Vehicle sliding prevention method and system, electronic equipment and computer readable storage medium

Abstract

The invention relates to the technical field of vehicle safety. The anti-sliding method comprises the steps that current load data, vehicle state data, road condition data and environment data of a target vehicle are collected in real time; determining a load level to which the target vehicle belongs based on the current load data; calling a car sliding risk prediction model corresponding to the load level; the current load data, the vehicle state data, the road condition data and the environment data are input into a vehicle sliding risk prediction model, the vehicle sliding risk prediction model outputs a vehicle sliding risk judgment result, and the vehicle sliding risk judgment result comprises any one of no risk, low risk, medium risk and high risk; when the vehicle sliding risk judgment result is low risk, medium risk or high risk, a matched vehicle sliding prevention control strategy is selected from a preset control strategy set according to the load grade and the vehicle sliding risk judgment result; and executing a corresponding anti-slip response action according to the anti-slip control strategy.

Description

Technical Field

[0001] This application relates to the field of vehicle safety technology, and in particular to a method, system, electronic device and computer-readable storage medium for preventing a vehicle from rolling away. Background Technology

[0002] With the continuous evolution of automotive electronic control technology, modern vehicles are generally equipped with intelligent devices such as electronic parking brake systems, which aim to effectively solve the problem of vehicle rollover that may occur during parking. Traditional control strategies mostly adopt standardized schemes based on fixed parameters. That is, after the sensor is triggered, the system will automatically apply a preset braking pressure or output a constant driving torque to maintain the static balance of the vehicle.

[0003] Existing control methods are mainly optimized for standard load conditions at the engineering design level. However, in actual operation, under heavy load conditions, the vehicle has a greater tendency to slide downhill due to its large mass. If the braking force is insufficient, it is very easy to cause a rollover accident. Conversely, if the forced braking strategy of high load matching is still used under no-load or light-load conditions, it may lead to excessive braking, tire lock-up, or even cause longitudinal oscillation or loss of control of the vehicle. Summary of the Invention

[0004] In view of this, embodiments of this application provide a vehicle anti-rollover method, system, electronic device, and computer-readable storage medium, which can effectively solve the problem that existing electronic parking controls mostly adopt standardized control strategies with fixed parameters, achieving anti-rollover based on preset braking pressure or drive torque. Such methods are mainly designed for standard load conditions and do not consider the impact of dynamic changes in vehicle load during actual operation. Under heavy load, the fixed braking force may be insufficient to resist the downward force, leading to rollover; while under no-load or light-load conditions, excessive braking force can easily cause tire lock-up, longitudinal oscillation, or handling instability, resulting in insufficient safety and control adaptability.

[0005] In a first aspect, embodiments of this application provide a method for preventing a vehicle from rolling away, the method comprising: Real-time collection of target vehicle's current load data, vehicle status data, road condition data, and environmental data; The load class of the target vehicle is determined based on the current load data; Call the runaway risk prediction model corresponding to the load level; The current load data, vehicle status data, road condition data, and environmental data are input into the runaway risk prediction model, and the runaway risk prediction model outputs a runaway risk assessment result, wherein the runaway risk assessment result includes any one of no risk, low risk, medium risk, and high risk. When the risk assessment result of the runaway vehicle is low risk, medium risk or high risk, a matching anti-runaway vehicle control strategy is selected from the preset control strategy set according to the load level and the risk assessment result of the runaway vehicle. The corresponding anti-runaway response action is executed according to the aforementioned anti-runaway control strategy.

[0006] In some embodiments, the load class is based on the ratio of the actual load capacity of the target vehicle to the rated load capacity of the target vehicle, and includes: unloaded class, light load class, medium load class and heavy load class.

[0007] In some embodiments, the anti-runaway control strategy set includes differentiated anti-runaway control strategies corresponding to different load levels. Within any load level, as the risk level of the runaway risk assessment result increases, the control intensity of the anti-runaway response action included in the matched anti-runaway control strategy increases. The anti-rollover response action includes at least one of the following: warning prompt, power output adjustment, active application of service brakes, and activation of the parking lock mechanism.

[0008] In some embodiments, when the actual load of the target vehicle is detected to increase, causing the load level to jump to a higher level, and the risk of runaway is determined to be low, medium or high, the corresponding anti-runaway response action is executed according to the anti-runaway control strategy corresponding to the increased load level.

[0009] In some embodiments, it also includes: Collect historical data from multiple vehicles under different load conditions. The historical data includes: historical driving data, historical load data, historical road condition data, historical environmental data, and manually labeled historical runaway event data. The historical data is preprocessed to obtain preprocessed data; The preprocessed data is divided into multiple data subsets according to each load level; For each of the aforementioned data subsets, feature variables related to the car slippage are extracted to form a corresponding car slippage-related feature set; Based on the various slippage-related feature sets, each machine learning algorithm is trained to obtain a slippage risk prediction model corresponding to each load level.

[0010] In some embodiments, training each machine learning algorithm based on each of the respective car-traveling association feature sets includes: Cross-validation was used to optimize the model hyperparameters.

[0011] In some embodiments, after executing the corresponding anti-runaway response action according to the anti-runaway control strategy, the method further includes: Collect real-time operating data and control strategy execution results of the target vehicle to form feedback data; The feedback data is classified according to the load level and stored in the corresponding database as new training samples. The slippage risk prediction model corresponding to each load level is updated online using the newly added training samples through incremental training.

[0012] Secondly, embodiments of this application provide a vehicle anti-rollover system, comprising: The data acquisition module collects real-time data on the target vehicle's current load, vehicle status, road conditions, and environmental conditions. The determination module determines the load class of the target vehicle based on the current load data; The module invokes the runaway risk prediction model corresponding to the load level. The input module inputs the current load data, vehicle status data, road condition data, and environmental data into the runaway risk prediction model, and the runaway risk prediction model outputs a runaway risk assessment result, wherein the runaway risk assessment result includes any one of no risk, low risk, medium risk, and high risk. The selection module selects a matching anti-runaway control strategy from a preset control strategy set based on the load level and the runaway risk determination result when the runaway risk determination result is low risk, medium risk, or high risk. The execution module performs corresponding anti-runaway response actions according to the anti-runaway control strategy.

[0013] Thirdly, embodiments of this application provide an electronic device, the electronic device comprising: a processor and a memory, the memory storing a computer program, and the processor executing the computer program to implement the vehicle anti-rollover method described in the first aspect above.

[0014] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the vehicle anti-rollover method described in the first aspect.

[0015] The embodiments of this application have the following beneficial effects: The anti-rollover method of this application includes: real-time collection of the target vehicle's current load data, vehicle status data, road condition data, and environmental data; determining the target vehicle's load level based on the current load data; calling the rollover risk prediction model corresponding to the load level; inputting the current load data, vehicle status data, road condition data, and environmental data into the rollover risk prediction model, and having the rollover risk prediction model output a rollover risk judgment result, wherein the rollover risk judgment result includes any one of no risk, low risk, medium risk, and high risk; when the rollover risk judgment result is low risk, medium risk, or high risk, selecting a matching anti-rollover control strategy from a preset control strategy set according to the load level and the rollover risk judgment result; and executing the corresponding anti-rollover response action according to the anti-rollover control strategy. This application achieves accurate identification and classification of slippage risk under different load conditions by collecting multi-dimensional data such as the vehicle's current load, status, road conditions, and environment in real time, combined with load level classification and corresponding slippage risk prediction models. Based on the linkage between risk level and load level, it matches differentiated anti-slippage control strategies, thereby dynamically adjusting warning prompts, power output limits, or braking force levels in low, medium, and high risk situations. This avoids the problems of insufficient braking force leading to slippage under heavy loads or excessive braking causing tire lock-up and instability under light / empty loads, which are common with traditional fixed parameter control strategies. It significantly improves the safety and control adaptability of commercial vehicles in complex slope environments during remote driving scenarios. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A first flowchart of a vehicle anti-rollover method according to an embodiment of this application is shown; Figure 2 This paper illustrates a second flowchart of a vehicle anti-rollover method according to an embodiment of this application. Figure 3 A schematic diagram of the third process of the vehicle anti-rollover method according to an embodiment of this application is shown; Figure 4 A schematic diagram of a vehicle anti-rollover system according to an embodiment of this application is shown. Detailed Implementation

[0018] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0019] The components of the embodiments of this application described and illustrated in the accompanying drawings can be arranged and designed in a variety of different configurations. Therefore, the following detailed description of the embodiments of this application provided in the drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0020] In the following text, the terms "comprising," "having," and their cognates, which may be used in various embodiments of this application, are intended only to indicate a particular feature, number, step, operation, element, component, or combination thereof, and should not be construed as primarily excluding the presence of one or more other features, numbers, steps, operations, elements, components, or combinations thereof, or adding the possibility of one or more combinations thereof. Furthermore, the terms "first," "second," "third," etc., are used only for distinguishing descriptions and should not be construed as indicating or implying relative importance.

[0021] Unless otherwise specified, all terms used herein (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which the various embodiments of this application pertain. Terms (such as those defined in commonly used dictionaries) shall be interpreted as having the same meaning as in their contextual meaning in the relevant technical field and shall not be construed as having an idealized or overly formal meaning, unless clearly defined in the various embodiments of this application.

[0022] The following detailed description of some embodiments of this application is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0023] Existing electronic parking brake controls often employ standardized control strategies with fixed parameters, relying on preset braking pressure or drive torque to prevent runaway. These methods are primarily designed for standard load conditions and do not consider the impact of dynamic changes in vehicle load during actual operation. Under heavy loads, the fixed braking force may be insufficient to resist the downward force, leading to runaway; while under no-load or light-load conditions, excessive braking force can easily cause tire lock-up, longitudinal oscillation, or handling instability, resulting in insufficient safety and control adaptability. This application provides a vehicle anti-rollover method, system, electronic device, and computer-readable storage medium. By collecting multi-dimensional data such as the vehicle's current load, status, road conditions, and environment in real time, and combining the load level classification with the corresponding rollover risk prediction model, this application achieves accurate identification and classification of rollover risk under different load conditions. Based on the linkage between risk level and load level, it matches differentiated anti-rollover control strategies, thereby dynamically adjusting warning prompts, power output limits, or braking force levels in low, medium, and high risk situations. This avoids the problems of insufficient braking force leading to rollover under heavy loads or excessive braking causing tire lock-up and instability under light / empty loads, which are common with traditional fixed parameter control strategies. This significantly improves the safety and control adaptability of commercial vehicles starting and stopping in complex slope environments under remote driving scenarios.

[0024] The following describes the method for preventing the vehicle from rolling away, using some specific examples.

[0025] Figure 1 This illustration shows a flowchart of a vehicle anti-rollback method according to an embodiment of this application. It is understood that this method can be applied to any type of vehicle, such as commercial vehicles, engineering vehicles, or special transport vehicles. Exemplarily, this application is applied to anti-rollback control of commercial vehicles in remote-controlled driving scenarios when driving on slopes or stationary. In such scenarios, the driver is off-cab and controls the vehicle via a remote terminal, unable to directly perceive the slope angle, road surface adhesion, and power response delay, resulting in a high risk of rollback. Simultaneously, the load capacity of commercial vehicles varies greatly, potentially differing by tens of tons from empty to fully loaded. This load difference significantly affects the vehicle's power requirements, braking performance, and anti-rollback capability. Under heavy load, the vehicle has high inertia, requiring greater driving or braking force to maintain stability during slope starts or braking; while under empty load, the axle load is reduced, the tire-to-ground contact force decreases, and the vehicle is more sensitive to changes in the road surface friction coefficient. Even slight fluctuations in power output or brake release can lead to slippage or rollback.

[0026] To address the aforementioned problems, this application proposes a vehicle anti-rollover method based on big data analysis and intelligent decision-making. This method dynamically identifies the risk of rollover based on the vehicle's real-time operating conditions and executes a matching prevention and control strategy. This method can be integrated into the vehicle's braking control system, deployed in the vehicle controller, or function as an independent module working collaboratively with other systems via an onboard communication network. Example, such as... Figure 1 As shown, the vehicle anti-rollover method includes S101-S106: S101 collects real-time data on the target vehicle's current load, vehicle status, road conditions, and environment.

[0027] Specifically, during the remote-controlled operation of the target commercial vehicle, vehicle status data is acquired through onboard sensors, including parameters such as driving speed, acceleration, gear status, braking pressure, and wheel speed. Actual load information is obtained through pressure sensors installed at the connection between the chassis and axle, and weight sensors at the bottom of the cargo box. Simultaneously, a camera positioned above the cargo box captures images of the cargo, and image recognition technology is used to determine the load distribution, identifying any uneven loading or center of gravity shift. This load distribution data is then incorporated into the current load data for subsequent risk assessment. Road condition data can be obtained by estimating the gradient through the onboard inertial navigation system or by accessing a high-precision map platform via 4G / 5G communication to obtain information such as road gradient and curvature. Environmental data is obtained by connecting to a third-party meteorological service platform, including real-time wind speed, rainfall, temperature, and road surface slippage—factors affecting adhesion performance.

[0028] S102, determine the load class of the target vehicle based on the current load data.

[0029] After completing multi-source data collection, the system calculates the ratio of the actual load to the rated load based on the current load data to determine the load class of the target vehicle. The load class is divided based on the ratio of the target vehicle's actual load to its rated load, and includes: empty load class, light load class, medium load class, and heavy load class. These can be set according to actual application conditions. For example, the empty load class corresponds to an actual load of no more than 10% of the rated load, suitable for situations where no goods are loaded or only a small amount of onboard equipment is carried; the light load class corresponds to more than 10% but no more than 30% of the rated load, commonly seen in short-distance transportation or partially loaded scenarios; the medium load class corresponds to more than 30% but no more than 70% of the rated load, representing the majority of load conditions in regular operations; and the heavy load class corresponds to an actual load exceeding 70% of the rated load, typically occurring in fully loaded long-distance transportation or extreme load-bearing operations.

[0030] Understandably, this grading standard has good scalability, allowing for adjustments to the threshold range based on different vehicle models or usage scenarios. For example, a higher starting point for heavy-duty dump trucks in mining areas can be defined separately to adapt to diverse application needs. This grading method fully considers the nonlinear changes in vehicle dynamics under different load conditions. The system can accurately perceive the vehicle's current load status, providing reliable input for subsequent risk prediction and response actions based on grade-specific models. This effectively avoids the problems of over- or under-control caused by ignoring load differences in traditional uniform strategies, significantly improving the safety and intelligence level of remote-controlled driving.

[0031] S103, invoke the runaway risk prediction model corresponding to the load level.

[0032] After determining the load level of the target vehicle, the system loads the corresponding rollover risk prediction model from the cloud or the vehicle's local model library based on the obtained level result. These models are not general-purpose algorithms, but level-specific models specially trained based on a large amount of real vehicle operation data, ensuring that they can accurately capture the key influencing factors and their operating patterns that lead to rollover under different load conditions.

[0033] For example, in one implementation, such as Figure 2 As shown, S103 includes the following sub-steps: S201 collects historical data from multiple vehicles under different load conditions. The historical data includes: historical driving data, historical load data, historical road condition data, historical environmental data, and historical runaway event data that has been manually labeled.

[0034] The model's construction begins with the systematic collection and integration of historical data, achieved through the establishment of a commercial vehicle big data collection and analysis platform. This platform utilizes 4G and Bluetooth communication to establish a stable connection with each vehicle's T-BOX, receiving real-time historical driving data including speed, acceleration, gear status, power output parameters, and braking pressure. Simultaneously, it acquires historical load data such as actual load weight, load distribution, and load type through pressure sensors, weight sensors, and cargo box cameras deployed on the vehicle. Furthermore, the platform integrates with third-party high-precision map services to obtain historical road condition data such as road slope, smoothness, and curvature, and connects to a meteorological service platform to collect historical environmental data such as wind speed, rain / snow weather, and temperature. Notably, all past runaway events have been manually retrospectively labeled, forming a complete historical runaway event dataset containing the load status at the time of the incident, vehicle dynamics, road conditions, and environmental parameters, constituting a multi-dimensional, high-quality raw database.

[0035] S202, preprocess the historical data to obtain preprocessed data.

[0036] After the data aggregation is completed, the system enters the preprocessing stage. The system cleans and standardizes the collected historical data, removes outliers caused by signal interference, sensor failure or transmission errors, removes duplicate records and irrelevant fields, and retains time series segments that are strongly related to the risk of runaway vehicles, resulting in preprocessed data with a clear structure and high consistency.

[0037] S203 divides the preprocessed data into multiple data subsets according to each load level.

[0038] Based on the defined load class classification standards—namely, unloaded, lightly loaded, medium-loaded, and heavily loaded—the preprocessed data is divided into four independent subsets according to the actual load range. Each subset represents a set of typical working conditions within a specific load range, laying the foundation for subsequent hierarchical modeling.

[0039] S204. For each data subset, extract the feature variables related to the car slippage to form the corresponding car slippage-related feature set.

[0040] For each data subset, the system further extracts feature variables closely related to rollback behavior, such as the matching degree between the rate of increase of driving force and the timing of brake release during hill start, the slope of the change in the trend of small displacements when the vehicle is stationary, the impact of center of gravity shift on adhesion under different load distributions, and the degree of abrupt change in wheel speed difference on wet and slippery surfaces. These features are combined to form a rollback-related feature set specific to each level. These feature sets fully reflect the essential driving forces of rollback under various load conditions, avoiding the risk of misjudgment caused by using a single parameter such as slope or vehicle speed for judgment in traditional methods.

[0041] S205, based on each set of features associated with slippage, trains each machine learning algorithm to obtain a slippage risk prediction model corresponding to each load level.

[0042] The system employs machine learning algorithms to model the feature sets of each load level. These include gradient boosting decision tree algorithms and convolutional neural network algorithms. Gradient boosting decision tree algorithms excel at handling structured features and capturing nonlinear relationships, making them suitable for classifying risk scenarios with well-defined rules. Convolutional neural network algorithms, on the other hand, can effectively mine deep patterns in time series data, making them suitable for analyzing the evolution of potential risks in continuous dynamic processes.

[0043] For each load class, a dedicated risk prediction model is trained independently to ensure that the model output closely reflects the actual response characteristics of vehicles at that load class. Cross-validation is used during training to optimize model hyperparameters, preventing overfitting and improving generalization ability. Ultimately, this ensures that each model achieves a runaway event prediction accuracy of 93% or higher on an independent test set. After training, the models are deployed to edge computing units or cloud servers, supporting remote access and rapid inference. When the system enters real-time operation, simply selecting the corresponding model based on the current vehicle's load class enables high-precision, low-latency risk assessment.

[0044] S104: Input the current load data, vehicle status data, road condition data, and environmental data into the runaway risk prediction model, and the runaway risk prediction model outputs the runaway risk judgment result.

[0045] The risk assessment result for runaway vehicles includes any one of the following: no risk, low risk, medium risk, and high risk.

[0046] Based on risk patterns learned during prior training under various load conditions, the model classifies and judges whether there is a tendency for the vehicle to slip under the current operating condition, outputting a result of no risk, low risk, medium risk, or high risk. This process achieves intelligent mapping from raw perceived data to semantic risk levels, ensuring that the judgment does not rely on a single indicator but integrates the synergistic relationship of multiple factors. For example, when a commercial vehicle under medium load releases the brake pedal on a slippery slope, if the system detects that the drive wheel torque increases too quickly while the vehicle body has not moved, it may be judged as medium risk; if this is accompanied by a slight negative increase in the rear axle wheel speed, it is upgraded to high risk.

[0047] S105: When the risk assessment result of runaway is low risk, medium risk or high risk, a matching anti-runaway control strategy is selected from the preset control strategy set according to the load level and the runaway risk assessment result.

[0048] The system selects a matching anti-rollover control strategy from a preset set of control strategies based on the combination of the current load level and the rollover risk assessment result. The anti-rollover control strategy set adopts a hierarchical design principle, including differentiated anti-rollover control strategies corresponding to different load levels. Different response logic is configured for different load levels. Within any load level, as the risk level of the rollover risk assessment result increases, the control intensity of the anti-rollover response action included in the matched anti-rollover control strategy increases progressively. This gradient response mechanism not only avoids runaway of empty vehicles due to excessive braking, but also prevents protection failure of heavily loaded vehicles due to insufficient braking force, fully adapting to the real-world needs of commercial vehicles with a wide load range and nonlinear dynamic characteristics.

[0049] The anti-rollover response includes at least one of the following: warning notification, power output adjustment, active application of service brakes, and activation of the parking lock mechanism. The specific execution methods are flexibly combined based on the strategy settings. For example, under no-load conditions, a warning is triggered only to alert the operator at low risk; at medium risk, power output is automatically reduced by 20% to suppress drive wheel slippage; at high risk, first-level braking is activated, applying braking torque equivalent to 30% of the rated power. Under light-load conditions, a warning is also issued at low risk; power output is reduced by 30% at medium risk; and second-level braking is triggered at high risk, increasing braking force to 50% of the rated power. For medium-load conditions, a warning and a 10% reduction in power output are implemented as pre-emptive intervention measures at low risk; power is further reduced by 40% at medium risk; and third-level braking is activated at high risk, applying 70% of the rated power. Under heavy load conditions, due to the vehicle's large inertia and strong downward force, the control strategy is more stringent. In low-risk situations, an early warning is issued and the power is reduced by 20% in advance to prepare for emergencies; in medium-risk situations, the power output is reduced by 50%; and in high-risk situations, 100% of the rated power is applied immediately, and the parking lock mechanism is activated simultaneously to achieve double locking and prevent the vehicle from rolling away to the greatest extent possible.

[0050] S106, execute the corresponding anti-runaway response action according to the anti-runaway control strategy.

[0051] The system executes all anti-runaway response actions in the selected control strategy to complete closed-loop control.

[0052] Furthermore, this embodiment also introduces a dynamic load upgrade response mechanism. When the actual load of the target vehicle increases, causing the load level to jump to a higher level, and the rollback risk assessment result is low, medium, or high risk, the corresponding anti-rollback response action is executed according to the anti-rollback control strategy corresponding to the upgraded load level. Specifically, when the actual load of the target vehicle increases due to loading or other reasons while stationary on a slope, the system continuously monitors its load change trend. Once it is identified that the load exceeds the threshold, causing the load level to jump to a higher level, and the current rollback risk assessment result is low, medium, or high risk, the system immediately switches to the anti-rollback control strategy corresponding to the upgraded level and executes the corresponding response action according to the risk level of the new level.

[0053] For example, if a bus that was originally in a medium-load category begins loading cargo while parked on a slope, the system detects a significant increase in axle load and determines that it has entered a heavy-load category. Even if there is no obvious displacement at this time, due to the potential risk, the system immediately reduces power and strengthens braking preparation in advance according to the low-risk strategy of the heavy-load category. If the vehicle shows a slight tendency to move backward at this time, it quickly upgrades to a high-risk response, directly applies full braking and locks the parking mechanism, effectively preventing the problem of lag in response of traditional control systems caused by sudden changes in load.

[0054] For example, when a heavy-duty bus pauses on an icy slope and prepares to start, the system invokes a dedicated heavy-duty model. Combining this model with real-time slope, tire pressure, residual braking force, and power request curves, it proactively identifies the potential for rollback due to insufficient driving force and automatically issues a warning or intervenes directly. This process shifts from passive response to proactive prediction, significantly improving driving safety under complex conditions. Furthermore, this strategy system boasts excellent scalability, allowing for further refinement of risk level classifications or optimization of control parameters based on vehicle characteristics, regional climate, or operational habits. It also supports OTA remote updates and personalized configurations to meet the needs of diverse application scenarios.

[0055] This solution constructs a scientific, reasonable, safe and reliable vehicle anti-rollover control system through a technical chain of precise classification, intelligent judgment, gradient response and dynamic adaptation. It effectively solves the problems of existing technologies ignoring load differences and rigid control strategies, and is especially suitable for the safety protection of commercial vehicles in new operating modes such as unmanned driving and remote control.

[0056] To further improve the accuracy and long-term adaptability of the runaway vehicle risk prediction model, in one embodiment, based on the above embodiments, the deployed risk prediction model can be updated periodically or through event-triggered online updates. This update process is achieved through closed-loop feedback learning; for example, in one implementation, such as... Figure 3 As shown, specifically including S301-S303: S301 collects real-time operating data of the target vehicle and the results of control strategy execution to form feedback data.

[0057] Real-time operational data includes dynamic records of the vehicle's load status, slope information, environmental conditions, power and braking parameters before and after the anti-rollover response is executed. The control strategy execution results include evaluation information such as whether the intervention successfully prevented the vehicle from rolling back, whether the response timing was timely, and whether the selected control intensity was suitable for the current operating conditions. For example, if the vehicle still moves slightly backward after a high-risk braking action, it is judged as a partial failure; if the vehicle is completely stable and has no displacement, it is recorded as a successful case.

[0058] S302, the feedback data is classified according to load level and stored in the corresponding database as new training samples.

[0059] The system categorizes and aggregates feedback data according to the current vehicle's load class, ensuring that new samples accurately match the training domain of the corresponding model. For example, a successful braking event occurring under heavy load conditions will have all its data categorized as Heavy Load-High Risk-Effective Response and stored in a dedicated historical database for heavy load classes, serving as new training samples for subsequent model optimization. This classification mechanism ensures that experience accumulated under different load conditions is used for independent modeling, preventing light load data from interfering with heavy load models and maintaining the professionalism and specificity of models at each level. All feedback data is uploaded in real-time to the cloud-based big data platform via 4G communication through the in-vehicle T-BOX, or it can be temporarily stored on a local edge server and then synchronized in batches, ensuring the integrity and reliability of data transmission.

[0060] S303 uses incremental training to update the runaway risk prediction model for each load level online using newly added training samples.

[0061] Incremental training is a highly efficient machine learning optimization method. Its key feature is that it gradually integrates new samples into the existing model without discarding existing knowledge, eliminating the need for retraining from scratch and significantly reducing computational resource consumption and update latency. This process can be executed centrally in the cloud or locally on an onboard controller with sufficient computing power. The updated model can adapt more quickly to new road conditions, climate changes, or evolving driving habits that bring new risk patterns. For example, in a region where continuous rainfall causes multiple instances of empty vehicles skidding when starting on a gentle slope, the original model may not have fully covered such slippery, low-adhesion scenarios. However, as multiple similar events are labeled and reported, the empty vehicle classification model continuously strengthens its ability to identify such conditions through incremental training. Ultimately, under subsequent similar conditions, it can issue early warnings and activate corresponding control strategies, significantly improving the success rate of protection.

[0062] A complete closed-loop logic chain of perception, decision-making, execution, feedback, and optimization has been constructed, enabling the entire anti-runaway system to have self-evolution capabilities. Compared with the limitations of traditional static models that remain fixed once deployed, this application achieves continuous improvement in model performance, making it particularly suitable for the diverse and long-term application needs of large-scale commercial vehicle fleet operations.

[0063] Furthermore, this mechanism supports flexible configuration of the update frequency. It can be set as a scheduled task, such as daily batch updates at midnight, or as an event-driven mode, triggering the training process immediately when a certain level accumulates a certain number of valid feedback samples, further improving response efficiency. In this way, the system not only provides high-precision risk assessment in the initial stage but also continuously absorbs real-world experience during long-term use, enhancing its generalization ability to complex edge scenarios. This comprehensively improves the safety, intelligence, and reliability of commercial vehicles in remote-controlled environments when driving on slopes or stationary.

[0064] like Figure 4 As shown, based on the method of the above embodiments, this embodiment provides a vehicle anti-rollover system. Exemplarily, the vehicle anti-rollover system 100 includes: The data acquisition module 110 collects real-time data on the target vehicle's current load, vehicle status, road conditions, and environmental conditions. The determination module 120 determines the load class of the target vehicle based on the current load data; Module 130 is invoked to call the runaway risk prediction model corresponding to the load level; Input module 140 inputs current load data, vehicle status data, road condition data and environmental data into the runaway risk prediction model, and the runaway risk prediction model outputs the runaway risk judgment result, which includes any one of no risk, low risk, medium risk and high risk. Select module 150. When the risk assessment result of runaway is low risk, medium risk or high risk, select a matching anti-runaway control strategy from the preset control strategy set according to the load level and the runaway risk assessment result. The execution module 160 executes the corresponding anti-runaway response actions according to the anti-runaway control strategy.

[0065] It is understood that the system in this embodiment corresponds to the method in the above embodiments, and the options in the above embodiments are also applicable to this embodiment, so they will not be described again here.

[0066] This application also provides an electronic device, exemplary of which includes a processor and a memory, wherein the memory stores a computer program, and the processor executes the computer program to cause the device to perform the functions of the various modules in the above-described vehicle anti-rollover method or vehicle anti-rollover system. It is understood that the electronic device can be a vehicle or other electronic devices, and no limitation is made herein.

[0067] The processor can be an integrated circuit chip with signal processing capabilities. The processor can be a general-purpose processor, including at least one of a Central Processing Unit (CPU), Graphics Processing Unit (GPU), Network Processor (NP), Digital Signal Processor (DSP), Application-Specific Integrated Circuit (ASIC), Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The general-purpose processor can be a microprocessor or any conventional processor, capable of implementing or executing the methods, steps, and logic block diagrams disclosed in the embodiments of this application.

[0068] Memory can be, but is not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), and Electrically Erasable Programmable Read-Only Memory (EEPROM). Memory is used to store computer programs, and the processor can execute these programs upon receiving execution instructions.

[0069] This application also provides a computer-readable storage medium for storing computer programs used in the aforementioned terminal devices. For example, the computer-readable storage medium may include, but is not limited to, various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0070] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that, in alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and combinations of blocks in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0071] In addition, the functional modules or units in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0072] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a smartphone, personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application.

[0073] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for preventing a vehicle from rolling away, characterized in that, The method includes: Real-time collection of target vehicle's current load data, vehicle status data, road condition data, and environmental data; The load class of the target vehicle is determined based on the current load data; Call the runaway risk prediction model corresponding to the load level; The current load data, vehicle status data, road condition data, and environmental data are input into the runaway risk prediction model, and the runaway risk prediction model outputs a runaway risk assessment result, wherein the runaway risk assessment result includes any one of no risk, low risk, medium risk, and high risk. When the risk assessment result of the runaway vehicle is low risk, medium risk or high risk, a matching anti-runaway vehicle control strategy is selected from the preset control strategy set according to the load level and the risk assessment result of the runaway vehicle. The corresponding anti-runaway response action is executed according to the aforementioned anti-runaway control strategy.

2. The vehicle anti-rollover method according to claim 1, characterized in that, The load rating is based on the ratio of the actual load capacity of the target vehicle to the rated load capacity of the target vehicle, and includes: unloaded rating, light load rating, medium load rating and heavy load rating.

3. The vehicle anti-rollover method according to claim 1, characterized in that, The anti-runaway control strategy set includes differentiated anti-runaway control strategies corresponding to different load levels. Within any load level, as the risk level of the runaway risk assessment result increases, the control intensity of the anti-runaway response action included in the matched anti-runaway control strategy increases. The anti-rollover response action includes at least one of the following: warning prompt, power output adjustment, active application of service brakes, and activation of the parking lock mechanism.

4. The vehicle anti-rollover method according to claim 1, characterized in that, Also includes: When the actual load of the target vehicle is detected to increase, causing the load level to jump to a higher level, and the risk of runaway is determined to be low, medium or high, the corresponding anti-runaway response action is executed according to the anti-runaway control strategy corresponding to the increased load level.

5. The vehicle anti-rollover method according to claim 1, characterized in that, Also includes: Collect historical data from multiple vehicles under different load conditions. The historical data includes: historical driving data, historical load data, historical road condition data, historical environmental data, and manually labeled historical runaway event data. The historical data is preprocessed to obtain preprocessed data; The preprocessed data is divided into multiple data subsets according to each load level; For each of the aforementioned data subsets, feature variables related to the car slippage are extracted to form a corresponding car slippage-related feature set; Based on the various slippage-related feature sets, each machine learning algorithm is trained to obtain a slippage risk prediction model corresponding to each load level.

6. The vehicle anti-rollover method according to claim 5, characterized in that, The step of training each machine learning algorithm based on each of the aforementioned car-related feature sets includes: Cross-validation was used to optimize the model hyperparameters.

7. The vehicle anti-rollover method according to claim 1, characterized in that, After executing the corresponding anti-runaway response action according to the anti-runaway control strategy, the method further includes: Collect real-time operating data and control strategy execution results of the target vehicle to form feedback data; The feedback data is classified according to the load level and stored in the corresponding database as new training samples. The slippage risk prediction model corresponding to each load level is updated online using the newly added training samples through incremental training.

8. A vehicle anti-rollover system, characterized in that, include: The data acquisition module collects real-time data on the target vehicle's current load, vehicle status, road conditions, and environmental conditions. The determination module determines the load class of the target vehicle based on the current load data; The module invokes the runaway risk prediction model corresponding to the load level. The input module inputs the current load data, vehicle status data, road condition data, and environmental data into the runaway risk prediction model, and the runaway risk prediction model outputs a runaway risk assessment result, wherein the runaway risk assessment result includes any one of no risk, low risk, medium risk, and high risk. The selection module selects a matching anti-runaway control strategy from a preset control strategy set based on the load level and the runaway risk determination result when the runaway risk determination result is low risk, medium risk, or high risk. The execution module performs corresponding anti-runaway response actions according to the anti-runaway control strategy.

9. An electronic device, characterized in that, The electronic device includes a processor and a memory, the memory storing a computer program, and the processor executing the computer program to implement the vehicle anti-rollover method according to any one of claims 1-7.

10. A computer-readable storage medium storing a computer program thereon, characterized in that, When the computer program is run by the processor, it performs the steps of the vehicle anti-runaway method according to any one of claims 1-7.

Images