby cycle (i.e., the time elapsed since the power-off moment before the node has entered sleep mode).

[0051] Then, the real-time monitored behavioral features are compared with the corresponding dormant behavioral features in the normal mode in the behavioral fingerprint database, and the behavioral deviation corresponding to each behavioral feature is calculated to evaluate the dormant deviation degree of the target node in real time.

[0052] This embodiment establishes a multi-dimensional behavioral fingerprint database for each node, including the non-sleep duration curve, network management message characteristics, and environmental correlation characteristics, and quantifies the degree to which a node deviates from the normal pattern through behavioral deviations. This multi-dimensional quantitative assessment method can identify precursory features of an impending node anomaly (such as a gradually increasing non-sleep duration and abnormal fluctuations in NM message frequency), enabling early risk identification.

[0053] In some embodiments of this application, the actual behavioral characteristics also include network management message characteristics and environmental association characteristics of the target node from the moment of power-off. The sleep deviation degree is determined based on the behavioral deviation of multiple actual behavioral characteristics, including: determining the feature weights corresponding to multiple actual behavioral characteristics; and performing a weighted calculation on the behavioral deviation of multiple actual behavioral characteristics based on the feature weights corresponding to multiple actual behavioral characteristics to obtain the sleep deviation degree.

[0054] Specifically, after the vehicle enters sleep mode after power-off, multiple actual behavioral characteristics of each target node are acquired, and the behavioral deviation between each actual behavioral characteristic and the corresponding sleep behavior characteristic is calculated. Taking the non-sleep duration curve as an example, assuming that the standard non-sleep duration curve in the behavioral fingerprint database of the target node (such as a gateway) shows that it enters sleep mode within 10 seconds (T_normal) after power-off in 90% of its historical normal standby cycles, then this embodiment uses the following formula to calculate the behavioral deviation corresponding to the non-sleep duration curve: Behavioral deviation = (T_current - T_normal) / T_normal × 100% Wherein, T_normal is the standard non-sleep duration determined by the standard sleep duration curve; T_current is the "non-sleep duration" (in seconds) of the node in the current cycle from the moment of power-off. Since the value of T_current increases in real time, the corresponding behavior deviation also changes dynamically.

[0055] For example, Table 1 shows an example of calculating the dormancy deviation using only the non-dormant duration curve for risk identification.

[0056] Table 1

[0057] The thresholds (30%, 70%) and standard sleep time (T_normal=10 seconds) mentioned above are all configurable initial values, which can be dynamically adjusted according to the fingerprint database data in actual applications.

[0058] The behavioral deviations corresponding to each actual behavioral feature are calculated in the same way. Then, by introducing multi-dimensional weights, the behavioral deviations of multiple actual behavioral features are weighted and fused to obtain the sleep deviation degree of the target node. Taking multiple actual behavioral features, including the non-sleep duration curve and network management message features, as an example, the sleep time deviation is calculated based on the non-sleep duration curve, and the NM message frequency deviation is calculated based on the network management message features. The sleep deviation degree = (sleep time deviation × weight 1) + (NM message frequency deviation × weight 2).

[0059] This embodiment uses a weighted fusion of behavioral deviations from multiple actual behavioral features as the dormancy deviation of the target node, which reduces the computational resource requirements and improves computational efficiency while ensuring computational accuracy.

[0060] In some embodiments of this application, the vehicle control method further includes: acquiring at least one historical behavior sequence data of a target node, wherein the historical behavior sequence data characterizes the actual behavior characteristics of the target node during the process of successfully entering a hibernation state within a historical standby period; determining current behavior sequence data based on the actual behavior characteristics of the target node, wherein the current behavior sequence data characterizes the actual behavior characteristics of the target node during the process of successfully entering a hibernation state within the current standby period; predicting the anomalous probability of the target node based on the current behavior sequence data and at least one historical behavior sequence data; and adjusting a preset risk threshold and / or feature weights based on the anomalous probability.

[0061] In other words, historical behavior sequence data and current behavior sequence data of the vehicle over multiple periods are retrieved. For example, the last five standby cycles, with sleep durations of 10, 11, 13, 16, and 20 seconds respectively, are used to analyze the power-down behavior trend of the target node and predict the probability of anomalies occurring at the target node. Based on the prediction results, preset risk thresholds and / or feature weights are adjusted. For example, when the probability of anomalies in the target node increases, the preset risk threshold is lowered, or the weights of various features are adjusted, so that the pre-treatment strategy matches the actual state of the vehicle and ensures the effectiveness of the control strategy.

[0062] For example, a prediction model can be configured to predict the probability of anomalies in a target node based on multiple sets of behavioral sequence data. This prediction model can be configured in the cloud or on the vehicle itself, with no specific restrictions.

[0063] Taking a predictive model configured in the cloud as an example, monitoring data and pre-treatment records from all vehicle nodes are reported in real time to the cloud-based big data platform via the TBOX (Telematics Box, onboard telematics control unit). The cloud platform then performs the following analyses: 1. Behavioral fingerprint optimization: Aggregate massive vehicle data and continuously optimize the behavioral fingerprint database of each node under normal conditions (such as different vehicle models and different batches of nodes may have different normal behavioral patterns).

[0064] 2. Predictive Model Definition: Time-series prediction models, anomaly detection models, etc., can be used as prediction models. The input is multi-period behavior sequence data reported by vehicles (such as the non-dormant duration sequence of the past N periods, NM message frequency sequence, etc.), and the output is the risk probability or prediction score of an anomaly occurring within a specific future time window. This model differs from the rule-based deviation judgment at the vehicle end; it learns anomaly precursor patterns from massive amounts of data to achieve early identification of trend risks.

[0065] 3. Predictive Model Training: Machine learning algorithms (such as time series prediction and anomaly detection models) are used to analyze behavioral characteristics preceding anomalies, identify early warning patterns, and continuously optimize the accuracy of risk prediction. Specifically, this can involve analyzing trends such as the gradual increase in the duration of a node's non-dormant state and the patterns of abnormal fluctuations in NM message frequency.

[0066] 4. Model Deployment: The optimized prediction model or optimized decision parameters are deployed to the vehicle via vehicle-to-cloud communication (upgrades can be made via OTA (Over-The-Air) or real-time synchronization), enabling the vehicle's prediction capabilities to continuously evolve.

[0067] The application methods after the model is distributed include any of the following: (1) Parameter optimization mode: The cloud model outputs the optimized decision parameters (such as deviation threshold and weight coefficient of each feature). After receiving the results, the vehicle updates the local behavior deviation calculation formula and risk classification rules. (2) Edge-side inference mode: The cloud distributes the trained lightweight prediction model (such as random forest or logistic regression model) to the vehicle. The vehicle collects behavioral sequence data of the current period and historical periods in real time, inputs it into the local model, and the model outputs a risk score. When the risk score exceeds the preset threshold, parameter optimization measures are actively triggered to adjust the deviation threshold and / or the weight coefficients.

[0068] Thus, through the above methods, the vehicle always possesses the latest knowledge from the cloud, forming a continuous evolutionary closed loop of cloud training and vehicle-side inference.

[0069] In some embodiments of this application, pre-processing is performed on the target node when the sleep deviation reaches a preset risk threshold, including: generating and sending a diagnostic request to the target node when the sleep deviation reaches the preset risk threshold, wherein the target node reads self-diagnostic data based on the diagnostic request and generates a self-diagnostic feedback signal based on the self-diagnostic data; if no self-diagnostic feedback signal is received within a preset time, or if a self-diagnostic feedback signal is received within a preset time and it is determined based on the self-diagnostic feedback signal that the target node has a non-sleep record and / or an abnormal wake-up source, controlling the target node to reset and restart; if a self-diagnostic feedback signal is received within a preset time and it is determined based on the self-diagnostic feedback signal that the target node does not have a non-sleep record and an abnormal wake-up source, continuously monitoring the sleep deviation, and controlling the target node to reset and restart when the sleep deviation reaches a preset reset threshold, wherein the preset reset threshold is greater than the preset risk threshold.

[0070] Specifically, taking a preset risk threshold of 30% and a preset reset threshold of 70% as an example, the node risk level is divided into three levels according to the dormancy deviation, and corresponding pre-treatment measures are implemented, as shown in Table 2.

[0071] Table 2

[0072] The risk level classification uses a threshold-triggered mechanism, rather than a waiting-within-a-range mechanism. During this vehicle power-off process, the dormancy deviation of each target node is continuously calculated. When it first reaches or exceeds a preset threshold for a certain risk level, the system immediately triggers the corresponding action. For example, when the deviation first reaches 30%, a medium-risk action is immediately triggered, instead of waiting until 69%; when the deviation first reaches 70%, a high-risk action is immediately triggered. This design ensures that the system can intervene at the earliest point when node behavior abnormally deteriorates.

[0073] Taking Table 2 as an example, when the vehicle enters sleep mode, the vehicle's master control node acquires the actual behavioral characteristics of each target node in order to calculate the sleep deviation of each target node.

[0074] When the calculated dormancy deviation is less than 30%, the master node considers the target node to be in a low-risk range, does not take active intervention measures, marks it as a concern, and extends the monitoring period.

[0075] When the calculated hibernation deviation reaches 30%, the master node considers the target node to have entered the medium-risk zone and triggers a diagnostic probe when the target node first enters the medium-risk zone; that is, the master node sends a diagnostic request to the target node. After receiving the diagnostic request, the target node reads the non-hibernation self-diagnostic data recorded internally. This non-hibernation self-diagnostic data may include the timestamp of the non-hibernation occurrence, the wake-up source, the maintenance source, etc., and generates a self-diagnostic feedback signal based on the self-diagnostic data, which is then sent to the master node.

[0076] If the target node does not respond or returns a negative response code within 2 seconds, that is, if the master node does not receive a self-diagnostic feedback signal from the target node within the preset time, the target node is determined to be seriously abnormal, immediately upgraded to high risk, and a preventive reset is performed. If the target node returns self-diagnostic data normally and the data content indicates the existence of a non-sleep record or an abnormal wake-up source, that is, if the master node receives a self-diagnostic feedback signal within a preset time and determines that the target node has a non-sleep record and / or an abnormal wake-up source based on the self-diagnostic feedback signal, it will be upgraded to high risk and a preventive reset will be performed. If the master node receives a self-diagnostic feedback signal within a preset time and determines, based on the self-diagnostic feedback signal, that the target node has no non-sleep records or abnormal wake-up sources, then it determines that the current deviation may originate from occasional fluctuations. The master node will no longer repeatedly send probes but will continue to monitor the deviation changes in real time. If the deviation subsequently falls back to the low-risk range, the risk level will be downgraded accordingly; if the deviation continues to increase and reaches or exceeds 70% for the first time, it will be immediately upgraded to high risk and a preventative reset will be performed.

[0077] Among them, preventive reset refers to the master node sending a soft reset command to enable the target node to restart and return to normal status before the abnormality occurs.

[0078] This embodiment uses real-time monitoring of sleep deviation to predict and classify the risks of target nodes. The prediction process does not refer to predicting the duration of a single sleep cycle, but rather to continuously monitoring the node's behavior characteristics from the moment of power-off within a single standby cycle. As time increases, the deviation from the behavior fingerprint database is calculated in real time. Before a node reaches the anomaly judgment criteria, a pre-set threshold is used to predict its potential for anomalies, and graded intervention measures are taken to achieve predictive power management.

[0079] In some embodiments of this application, after the target node is reset and restarted, the vehicle control method further includes: acquiring the actual behavior characteristics of the target node after reset within a preset reset verification window; determining the reset sleep deviation degree based on the actual behavior characteristics after reset and the behavior fingerprint database; and controlling the target node to power off when the reset sleep deviation degree is greater than or equal to a preset risk threshold.

[0080] In other words, after the pre-processing measures are implemented and the target node completes its restart, the master control node enters a verification window period to continuously monitor the network behavior of the target node. The preset verification window duration after reset can be set according to actual conditions, such as 60 seconds.

[0081] If the deviation of the target node does not return to the normal range within the verification window after the preset reset, i.e. the dormancy deviation is greater than or equal to the first preset threshold, then the node is confirmed to be abnormal, and power-off protection measures for the target node are executed, such as power-off reset or complete power-off, to prevent the occurrence of battery depletion caused by abnormal dormancy.

[0082] In some embodiments of this application, the vehicle control method further includes: updating the sleep behavior characteristics based on the actual behavior characteristics after reset when the reset sleep deviation is less than a preset risk threshold.

[0083] In other words, if the deviation of the target node returns to the normal range within the preset reset verification window period, i.e., the sleep deviation is less than the first preset threshold, the target node is determined to have returned to normal. Then, the behavior fingerprint database of the node is updated based on the actual behavioral characteristics continuously acquired after the reset. Specifically, the successful pre-processing record can be included in historical data, and a weighted average fusion method can be used to generate the corresponding sleep behavior characteristics for subsequent power-down control. Assuming the target node took 11 seconds from power-down reset to sleep (the behavior fingerprint database originally showed 10 seconds), the 11 seconds determined in this acquisition is added to the statistical sample, and the distribution (e.g., mean, 90th percentile) is recalculated to allow the behavior fingerprint database to gradually approximate the vehicle's actual performance.

[0084] Furthermore, the successful record of this pre-treatment can be incorporated into historical data to optimize the aforementioned prediction model. In other words, after this power-down control, all node behavior monitoring data, pre-treatment records, and verification results can be reported in real-time to the cloud-based big data platform via TBOX. The cloud platform then performs anomaly probability prediction operations on the target node to optimize the deviation threshold and / or the weights of each feature.

[0085] As a specific embodiment of this application, the system architecture is as follows: Figure 2 As shown, this vehicle control method is executed by the master control node, with a preset risk threshold of 30% and a preset reset threshold of 70%. The vehicle control method is as follows: Figure 3 As shown, the following steps may be included: S101 monitors the network behavior of each node in real time when the vehicle is powered off and enters standby mode in order to obtain the actual behavioral characteristics of the nodes.

[0086] S102, compare the actual behavioral characteristics with the dormant behavioral characteristics in the behavioral fingerprint database, and calculate the dormancy deviation.

[0087] S103, determine whether the hibernation deviation is less than 30%. If yes, proceed to step S104; otherwise, proceed to step S105.

[0088] S104: Determine if the node is in a low-risk zone, mark it as a priority, and extend the monitoring period.

[0089] S105, determine whether the hibernation deviation is less than 70%. If yes, proceed to step S106; otherwise, proceed to step S108.

[0090] S106, determine that the node is in the medium-risk range, and send diagnostic probes to the node.

[0091] S107, determine if there is no response from the node or if the feedback data is abnormal. If yes, proceed to step S108; otherwise, proceed to step S113.

[0092] S108, if the node is determined to be in a high-risk zone, perform preventative reset control on the node to reset and restart it.

[0093] S109: After the node is reset and restarted, the node behavior is continuously monitored during the reset verification window, and the post-reset deviation is calculated.

[0094] S110, determine whether the deviation after reset is less than 30%. If yes, proceed to step S111; otherwise, proceed to step S112.

[0095] S111, Abnormal risk resolved, update behavioral fingerprint database (record this successful intervention), end this cycle processing.

[0096] S112, confirm that the node has entered a critically abnormal state, and execute power-off protection measures for the node.

[0097] S113, continue to monitor changes in deviation in real time.

[0098] This method first establishes a behavioral fingerprint database for each ECU's hibernation behavior through three stages: factory calibration, real-vehicle adaptive learning, and cloud-based aggregation optimization, recording the characteristics of its normal hibernation behavior. After the vehicle is powered off, the system monitors the actual network behavior of the ECU in real time, compares it with the behavioral fingerprint database, and calculates the deviation. When the deviation reaches a certain threshold, the system predicts that the ECU has an abnormal risk and takes different levels of pre-treatment measures according to the risk level: for low risk, it only marks the ECU as a concern; for medium risk, it sends a diagnostic probe (triggered only once per cycle) to read the node's internal non-hibernation self-diagnostic data, and upgrades the system if the node does not respond or the data indicates abnormality; for high risk, it proactively sends a reset command for preventative reset. After pre-treatment, the system enters a verification window period to confirm whether the node has returned to normal and feeds the verification results back to the cloud. The cloud aggregates massive amounts of data, continuously optimizes the behavioral fingerprint database and prediction model through machine learning, and issues updates through vehicle-cloud communication, forming a self-evolving closed loop.

[0099] Therefore, this control method has the following technical advantages compared to related technical solutions: (1) From passive response to active prevention, fundamentally reducing the occurrence rate of anomalies: By predicting risks and intervening in advance before anomalies occur, the power management is fundamentally improved compared to the control strategies of existing technologies that can only be dealt with after the fact.

[0100] (2) Significantly reduce user function interruption and power waste: Pre-treatment measures (such as preventive reset) are completed without the user's awareness, avoiding the node from entering an abnormal state, thus eliminating function interruption and eliminating unnecessary power consumption from the source.

[0101] (3) Identify the abnormal precursors of the target node and predict the abnormal probability of the target node in order to optimize the control parameters.

[0102] In summary, the vehicle control method according to the embodiments of this application acquires the actual behavioral characteristics of the target node in the vehicle during the current standby cycle after the vehicle is powered off. The actual behavioral characteristics include at least the duration of the target node's non-sleep state from the moment of power-off. The sleep deviation is determined in real time based on the actual behavioral characteristics and the behavioral fingerprint database of the target node. When the sleep deviation reaches a preset risk threshold, the target node is pre-processed to ensure that the target node successfully enters a sleep state before being identified as an abnormal wake-up node. This enables proactive prediction and management of node anomalies, anticipates risks and intervenes in advance before node anomalies occur, significantly reduces the probability of anomalies, and achieves successful sleep of the target node. This solves the technical problem in related technologies where passive response power management schemes continuously consume battery power and may interrupt user functions during node anomalies.

[0103] This application also provides a vehicle control device, such as... Figure 4 As shown, the device includes: an acquisition module 10, used to acquire the actual behavioral characteristics of a target node in the vehicle during the current standby cycle when the vehicle is powered off, the actual behavioral characteristics including at least the duration of the target node's non-sleep state from the moment of power-off; a determination module 20, used to determine the sleep deviation degree based on the actual behavioral characteristics and the target node's behavioral fingerprint database, the behavioral fingerprint database including the sleep behavior characteristics of the target node in a normal state during the standby cycle; and a control module 30, used to pre-process the target node when the sleep deviation degree reaches a preset risk threshold, so that the target node can successfully enter a sleep state before an anomaly occurs.

[0104] In some embodiments of this application, the number of feature types of actual behavioral features is multiple. The determining module 20 determines the dormancy deviation degree based on the actual behavioral features and the behavioral fingerprint database of the target node. Specifically, it is used to: identify the behavioral deviation of the corresponding actual behavioral feature based on the actual behavioral features and the corresponding dormancy behavioral features in the behavioral fingerprint database; and determine the dormancy deviation degree based on the behavioral deviation of multiple actual behavioral features.

[0105] In some embodiments of this application, the actual behavioral characteristics also include network management message characteristics and environmental association characteristics of the target node from the moment of power-off. The determining module 20 determines the sleep deviation degree based on the behavioral deviation of multiple actual behavioral characteristics. Specifically, it is used to: determine the feature weights corresponding to multiple actual behavioral characteristics; and perform weighted calculation on the behavioral deviation of multiple actual behavioral characteristics based on the feature weights corresponding to multiple actual behavioral characteristics to obtain the sleep deviation.

[0106] In some embodiments of this application, the control module 30 is further configured to: acquire at least one historical behavior sequence data of the target node, wherein the historical behavior sequence data characterizes the actual behavior characteristics of the target node during the process of successfully entering a hibernation state within a historical standby period; determine the current behavior sequence data based on the actual behavior characteristics of the target node, wherein the current behavior sequence data characterizes the actual behavior characteristics of the target node during the process of successfully entering a hibernation state within a historical standby period; predict the anomalous probability of the target node based on the current behavior sequence data and at least one historical behavior sequence data; and adjust the preset risk threshold and / or feature weights based on the anomalous probability.

[0107] In some embodiments of this application, the control module 30 pre-processes the target node when the sleep deviation reaches a preset risk threshold. Specifically, it is used to: generate and send a diagnostic request to the target node when the sleep deviation reaches the preset risk threshold, wherein the target node reads self-diagnostic data based on the diagnostic request and generates a self-diagnostic feedback signal based on the self-diagnostic data; if no self-diagnostic feedback signal is received within a preset time, or if a self-diagnostic feedback signal is received within a preset time and it is determined based on the self-diagnostic feedback signal that the target node has a non-sleep record and / or an abnormal wake-up source, the control module 30 controls the target node to reset and restart; if a self-diagnostic feedback signal is received within a preset time and it is determined based on the self-diagnostic feedback signal that the target node does not have a non-sleep record and an abnormal wake-up source, the control module 30 continuously monitors the sleep deviation and controls the target node to reset and restart when the sleep deviation reaches a preset reset threshold, wherein the preset reset threshold is greater than the preset risk threshold.

[0108] In some embodiments of this application, after the target node is reset and restarted, the control module 30 is further configured to: acquire the actual behavior characteristics of the target node after reset within a preset reset window; determine the reset sleep deviation degree based on the actual behavior characteristics after reset and the behavior fingerprint database; and control the target node to power off when the reset sleep deviation degree is greater than or equal to a preset risk threshold.

[0109] In some embodiments of this application, the control module 30 is further configured to: update the hibernation behavior characteristics according to the actual behavior characteristics after reset when the reset hibernation deviation is less than a preset risk threshold.

[0110] This application also provides an electronic device 100, please refer to... Figure 5 It includes a memory 110 and a processor 120, wherein the memory 110 is used to store computer programs; and the processor 120 is used to execute the programs stored in the memory 110 to implement the vehicle control method described in any embodiment of this application.

[0111] This application also provides a vehicle 1000, please refer to... Figure 6The vehicle 1000 includes the aforementioned electronic equipment 100.

[0112] In this application, "multiple" refers to two or more.

[0113] In this application, unless otherwise expressly defined, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0114] The terms “first,” “second,” “third,” “fourth,” etc., in this application (if present) are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0115] In this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, in this application, the character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0116] Unless otherwise specified, all steps in this application may be performed sequentially or randomly. For example, if the method includes steps A and B, it means that the method may include steps A and B performed sequentially, or it may include steps B and A performed sequentially. For example, if the method may also include step C, it means that step C may be added to the method in any order. For example, the method may include steps A, B, and C, or it may include steps A, C, and B, or it may include steps C, A, and B, etc.

[0117] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A vehicle control method, characterized in that, include: After the vehicle is powered off, the actual behavior characteristics of the target node in the vehicle during the current standby period are obtained. The actual behavior characteristics include at least the duration of the target node's non-sleep time from the moment of power-off. The hibernation deviation is determined based on the actual behavioral characteristics and the behavioral fingerprint database of the target node, wherein the behavioral fingerprint database includes the hibernation behavioral characteristics of the target node in a normal state during the standby period. When the hibernation deviation reaches a preset risk threshold, the target node is pre-processed to ensure that the target node successfully enters a hibernation state before an anomaly occurs.

2. The method according to claim 1, characterized in that, The actual behavioral characteristics have multiple feature types. The dormancy deviation is determined based on the actual behavioral characteristics and the behavioral fingerprint database of the target node, including: Identify behavioral deviations corresponding to actual behavioral characteristics based on actual behavioral characteristics and the corresponding dormant behavioral characteristics in the behavioral fingerprint database; The dormancy deviation degree is determined based on behavioral deviations of multiple actual behavioral characteristics.

3. The method according to claim 2, characterized in that, The actual behavioral characteristics also include the network management message characteristics and environmental association characteristics of the target node from the moment of power-off. The sleep deviation degree is determined based on the behavioral deviation of multiple actual behavioral characteristics, including: Determine the feature weights corresponding to the multiple actual behavioral features; The dormancy deviation is obtained by weighting the behavioral deviations of multiple actual behavioral features based on the feature weights corresponding to the multiple actual behavioral features.

4. The method according to claim 3, characterized in that, The method further includes: At least one historical behavior sequence data of the target node is obtained, wherein the historical behavior sequence data characterizes the actual behavior features of the target node during the process of successfully entering a hibernation state within a historical standby period; The current behavior sequence data is determined based on the actual behavior characteristics of the target node, wherein the current behavior sequence data characterizes the actual behavior characteristics of the target node during the process of successfully entering the hibernation state within the current standby cycle; Predict the anomaly probability of the target node based on the current behavior sequence data and at least one of the historical behavior sequence data; The preset risk threshold and / or the feature weights are adjusted based on the anomaly probability.

5. The method according to claim 1, characterized in that, When the dormancy deviation reaches a preset risk threshold, pre-processing is performed on the target node, including: When the dormancy deviation reaches the preset risk threshold, a diagnostic request is generated and sent to the target node, wherein the target node reads self-diagnostic data based on the diagnostic request and generates a self-diagnostic feedback signal based on the self-diagnostic data; If the self-diagnostic feedback signal is not received within a preset time, or if the self-diagnostic feedback signal is received within the preset time and it is determined from the self-diagnostic feedback signal that the target node has a non-sleep record and / or an abnormal wake-up source, the target node is controlled to reset and restart. If the self-diagnostic feedback signal is received within the preset time and it is determined from the self-diagnostic feedback signal that the target node does not have a non-sleep record or an abnormal wake-up source, the sleep deviation degree is continuously monitored, and when the sleep deviation degree reaches a preset reset threshold, the target node is controlled to reset and restart, wherein the preset reset threshold is greater than the preset risk threshold.

6. The method according to claim 5, characterized in that, After controlling the target node to reset and restart, the method further includes: Obtain the actual behavior characteristics of the target node after reset within the preset reset window; The reset dormancy deviation is determined based on the actual behavioral characteristics after reset and the behavioral fingerprint database. When the reset sleep deviation is greater than or equal to the preset risk threshold, the target node is controlled to be powered off for protection.

7. The method according to claim 6, characterized in that, The method further includes: When the reset hibernation deviation is less than the preset risk threshold, the hibernation behavior characteristics are updated according to the actual behavior characteristics after the reset.

8. A vehicle control device, characterized in that, include: The acquisition module is used to acquire the actual behavior characteristics of the target node in the vehicle during the current standby cycle after the vehicle is powered off. The actual behavior characteristics include at least the non-sleep duration of the target node from the moment of power-off. The determination module is used to determine the sleep deviation degree based on the actual behavioral characteristics and the behavioral fingerprint database of the target node, wherein the behavioral fingerprint database includes the sleep behavior characteristics of the target node in the standby period when it is in a normal state. The control module is used to pre-process the target node when the hibernation deviation reaches a preset risk threshold, so that the target node can successfully enter a hibernation state before the anomaly occurs.

9. An electronic device, characterized in that, Including processor and memory, among which Memory, used to store computer programs; A processor for executing a program stored in memory to implement the method described in any one of claims 1-7.

10. A vehicle, characterized in that, It includes the electronic device as described in claim 9.

Images