Vehicle control method and vehicle

Abstract

This disclosure relates to the field of intelligent driving technology for vehicles, and provides a vehicle control method and a vehicle. The method involves acquiring vehicle operating information and environmental data, determining a data frame sequence based on the vehicle operating information and environmental data, determining a wind speed gradient evolution sequence based on the data frame sequence, determining a target lateral response type based on system state parameters and the wind speed gradient evolution sequence, inputting the data frame sequence, wind speed gradient evolution sequence, and target lateral response type into a vehicle compensation control model, processing the data according to the vehicle compensation control model to obtain a target compensation control command, and controlling the vehicle's movement according to the target compensation control command. This disclosure, when performing pre-compensation control on the vehicle, comprehensively considers the data frame sequence corresponding to the actual operating data, the predicted wind speed gradient evolution sequence, and the calculated target lateral response type, avoiding problems such as vehicle lateral deviation caused by sudden changes in wind direction, and improving the stability of vehicle movement.

Description

Technical Field

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

[0002] With the rapid development of vehicle technology, vehicles have become an important means of transportation in people's daily lives. Currently, when vehicles are driving, especially when passing through valley exits or windy areas at high speeds, sudden changes in wind direction often cause a sharp increase in crosswind lateral force, which can lead to vehicle lateral deviation, steering deviation, or even lane departure, seriously threatening driving stability. Summary of the Invention

[0003] In view of this, the purpose of this disclosure is to propose a vehicle control method and a vehicle to solve the problem that when a vehicle is in motion, sudden changes in wind direction often lead to a sharp increase in crosswind lateral force, which in turn causes vehicle lateral deviation, steering deviation, or even lane departure, seriously threatening driving stability.

[0004] To achieve the above objectives, a first aspect of this disclosure provides a vehicle control method, the method comprising:

[0005] Acquire vehicle operation information and environmental data of the vehicle's environment, determine a data frame sequence based on the vehicle operation information and the environmental data, and determine a wind speed gradient evolution sequence based on the data frame sequence. Obtain vehicle system state parameters, and determine the target lateral response type based on the system state parameters and the wind speed gradient evolution sequence; A pre-trained vehicle control model is obtained, and the data frame sequence, the wind speed gradient evolution sequence, and the target lateral response type are input into the vehicle compensation control model. The vehicle compensation control model is processed to obtain a target compensation control command, and the vehicle is controlled to drive according to the target compensation control command.

[0006] 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 acquire vehicle operation information and environmental data of the vehicle's environment, determine a data frame sequence based on the vehicle operation information and the environmental data, and determine a wind speed gradient evolution sequence based on the data frame sequence. The type determination module is configured to acquire vehicle system state parameters and determine the target lateral response type based on the system state parameters and the wind speed gradient evolution sequence. The vehicle control module is configured to acquire a pre-trained vehicle control model, input the data frame sequence, the wind speed gradient evolution sequence, and the target lateral response type into the vehicle compensation control model, process the data according to the vehicle compensation control model to obtain a target compensation control command, and control the vehicle to drive according to the target compensation control command.

[0007] 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.

[0008] 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.

[0009] 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.

[0010] As can be seen from the above, this disclosure proposes a vehicle control method and a vehicle. The method involves acquiring vehicle operation information and environmental data of the vehicle's environment, determining a data frame sequence based on the vehicle operation information and the environmental data, and determining a wind speed gradient evolution sequence based on the data frame sequence. The wind speed gradient evolution sequence represents the future wind field evolution path, i.e., the predicted wind speed tensor, describing the wind field changes over a future time period after the current moment. The method also involves acquiring vehicle system state parameters, determining a target lateral response type based on the system state parameters and the wind speed gradient evolution sequence. The target lateral response type is calculated based on the vehicle's historical response to crosswind disturbances, and represents the vehicle's response in a future wind field change scenario as shown by the wind speed gradient evolution sequence. Finally, the method involves acquiring a pre-trained vehicle control model, inputting the data frame sequence, the wind speed gradient evolution sequence, and the target lateral response type into the vehicle compensation control model, processing the data frame sequence, obtaining a target compensation control command, and controlling the vehicle's movement according to the target compensation control command. The target compensation control command is determined by a pre-trained vehicle compensation control model. Because the vehicle compensation control model is trained on a large amount of data, the resulting target compensation control command is more accurate. Simultaneously, when performing pre-compensation control on the vehicle, the data frame sequence corresponding to the actual operating data, the predicted wind speed gradient evolution sequence, and the calculated target lateral response type are comprehensively considered to avoid problems such as vehicle yaw caused by sudden changes in wind direction, thereby improving vehicle stability. Attached Figure Description

[0011] 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.

[0012] 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

[0013] 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.

[0014] 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.

[0015] The following are definitions of terms used in this disclosure: Diffusion Model: The diffusion model is an advanced generative artificial intelligence algorithm that first adds noise to data (such as images) continuously until it is completely destroyed and becomes pure random noise. The model is then trained to learn how to gradually remove noise from this pure noise and accurately restore or create new content that meets the requirements in the process.

[0016] Online Dictionary Learning is an iterative algorithm that learns a sparse representation basis (i.e., a dictionary) from data step by step. It updates the dictionary online efficiently by processing only one or a small batch of data samples at a time, so that each sample can be represented by a linear combination of a few atoms in the dictionary. This enables efficient adaptive data representation in fields such as image denoising, compression, and feature extraction.

[0017] EPS: Electronic Power Steering (EPS) provides steering assistance to the driver directly through an electric motor. It can adjust the steering force in real time according to the vehicle speed. It is light and agile at low speeds and stable and precise at high speeds. It not only makes parking and maneuvering easier, but also effectively improves driving stability.

[0018] With the rapid development of vehicle technology, vehicles have become an important means of transportation in people's daily lives. Currently, intelligent vehicles often face crosswind interference caused by sudden changes in wind direction when traversing complex terrains such as highway exits of valleys or wind gaps where wind power is concentrated. This crosswind lateral force increases dramatically in an instant, causing the vehicle to veer, deviate from its steering course, or even drift out of its lane, greatly threatening driving safety.

[0019] To address this issue, most existing technologies rely on fixed wind speed thresholds or simplified wind field models to determine crosswind scenarios. This approach ignores the topological characteristics of wind field changes, making it difficult to accurately identify abrupt changes in wind speed gradients and the trends in wind field changes. Meanwhile, traditional control strategies often suffer from response lag, failing to provide pre-compensation control based on future wind disturbance trends, thus making it difficult to effectively cope with sudden crosswind disturbances.

[0020] Furthermore, multi-model systems in related technologies often operate independently and lack effective fusion mechanisms, resulting in insufficient stability and adaptability of control strategies, making it difficult to achieve precise crosswind disturbance control in complex terrain.

[0021] In conclusion, when facing crosswind disturbances in complex terrain, intelligent vehicles urgently need to introduce new technologies to improve their ability to identify wind field changes and the response speed of control strategies, thereby ensuring driving stability and safety.

[0022] Therefore, this embodiment proposes a vehicle control method, such as... Figure 1 As shown, the method includes: Step 101: Obtain vehicle operation information and environmental data of the vehicle's location; determine a data frame sequence based on the vehicle operation information and the environmental data; and determine a wind speed gradient evolution sequence based on the data frame sequence.

[0023] In specific implementation, vehicle operation information and environmental data of the vehicle's environment are acquired. The vehicle operation information includes vehicle attitude information and vehicle dynamic response data. The vehicle attitude information includes at least one of the following: instantaneous vehicle acceleration, yaw rate, sideslip angle, etc. The vehicle dynamic response data includes at least one of the following: wheel speed signal, steering wheel angle, yaw rate, lateral acceleration, and EPS control current, etc.

[0024] In this embodiment, the environmental data of the vehicle's environment includes at least one of the following: pressure changes measured in real time by a miniature barometric pressure array sensor mounted on the vehicle roof, instantaneous wind speed vector output by a laser particle scattering wind speed measurement module, and ambient air density disturbances detected by millimeter-wave radar.

[0025] After acquiring vehicle operation information and environmental data, a data frame sequence is determined based on the vehicle operation information and environmental data. After obtaining the data frame sequence, a wind speed gradient evolution sequence is determined based on the data frame sequence.

[0026] Specifically, the driver's manual correction actions and braking states during crosswinds are collected synchronously and used as priors for behavioral response feature modeling. All data are aligned using a unified timestamp from the master clock, aligning environmental data, vehicle operation information, and driver behavior data. Adaptive sliding window filtering is used to process high-frequency jitter and eliminate sampling errors, ultimately forming a standardized multimodal data frame sequence under wind disturbance scenarios. Each data frame forms a corresponding high-dimensional tensor.

[0027] Step 102: Obtain vehicle system state parameters, and determine the target lateral response type based on the system state parameters and the wind speed gradient evolution sequence.

[0028] In practice, vehicle system state parameters are acquired, whereby these parameters represent the state data of various vehicle systems at the current moment. Specifically, the vehicle system state parameters include at least one of the following: vehicle wheelbase, vehicle mass distribution, suspension stiffness calibration value, and steering system response parameters.

[0029] After acquiring the system state parameters, the target lateral response type is determined based on the system state parameters and the wind speed gradient evolution sequence. The target lateral response type is the vehicle's response mode in a future scenario where the wind field changes to the wind speed gradient evolution sequence, calculated based on the vehicle's historical response to crosswind disturbances.

[0030] Step 103: Obtain the pre-trained vehicle control model, input the data frame sequence, the wind speed gradient evolution sequence, and the target lateral response type into the vehicle compensation control model, process the data according to the vehicle compensation control model to obtain the target compensation control command, and control the vehicle to drive according to the target compensation control command.

[0031] In practice, a pre-trained vehicle control model is acquired, and the data frame sequence, the wind speed gradient evolution sequence, and the target lateral response type are input into the vehicle compensation control model. Specifically, the data frame sequence, the wind speed gradient evolution sequence, and the target lateral response type can be fused to obtain a target fusion vector.

[0032] The target fusion vector is then input into the vehicle compensation control model, which processes it to obtain the target compensation control command. The vehicle is then controlled according to the target compensation control command. The target compensation control command is in signal form, and the corresponding control signal is transmitted via a standard interface to the vehicle's underlying execution units, including the EPS module, braking system, and active suspension system, to ensure that the compensation behavior matches the vehicle's hardware capabilities.

[0033] In this embodiment, the training process of the vehicle compensation control model specifically includes: Step A: Obtain the first training dataset and the initial vehicle compensation control model, wherein the first training dataset includes historical fusion vectors and historical compensation control commands.

[0034] Step B: Input the training data from the first training dataset into the initial vehicle compensation control model for training, determine that the first preset training termination condition is met, and obtain the vehicle compensation control model.

[0035] In specific implementation, a first training dataset and an initial vehicle compensation control model are obtained, wherein the first training dataset includes historical fusion vectors and historical compensation control commands. Training data from the first training dataset is input into the initial vehicle compensation control model for training. Once a first preset training termination condition is met, the vehicle compensation control model is obtained.

[0036] 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 vehicle compensation control model for training, determining that the loss function of the initial vehicle compensation control model has converged to a first convergence threshold, or determining that the initial vehicle compensation control model has been iteratively trained to a first preset number of iterations.

[0037] 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 vehicle compensation control model for training: The first training dataset contains fifty sets of data, each set including a historical fusion vector and historical compensation control commands. The first preset training termination condition is that all data in the first training dataset has been input into the initial vehicle compensation control model for training. That is, when all fifty sets of data have been input into the initial vehicle compensation control model, there is no training data in the first training dataset that has not yet been input into the initial vehicle compensation control model. At this point, the initial vehicle compensation control model training is considered complete, and the vehicle compensation control model is obtained.

[0038] In another example, the first preset training termination condition is to determine that the loss function of the initial vehicle compensation control model converges to a first convergence threshold: Training data from the first training dataset is input into the initial vehicle compensation control model for training, and the training results are output. A loss function is determined based on the training results and the historical compensation control commands. The loss function may be of at least one of the following types: mean squared error loss function, cross-entropy loss function, logarithmic loss function, exponential loss function, squared loss function, or absolute value loss function. When the loss function converges to a first convergence threshold, a first preset training termination condition is satisfied, and the vehicle compensation control model is obtained.

[0039] In another example, the first preset training termination condition is to determine the initial vehicle compensation control model to be iteratively trained up to the first preset number of iterations.

[0040] The training data in the first training dataset is input into the initial vehicle compensation control 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 vehicle compensation control model is obtained.

[0041] The above scheme acquires vehicle operation information and environmental data of the vehicle's environment. Based on the vehicle operation information and environmental data, a data frame sequence is determined. Then, a wind speed gradient evolution sequence is determined based on the data frame sequence. This wind speed gradient evolution sequence represents the future wind field evolution path, i.e., the predicted wind speed tensor, describing the wind field changes over a future time period after the current moment. Vehicle system state parameters are acquired. Based on the system state parameters and the wind speed gradient evolution sequence, a target lateral response type is determined. This target lateral response type is calculated based on the vehicle's historical response to crosswind disturbances and represents the vehicle's response in the future wind field change scenario shown by the wind speed gradient evolution sequence. A pre-trained vehicle control model is acquired. The data frame sequence, the wind speed gradient evolution sequence, and the target lateral response type are input into the vehicle compensation control model. The model processes these parameters to obtain a target compensation control command, which controls the vehicle's movement. The target compensation control command is determined through a pre-trained vehicle compensation control model, which is trained on a large amount of data, resulting in a more accurate target compensation control command. Meanwhile, by comprehensively considering the data frame sequence corresponding to the actual operating data, the predicted wind speed gradient evolution sequence, and the calculated target lateral response type, the vehicle is pre-compensated for control, avoiding problems such as vehicle lateral deviation caused by sudden changes in wind direction, and improving the stability of vehicle driving.

[0042] In some embodiments, the wind speed gradient evolution sequence can be determined using a pre-trained wind speed gradient diffusion model. Specifically, step 101, determining the wind speed gradient evolution sequence based on the data frame sequence, includes: Step 1011: Map the data frame sequence to obtain a topological feature vector; Step 1012: Input the data frame sequence and the topological feature vector into the pre-trained wind speed gradient diffusion model, and process it through the wind speed gradient diffusion model to obtain the wind speed gradient evolution sequence.

[0043] In practice, the data frame sequence is mapped to extract structural change features of wind speed abrupt changes from continuous time series data, resulting in a topological feature vector. This topological feature vector provides boundary constraints and prior knowledge for subsequent wind speed gradient diffusion model prediction.

[0044] A pre-trained wind speed gradient diffusion model is obtained, and based on existing historical disturbance information, possible future wind field evolution paths are generated, thereby providing a pre-compensation time window for lateral stability control. Specifically, the data frame sequence and the topological feature vector are input into the wind speed gradient diffusion model, and processed by the model to obtain the wind speed gradient evolution sequence.

[0045] Specifically, the data frame sequence is first vector-field encoded to extract the distribution pattern of wind speed disturbances in the spatial region for each time slice, and a set of potential wind disturbance evolution trajectories is constructed through temporal nesting. The wind speed gradient diffusion model mainly adopts a joint modeling mechanism of forward disturbance process and reverse restoration process. The forward process simulates the expansion of wind disturbance evolution from a deterministic state to Gaussian noise disturbance, while the reverse process recovers the potential wind disturbance evolution path based on the number of diffusion steps.

[0046] In this embodiment, a conditional diffusion structure is adopted when training the wind speed gradient diffusion model. A topological feature vector is introduced as conditional information in each diffusion step to ensure that the generated wind speed gradient evolution sequence has a topological evolution law consistent with the real wind field. Simultaneously, a wind disturbance spatial consistency loss function is introduced to control the continuity of the generated wind speed vector field in local space and the interpretability of the macroscopic gradient, thereby improving the model's usability in actual physical wind fields.

[0047] Specifically, the topological feature vectors are mapped to time-dependent structural potential functions and injected into the noise scheduling and reverse denoising network of the wind speed gradient diffusion model. The topological phase transition intensity modulates the diffusion noise variance at each step, making the diffusion randomness more limited as the wind field structure approaches the abrupt boundary. A conditional normalization layer transforms the topological feature vectors into scale and offset parameters, dynamically reshaping the intermediate feature distribution to ensure that the generated wind speed gradient evolution sequence maintains connectivity and porosity consistent with the real topological evolution.

[0048] In this embodiment, the wind speed gradient evolution sequence describes the wind speed change trend and acceleration information at different locations and heights in tensor form, serving as the prediction basis for subsequently determining the target compensation control command.

[0049] The above scheme introduces a wind speed gradient diffusion model to generate future states, no longer relying on static thresholds to judge sudden wind disturbances, but has the ability to predict upcoming structural wind speed transitions, thus gaining critical response time for crosswind compensation strategies, i.e., target compensation control commands.

[0050] In some embodiments, the mapping process of the data frame sequence in step 1011 to obtain the topological feature vector includes: Step 10111: Perform sliding embedding processing on the data frame sequence to obtain high-dimensional point cloud data; Step 10112: Extract topological features from the high-dimensional point cloud data based on a preset algorithm to obtain the target topological structure; Step 10113: Quantize the target topology to obtain a topological feature vector.

[0051] In specific implementation, the data frame sequence is subjected to sliding embedding processing, that is, a data embedding subspace with a fixed time span is formed by sliding serial port, and the phase space trajectory is reconstructed by delayed embedding, so as to map the data frame sequence to a high-dimensional point cloud distribution and obtain high-dimensional point cloud data.

[0052] Specifically, the original one-dimensional time series is embedded into a higher-dimensional space through time delay, forming a trajectory composed of point clouds. This is equivalent to transforming the evolution of time into a distribution of spatial geometric shapes, allowing the dynamic system characteristics hidden in the time series signal (such as periodicity and chaos) to be manifested as specific geometric structures (such as tori and bifurcations) in the higher-dimensional space.

[0053] To capture the process of wind field disturbances changing from stable to drastic, a persistent cohomology algorithm is introduced. This algorithm generates a persistent map by calculating the lifetime of the Betti number at different scales. Specifically, the algorithm extracts topological features from the high-dimensional point cloud data to obtain the target topological structure. Segmental analysis using a sliding window approach reveals that when wind field disturbances are small, the point cloud structure may be tightly clustered, exhibiting high aggregation. When disturbances intensify or abruptly change, new branches, ring structures, and even voids appear in the point cloud. This process can clearly identify local extrema corresponding to gradual and abrupt changes in wind speed, changes in connecting branches, and the generation of high-dimensional voids, thereby establishing a quantitative mapping between the disturbance potential and the spatial evolution pattern.

[0054] The target topology is quantized to obtain a topological feature vector. This topological feature vector includes structural indices such as topological entropy in the perturbation dimension, persistent bar chart distribution, and feature point clustering. These topological features not only reveal the spatial continuity and abrupt changes in wind speed gradients but also provide perturbation boundary conditions for subsequent wind speed gradient diffusion models.

[0055] The above approach, through vehicle wind disturbance topology modeling, provides a model-free, data-driven structural insight capability. It does not rely on complex fluid dynamics equations but directly extracts features with generalization ability and structural interpretability from the geometric structure of the data itself. This transforms abstract physical disturbances into concrete mathematical language, ensuring high generalization ability and structural interpretability when facing complex wind field disturbances.

[0056] In some embodiments, a pre-trained lateral response type identification model is introduced, and then the topological feature vector, system state parameters, and wind speed gradient evolution sequence are analyzed using the lateral response type identification model to obtain the target lateral response type. Specifically, determining the target lateral response type based on the system state parameters and the wind speed gradient evolution sequence in step 102 includes: Step 1021: Obtain the pre-trained horizontal type recognition model; Step 1022: Input the topological feature vector, the system state parameters and the wind speed gradient evolution sequence into the lateral type identification model, and process them through the lateral type identification model to obtain the target lateral response type.

[0057] In specific implementation, a pre-trained lateral type recognition model is obtained, wherein the training process of the lateral type recognition model specifically includes: Obtain a second training dataset and an initial lateral type recognition model, wherein the second training dataset includes historical topological feature vectors, historical system state parameters, historical wind speed gradient evolution sequences, and historical lateral response types. Input the training data from the second training dataset into the initial lateral type recognition model for training, and determine that a second preset training termination condition is met to obtain the lateral type recognition model.

[0058] The second preset training termination condition includes at least one of the following: determining that all data in the second training dataset has been input into the initial lateral type recognition model for training, determining that the loss function of the initial lateral type recognition model has converged to the second convergence threshold, or determining that the initial lateral type recognition model has been iteratively trained to the second preset number of iterations.

[0059] For example, the second preset training termination condition is to determine that all data in the second training dataset has been input into the initial lateral type recognition model for training: The second training dataset contains fifty sets of data, each set including historical topological feature vectors, historical system state parameters, historical wind speed gradient evolution sequences, and historical lateral response types. The second preset training termination condition is determined when all data in the second training dataset has been input into the initial lateral type recognition model. Specifically, when all fifty sets of data have been input into the initial lateral type recognition model, there is no longer any training data in the second training dataset that has not yet been input into the initial lateral type recognition model. At this point, the initial lateral type recognition model training is considered complete, and the lateral type recognition model is obtained.

[0060] In another example, the second preset training termination condition is to determine that the loss function of the initial lateral type recognition model converges to a second convergence threshold: The training data from the second training dataset is input into the initial lateral type recognition model for training, and the training results are output. A loss function is determined based on the training results and the historical lateral response 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 second convergence threshold, it is determined that the second preset training termination condition is met, and the lateral type recognition model is obtained.

[0061] In another example, the second preset training termination condition is to determine the initial lateral type recognition model to be iterated and trained to the second preset number of iterations.

[0062] The training data in the second training dataset is input into the initial horizontal type recognition model for iterative training. The number of iterations is recorded. When the number of iterations is equal to the second preset number of iterations, the second preset training termination condition is met, and the horizontal type recognition model is obtained.

[0063] After obtaining the trained lateral type recognition model, the topological feature vector, the system state parameters, and the wind speed gradient evolution sequence are input into the lateral type recognition model. The model processes these elements to obtain the target lateral response type. The target lateral response type is a set of sparse coding coefficients and corresponding optimal response atom indices, used to describe the possible lateral dynamic response type and intensity of the vehicle under the current predicted wind disturbance conditions.

[0064] In this embodiment, the lateral type recognition model is a neural network structure model. Preferably, the lateral type recognition model is a model employing an online dictionary learning mechanism. That is, an online dictionary learning module is constructed to learn the lateral response behavior of vehicles under crosswind disturbance conditions, which is used to map the predicted wind speed gradient evolution results into typical response patterns that can be directly used for control decisions.

[0065] Specifically, based on historical real-vehicle test data and high-fidelity wind field simulation data, an initial set of crosswind disturbance response samples is constructed. Each sample contains the correspondence between wind speed gradient change patterns and vehicle yaw rate, sideslip angle, and steering wheel compensation behavior. An online dictionary learning mechanism is employed to represent the high-dimensional wind disturbance input as a linear combination of a set of sparse coefficients and several response atom vectors. These response atoms represent typical crosswind disturbance response types, i.e., lateral response types, such as abrupt lateral thrust, gradual lateral shift, or alternating disturbances.

[0066] The dictionary update process continues during vehicle operation. When the system detects a decrease in the atomic matching degree between the current wind disturbance state and the existing dictionary, it automatically introduces new samples and incrementally updates the dictionary to adapt to the wind disturbance characteristics under different terrains, vehicle models, and load conditions.

[0067] The above scheme uses dictionary representation to transform the complex and continuous wind disturbance evolution results into a finite and interpretable set of response modes, providing a structured and robust input for subsequent multi-model fusion and control signal generation, while significantly reducing the real-time computation burden and ensuring that the system has a rapid response capability in sudden crosswind scenarios.

[0068] In some embodiments, before inputting the data frame sequence, wind speed gradient evolution sequence, and target lateral response type into the vehicle compensation control model, the data frame sequence, wind speed gradient evolution sequence, and target lateral response type can be fused, and then the fused vector can be input into the vehicle compensation control model for analysis. That is, step 103, which involves inputting the data frame sequence, the wind speed gradient evolution sequence, and the target lateral response type into the vehicle compensation control model, and processing them according to the vehicle compensation control model to obtain the target compensation control command, includes: Step 1031: Determine the target fusion weight, and perform fusion processing on the data frame sequence, the topological feature vector, the wind speed gradient evolution sequence, and the target lateral response type based on the target fusion weight to obtain the target fusion vector; Step 1032: Input the target fusion vector into the vehicle compensation control model, process it according to the vehicle compensation control model, and obtain the target compensation control command.

[0069] In specific implementation, the target fusion weight is determined, and the data frame sequence, the topological feature vector, the wind speed gradient evolution sequence, and the target lateral response type are fused based on the target fusion weight to obtain the target fusion vector.

[0070] After determining the target fusion vector, it is input into the vehicle compensation control model. The model processes the vector to obtain the target compensation control command. Specifically, firstly, the risk label in the target fusion vector is parsed using a disturbance level decoder to determine the current wind disturbance intensity level and the stability margin range of the vehicle control. Based on the level, a predefined set of control rules and a compensation model library are invoked. Then, the corresponding response model template is called based on the dictionary atomic index to generate preliminary control suggestions, including steering angle fine-tuning, lateral torque compensation, and suspension stiffness adjustment suggestions.

[0071] In this embodiment, the target wind disturbance intensity is determined based on the environmental data, that is, by searching a database based on the environmental data. The database stores the correspondence between environmental data and wind disturbance intensity.

[0072] Based on the target wind disturbance intensity, vehicle dynamic deviation trend, and current control execution delay, the output priority of control signals is dynamically allocated. For example, under high-speed sudden wind disturbance, the system prioritizes activating the EPS motor to quickly adjust the path, while in medium-to-low speed continuous wind disturbance environments, it may enhance the lateral support force of the suspension through chassis linkage to mitigate yaw tendency. This priority mechanism uses control response timeliness as the core indicator, constructing a mapping strategy table between the vehicle state space and control actions to achieve controllable, predictable, and adjustable intervention.

[0073] The above scheme, through the integration of control strategies, not only achieves seamless connection from perception to control, but also improves the interpretability and credibility of control commands based on multi-source decision-making. Especially in the case of sudden wind disturbance, it has millisecond-level intervention capability and strong dynamic adaptability.

[0074] In some embodiments, the target fusion weights include a first target weight corresponding to the data frame sequence, a second target weight corresponding to the topological feature vector, a third target weight corresponding to the wind speed gradient evolution sequence, and a fourth target weight corresponding to the target lateral response type. The determination of the target fusion weights in step 1031 specifically includes: Step 10311: Determine the target wind disturbance intensity based on the environmental data; Step 10312: In response to the target wind disturbance intensity being greater than or equal to a preset intensity threshold, determine the third target weight corresponding to the wind speed gradient evolution sequence based on the target wind disturbance intensity; Step 10313: Determine the first target weight corresponding to the data frame sequence, the second target weight corresponding to the topological feature vector, and the fourth target weight corresponding to the target lateral response type based on the third target weight.

[0075] In practice, the target wind disturbance intensity is determined based on the environmental data; that is, the target wind disturbance intensity is determined by searching a database based on the environmental data. The database stores the correspondence between environmental data and wind disturbance intensity.

[0076] The target wind disturbance intensity is compared with a preset intensity threshold. If the target wind disturbance intensity is greater than or equal to the preset intensity threshold, it indicates that the vehicle is in a high-intensity sudden wind field. At this time, the third target weight corresponding to the wind speed gradient evolution sequence is determined according to the target wind disturbance intensity. That is, the database is searched according to the target wind disturbance intensity to determine the third target weight corresponding to the target wind disturbance intensity. The database stores the correspondence between the target wind disturbance intensity and the third target weight.

[0077] It is understandable that in high-intensity abrupt wind fields, the weight of diffusion prediction results is increased, so the weight of the third target determined at this time is greater than the third initial weight corresponding to the wind speed gradient evolution sequence.

[0078] After determining the third target weight corresponding to the wind speed gradient evolution sequence, the first target weight corresponding to the data frame sequence, the second target weight corresponding to the topological feature vector, and the fourth target weight corresponding to the target lateral response type are determined based on the third target weight.

[0079] Specifically, the first initial weight corresponding to the data frame sequence, the second initial weight corresponding to the topological feature vector, and the fourth initial weight corresponding to the target lateral response type are obtained. The first initial weight, the second initial weight, and the fourth initial weight are summed to obtain the first weight sum value.

[0080] A first ratio is determined based on the first initial weight and the sum of the first weights; a second ratio is determined based on the second initial weight and the sum of the first weights; and a fourth ratio is determined based on the fourth initial weight and the sum of the first weights. The difference between the third target weight and the third initial weight is calculated to obtain the first weight difference.

[0081] The first ratio is multiplied by the first weight difference to obtain a first weight change. This first weight change is then summed with the first initial weight to obtain a first target weight. The second ratio is multiplied by the first weight difference to obtain a second weight change. This second weight change is then summed with the first initial weight to obtain a second target weight. Finally, the fourth ratio is multiplied by the first weight difference to obtain a fourth weight change. This fourth weight change is then summed with the first initial weight to obtain a fourth target weight.

[0082] In some embodiments, the target fusion weights include a first target weight corresponding to the data frame sequence, a second target weight corresponding to the topological feature vector, a third target weight corresponding to the wind speed gradient evolution sequence, and a fourth target weight corresponding to the target lateral response type. Step 10311, after determining the target wind disturbance intensity based on the environmental data, specifically further includes: Step 1031A: In response to the target wind disturbance intensity being less than a preset intensity threshold, the third target weight corresponding to the wind speed gradient evolution sequence is determined to be a preset minimum weight threshold. Step 1031B: Determine the fourth target weight corresponding to the target lateral response type based on the target wind disturbance intensity; Step 1031C: Determine the first target weight corresponding to the data frame sequence and the second target weight corresponding to the topological feature vector based on the third target weight and the fourth target weight.

[0083] In practice, the target wind disturbance intensity is compared with a preset intensity threshold. If the target wind disturbance intensity is determined to be less than the preset intensity threshold, it indicates that the wind disturbance intensity is low, i.e., the wind disturbance is evolving stably. At this point, a preset minimum weight threshold is obtained, and the third target weight corresponding to the wind speed gradient evolution sequence is determined as the minimum weight threshold.

[0084] Based on the target wind disturbance intensity, the fourth target weight corresponding to the target lateral response type is determined. That is, the database is searched based on the target wind disturbance intensity to determine the fourth target weight corresponding to the target wind disturbance intensity. The database stores the correspondence between the target wind disturbance intensity and the fourth target weight.

[0085] It is understandable that the dictionary atomic matching weights are increased to enhance control stability during the stable evolution of wind disturbance. Therefore, the fourth target weight corresponding to the target lateral response type is greater than the fourth initial weight corresponding to the target lateral response type.

[0086] After determining the third target weight and the fourth target weight, the first target weight corresponding to the data frame sequence and the second target weight corresponding to the topological feature vector are determined based on the third target weight and the fourth target weight.

[0087] Specifically, the difference between the third target weight and the third initial weight is calculated to obtain the third weight difference. The difference between the fourth target weight and the fourth initial weight is calculated to obtain the fourth weight difference. The third weight difference and the fourth weight difference are then summed to obtain the target weight difference.

[0088] The first initial weight and the second initial weight are summed to obtain the target weight sum. The ratio of the first initial weight to the target weight sum is calculated to obtain the fifth ratio. The ratio of the second initial weight to the target weight sum is calculated to obtain the sixth ratio.

[0089] The fifth ratio is multiplied by the difference between the target weights to obtain the fifth weight change. This fifth weight change is then summed with the first initial weight to obtain the first target weight. The sixth ratio is multiplied by the difference between the target weights to obtain the sixth weight change. This sixth weight change is then summed with the second initial weight to obtain the second target weight.

[0090] In some embodiments, step 1031, which involves fusing the data frame sequence, the topological feature vector, the wind speed gradient evolution sequence, and the target lateral response type based on the target fusion weight to obtain a target fusion vector, specifically includes: Step 1031a: The data frame sequence, the topological feature vector, the wind speed gradient evolution sequence, and the target lateral response type are mapped and aligned to obtain the data embedding vector, the topological embedding vector, the wind speed gradient embedding vector, and the response type embedding vector. Step 1031b: Input the data embedding vector, the topology embedding vector, the wind speed gradient embedding vector, the response type embedding vector, and the target fusion weight into a pre-trained vector fusion model, and output the target fusion vector through the vector fusion model.

[0091] In specific implementation, the data frame sequence, the topological feature vector, the wind speed gradient evolution sequence, and the target lateral response type are mapped and aligned to obtain the data embedding vector, the topological embedding vector, the wind speed gradient embedding vector, and the response type embedding vector.

[0092] Specifically, a linear transformation is performed on the data frame sequence to map it to a preset semantic space, resulting in a data embedding vector. A linear transformation is also performed on the topological feature vector to map it to the preset semantic space, resulting in a topological embedding vector. A linear transformation is performed on the wind speed gradient evolution sequence to map it to the preset semantic space, resulting in a wind speed gradient embedding vector. Finally, a linear transformation is performed on the target lateral response type to map it to the preset semantic space, resulting in a response type embedding vector.

[0093] The aligned data embedding vector, topological embedding vector, wind speed gradient embedding vector, and response type embedding vector are further fused using a multi-head attention mechanism, enabling the model to identify deep relationships and causal connections among them. Specifically, the data embedding vector, topological embedding vector, wind speed gradient embedding vector, response type embedding vector, and target fusion weights are input into a pre-trained vector fusion model, which processes the data and outputs the target fusion vector.

[0094] In this embodiment, a lightweight graph structure generation network is introduced to construct a spatiotemporal dependency graph between models in the embedding space. The current vehicle position, predicted trajectory, and historical response patterns are combined into directed graph nodes, and edge weight training reflects the information flow path between models. Through graph update mechanism and embedding weight adjustment mechanism, the system dynamically generates the optimal information fusion path in each control cycle, realizing deep semantic fusion of the three models, and thus obtaining the target fusion vector.

[0095] The training process of the vector fusion model specifically includes: Obtain a third training dataset and an initial vector fusion model. The third training dataset includes historical data frame sequences, historical topological feature vectors, historical wind speed gradient evolution sequences, historical lateral response types, and historical fusion vectors. Input the training data from the third training dataset into the initial vector fusion model for training. Determine if a third preset training termination condition is met to obtain the vector fusion model.

[0096] The third preset training termination condition includes at least one of the following: determining that all data in the third training dataset has been input into the initial vector fusion model for training, determining that the loss function of the initial vector fusion model has converged to the third convergence threshold, or determining that the initial vector fusion model has been iteratively trained to the third preset number of iterations.

[0097] For example, the third preset training termination condition is to determine that all data in the third training dataset has been input into the initial vector fusion model for training: The third training dataset contains fifty sets of data. Each set includes a historical data frame sequence, a historical topological feature vector, a historical wind speed gradient evolution sequence, a historical lateral response type, and a historical fusion vector. The third preset training termination condition is determined when all data in the third training dataset has been input into the initial vector fusion model for training. That is, when all fifty sets of data have been input into the initial vector fusion model, there is no training data in the third training dataset that has not yet been input into the initial vector fusion model. At this point, the initial vector fusion model training is considered complete, and the vector fusion model is obtained.

[0098] In another example, the third preset training termination condition is that the loss function of the initial vector fusion model converges to the third convergence threshold: The training data from the third training dataset is input into the initial vector fusion model for training, and the training results are output. A loss function is determined based on the training results and the historical fused vectors. 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 third convergence threshold, the third preset training termination condition is satisfied, and the vector fusion model is obtained.

[0099] In another example, the third preset training termination condition is to determine the initial vector fusion model iteratively train to the third preset number of iterations.

[0100] The training data in the third training dataset is input into the initial vector fusion model for iterative training. The number of iterations is recorded. When the number of iterations is equal to the third preset number of iterations, the third preset training termination condition is met, and the vector fusion model is obtained.

[0101] The above scheme fuses data embedding vectors, topology embedding vectors, wind speed gradient embedding vectors, and response type embedding vectors to obtain a target fusion vector. This allows the subsequent vehicle compensation control model to automatically identify the correlation between vectors obtained from different models when analyzing and processing the target fusion vector, thereby improving the accuracy of determining the target compensation control command.

[0102] 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.

[0103] 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.

[0104] Based on the same inventive concept, corresponding to any of the above-described embodiments, this disclosure also provides a vehicle control device.

[0105] refer to Figure 2 , Figure 2 The vehicle control device, as described in this embodiment, includes: The data acquisition module 201 is configured to acquire vehicle operation information and environmental data of the vehicle's environment, determine a data frame sequence based on the vehicle operation information and the environmental data, and determine a wind speed gradient evolution sequence based on the data frame sequence. The type determination module 202 is configured to acquire vehicle system state parameters and determine the target lateral response type based on the system state parameters and the wind speed gradient evolution sequence. The vehicle control module 203 is configured to acquire a pre-trained vehicle control model, input the data frame sequence, the wind speed gradient evolution sequence, and the target lateral response type into the vehicle compensation control model, process the data according to the vehicle compensation control model to obtain a target compensation control command, and control the vehicle to drive according to the target compensation control command.

[0106] In some embodiments, the data acquisition module 201 is specifically configured as follows: The data frame sequence is mapped to obtain a topological feature vector; The data frame sequence and the topological feature vector are input into a pre-trained wind speed gradient diffusion model, and the wind speed gradient evolution sequence is obtained through processing by the wind speed gradient diffusion model.

[0107] In some embodiments, the data acquisition module 201 is specifically configured as follows: The data frame sequence is subjected to sliding embedding processing to obtain high-dimensional point cloud data; Based on a preset algorithm, topological features are extracted from the high-dimensional point cloud data to obtain the target topological structure; The target topology is quantized to obtain a topological feature vector.

[0108] In some embodiments, the type determination module 202 is specifically configured to: Obtain the pre-trained horizontal type recognition model; The topological feature vector, the system state parameters, and the wind speed gradient evolution sequence are input into the lateral type identification model. After processing by the lateral type identification model, the target lateral response type is obtained.

[0109] In some embodiments, the vehicle control module 203 is specifically configured as follows: Determine the target fusion weight, and perform fusion processing on the data frame sequence, the topological feature vector, the wind speed gradient evolution sequence, and the target lateral response type based on the target fusion weight to obtain the target fusion vector; The target fusion vector is input into the vehicle compensation control model, and the target compensation control command is obtained by processing the vehicle compensation control model.

[0110] In some embodiments, the target fusion weights include a first target weight corresponding to the data frame sequence, a second target weight corresponding to the topological feature vector, a third target weight corresponding to the wind speed gradient evolution sequence, and a fourth target weight corresponding to the target lateral response type. The vehicle control module 203 is specifically configured as follows: The target wind disturbance intensity is determined based on the environmental data. In response to the target wind disturbance intensity being greater than or equal to a preset intensity threshold, the third target weight corresponding to the wind speed gradient evolution sequence is determined based on the target wind disturbance intensity; The first target weight corresponding to the data frame sequence, the second target weight corresponding to the topological feature vector, and the fourth target weight corresponding to the target lateral response type are determined based on the third target weight.

[0111] In some embodiments, the vehicle control module 203 is specifically configured as follows: In response to the target wind disturbance intensity being less than a preset intensity threshold, the third target weight corresponding to the wind speed gradient evolution sequence is determined to be a preset minimum weight threshold; Based on the target wind disturbance intensity, determine the fourth target weight corresponding to the target lateral response type; Based on the third target weight and the fourth target weight, the first target weight corresponding to the data frame sequence and the second target weight corresponding to the topological feature vector are determined.

[0112] In some embodiments, the vehicle control module 203 is specifically configured as follows: The data frame sequence, the topological feature vector, the wind speed gradient evolution sequence, and the target lateral response type are mapped and aligned to obtain the data embedding vector, the topological embedding vector, the wind speed gradient embedding vector, and the response type embedding vector. The data embedding vector, the topology embedding vector, the wind speed gradient embedding vector, the response type embedding vector, and the target fusion weight are input into a pre-trained vector fusion model. The target fusion vector is then output after processing by the vector fusion model.

[0113] In some embodiments, the apparatus further includes a model training module, which is specifically configured to: Obtain the third training dataset and the initial vector fusion model, wherein the third training dataset includes historical data frame sequences, historical topological feature vectors, historical wind speed gradient evolution sequences, historical lateral response types, and historical fusion vectors; The training data in the third training dataset is input into the initial vector fusion model for training. Once the third preset training termination condition is met, the vector fusion model is obtained.

[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: Acquire vehicle operation information and environmental data of the vehicle's environment, determine a data frame sequence based on the vehicle operation information and the environmental data, and determine a wind speed gradient evolution sequence based on the data frame sequence. Obtain vehicle system state parameters, and determine the target lateral response type based on the system state parameters and the wind speed gradient evolution sequence; A pre-trained vehicle control model is obtained, and the data frame sequence, the wind speed gradient evolution sequence, and the target lateral response type are input into the vehicle compensation control model. The vehicle compensation control model is processed to obtain a target compensation control command, and the vehicle is controlled to drive according to the target compensation control command.

2. The method according to claim 1, characterized in that, Determining the wind speed gradient evolution sequence based on the data frame sequence includes: The data frame sequence is mapped to obtain a topological feature vector; The data frame sequence and the topological feature vector are input into a pre-trained wind speed gradient diffusion model, and the wind speed gradient evolution sequence is obtained through processing by the wind speed gradient diffusion model.

3. The method according to claim 2, characterized in that, The mapping process of the data frame sequence to obtain the topological feature vector includes: The data frame sequence is subjected to sliding embedding processing to obtain high-dimensional point cloud data; Based on a preset algorithm, topological features are extracted from the high-dimensional point cloud data to obtain the target topological structure; The target topology is quantized to obtain a topological feature vector.

4. The method according to claim 2, characterized in that, Determining the target lateral response type based on the system state parameters and the wind speed gradient evolution sequence includes: Obtain the pre-trained horizontal type recognition model; The topological feature vector, the system state parameters, and the wind speed gradient evolution sequence are input into the lateral type identification model. After processing by the lateral type identification model, the target lateral response type is obtained.

5. The method according to claim 1, characterized in that, The step involves inputting the data frame sequence, the wind speed gradient evolution sequence, and the target lateral response type into the vehicle compensation control model, and processing them according to the vehicle compensation control model to obtain the target compensation control command, including: Determine the target fusion weight, and perform fusion processing on the data frame sequence, the topological feature vector, the wind speed gradient evolution sequence, and the target lateral response type based on the target fusion weight to obtain the target fusion vector; The target fusion vector is input into the vehicle compensation control model, and the target compensation control command is obtained by processing the vehicle compensation control model.

6. The method according to claim 5, characterized in that, The target fusion weights include a first target weight corresponding to the data frame sequence, a second target weight corresponding to the topological feature vector, a third target weight corresponding to the wind speed gradient evolution sequence, and a fourth target weight corresponding to the target lateral response type. The determination of the target fusion weights includes: The target wind disturbance intensity is determined based on the environmental data. In response to the target wind disturbance intensity being greater than or equal to a preset intensity threshold, the third target weight corresponding to the wind speed gradient evolution sequence is determined based on the target wind disturbance intensity; The first target weight corresponding to the data frame sequence, the second target weight corresponding to the topological feature vector, and the fourth target weight corresponding to the target lateral response type are determined based on the third target weight.

7. The method according to claim 6, characterized in that, After determining the target wind disturbance intensity based on the environmental data, the process also includes: In response to the target wind disturbance intensity being less than a preset intensity threshold, the third target weight corresponding to the wind speed gradient evolution sequence is determined to be a preset minimum weight threshold; Based on the target wind disturbance intensity, determine the fourth target weight corresponding to the target lateral response type; Based on the third target weight and the fourth target weight, the first target weight corresponding to the data frame sequence and the second target weight corresponding to the topological feature vector are determined.

8. The method according to claim 5, characterized in that, The process of fusing the data frame sequence, the topological feature vector, the wind speed gradient evolution sequence, and the target lateral response type based on the target fusion weight to obtain the target fusion vector includes: The data frame sequence, the topological feature vector, the wind speed gradient evolution sequence, and the target lateral response type are mapped and aligned to obtain the data embedding vector, the topological embedding vector, the wind speed gradient embedding vector, and the response type embedding vector. The data embedding vector, the topology embedding vector, the wind speed gradient embedding vector, the response type embedding vector, and the target fusion weight are input into a pre-trained vector fusion model. The target fusion vector is then output after processing by the vector fusion model.

9. The method according to claim 8, characterized in that, The training process of the vector fusion model includes: Obtain the third training dataset and the initial vector fusion model, wherein the third training dataset includes historical data frame sequences, historical topological feature vectors, historical wind speed gradient evolution sequences, historical lateral response types, and historical fusion vectors; The training data in the third training dataset is input into the initial vector fusion model for training. Once the third preset training termination condition is met, the vector fusion model is obtained.

10. A vehicle, characterized in that, include: Memory, used to store executable programs; processor; When the executable program is executed by the processor, the method as described in any one of claims 1-9 is implemented.

Images