Vehicle control method, electronic device, and vehicle

Abstract

The present disclosure relates to the technical field of vehicle intelligent driving, and provides a vehicle control method, an electronic device and a vehicle. The vehicle control method comprises the following steps: determining that a vehicle is in an off-road mode; acquiring vehicle operation data; determining a state mutation confidence according to the vehicle operation data; acquiring environment data of an environment in which the vehicle is located; determining a target road phase type according to the environment data and the state mutation confidence; determining a target state label and a first confidence corresponding to the target state label according to the state mutation confidence and the target road phase type; and controlling the vehicle to exit the off-road mode in response to the target state label and the first confidence satisfying a preset mode switching condition. The present disclosure guarantees the accuracy, smoothness and controllability in the automatic exiting process of the off-road mode, improves the controllability and stability of the vehicle, and improves the driving experience of a user.

Description

Technical Field

[0001] This disclosure relates to the field of intelligent driving technology for vehicles, and in particular to a vehicle control method, electronic device, and vehicle. Background Technology

[0002] With the rapid development of vehicle technology, vehicles have become an important means of transportation in people's daily lives. To facilitate driving, vehicles are equipped with an off-road mode to assist users in better controlling the vehicle in off-road scenarios.

[0003] Currently, both entering and exiting off-road mode require manual switching by the driver. When the vehicle leaves off-road terrain and enters ordinary roads, the system continues to operate in off-road mode, leading to problems such as sluggish braking response, which affects driving comfort and handling stability. Summary of the Invention

[0004] In view of this, the purpose of this disclosure is to provide a vehicle control method, electronic device and vehicle to solve the problem that the system continues to operate in off-road mode after the current vehicle leaves off-road terrain and enters ordinary road, resulting in sluggish braking response and affecting driving comfort and handling stability.

[0005] To achieve the above objectives, a first aspect of this disclosure provides a vehicle control method, the method comprising:

[0006] Determine that the vehicle is in off-road mode, acquire vehicle operation data, and determine the confidence level of state change based on the vehicle operation data; Acquire environmental data of the vehicle's surroundings, and determine the target road phase type based on the environmental data and the state change confidence level; Based on the state change confidence level and the target road stage type, determine the target state label and the first confidence level corresponding to the target state label; In response to the target status label and the first confidence level satisfying the preset mode switching conditions, the vehicle is controlled to exit the off-road mode.

[0007] Based on the same inventive concept, a second aspect of this disclosure provides a vehicle control device, comprising: The data acquisition module is configured to determine that the vehicle is in off-road mode, acquire vehicle operation data, and determine the confidence level of state change based on the vehicle operation data. The road type determination module is configured to acquire environmental data of the vehicle's environment and determine the target road stage type based on the environmental data and the state change confidence level. The status label determination module is configured to determine a target status label and a first confidence level corresponding to the target status label based on the status change confidence level and the target road stage type. The vehicle control module is configured to control the vehicle to exit off-road mode in response to the target status label and the first confidence level satisfying a preset mode switching condition.

[0008] Based on the same inventive concept, a third aspect of this disclosure proposes an electronic device including a memory, a processor, and a computer program stored in the memory and executable by the processor, wherein the processor implements the vehicle control method as described above when executing the computer program.

[0009] Based on the same inventive concept, a fourth aspect of this disclosure provides a non-transitory computer-readable storage medium that stores computer instructions for causing a computer to perform the vehicle control method as described above.

[0010] Based on the same inventive concept, the fifth aspect of this disclosure provides a vehicle including the vehicle control device described in the second aspect, the electronic device described in the third aspect, or the storage medium described in the fourth aspect.

[0011] As can be seen from the above, this disclosure proposes a vehicle control method, electronic device, and vehicle. The method determines that the vehicle is in off-road mode, acquires vehicle operating data, and determines a state change confidence level based on the vehicle operating data. The state change confidence level represents the probability of a mode change, i.e., the probability of switching from off-road mode to another mode. Environmental data of the vehicle's environment is acquired, and a target road stage type is determined based on the environmental data and the state change confidence level. The target road stage type is the current stage type of the vehicle, i.e., whether it is an off-road road, another road, or in a switching state. The determination of the target road stage type comprehensively considers environmental factors and the state change probability determined based on vehicle operating information, making the determination of the target road stage type more accurate. Based on the state change confidence level and the target road stage type, a target state label and a first confidence level corresponding to the target state label are determined. The target state label indicates whether the vehicle should exit off-road mode; that is, a judgment result on whether to exit off-road mode is obtained based on the state change confidence level and the target road stage type. If the target status label and the first confidence level meet the preset mode switching conditions, it indicates that the vehicle has left the off-road section and should be controlled to exit the off-road mode. With the confidence level as the core, the accuracy, smoothness and controllability of the automatic exit process of the off-road mode are ensured, thereby improving the vehicle's handling and stability and enhancing the user's driving experience. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in this disclosure or related technologies, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a flowchart of a vehicle control method according to an embodiment of the present disclosure; Figure 2 This is a structural block diagram of a vehicle control device according to an embodiment of the present disclosure; Figure 3 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present disclosure. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of this disclosure clearer, the following detailed description is provided in conjunction with specific embodiments and the accompanying drawings.

[0015] It should be noted that, unless otherwise defined, the technical or scientific terms used in the embodiments of this disclosure should have the ordinary meaning understood by one of ordinary skill in the art to which this disclosure pertains. The terms "first," "second," and similar terms used in the embodiments of this disclosure do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0016] The following are definitions of terms used in this disclosure: EventNet-Lite: EventNet-Lite is a lightweight neural network architecture designed for resource-constrained scenarios. Its core idea is to introduce an event-driven sparse computing mechanism, which activates neurons for inference only in regions of images or videos where key dynamic changes occur (such as object movement or sudden changes in lighting). This significantly reduces computational overhead and power consumption while maintaining high recognition accuracy, making it particularly suitable for deployment in mobile terminals, embedded devices, and real-time response systems.

[0017] TDE-GNN: Temporal Dynamic Graph Neural Network (TDE-GNN) is a novel deep learning model designed specifically for dynamic graphs. Its core innovation lies in treating the edges in the graph as dynamic entities that evolve over time, rather than just fixed connections between nodes. This is achieved by functionally encoding the timestamp information of nodes and edges and integrating it into the message passing process of the graph neural network.

[0018] RoadPhaseSwitch-Net: RoadPhaseSwitch-Net is a structured recognition model designed specifically for road phase transition recognition tasks. It achieves automated identification and intelligent analysis of road lifecycle changes by deeply mining the dynamic evolution patterns of road elements over time.

[0019] GPS: GPS, or Global Positioning System, uses at least 24 satellites in orbit to continuously broadcast ranging signals and time information to Earth. By receiving these signals and calculating the distance to multiple satellites, user-end receiving devices can determine their own three-dimensional position, speed, and time in real time with high accuracy.

[0020] With the rapid development of vehicle technology, vehicles have become an important means of transportation in people's daily lives. To facilitate driving, vehicles are equipped with off-road modes to assist users in better controlling the vehicle in off-road scenarios. Among related technologies, off-road driving assistance systems have significant limitations in mode switching, and their operating logic is mainly based on the driver's active judgment and manual intervention. When the vehicle is traveling on unpaved roads, such as mud, sand, or steep slopes, the driver will manually select the corresponding off-road mode according to the road conditions. The vehicle's electronic control system will then adjust parameters such as engine response, transmission shift logic, drive force distribution, suspension stiffness, and the timing of electronic stability program intervention to adapt to the challenges of harsh terrain.

[0021] However, the activation and deactivation of this mode rely entirely on human subjective decision-making. Once the vehicle completes the off-road section and re-enters a paved road, if the driver, due to negligence or a lack of awareness of mode switching, fails to manually revert to the standard driving mode in time, the vehicle will continue to drive on the road with the parameters set in the off-road mode. This parameter mismatch directly leads to a series of negative consequences. The delayed response of the braking system, designed to prevent lock-up on soft surfaces, will increase braking distance on the road. The reduced steering wheel effort, designed to reduce kickback on bumpy roads, will feel excessively playy and vague on the road. The raised body and softened damping of the air suspension, designed for off-road capability, will cause increased body sway and decreased handling on the road, severely impacting driving comfort during high-speed cruising and handling stability during emergency avoidance.

[0022] Meanwhile, most off-road assist systems rely on a single sensor source to assist in selecting driving modes. For example, some systems use a front-facing camera for image recognition to determine if there are tire tracks, puddles, or sand textures ahead. Other systems rely on an inertial measurement unit (IMU) to monitor the vehicle's pitch and roll angles, activating or maintaining off-road mode based on detected large tilt angles. This single-sensor approach works effectively in typical off-road obstacles, but its reliability is significantly reduced in blurred boundary scenarios such as the transition zone between off-road and on-road terrain.

[0023] For example, when a vehicle is driving on a hardened road surface covered with a thin layer of sand, the camera might determine that off-road mode is still required based on the sand's texture, while inertial data might indicate that the vehicle has returned to a level position. If the system relies solely on one set of data, it will cause a recognition conflict or error. This kind of scene misjudgment caused by a single sensing source and low fusion makes the system often react too late or too late when faced with gradual changes in road conditions, failing to accurately capture the critical points of fundamental changes in the driving environment.

[0024] To address the aforementioned problems, this embodiment proposes a vehicle control method, such as... Figure 1 As shown, the method includes: Step 101: Determine that the vehicle is in off-road mode, acquire vehicle operation data, and determine the confidence level of state change based on the vehicle operation data.

[0025] In specific implementation, if the vehicle is currently in off-road mode, i.e. off-road mode is activated, vehicle operation data is acquired. The vehicle operation data includes at least one of the following: braking pressure, steering angular velocity, accelerator pedal opening, suspension extension and retraction status, longitudinal acceleration, yaw rate, four-wheel vertical load, wheel speed difference, engine load rate, GPS altitude change rate, terrain tilt angle, etc.

[0026] In this embodiment, the vehicle's off-road mode is a special driving mode set by comprehensively adjusting power output, transmission shift logic, four-wheel drive system, differential lock, and electronic stability system (ESP) through the electronic system. After activating the off-road mode, the vehicle will prioritize providing a more linear throttle response and more sustained low-gear high torque output, while limiting premature intervention of the electronic stability system, in order to enhance traction and passability on unpaved roads such as sand, mud, rocks, or steep slopes, helping the driver to cope with complex terrain more easily.

[0027] In this embodiment, a data acquisition channel with the CAN bus as its core is constructed, focusing on the multi-source fusion of key vehicle dynamic response signals. The data source is the high-frequency signal corresponding to the vehicle operation data, and the sampling frequency is maintained between 50Hz and 100Hz to ensure the capture of subtle behavioral changes during the transition from off-road to highway.

[0028] After acquiring vehicle operation data, the confidence level of state change is determined based on the vehicle operation data. The confidence level of state change represents the probability that the vehicle has a mode change, that is, the probability of switching from off-road mode to other modes.

[0029] In this embodiment, the state change confidence score is used to quantify whether there is a high probability of an off-road end feature. That is, if the state change confidence score is greater than a preset change confidence score threshold, it indicates that there is a high probability of an off-road end feature, meaning that the vehicle state has changed abruptly. If the state change confidence score is less than or equal to the preset change confidence score threshold, it indicates that there is no high probability of an off-road end feature, meaning that the vehicle is still in off-road mode or is switching between off-road mode and other modes.

[0030] Step 102: Obtain environmental data of the vehicle's environment, and determine the target road stage type based on the environmental data and the state change confidence level.

[0031] In specific implementation, environmental data of the vehicle's environment is acquired. The environmental data is used to describe the environmental state of the vehicle's environment. Specifically, the environmental data includes at least one of the following: road surface roughness estimation, road surface slope, road surface adhesion coefficient, GPS elevation difference rate, and roadside line edge recognition results within the camera frame.

[0032] The target road stage type is determined based on the environmental data and the confidence level of the state change. The target road stage type refers to the current stage of the vehicle, i.e., whether it is an off-road road, another road, or in a transition state.

[0033] In this embodiment, the target road phase type includes a first road type, a second road type, and a road switching type. Specifically, the first road type is a road type other than off-road roads, such as a highway. The second road type is an off-road road type, and the road switching type indicates that the vehicle is switching from an off-road road to another road type.

[0034] Step 103: Determine the target state label and the first confidence level corresponding to the target state label based on the state change confidence level and the target road stage type.

[0035] In specific implementation, a target state label and a first confidence level corresponding to the target state label are determined based on the state change confidence level and the target road stage type. The target state label indicates whether the vehicle should exit off-road mode, that is, the judgment result of whether to exit off-road mode is obtained based on the state change confidence level and the target road stage type.

[0036] In this embodiment, the first confidence level corresponding to the target state label represents the credibility of the target state label determined based on the confidence level of state change and the target road stage type.

[0037] Step 104: In response to the target status label and the first confidence level satisfying the preset mode switching conditions, control the vehicle to exit the off-road mode.

[0038] In practice, the system determines whether the preset mode switching conditions are met based on the target status label and its corresponding first confidence level. If the preset mode switching conditions are not met, it means that the vehicle is still driving in an off-road environment. In this case, the vehicle should remain in off-road mode and should not be switched to other modes.

[0039] If the preset mode switching conditions are met, it indicates that the vehicle has left the off-road scenario, and the vehicle is then controlled to exit off-road mode. If the vehicle is determined to be in a highway environment, the vehicle domain controller sends a command to the power domain, braking domain, steering domain, and other subsystems to restore the corresponding highway mode control logic, such as linear steering response curves, anti-lock braking logic for normal roads, and torque mapping strategy for the accelerator pedal.

[0040] In this embodiment, the Highway mode (also known as Standard mode, Comfort mode, or Normal mode) is the vehicle's default and most commonly used driving setting, designed to balance power output, fuel economy, and ride smoothness on paved roads (such as city roads and highways). In Highway mode, the vehicle's throttle response, transmission shift logic, suspension stiffness, and steering assist are all tuned to a neutral state, which can meet the acceleration needs of daily commuting, filter road bumps to ensure ride comfort, and simultaneously achieve optimal energy consumption.

[0041] In this embodiment, after the vehicle exits off-road mode, the exit event will generate a log in the vehicle recording system, marking the exit time, trigger source, snapshots of each model output, and other information, and storing it in the vehicle database for subsequent analysis.

[0042] The above scheme determines that the vehicle is in off-road mode, acquires vehicle operation data, and determines a state change confidence level based on the vehicle operation data. This state change confidence level represents the probability of a mode change, i.e., the probability of switching from off-road mode to another mode. Environmental data of the vehicle's environment is acquired, and the target road stage type is determined based on the environmental data and the state change confidence level. The target road stage type is the current stage of the vehicle, i.e., whether it is an off-road road, another road, or in a switching state. The determination of the target road stage type comprehensively considers environmental factors and the state change probability determined based on vehicle operation information, making the determination of the target road stage type more accurate. Based on the state change confidence level and the target road stage type, a target state label and its corresponding first confidence level are determined. The target state label indicates whether the vehicle should exit off-road mode; that is, the judgment result of whether to exit off-road mode is obtained based on the state change confidence level and the target road stage type. If the target status label and the first confidence level meet the preset mode switching conditions, it indicates that the vehicle has left the off-road section and should be controlled to exit the off-road mode. With the confidence level as the core, the accuracy, smoothness and controllability of the automatic exit process of the off-road mode are ensured, thereby improving the vehicle's handling and stability and enhancing the user's driving experience.

[0043] In some embodiments, a stage embedding vector can be determined first based on the state change confidence level and stage annotation data, and then the road stage type can be determined based on the stage embedding vector and environmental data using a pre-trained stage type recognition model. Specifically, step 102, which involves obtaining environmental data of the vehicle's environment and determining the target road stage type based on the environmental data and the state change confidence level, includes: Step 1021: Obtain the stage annotation data corresponding to the historical operation data, and determine the stage embedding vector based on the state change confidence and the stage annotation data; Step 1022: Determine the behavioral feature vector based on the vehicle operation data and the confidence level of the state change; Step 1023: Input the stage embedding vector, the behavior feature vector and the environmental data into the pre-trained stage type recognition model, and process them through the stage type recognition model to obtain the target road stage type.

[0044] In practice, historical operational data is acquired, and the corresponding stage labeling data is determined. A stage embedding vector is then determined based on the state change confidence level and the stage labeling data. The historical operational data refers to operational data within a preset time period prior to the current moment, and the stage labeling data consists of pre-annotated tags indicating whether the vehicle is in off-road mode within the preset time period prior to the current moment.

[0045] In this embodiment, to ensure the accuracy of the stage annotation data, the vehicle switching signal and manual scene verification are used to identify whether the vehicle has activated the off-road mode, thereby obtaining the stage annotation data corresponding to the historical operation data.

[0046] Based on the vehicle operation data and the confidence level of the state change, a behavioral feature vector is determined. This behavioral feature vector describes the high-order correlations between vehicle operation data. Specifically, the method for determining the behavioral feature vector includes: A graph structure is constructed based on the vehicle operation data and the confidence level of state abrupt changes. Specifically, vehicle operation data for each moment within a preset time period before the current moment is abstracted as graph nodes, where each graph node is actually a multi-dimensional feature vector. A graph structure is constructed chronologically based on multiple graph nodes, connecting the originally isolated time window nodes through edges to form a topological network. Following the logic of chronological order, the edges between nodes represent the interdependence and temporal continuity between events. The edge weights are jointly calculated by the degree of abrupt change, signal coupling strength, and time interval, forming a graph topological network with behavioral driving significance.

[0047] After obtaining the graph structure, the nodes within the graph structure are aggregated to obtain a behavioral feature vector. Specifically, for the graph node corresponding to the current time step, a multi-head graph attention mechanism is used to obtain the importance of different neighboring nodes in influencing the behavior of this node, i.e., the weight value. The features of the neighboring nodes are weighted and summarized, and then fused with the features of the current graph node. After multiple aggregations, the graph node corresponding to the current time step obtains a new vector representation containing rich contextual information, i.e., the behavioral feature vector.

[0048] A pre-trained stage type recognition model is obtained. The stage embedding vector, the behavior feature vector, and the environmental data are input into the stage type recognition model. The target road stage type is obtained by processing the stage type recognition model.

[0049] Specifically, the training method of the stage type recognition model includes: Step a, obtain the first training dataset and the initial stage type recognition model, wherein the first training dataset includes historical stage embedding vectors, historical behavior feature vectors, historical environmental data and historical road stage types.

[0050] Step b: Input the training data in the first training dataset into the initial stage type recognition model for training, determine that the first preset training termination condition is met, and obtain the stage type recognition model.

[0051] In specific implementation, a first training dataset and an initial stage type recognition model are obtained, wherein the first training dataset includes historical stage embedding vectors, historical behavioral feature vectors, historical environmental data, and historical road stage types. The training data from the first training dataset is input into the initial stage type recognition model for training. Once a first preset training termination condition is met, the stage type recognition model is obtained.

[0052] The first preset training termination condition includes at least one of the following: determining that all data in the first training dataset has been input into the initial stage type recognition model for training, determining that the loss function of the initial stage type recognition model has converged to a first convergence threshold, or determining that the initial stage type recognition model has been iteratively trained to a first preset number of iterations.

[0053] For example, the first preset training termination condition is to determine that all data in the first training dataset has been input into the initial stage type recognition model for training: The first training dataset contains fifty sets of data, each set including historical stage embedding vectors, historical behavioral feature vectors, historical environmental data, and historical road stage types. The first preset training termination condition is determined when all data in the first training dataset has been input into the initial stage type recognition model for training. That is, when all fifty sets of data have been input into the initial stage type recognition model, there is no training data in the first training dataset that has not yet been input into the initial stage type recognition model. At this point, the initial stage type recognition model training is considered complete, and the stage type recognition model is obtained.

[0054] In another example, the first preset training termination condition is to determine that the loss function of the initial stage type recognition model converges to a first convergence threshold: The training data from the first training dataset is input into the initial stage type recognition model for training, and the training results are output. A loss function is determined based on the training results and the historical road stage types. The loss function may include at least one of the following: mean squared error loss function, cross-entropy loss function, logarithmic loss function, exponential loss function, squared loss function, or absolute value loss function, etc. When the loss function converges to a first convergence threshold, a first preset training termination condition is satisfied, and the stage type recognition model is obtained.

[0055] In another example, the first preset training termination condition is to determine the initial stage type recognition model iteratively train to the first preset number of iterations.

[0056] The training data in the first training dataset is input into the initial stage type recognition model for iterative training. The number of iterations is recorded. When the number of iterations is equal to the first preset number of iterations, the first preset training termination condition is met, and the stage type recognition model is obtained.

[0057] In this embodiment, the stage type recognition model is a neural network structure model. Preferably, the stage type recognition model is the RoadPhaseSwitch-Net model. RoadPhaseSwitch-Net is a structured recognition model specifically designed for road stage transition recognition tasks. Its purpose is to further compress and classify the stage embedding vectors and temporal behavior graph features generated by the preceding modules to form a high-confidence discrimination result for off-road road switching.

[0058] Specifically, the stage embedding vector, behavioral feature vector, and environmental data are preprocessed and concatenated to obtain input features. These input features are then fed into the backbone of the stage type recognition model. The backbone employs a stage-guided deep attention classifier, which specifically includes a multi-scale stage group encoder, a similarity-guided classifier, and a confidence learning submodule.

[0059] The multi-scale stage group encoder divides input features into three dimensions: event mutation frequency, graph node density, and vehicle operating state. It extracts the stage feature differences under different dimensions and captures the intersection of off-road state decay and on-road behavior manifestation. The similarity-guided classifier uses a cosine matching mechanism based on the reference stage center vector to calculate the similarity score between the current state and the off-road center and the on-road center, and determines the transition interval of the current window through dynamic boundary reasoning. The confidence learning submodule takes the classification stability of continuous windows as input, establishes a confidence evaluation curve, quantifies the consistency of the model's output on the sliding window sequence, and eliminates short-term misclassification results caused by environmental fluctuations.

[0060] In this embodiment, during the initialization of the reference stage center vector, high-purity samples are extracted from manually confirmed stable off-road and stable highway segments. Only samples with consistent multi-window performance and stable mutation intensity are retained. Robust prototype extraction is used to generate the off-road center and highway center. At the same time, physical consistency verification constraints are used, such as perturbation energy and graph edge weight density should correspond to high / low levels respectively, to avoid semantic offset, thereby obtaining the off-road medium-sized and highway centers corresponding to the reference stage center vector.

[0061] Simultaneously, the reference stage center vector can be updated. Specifically, momentum-based sliding updates are used, and updates are only included when the model makes consecutive high-confidence decisions across multiple windows and is far from the switching boundary. The model slowly approximates the new sample prototype with a small learning rate and sets a minimum center spacing constraint to prevent the two-stage representation from collapsing, thus achieving long-term adaptive and stable evolution.

[0062] The above scheme introduces a dual mechanism of road state semantic understanding and behavior phase continuity analysis in its structure. This enables it to stably identify phase transition trends even when road switching boundaries are ambiguous and driving behavior has not converged significantly, ensuring that the entire identification system accurately captures and avoids misjudgments of off-road exit timing.

[0063] In some embodiments, the step 1021 of determining the stage embedding vector based on the state change confidence and the stage annotation data specifically includes: Step 10211: Perform sliding alignment processing on the vehicle operation data to obtain the target operation data; Step 10212: Input the target running data, the state change confidence and the stage labeling data into the pre-trained time-aware model, and process them through the time-aware model to obtain the stage embedding vector.

[0064] In practice, due to issues such as sensor operating frequency and clock drift, multi-source data undergoes unified alignment processing, specifically a sliding alignment process is applied to the vehicle operation data to obtain the target operation data. Specifically, a fusion method based on timestamp resampling and linear interpolation is used to align all signals to a unified time series axis, and a sliding window sample structure is constructed. Each window covers a duration of 2-5 seconds, forming stable time-series sample segments, thus obtaining the target operation data.

[0065] A pre-trained time-aware model is obtained, and the target running data, the state change confidence, and the stage labeling data are input into the time-aware model. The time-aware model processes the data to obtain the stage embedding vector.

[0066] In this embodiment, the time-series perception model is a neural network model. Preferably, the time-series perception model adopts the TDE-GNN model structure. The TDE-GNN model is designed to describe the dynamic topological structure of driving behavior during the transition from off-road to highway scenarios. Its core objective is to construct a time-series driven behavior graph structure to express the evolution path and scene stage changes between multiple source events.

[0067] Specifically, the temporal awareness model employs a multi-layer graph neural network structure, with temporal embedding weights introduced in each layer. A temporal encoder enhances the nodes' ability to perceive the stage context, enabling the model to understand the transitional pattern of driving behavior from drastic changes to a gradual stabilization. In the embedding learning part, stage label supervision is introduced during the aggregation stage of the graph neural network. A weakly supervised method maps the graph structure to a low-dimensional stage vector space, forming an embedding representation that can be used for stage classification.

[0068] In this embodiment, during the graph update process, three types of structural changes are focused on: the disappearance of abnormally high-frequency mutation nodes, the convergence of edge weights to low intensity, and the stabilization of flow paths between node clusters. These three types are key signs of off-road behavior converging into highway driving behavior.

[0069] The above approach, using the trained temporal perception model, breaks through the limitations of traditional temporal modeling in expressing nonlinear transitional behavior. It can accurately capture the continuous evolution of vehicle behavior from off-road to highway, improving the accuracy of subsequent judgments on whether to exit the off-road model.

[0070] In some embodiments, when determining the target state label and the first confidence level corresponding to the target state label, it can be determined jointly based on the off-road attenuation magnitude corresponding to the stage embedding vector, the second confidence level corresponding to the target road stage type, and the state change confidence level corresponding to the vehicle operation data. That is, step 103, which involves determining the target state label and the first confidence level corresponding to the target state label based on the state change confidence level and the target road stage type, specifically includes: Step 1031: Based on the pre-trained graph neural network, the stage embedding vector is aggregated and compressed to obtain the off-road attenuation magnitude; Step 1032: Obtain the second confidence level corresponding to the target road stage type; determine the target state label and the first confidence level corresponding to the target state label based on the state change confidence level, the second confidence level and the off-road attenuation magnitude.

[0071] In practice, a pre-trained graph neural network is obtained, and the road network or terrain grid is treated as a graph structure. Through the flow of graph nodes and edges, neighborhood information is aggregated and redundant features are compressed. That is, the stage embedding vector is aggregated and compressed to obtain the off-road attenuation magnitude. Here, the off-road attenuation magnitude is an index that changes over time, quantifying the decrease in the intensity of off-road behavior.

[0072] Obtain the second confidence level corresponding to the target road stage type, wherein the second confidence level and the target road stage type are both parameters output by the stage type recognition model. That is, the stage embedding vector, the behavior feature vector and the environmental data are input into the pre-trained stage type recognition model, processed by the stage type recognition model, and the target road stage type and the second confidence level corresponding to the target road stage type are output. The second confidence level is used to describe the accuracy of the target road stage type output by the stage type recognition model.

[0073] Based on the state change confidence level, the second confidence level, and the off-road attenuation magnitude, the target state label and the first confidence level corresponding to the target state label are determined.

[0074] Specifically, the method for determining the target state label and the first confidence level corresponding to the target state label includes: Step A: In response to the state change confidence being greater than a preset change confidence threshold and the off-road attenuation amplitude being less than a preset amplitude threshold, the target state label is determined to be the first state; Step B: Obtain a preset weight, and perform weighted processing on the confidence level of the state change and the off-road attenuation magnitude based on the preset weight to obtain the first confidence level corresponding to the target state label.

[0075] In practice, the confidence level of the state change is compared with a preset change confidence threshold, and the off-road attenuation magnitude is compared with a preset magnitude threshold. If the confidence level of the state change is greater than the preset change confidence threshold, and the off-road attenuation magnitude is less than the preset magnitude threshold, it indicates that the vehicle is in a high-probability off-road end characteristic, that is, the vehicle state has changed abruptly, and the attenuation of the off-road behavior intensity tends to stabilize. At this point, it can be considered that the vehicle has left the off-road section and has basically completed the road switch, that is, it is not in the process of switching from an off-road road to another road. At this time, the target state label is determined as the first state, which is allowed to exit the off-road mode.

[0076] Obtain a preset weight, and perform weighted processing on the confidence level of the state change and the off-road attenuation magnitude based on the preset weight. That is, when determining the first confidence level, the values ​​of the confidence level of the state change and the values ​​of the off-road attenuation magnitude are considered comprehensively. The weighted result is the first confidence level corresponding to the target state label.

[0077] Specifically, the method for determining the target state label and the first confidence level corresponding to the target state label includes: Step a: Obtain the historical road stage type and historical confidence level of the previous time step at the current time step; Step b: In response to the fact that both the historical road phase type and the target road phase type are the first road type, and the second confidence level is greater than or equal to the historical confidence level, the target state label is determined to be the first state; Step c: Use the second confidence level as the first confidence level corresponding to the target state label.

[0078] In practice, the historical road type and historical confidence level of the previous moment are obtained, and the historical confidence level is compared with the second confidence level corresponding to the target road stage type.

[0079] If it is determined that both the historical road phase type and the target road phase type are the first road type, and the second confidence level is greater than or equal to the historical confidence level, that is, at this time, multiple consecutive moments are all road types other than off-road roads, and the confidence level continues to rise.

[0080] At this point, it can be determined that the scene transition is complete, meaning the vehicle has left the off-road section and the road transition is essentially complete. It is no longer in the process of switching from an off-road road to another road. Therefore, the target state label is set to the first state, which indicates permission to exit off-road mode. Furthermore, the second confidence level corresponding to the target road stage type is used as the first confidence level corresponding to the target state label.

[0081] In this embodiment, if there is a short-term conflict between the off-road attenuation magnitude corresponding to the stage embedding vector, the second confidence level corresponding to the target road stage type, and the confidence level of the state change corresponding to the vehicle operation data, but the overall trend is consistent, fault tolerance judgment can be made through time redundancy sliding window to avoid erroneous switching.

[0082] Specifically, the time-redundant sliding window employs two time ranges: a basic recognition window of approximately 3 seconds and a redundant judgment window covering the most recent 9 seconds, i.e., three consecutive basic windows, used to confirm whether the trend is truly stable. The voting mechanism is not a simple majority vote; instead, newer results have higher weight because the latest state is closer to the actual road conditions. A trend reward is also incorporated: if the confidence level of a stage continuously increases, or the behavioral mutation continuously weakens, the passing weight is further increased. Ultimately, as long as the overall judgment over the recent period is stable and the direction of change is consistent, the switch is confirmed. By using both time freshness and behavioral convergence trend as judgment criteria, the system can filter out short-term fluctuations and react faster and more stably to real, continuous changes.

[0083] In some embodiments, after determining that the preset mode switching conditions are met, vehicle status data can also be acquired. Furthermore, the vehicle can only be controlled to exit the off-road mode after the vehicle status data meets the preset off-road release conditions. That is, step 104, which involves controlling the vehicle to exit the off-road mode in response to the target status label and the first confidence level meeting the preset mode switching conditions, specifically includes: Step 1041: Obtain the historical state label corresponding to each time point within the preset time period of the current time, and the third confidence level corresponding to each historical state label; Step 1042: In response to the fact that both the target state label and the historical state label are in the first state and the first confidence level is greater than or equal to the third confidence level, obtain vehicle state data; Step 1043: In response to the vehicle status data meeting the preset off-road release conditions, control the vehicle to exit off-road mode.

[0084] In practice, the historical state label corresponding to each time point within the preset time period of the current time is obtained, and the third confidence level corresponding to each historical state label is compared with the first confidence level corresponding to the target state label.

[0085] If both the target status label and the historical status label are in the first state, and the first confidence level is greater than or equal to the third confidence level, then the soft exit condition is met, and a soft exit command is generated. Vehicle status data is acquired, wherein the vehicle status data is the vehicle's current status data, and the vehicle status data includes at least one of the following: suspension height, vehicle driving mode, etc.

[0086] If the vehicle status data meets the preset off-road release conditions, it indicates that the hard exit condition is met, triggering a hard exit signal. That is, at this time, it is determined that the vehicle meets the preset mode switching conditions, and the vehicle is controlled to exit the off-road mode.

[0087] In this embodiment, the preset off-road disengagement conditions include at least one of the following: whether the suspension height has been restored to the default value, whether the low-speed four-wheel drive mode has been disengaged, and whether the driver has not performed any off-road mode locking operation.

[0088] In this embodiment, if user intervention is detected during the process of determining whether the vehicle meets the preset mode switching conditions, the mode switching is stopped and the control is returned to the user.

[0089] By employing the above-mentioned solution, which combines soft and hard exit conditions and uses a combined soft and hard mechanism, frequent mode switching caused by misjudgment is avoided. This ensures the accuracy, smoothness, and controllability of the system during the automatic exit process of off-road mode, thereby improving the user's driving experience.

[0090] In some embodiments, a pre-trained state change detection model can be used to analyze vehicle operating data to obtain a state change confidence level. Specifically, determining the state change confidence level based on the vehicle operating data in step 101 includes: Step 1011: Obtain the pre-trained state mutation detection model, wherein the state mutation detection model includes an input layer, an intermediate layer and an output layer; Step 1012: Each piece of vehicle operation data is input into the state change detection model through the input layer, and each piece of operation data is encoded to obtain multiple operation vectors; Step 1013: Determine the target weight corresponding to each running vector based on the intermediate layer, and take the running vector with the target weight greater than the preset weight threshold as the target running vector; Step 1014: Determine the state mutation confidence based on the target weight corresponding to each target running vector, and output the state mutation confidence through the output layer.

[0091] In specific implementation, a pre-trained state change detection model is obtained, which includes an input layer, an intermediate layer, and an output layer. Each piece of vehicle operation data is input into the state change detection model through the input layer, and each piece of operation data is encoded to obtain multiple operation vectors.

[0092] Specifically, vehicle operation data is sliced ​​according to fixed time windows, and multi-source data within each window are spatiotemporally aligned to ensure that each sample contains synchronous multimodal information. Within each aligned time window, each piece of operation data is encoded separately to obtain multiple operation vectors.

[0093] The target weight corresponding to each running vector is determined based on the intermediate layer, and running vectors with target weights greater than a preset weight threshold are used as target running vectors. Specifically, the input running vectors are allocated to different feature sub-channels according to their physical sources, ensuring that the calculation of each physical quantity is independent and preserving the original physical characteristics. The physical sources include at least one of the following: braking, steering, acceleration, vertical load, and GPS.

[0094] Each channel's internal network structure processes its corresponding time series, capturing the temporal dependence of each physical quantity within a window. An attention mechanism is introduced to dynamically evaluate the outputs of all sub-channels. Based on the context of the current window, a target weight is determined for each running vector. Each target weight is compared with a preset weight threshold, and running vectors with target weights greater than the preset weight threshold are selected as target running vectors.

[0095] The target weights corresponding to each target running vector are fused together, that is, the target weights corresponding to each target running vector are summed to obtain the state change confidence, and the state change confidence is output through the output layer.

[0096] In this embodiment, the state change detection model is a neural network structure model. Preferably, the state change detection model is the EventNet-Lite model. The EventNet-Lite model is a lightweight event-driven recognition structure specifically designed to capture key behavioral change features of a vehicle during the transition from off-road driving to on-road driving. The core of the model design is to use sudden changes in high-frequency CAN signals as trigger sources, focusing on the event-driven state transition process rather than relying on traditional equal-interval time series modeling methods.

[0097] In this embodiment, to ensure that the off-road automatic exit recognition system has high robustness and adaptability under different vehicle platforms, driving styles and terrain environments, the system acquires sliding window data recorded during long-term vehicle operation, fused decision results, control domain feedback logs, driver manual intervention records and road label information, and forms a training data pool through a local or cloud platform.

[0098] Specifically, an offline real-vehicle playback system is built to reproduce historical driving segments of the vehicle in chronological order. Key nodes, such as misjudged exit, delayed exit, and forced driver switching, are labeled and corrected to form real-label samples, which are used to supervise and fine-tune the weight configuration of each sub-model.

[0099] For the state mutation detection model, the focus is on optimizing the event mutation sensitivity parameter and adjusting the event threshold and triggering strategy. For the time-series awareness model, the node feature weight distribution and graph evolution learning rate are dynamically adjusted. For the stage type recognition model, the stage center vector is reconstructed based on the actual road condition distribution to improve the stability of the classification boundary.

[0100] Meanwhile, after each vehicle triggers the exit judgment, the system automatically records the environmental and behavioral data before and after the event, and performs post-effect analysis based on the actual exit control execution results. If there are cases where the confidence level is high but the exit is not successful, or where the confidence level is low but the exit is triggered erroneously, they are marked as learning samples and added to the fine-tuning task queue.

[0101] In terms of parameter tuning strategy, a collaborative adjustment mechanism between modules is adopted. For example, when the event model frequently outputs high mutation scores while the structural model is identified as stable, the mutation sensitivity of the state mutation detection model will be reduced to improve the consistency and robustness of the overall system. All tuning and optimization processes retain iterative version control, support rollback and comparative analysis, and ensure the traceability and reliability of the system upgrade process.

[0102] The above solution enables the system to have long-term adaptive evolution capabilities, supporting the algorithm to gradually stabilize under different vehicle series and driving scenarios. It effectively solves the problem of fluctuation in recognition accuracy caused by sample bias, driving style differences or sensor noise in the actual vehicle deployment of off-road automatic exit recognition, and comprehensively improves the engineering implementation value and continuous optimization capability of the patent solution.

[0103] It should be noted that the method of this disclosure embodiment can be executed by a single device, such as a computer or server. The method of this embodiment can also be applied to a distributed scenario, where multiple devices cooperate to complete the task. In such a distributed scenario, one of these devices may execute only one or more steps of the method of this disclosure embodiment, and the multiple devices will interact with each other to complete the method described.

[0104] It should be noted that the above description describes some embodiments of this disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in a different order than that shown in the above embodiments and still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0105] Based on the same inventive concept, corresponding to any of the above-described embodiments, this disclosure also provides a vehicle control device.

[0106] refer to Figure 2 , Figure 2 The vehicle control device, as described in this embodiment, includes: The data acquisition module 201 is configured to determine that the vehicle is in off-road mode, acquire vehicle operation data, and determine the confidence level of state change based on the vehicle operation data. The road type determination module 202 is configured to acquire environmental data of the vehicle's environment and determine the target road stage type based on the environmental data and the state change confidence level. The status label determination module 203 is configured to determine a target status label and a first confidence level corresponding to the target status label based on the status change confidence level and the target road stage type. The vehicle control module 204 is configured to control the vehicle to exit the off-road mode in response to the target status label and the first confidence level satisfying the preset mode switching condition.

[0107] In some embodiments, the road type determination module 202 is specifically configured to: Obtain the stage annotation data corresponding to the historical operation data, and determine the stage embedding vector based on the state change confidence and the stage annotation data; Based on the vehicle operation data and the confidence level of the state change, a behavioral feature vector is determined; The stage embedding vector, the behavior feature vector, and the environmental data are input into a pre-trained stage type recognition model. The target road stage type is obtained by processing the stage type recognition model.

[0108] In some embodiments, the road type determination module 202 is specifically configured to: The vehicle operation data is subjected to sliding alignment processing to obtain the target operation data; The target running data, the state change confidence, and the stage labeling data are input into a pre-trained time-aware model. The time-aware model processes the data to obtain the stage embedding vector.

[0109] In some embodiments, the status tag determination module 203 is specifically configured as follows: Based on the pre-trained graph neural network, the stage embedding vector is aggregated and compressed to obtain the off-road attenuation magnitude. Obtain the second confidence level corresponding to the target road stage type, and determine the target state label and the first confidence level corresponding to the target state label based on the state change confidence level, the second confidence level and the off-road attenuation magnitude.

[0110] In some embodiments, the status tag determination module 203 is specifically configured as follows: In response to the state change confidence being greater than a preset change confidence threshold and the off-road attenuation amplitude being less than a preset amplitude threshold, the target state label is determined to be the first state; Obtain a preset weight, and perform weighted processing on the confidence level of the state change and the off-road attenuation magnitude based on the preset weight to obtain the first confidence level corresponding to the target state label.

[0111] In some embodiments, the status tag determination module 203 is specifically configured as follows: Obtain the historical road stage type and historical confidence level of the previous time step at the current moment; In response to the fact that both the historical road phase type and the target road phase type are the first road phase type, and the second confidence level is greater than or equal to the historical confidence level, the target state label is determined to be the first state; The second confidence level is used as the first confidence level corresponding to the target state label.

[0112] In some embodiments, the vehicle control module 204 is specifically configured as follows: Obtain the historical state label corresponding to each time point within the preset time period of the current time, and the third confidence level corresponding to each historical state label; In response to the fact that both the target state label and the historical state label are in the first state, and the first confidence level is greater than or equal to the third confidence level, vehicle state data is obtained; In response to the vehicle status data meeting the preset off-road disengagement conditions, it is determined that the preset mode switching conditions are met, and the vehicle is controlled to exit the off-road mode.

[0113] In some embodiments, the data acquisition module 201 is specifically configured as follows: Obtain a pre-trained state mutation detection model, wherein the state mutation detection model includes an input layer, an intermediate layer, and an output layer; Each piece of vehicle operation data is input into the state change detection model through the input layer, and each piece of operation data is encoded to obtain multiple operation vectors. Based on the intermediate layer, the target weight corresponding to each running vector is determined, and the running vector with the target weight greater than the preset weight threshold is taken as the target running vector; The confidence level of state mutation is determined based on the target weight corresponding to each target running vector, and the confidence level of state mutation is output through the output layer.

[0114] For ease of description, the above apparatus is described in terms of its functions, divided into various modules. Of course, in implementing this disclosure, the functions of each module can be implemented in one or more software and / or hardware.

[0115] The apparatus of the above embodiments is used to implement the corresponding vehicle control method in any of the foregoing embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0116] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, this disclosure also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the vehicle control method described in any of the above embodiments.

[0117] Figure 3 This embodiment illustrates a more specific hardware structure of an electronic device. The device may include a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, memory 1020, input / output interface 1030, and communication interface 1040 are interconnected internally via the bus 1050.

[0118] The processor 1010 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0119] The memory 1020 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1020 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program code is stored in the memory 1020 and is called and executed by the processor 1010.

[0120] The input / output interface 1030 is used to connect input / output modules to realize information input and output. Input / output modules can be configured as components within the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touchscreens, microphones, various sensors, etc., while output devices may include displays, speakers, vibrators, indicator lights, etc.

[0121] The communication interface 1040 is used to connect a communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0122] Bus 1050 includes a pathway for transmitting information between various components of the device, such as processor 1010, memory 1020, input / output interface 1030, and communication interface 1040.

[0123] It should be noted that although the above-described device only shows the processor 1010, memory 1020, input / output interface 1030, communication interface 1040, and bus 1050, in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the embodiments of this specification, and not necessarily all the components shown in the figures.

[0124] The electronic devices described above are used to implement the corresponding vehicle control methods in any of the foregoing embodiments and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0125] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, this disclosure also provides a non-transitory computer-readable storage medium that stores computer instructions for causing the computer to execute the vehicle control method as described in any of the above embodiments.

[0126] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0127] The computer instructions stored in the storage medium of the above embodiments are used to cause the computer to execute the vehicle control method as described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0128] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, this application also provides a vehicle, including the vehicle control device, the electronic device, and the computer-readable storage medium in the above embodiments, wherein the vehicle device implements the vehicle control method described in any of the above embodiments.

[0129] The vehicles described in the above embodiments are used to implement the vehicle control method described in any of the foregoing embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0130] It is understood that before using the technical solutions of the various embodiments in this disclosure, users will be informed of the type, scope of use, and usage scenarios of the personal information involved in an appropriate manner, and user authorization will be obtained.

[0131] For example, upon receiving a user's active request, a prompt message is sent to the user to explicitly inform them that the requested operation will require the acquisition and use of the user's personal information. This allows the user to independently choose, based on the prompt message, whether to provide personal information to the software or hardware such as electronic devices, applications, servers, or storage media performing the operations of this disclosed technical solution.

[0132] As an optional but not limited implementation, in response to a user's active request, sending a prompt message to the user can be done via a pop-up window, where the prompt message can be presented in text format. Furthermore, the pop-up window can also include a selection control allowing the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0133] It is understood that the above notification and user authorization process are merely illustrative and do not constitute a limitation on the implementation of this disclosure. Other methods that comply with relevant laws and regulations may also be applied to the implementation of this disclosure.

[0134] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of this disclosure (including the claims) is limited to these examples; within the framework of this disclosure, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of the embodiments of this disclosure as described above, which are not provided in detail for the sake of brevity.

[0135] Additionally, to simplify the description and discussion, and to avoid obscuring the embodiments of this disclosure, the provided drawings may or may not show well-known power / ground connections to integrated circuit (IC) chips and other components. Furthermore, the apparatus may be shown in block diagram form to avoid obscuring the embodiments of this disclosure, and this also takes into account the fact that the details of implementation of these block diagram apparatuses are highly dependent on the platform on which the embodiments of this disclosure will be implemented (i.e., these details should be fully understood by those skilled in the art). While specific details (e.g., circuits) have been set forth to describe exemplary embodiments of this disclosure, it will be apparent to those skilled in the art that the embodiments of this disclosure can be implemented without these specific details or with variations thereof. Therefore, these descriptions should be considered illustrative rather than restrictive.

[0136] Although this disclosure has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed.

[0137] This disclosure is intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A vehicle control method, characterized in that, include: Determine that the vehicle is in off-road mode, acquire vehicle operation data, and determine the confidence level of state change based on the vehicle operation data; Acquire environmental data of the vehicle's surroundings, and determine the target road phase type based on the environmental data and the state change confidence level; Based on the state change confidence level and the target road stage type, determine the target state label and the first confidence level corresponding to the target state label; In response to the target status label and the first confidence level satisfying the preset mode switching conditions, the vehicle is controlled to exit the off-road mode.

2. The method according to claim 1, characterized in that, The step of acquiring environmental data of the vehicle's location and determining the target road stage type based on the environmental data and the state change confidence level includes: Obtain the stage annotation data corresponding to the historical operation data, and determine the stage embedding vector based on the state change confidence and the stage annotation data; Based on the vehicle operation data and the confidence level of the state change, a behavioral feature vector is determined; The stage embedding vector, the behavior feature vector, and the environmental data are input into a pre-trained stage type recognition model. The target road stage type is obtained by processing the stage type recognition model.

3. The method according to claim 2, characterized in that, The step of determining the stage embedding vector based on the state change confidence and the stage annotation data includes: The vehicle operation data is subjected to sliding alignment processing to obtain the target operation data; The target running data, the state change confidence, and the stage labeling data are input into a pre-trained time-aware model. The time-aware model processes the data to obtain the stage embedding vector.

4. The method according to claim 2, characterized in that, The step of determining the target state label and the first confidence level corresponding to the target state label based on the state change confidence level and the target road stage type includes: Based on the pre-trained graph neural network, the stage embedding vector is aggregated and compressed to obtain the off-road attenuation magnitude. Obtain the second confidence level corresponding to the target road stage type, and determine the target state label and the first confidence level corresponding to the target state label based on the state change confidence level, the second confidence level and the off-road attenuation magnitude.

5. The method according to claim 4, characterized in that, The step of determining the target state label and the first confidence level corresponding to the target state label based on the state change confidence level, the second confidence level, and the off-road attenuation magnitude includes: In response to the state change confidence being greater than a preset change confidence threshold and the off-road attenuation amplitude being less than a preset amplitude threshold, the target state label is determined to be the first state; Obtain a preset weight, and perform weighted processing on the confidence level of the state change and the off-road attenuation magnitude based on the preset weight to obtain the first confidence level corresponding to the target state label.

6. The method according to claim 4, characterized in that, The step of determining the target state label and the first confidence level corresponding to the target state label based on the state change confidence level, the second confidence level, and the off-road attenuation magnitude includes: Obtain the historical road stage type and historical confidence level of the previous time step at the current moment; In response to the fact that both the historical road phase type and the target road phase type are the first road phase type, and the second confidence level is greater than or equal to the historical confidence level, the target state label is determined to be the first state; The second confidence level is used as the first confidence level corresponding to the target state label.

7. The method according to claim 1, characterized in that, The step of controlling the vehicle to exit off-road mode in response to the target state label and the first confidence level satisfying the preset mode switching condition includes: Obtain the historical state label corresponding to each time point within the preset time period of the current time, and the third confidence level corresponding to each historical state label; In response to the fact that both the target state label and the historical state label are in the first state, and the first confidence level is greater than or equal to the third confidence level, vehicle state data is obtained; In response to the vehicle status data meeting the preset off-road disengagement conditions, it is determined that the preset mode switching conditions are met, and the vehicle is controlled to exit the off-road mode.

8. The method according to claim 1, characterized in that, Determining the confidence level of state change based on the vehicle operation data includes: Obtain a pre-trained state mutation detection model, wherein the state mutation detection model includes an input layer, an intermediate layer, and an output layer; Each piece of vehicle operation data is input into the state change detection model through the input layer, and each piece of operation data is encoded to obtain multiple operation vectors. Based on the intermediate layer, the target weight corresponding to each running vector is determined, and the running vector with the target weight greater than the preset weight threshold is taken as the target running vector; The confidence level of state mutation is determined based on the target weight corresponding to each target running vector, and the confidence level of state mutation is output through the output layer.

9. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the method as described in any one of claims 1 to 8.

10. A vehicle, characterized in that, The vehicle includes the electronic equipment as described in claim 9.

Images