Vehicle control method and device based on driving style recognition, equipment and medium

Abstract

The application discloses a vehicle control method and device based on driving style recognition, equipment and medium, the method comprises: obtaining real-time driving data in the vehicle driving process;Real-time driving data is input into the pre-trained driving style recognition model, and the driving style of the driver is obtained, wherein the driving style recognition model is obtained according to the original driving data of the driver with different driving styles under different preset working conditions;The driving style of the driver is used to update the vehicle dynamics parameters, and the updated vehicle dynamics parameters are obtained;According to the vehicle updated dynamics parameters, the target driving torque of each wheel of the vehicle is calculated, and the vehicle is controlled based on the target driving torque, which realizes the dynamic matching of the vehicle dynamics response characteristics and the driving style of the driver, effectively improves the coordination of man and vehicle, driving experience and vehicle control stability.

Description

Technical Field

[0001] This application relates to the field of vehicle control, specifically to a vehicle control method, device, equipment, and medium based on driving style recognition. Background Technology

[0002] Driving style, as a key characteristic representing a driver's operating habits, risk preferences, and expected vehicle response, is a crucial input variable for achieving personalized vehicle control. Accurately identifying driving style and adjusting vehicle control parameters accordingly is of great significance for improving vehicle handling stability, ride comfort, and driving safety.

[0003] Existing technologies are mainly divided into two categories: rule-based thresholding and data-driven approaches. Rule-based thresholding methods determine style by setting fixed thresholds for physical quantities, but they struggle to adapt to different vehicle models and road conditions, exhibiting poor generalization ability and a high false positive rate. Data-driven approaches, while incorporating neural networks, still have drawbacks: firstly, the training data lacks effective filtering, containing a large amount of invalid operating condition noise, reducing model robustness; secondly, feature extraction often focuses only on driver input, ignoring vehicle dynamics response, failing to accurately reflect the closed-loop characteristics between the driver and vehicle; and thirdly, the recognition results are disconnected from the underlying control parameters, making true adaptive control difficult to achieve.

[0004] Therefore, how to accurately identify the driver's driving style in order to achieve precise adaptive adjustment of vehicle control parameters is a technical problem that urgently needs to be solved in the field of vehicle control. Summary of the Invention

[0005] This application provides a vehicle control method, device, equipment, and medium based on driving style recognition, which can solve the technical problems existing in the prior art where the control parameters of vehicle driving assistance systems are fixed and cannot be dynamically and adaptively adjusted according to the driver's personalized driving style, resulting in a mismatch between vehicle dynamic response characteristics and driver's operating habits, poor driving experience, and insufficient human-vehicle coordination.

[0006] In a first aspect, embodiments of this application provide a vehicle control method based on driving style recognition, the vehicle control method based on driving style recognition including: Acquire real-time driving data during vehicle operation; The real-time driving data is input into a pre-trained driving style recognition model to obtain the driver's driving style. The driving style recognition model is trained based on the original driving data of multiple drivers with different driving styles under different preset conditions. The vehicle dynamics parameters are updated based on the driver's driving style to obtain updated vehicle dynamics parameters; The target driving torque for each wheel of the vehicle is calculated based on the updated vehicle dynamics parameters, and the vehicle is controlled based on the target driving torque.

[0007] Secondly, embodiments of this application provide a vehicle control device based on driving style recognition, the vehicle control device based on driving style recognition comprising: The acquisition module is used to acquire real-time driving data during vehicle operation; The recognition module is used to input the real-time driving data into a pre-trained driving style recognition model to obtain the driver's driving style. The driving style recognition model is trained based on the original driving data of multiple drivers with different driving styles under different preset working conditions. An update module is used to update the vehicle dynamics parameters according to the driver's driving style to obtain updated vehicle dynamics parameters. The control module is used to calculate the target driving torque of each wheel of the vehicle based on the updated vehicle dynamics parameters, and to control the vehicle based on the target driving torque.

[0008] Thirdly, embodiments of this application provide a vehicle control device based on driving style recognition, including a processor, a memory, and a vehicle control program based on driving style recognition stored in the memory and executable by the processor, wherein when the vehicle control program based on driving style recognition is executed by the processor, it implements the steps of the vehicle control method based on driving style recognition as described in any of the preceding claims.

[0009] Fourthly, embodiments of this application provide a vehicle control program based on driving style recognition stored on a computer-readable storage medium, wherein when the vehicle control program based on driving style recognition is executed by a processor, it implements the steps of the vehicle control method based on driving style recognition as described in any of the preceding claims.

[0010] The vehicle control method, apparatus, device, and medium based on driving style recognition provided in this application acquire real-time driving data during vehicle operation; input the real-time driving data into a pre-trained driving style recognition model to obtain the driver's driving style, wherein the driving style recognition model is trained based on the original driving data of multiple drivers with different driving styles under different preset working conditions; update the vehicle dynamic parameters according to the driver's driving style to obtain updated vehicle dynamic parameters; calculate the target driving torque of each wheel of the vehicle based on the updated vehicle dynamic parameters, and control the vehicle based on the target driving torque. This solves the technical problem in related technologies where the dynamic control parameters of vehicle driving assistance systems are fixed and cannot be adaptively adjusted according to the driver's personalized driving style, resulting in a mismatch between vehicle response characteristics and driver's operating habits, poor driving experience, and insufficient human-vehicle coordination. It achieves dynamic matching between vehicle dynamic response characteristics and driver's driving style, enabling vehicle control to actively adapt to the personalized operating needs of different drivers, effectively improving human-vehicle coordination, driving experience, and vehicle handling stability. Attached Figure Description

[0011] Figure 1 This is a flowchart illustrating an embodiment of the vehicle control method based on driving style recognition according to this application; Figure 2 This is a schematic diagram of the functional modules of a vehicle control device based on driving style recognition according to an embodiment of this application; Figure 3 This is a schematic diagram of the hardware structure of a vehicle control device based on driving style recognition involved in the embodiments of this application. Detailed Implementation

[0012] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0013] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0014] In a first aspect, embodiments of this application provide a vehicle control method based on driving style recognition.

[0015] In one embodiment, reference is made to Figure 1 , Figure 1This is a flowchart illustrating the first embodiment of the vehicle control method based on driving style recognition according to this application. Figure 1 As shown, the vehicle control method based on driving style recognition includes: Step S1: Obtain real-time driving data during vehicle operation.

[0016] The driving data in this embodiment includes steering wheel angle, yaw rate, longitudinal vehicle speed, lateral vehicle speed, and road surface adhesion coefficient. Real-time driving data refers to driving data collected in real time by sensors according to a preset sampling period during vehicle operation, used for current driving style identification.

[0017] Step S2: Input the real-time driving data into the pre-trained driving style recognition model to obtain the driver's driving style. The driving style recognition model is trained based on the original driving data of multiple drivers with different driving styles under different preset conditions.

[0018] In one embodiment, training a driving style recognition model includes the following steps: Step S201: Collect raw driving data from multiple drivers with different driving styles under typical steering conditions, different vehicle speeds, and different road surface adhesion coefficients to form a raw dataset.

[0019] Specifically, multiple typical steering conditions are pre-designed, including at least: single lane change, double lane change, and curves with different radii of curvature (e.g., fixed-radius curves and variable-radius curves). Simultaneously, multiple combinations of driving parameters are set for each typical steering condition. These driving parameters include at least: multiple vehicle speed ranges (e.g., 30-80 km / h, with speed intervals of 10 km / h) and road adhesion coefficients corresponding to different road surface adhesion conditions (e.g., dry asphalt pavement, wet and slippery pavement, low-adhesion pavement, etc.).

[0020] Multiple drivers with different driving habits and driving styles were selected and categorized into at least three driving styles based on questionnaires, driving experience, and subjective evaluations: conservative drivers, balanced drivers, and aggressive drivers. Drivers with different driving styles were stratified by at least one of the following methods: driving experience, age group, and gender, with a balanced distribution of sample sizes across each driving style.

[0021] In a simulator or real-vehicle environment, drivers with different driving styles complete preset typical steering condition tests, and driving data during the experiment is collected and recorded as raw driving data. Raw driving data includes at least: steering input data, vehicle motion state data, and road surface adhesion state data. Steering input data includes at least the steering wheel angle; vehicle motion state data includes at least yaw rate, longitudinal vehicle speed, lateral vehicle speed, and lateral acceleration; and road surface adhesion state data includes the road surface adhesion coefficient.

[0022] The collected raw driving data is associated and labeled with corresponding driver style tags and typical steering conditions to build a personalized driving database with multiple conditions and driving style tags.

[0023] Specifically, driving style labels can be generated through at least one of the following methods: obtaining subjective style evaluations through driver questionnaires; obtaining objective style evaluations through expert scoring of driving behavior data; and obtaining automatic style labels based on the comparison of driving behavior indicators with preset style thresholds.

[0024] Step S202: Select valid steering segments from the original dataset, extract the driving style feature parameters of each valid steering segment, and concatenate them to generate the corresponding style feature vector.

[0025] The constructed raw driving dataset undergoes preprocessing, including outlier removal, noise filtering, and time synchronization and alignment. Based on signals such as steering wheel angle and yaw rate, valid steering segments that meet preset conditions are identified and filtered out, while non-steering or low-information data are removed. The criteria for determining valid steering segments include: the rate of change of steering wheel angle is greater than a preset rate of change threshold, the steering duration is within a preset time range, and the lateral acceleration amplitude falls within a preset normal handling range.

[0026] Driving style feature parameters that characterize the driver's handling characteristics are extracted from each effective steering segment. In this embodiment, the driving style feature parameters include at least: steering entropy, peak steering wheel angular velocity, steering return delay time, yaw rate overshoot, and response settling time.

[0027] Among these, driving style characteristic parameters can be calculated from driving data. Steering entropy characterizes the complexity and smoothness of driver operation; conservative drivers typically have lower steering entropy, while aggressive drivers typically have higher steering entropy. Peak steering wheel angular velocity characterizes the speed at which the driver turns the steering wheel. Steering return-to-center delay time refers to the time interval from when the steering wheel angle reaches its peak to when it begins to return to center (change of sign of the angular velocity), characterizing the driver's perception and coordination with vehicle dynamics. Yaw rate overshoot characterizes the degree to which the vehicle's response exceeds the driver's input. Response settling time is the time required from the start of steering input for the yaw rate to reach its steady-state value, characterizing the speed of the vehicle's response.

[0028] Step S203: Associate the corresponding style feature vectors and driving style labels to form a labeled dataset.

[0029] The driving style feature parameters are normalized in terms of dimensions and outlier truncation is performed. They are then concatenated in a preset dimensional order to generate a style feature vector, which is then associated with the corresponding driving style label to form a low-dimensional driving feature dataset for model training, denoted as the labeled dataset.

[0030] Step S204: Using the style feature vectors in the labeled dataset as input and the corresponding driving style labels as output, train the initial driving style recognition model until the network converges to obtain the pre-trained driving style recognition model.

[0031] For example, this embodiment selects a back propagation neural network (BP) as the driving style recognition model. The BP neural network can learn and store a large number of input-output pattern mappings. Using the steepest descent method, it continuously adjusts the network's weights and thresholds through backpropagation to minimize the sum of squared errors. The construction process of the BP neural network includes: (1) Determine the number of nodes n in the input layer, q in the hidden layer, and m in the output layer based on the input-output sequence after dimensionality reduction. Then initialize the connection weights between neurons in the input layer and hidden layer, and between neurons in the output layer and hidden layer. Finally, initialize the biases of the hidden layer and the output layer, and determine the learning rate and the nonlinear activation function of the neurons. Among them, the number of nodes in the input layer corresponds to the dimension of the style feature vector (e.g., 5 dimensions, corresponding to 5 types of feature parameters), and the number of nodes in the output layer corresponds to the number of driving style categories (e.g., 3 categories: conservative, balanced, and aggressive).

[0032] (2) Hidden layer output calculation: Based on the input vector x, the connection weights V between the input layer and the hidden layer, and the hidden layer bias a, calculate the hidden layer output. :

[0033] in, is the nonlinear activation function of the hidden layer.

[0034] (3) Calculate the network prediction output based on the hidden layer output. Connect the weights W and the output layer bias b, and calculate the network's predicted output O:

[0035] (4) Calculation of network prediction error: Based on the network prediction output O and the network expected output Y, calculate the network prediction error e:

[0036] (5) Update connection weights: Update network connection weights V and W based on network prediction error e.

[0037] Where η is a positive learning rate.

[0038] (6) Bias update: Update the hidden layer bias and output layer bias according to the network prediction error.

[0039] (7) Calculation of network average error: Calculate the network average error based on the network prediction error.

[0040] The training iteration determines whether the algorithm iteration process can end based on the average network error calculated in step (7). If the average network error is less than a preset threshold (e.g., 0.001) or the maximum number of iterations (e.g., 1000) is reached, the training ends and the improved driving style recognition model is obtained; otherwise, the process returns to step (2) to continue iterating.

[0041] In one embodiment, a mapping relationship between driving style and vehicle dynamic parameters can be constructed based on the correspondence between driving style labels and corresponding calibrated vehicle target dynamic parameters in the labeled dataset, and the mapping relationship can be stored as a style-parameter mapping table.

[0042] Each driving style label is associated with a vehicle target dynamics parameter. The vehicle target dynamics parameter is obtained based on the vehicle test data corresponding to each driving style category. Specifically, it is obtained by statistically analyzing the average or median of the vehicle dynamics parameters under each style category, which is then used as the vehicle target dynamics parameter for that driving style category.

[0043] In this embodiment, the vehicle dynamics parameters include the understeer gradient, the natural frequency of the desired yaw rate system, and the damping ratio of the desired yaw rate system.

[0044] Furthermore, after the driving style recognition model is trained, the live model can be used to identify the current driver's driving style. Step S2 involves inputting the real-time driving data into the pre-trained driving style recognition model to obtain the driver's driving style, specifically including the following steps: Step S211: Perform time alignment processing on the real-time driving data and generate a driving data sequence with a unified time reference.

[0045] Specifically, the real-time driving data contains the same parameters as the original driving data mentioned above, including steering wheel angle, yaw rate, longitudinal vehicle speed, lateral vehicle speed, and road adhesion coefficient. This embodiment can employ an interpolation algorithm based on hardware timestamps to align the real-time driving data. A hardware timestamp is appended to each real-time driving data point. Using a preset unified time reference as the target sequence, the interpolation algorithm resamples data with different sampling frequencies to the unified time reference, resulting in time-aligned real-time driving data. The time-aligned real-time driving data is then low-pass filtered and outlier removed to finally generate a real-time driving data sequence with a unified time reference, which is used for subsequent effective steering segment extraction and feature extraction.

[0046] This embodiment ensures the accurate correspondence of real-time driving data on the timeline through the above processing, thus guaranteeing the accuracy of subsequent calculations of driving style characteristic parameters.

[0047] Step S212: Extract valid steering segments from the driving data sequence, extract the driving style feature parameters of each valid steering segment, and concatenate them to generate the corresponding style feature vector.

[0048] Specifically, the validity of the driving data sequence is first determined, including signal integrity and steering behavior assessment. Signal integrity assessment detects data frame loss rate, signal-to-noise ratio, and sensor fault indicators. Steering behavior assessment is based on whether the steering wheel angle change rate, steering duration, and lateral acceleration amplitude fall within a preset normal operating range.

[0049] When the effective steering conditions are met, continuous data segments are extracted from the driving data sequence along the time axis to obtain effective steering segments. Multiple extracted effective steering segments are then stored in a buffer in chronological order to form a set of effective steering segments. When the effective steering conditions are not met, a silence flag is output, the control flow enters a silence state, the control parameters of the previous cycle are maintained, and subsequent learning processes are paused.

[0050] For each valid steering segment in the set of valid steering segments, driving style feature parameters that characterize the driver's handling characteristics are extracted online. These driving style feature parameters include at least steering maneuver entropy, peak steering wheel angular velocity, steering return delay time, yaw rate overshoot, and response settling time. The extracted feature parameters are then subjected to dimensional normalization and outlier truncation. The normalized parameters are concatenated according to a preset dimensional order to obtain the corresponding style feature vector.

[0051] For example, the dimensional normalization in this embodiment can be achieved using the Z-Score standardization method, which converts each feature parameter into standard normal distribution data with a mean of 0 and a standard deviation of 1. Outlier truncation is used to truncate feature values ​​that exceed a preset range (e.g., ±3 times the standard deviation) to the boundary value.

[0052] Through the above processing, the generated style feature vector has a unified dimension and size, which makes it easy to input into the driving style recognition model for online recognition.

[0053] Step S213: Input the style feature vector into the pre-trained driving style recognition model to obtain the single driving style recognition result and recognition confidence level corresponding to each effective steering segment.

[0054] Specifically, the single driving style recognition result includes conservative, balanced, and aggressive styles. The pre-trained driving style recognition model outputs a recognition confidence score for each driving style category, and the style category with the highest recognition confidence score is taken as the current single style recognition result.

[0055] Step S214: Perform fusion processing on the single driving style recognition results corresponding to multiple consecutive valid steering segments within the preset sliding window to obtain the fused driving style result.

[0056] Specifically, a valid flag bit is generated for the corresponding single driving style recognition result based on the recognition confidence level. The valid flag bit indicates whether the confidence level of the current single recognition result has reached the output threshold. The valid flag bit is valid when the highest confidence score is greater than or equal to a preset confidence threshold; it is invalid when the highest confidence score is less than the preset confidence threshold. For example, the output threshold can be set to 0.7.

[0057] Furthermore, a style sliding window is constructed to statistically analyze multiple consecutive valid steering segments. Weighted majority voting or state smoothing fusion is performed on the single style recognition results within the window that satisfy the valid flag, resulting in a fused style result. The fusion algorithm assigns a higher time decay weight to recent recognition results to strengthen the representation of the current driving intention.

[0058] For example, the time decay weight can be set according to an exponential decay function, so that the weight of the most recent 1-2 recognition results within the sliding window exceeds 50%. Through weighted majority voting, the random errors of a single recognition can be effectively filtered out, improving the stability of driving style recognition.

[0059] Step S215: Determine the driver's driving style based on the fused driving style results.

[0060] Specifically, when the category consistency ratio or overall confidence score of the fusion results within the sliding window continuously reaches a preset convergence threshold, a style convergence flag is set to suppress style recognition jumps caused by short-term road condition disturbances or accidental operations. The category consistency ratio refers to the proportion of the same driving style category in multiple single style recognition results within the sliding window. The overall confidence score refers to the weighted average of the model output confidence scores of multiple single style recognition results within the sliding window.

[0061] For example, the preset convergence threshold can be set as follows: when the category consistency ratio is ≥80% or the overall confidence level is ≥0.85, and the above threshold is reached for 3 consecutive control cycles, the style convergence flag is set to the convergence state. When the style convergence flag is in the convergence state, the fused driving style result is used as the current driver's driving style; when the style convergence flag is not in the convergence state, the driving style of the previous cycle remains unchanged.

[0062] Step S3: Update the vehicle dynamics parameters according to the driver's driving style to obtain updated vehicle dynamics parameters.

[0063] In one embodiment, step S3 specifically includes the following steps: Step S301: Query the pre-established style-parameter mapping table according to the current driver's driving style to obtain the corresponding vehicle target dynamic parameters.

[0064] For example, the current driver's driving style is standardized and encoded into a numerical driving style code value. Based on the driving style code, the corresponding vehicle target control parameters are read from a pre-established style-parameter mapping table. The vehicle target control parameters include at least: understeer gradient, natural frequency of the desired yaw rate system, and damping ratio of the desired yaw rate system.

[0065] This embodiment can match corresponding vehicle dynamics parameters to drivers with different driving styles by querying the style-parameter mapping table, so that conservative drivers can obtain more stable vehicle dynamic characteristics and aggressive drivers can obtain more sensitive steering response.

[0066] Step S302: Perform physical constraints and safety boundary judgments on the vehicle target dynamic parameters to obtain the verified vehicle target dynamic parameters.

[0067] For example, physical constraints include limiting the range of values ​​for each vehicle target dynamic parameter. For instance, reasonable ranges are set for the understeer gradient, the natural frequency of the desired yaw rate system, and the damping ratio of the desired yaw rate system. When the retrieved vehicle target parameter exceeds the corresponding range, it is truncated to the boundary value. Safety boundary judgment includes checking the damping ratio within a stable range to ensure that the vehicle yaw response has sufficient stability margin and avoid the risk of vehicle oscillation due to an excessively low damping ratio. After the above physical constraints and safety boundary judgment, the verified target dynamic parameters are obtained and used as input for subsequent incremental smooth update calculations, ensuring the safety of parameter updates.

[0068] Step S303: Calculate the difference between the verified vehicle target dynamic parameters and the currently effective vehicle dynamic parameters to obtain the parameter increment.

[0069] Step S304: Smooth the parameter increment based on the rate of change limit and smoothing coefficient to obtain the processed parameter increment.

[0070] In this embodiment, the smoothing coefficient ranges from 0.1 to 0.3 (i.e., the smoothing coefficient α∈[0.1,0.3]). By introducing a rate-of-change constraint and a smoothing coefficient, the updated control parameter set and the corresponding parameter update flag are calculated, achieving a gradual evolution of the control objective rather than a step replacement.

[0071] Step S305: The currently effective vehicle dynamics parameters are summed with the processed parameter increments to obtain the updated vehicle dynamics parameters, which are used as the effective parameters for the current control cycle. Based on the updated vehicle dynamics parameters, a desired yaw rate transfer function model is constructed as the reference control target for the torque vector controller.

[0072] Step S4: Calculate the target driving torque of each wheel of the vehicle based on the updated vehicle dynamics parameters, and control the vehicle based on the target driving torque.

[0073] Specifically, in this embodiment, a desired yaw rate trajectory can be generated based on the updated dynamic parameters, and the target driving torque of each wheel of the vehicle can be calculated based on the desired yaw rate trajectory.

[0074] In one embodiment, step S4 specifically includes: configuring the model parameters of the vehicle's desired yaw rate transfer function model based on the updated vehicle dynamics parameters; inputting the steering wheel angle into the desired yaw rate transfer function model to obtain the desired yaw rate trajectory; and calculating the target drive torque distribution ratio of each drive wheel based on the updated vehicle dynamics parameters and the desired yaw rate trajectory, combined with the real-time driving data, to obtain the target drive torque of each wheel.

[0075] For example, when the vehicle update control parameter set is calculated, a corresponding parameter update flag is generated. When the parameter update flag is valid, the vehicle update control parameter set is sent to the vehicle torque vector controller via the on-board high-speed network. Based on the received vehicle update control parameter set, the torque vector controller, combined with the real-time vehicle dynamics state and target trajectory, calculates the target drive torque distribution ratio of each drive wheel and outputs torque commands to the underlying actuators to complete differential torque control.

[0076] Through the above steps, the updated dynamic parameters from the upper layer are transformed into torque commands for the lower-level actuators, thereby achieving adaptive torque control of the vehicle based on driving style.

[0077] Explanatoryly, in this module, vehicle dynamics parameters refer to the characteristic parameters in the desired yaw rate transfer function. The desired yaw rate transfer function is:

[0078] In the above formula, For steady-state yaw rate gain, The system's inherent frequency, Let be the system damping ratio. For the steady-state yaw rate gain, it is determined by the following formula:

[0079] The formula includes the understeer gradient K. If K > 0, it indicates that the vehicle has understeer characteristics; if K = 0, it indicates that the vehicle has neutral steering characteristics; if K < 0, it indicates that the vehicle has oversteer characteristics.

[0080] In the traditional calculation of desired yaw rate, the above parameters can be obtained from the vehicle's inherent parameters. However, since the parameters are fixed, the desired yaw rate cannot be adapted to the driving needs of different drivers, resulting in a single dynamic response characteristic of the controlled vehicle, which cannot meet the functional requirements of the vehicle adapting to the driver.

[0081] In this embodiment, a style sliding window approach is used to smoothly switch the desired yaw response (vehicle operating parameter). The number of driving style recognitions within the style sliding window is denoted as . The result of a single driving style recognition is The result of the blended driving style output by the sliding window is:

[0082] As the number of effective steering attempts by the driver increases, the fused driving style results based on the style sliding window will continuously evolve and update. The current update cycle of the style sliding window is m. As the driving style converges, the vehicle operating parameters will also converge, thus completing the self-evolutionary update of the vehicle operating parameters for the current driver's driving style.

[0083] Let the result of the fused driving style corresponding to the m-th iteration be . After a style sliding window cycle, the driver style recognition result is updated to... The incremental update of vehicle operating parameters can then be expressed as:

[0084] The description of vehicle operation under the current style sliding window is as follows: The updated vehicle operating parameters will then be updated. It can be represented as: .

[0085] In a preferred embodiment, the method further includes state machine management and functional safety control steps.

[0086] Specifically, it monitors the integrity of input signals, system operating status, identification confidence thresholds, and function exit commands in real time. Based on preset state machine logic, it switches between normal operating state, silent state, and fault state.

[0087] The silent state is triggered by the failure of valid steering segment determination or parameter validity verification; the fault state is triggered by sensor disconnection, computing power overload, or communication timeout. Specifically, if the real-time driving data signal is incomplete or does not meet the preset valid steering conditions, the system enters the silent state; if a sensor fault, computing power overload, or communication timeout is detected, the system enters the fault state.

[0088] When entering the silent state or the fault state, updates to the dynamic parameters are paused, and the last effective dynamic parameters are retained or switched to the preset default dynamic parameters. Specifically, this includes: immediately freezing the updates of desired control parameters, stopping style feature extraction, model identification, and parameter distribution processes; maintaining vehicle control stability, switching the torque controller to the default balance parameters, or maintaining the last effective safety parameters to ensure driving safety.

[0089] The vehicle control method based on driving style recognition provided in this embodiment adjusts the control target by identifying the driver's driving style online, thereby achieving online self-evolutionary updates of the control gain and realizing "vehicle adapting to driver." This part is performed online. First, real-time driving data is collected throughout the lifecycle. Effective driving feature parameters are extracted by collecting effective driving data of the current driver under typical steering conditions. Then, online analysis of the driver's personalized driving style is performed. Based on the effective feature parameters of the current driver's driving data, an offline-trained neural network is used to identify the driving style online. Finally, the control system undergoes online self-evolution. Based on the real-time online estimation results of the driving style, the control gain that matches the current driver's driving style is identified using the offline-constructed driving style and control parameter matching mapping relationship, thereby achieving online self-evolutionary updates of the control gain. Through a closed-loop mechanism of offline mapping relationship construction - online style identification - dynamic parameter adjustment, the online self-evolution of the control gain (understeering gradient, natural frequency, damping ratio) is achieved, balancing real-time response and driving stability. This allows the vehicle dynamics to accurately respond to and adapt to the driver's needs, thereby reducing driving risks and improving driving stability in different scenarios.

[0090] Secondly, embodiments of this application also provide a vehicle control method apparatus based on driving style recognition.

[0091] In one embodiment, reference is made to Figure 3 , Figure 3 This is a functional module diagram of an embodiment of the vehicle control method device based on driving style recognition according to this application. Figure 3 As shown, the vehicle control method apparatus based on driving style recognition includes: The acquisition module is used to acquire real-time driving data during vehicle operation; The recognition module is used to input the real-time driving data into a pre-trained driving style recognition model to obtain the driver's driving style. The driving style recognition model is trained based on the original driving data of multiple drivers with different driving styles under different preset working conditions. An update module is used to update the vehicle dynamics parameters according to the driver's driving style to obtain updated vehicle dynamics parameters. The control module is used to calculate the target driving torque of each wheel of the vehicle based on the updated vehicle dynamics parameters, and to control the vehicle based on the target driving torque.

[0092] Furthermore, in one embodiment, the device further includes a training module, which is used for: Original driving data were collected from multiple drivers with different driving styles under typical steering conditions, different vehicle speeds, and different road surface adhesion coefficients to form the original dataset. Valid steering segments are selected from the original dataset, and driving style feature parameters of each valid steering segment are extracted and concatenated to generate the corresponding style feature vector. Associate the corresponding style feature vectors with driving style labels to form a labeled dataset; Using style feature vectors from the labeled dataset as input and corresponding driving style labels as output, the initial driving style recognition model is trained until the network converges, resulting in a pre-trained driving style recognition model.

[0093] Furthermore, in one embodiment, the device further includes a mapping module, which is used for: Based on the correspondence between driving style labels and corresponding vehicle target dynamic parameters in the labeled dataset, a mapping relationship between driving style and vehicle dynamic parameters is constructed, and the mapping relationship is stored as a style-parameter mapping table.

[0094] Furthermore, in one embodiment: The typical steering conditions include: single lane change condition, double lane change condition, and curve conditions with different radii of curvature; The driving styles include conservative driving style, balanced driving style, and aggressive driving style; The drivers with different driving styles are stratified and sampled according to at least one of driving experience, age group and gender, and the number of samples in each driving style category is evenly distributed. Driving data should include at least steering wheel angle, yaw rate, longitudinal vehicle speed, lateral vehicle speed, and road surface adhesion coefficient; The driving style characteristic parameters include at least steering entropy, peak steering wheel angular velocity, steering return delay time, yaw rate overshoot, and response setup time. The dynamic parameters include at least one of the following: understeering gradient, natural frequency of the system at the desired yaw rate, and damping ratio of the system at the desired yaw rate.

[0095] Furthermore, in one embodiment, the identification module is also used for: The real-time driving data is time-aligned, and a driving data sequence with a unified time reference is generated. Effective steering segments are extracted from the driving data sequence, and driving style feature parameters of each effective steering segment are extracted and concatenated to generate the corresponding style feature vector. The style feature vector is input into a pre-trained driving style recognition model to obtain the single driving style recognition result and recognition confidence for each effective steering segment; The single driving style recognition results corresponding to multiple consecutive valid steering segments within a preset sliding window are fused to obtain a fused driving style result. The current driver's driving style is determined based on the results of the fused driving style analysis.

[0096] Furthermore, in one embodiment, the update module is also used to: Based on the current driver's driving style, a pre-established style-parameter mapping table is queried to obtain the corresponding vehicle target dynamic parameters; Physical constraints and safety boundary judgments are performed on the vehicle target dynamic parameters to obtain the verified vehicle target dynamic parameters; The difference between the verified vehicle target dynamics parameters and the currently effective vehicle dynamics parameters is calculated to obtain the parameter increment; The parameter increment is smoothed based on the rate of change limit and the smoothing coefficient to obtain the processed parameter increment. The updated vehicle dynamics parameters are obtained by summing the currently effective vehicle dynamics parameters with the processed parameter increments.

[0097] Furthermore, in one embodiment, the control module is also used for: The model parameters of the vehicle's desired yaw rate transfer function model are configured based on the vehicle's updated dynamic parameters. Input the steering wheel angle into the desired yaw rate transfer function model to obtain the desired yaw rate trajectory; Based on the updated vehicle dynamics parameters and the desired yaw rate trajectory, combined with the real-time driving data, the target drive torque distribution ratio of each drive wheel is calculated to obtain the target drive torque of each wheel.

[0098] The functions of each module in the above-mentioned vehicle control device based on driving style recognition correspond to the steps in the above-mentioned vehicle control method embodiment based on driving style recognition, and their functions and implementation processes will not be described in detail here.

[0099] Thirdly, embodiments of this application provide a vehicle control device based on driving style recognition. The vehicle control device based on driving style recognition can be a personal computer (PC), a laptop computer, a server, or other device with data processing capabilities.

[0100] Reference Figure 3 , Figure 3 This is a schematic diagram of the hardware structure of a vehicle control device based on driving style recognition, as described in an embodiment of this application. In this embodiment, the vehicle control device based on driving style recognition may include a processor, a memory, a communication interface, and a communication bus.

[0101] The communication bus can be of any type and is used to interconnect the processor, memory, and communication interface.

[0102] The communication interface includes input / output (I / O) interfaces, physical interfaces, and logical interfaces used for interconnecting components within the driving style recognition-based vehicle control device, as well as interfaces used for interconnecting the driving style recognition-based vehicle control device with other devices (such as other computing devices or user equipment). Physical interfaces can be Ethernet interfaces, fiber optic interfaces, ATM interfaces, etc.; user equipment can be displays, keyboards, etc.

[0103] Memory can be various types of storage media, such as random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), flash memory, optical storage, hard disk, programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), etc.

[0104] The processor can be a general-purpose processor, which can call the vehicle control program based on driving style recognition stored in memory and execute the vehicle control method based on driving style recognition provided in the embodiments of this application. For example, the general-purpose processor can be a central processing unit (CPU). The method executed when the vehicle control program based on driving style recognition is called can be referred to in the various embodiments of the vehicle control method based on driving style recognition in this application, and will not be repeated here.

[0105] Those skilled in the art will understand that Figure 3 The hardware structure shown does not constitute a limitation of this application and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0106] Fourthly, embodiments of this application also provide a computer-readable storage medium.

[0107] The present application provides a computer-readable storage medium storing a vehicle control program based on driving style recognition, wherein when the vehicle control program based on driving style recognition is executed by a processor, it implements the steps of the vehicle control method based on driving style recognition as described above.

[0108] The method implemented when the vehicle control program based on driving style recognition is executed can be referred to in various embodiments of the vehicle control method based on driving style recognition in this application, and will not be repeated here.

[0109] It should be noted that the sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0110] The terms "comprising" and "having," and any variations thereof, in the specification, claims, and accompanying drawings of this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such process, method, product, or apparatus. The terms "first," "second," and "third," etc., are used to distinguish different objects, etc., and do not indicate a sequence, nor do they limit "first," "second," and "third" to different types.

[0111] In the description of the embodiments of this application, terms such as "exemplary," "for example," or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "exemplary," "for example," or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary," "for example," or "for instance" is intended to present the relevant concepts in a concrete manner.

[0112] In the description of the embodiments of this application, unless otherwise stated, " / " means "or". For example, A / B can mean A or B. The "and / or" in the text is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist simultaneously, and B exists alone. In addition, in the description of the embodiments of this application, "multiple" means two or more.

[0113] In some processes described in the embodiments of this application, multiple operations or steps are included in a specific order. However, it should be understood that these operations or steps may not be executed in the order they appear in the embodiments of this application, or they may be executed in parallel. The sequence number of the operation is only used to distinguish different operations, and the sequence number itself does not represent any execution order. In addition, these processes may include more or fewer operations, and these operations or steps may be executed sequentially or in parallel, and these operations or steps may be combined.

[0114] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device to execute the methods described in the various embodiments of this application.

[0115] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A vehicle control method based on driving style recognition, characterized in that, The vehicle control method based on driving style recognition includes: Acquire real-time driving data during vehicle operation; The real-time driving data is input into a pre-trained driving style recognition model to obtain the driver's driving style. The driving style recognition model is trained based on the original driving data of multiple drivers with different driving styles under different preset conditions. The vehicle dynamics parameters are updated based on the driver's driving style to obtain updated vehicle dynamics parameters; The target driving torque for each wheel of the vehicle is calculated based on the updated vehicle dynamics parameters, and the vehicle is controlled based on the target driving torque.

2. The vehicle control method based on driving style recognition as described in claim 1, characterized in that, The training of the driving style recognition model includes: Original driving data were collected from multiple drivers with different driving styles under typical steering conditions, different vehicle speeds, and different road surface adhesion coefficients to form the original dataset. Valid steering segments are selected from the original dataset, and driving style feature parameters of each valid steering segment are extracted and concatenated to generate the corresponding style feature vector. Associate the corresponding style feature vectors with driving style labels to form a labeled dataset; Using style feature vectors from the labeled dataset as input and corresponding driving style labels as output, the initial driving style recognition model is trained until the network converges, resulting in a pre-trained driving style recognition model.

3. The vehicle control method based on driving style recognition as described in claim 2, characterized in that: The method also includes: Based on the correspondence between driving style labels and corresponding vehicle target dynamic parameters in the labeled dataset, a mapping relationship between driving style and vehicle dynamic parameters is constructed, and the mapping relationship is stored as a style-parameter mapping table.

4. The vehicle control method based on driving style recognition as described in claim 2, characterized in that: The typical steering conditions include: single lane change condition, double lane change condition, and curve conditions with different radii of curvature; The driving styles include conservative driving style, balanced driving style, and aggressive driving style; The drivers with different driving styles are stratified and sampled according to at least one of driving experience, age group and gender, and the number of samples in each driving style category is evenly distributed. Driving data should include at least steering wheel angle, yaw rate, longitudinal vehicle speed, lateral vehicle speed, and road surface adhesion coefficient; The driving style characteristic parameters include at least steering entropy, peak steering wheel angular velocity, steering return delay time, yaw rate overshoot, and response setup time. The dynamic parameters include at least one of the following: understeering gradient, natural frequency of the system at the desired yaw rate, and damping ratio of the system at the desired yaw rate.

5. The vehicle control method based on driving style recognition as described in claim 1, characterized in that, The real-time driving data is input into a pre-trained driving style recognition model to obtain the driver's driving style, including: The real-time driving data is time-aligned, and a driving data sequence with a unified time reference is generated. Effective steering segments are extracted from the driving data sequence, and driving style feature parameters of each effective steering segment are extracted and concatenated to generate the corresponding style feature vector. The style feature vector is input into a pre-trained driving style recognition model to obtain the single driving style recognition result and recognition confidence for each effective steering segment; The single driving style recognition results corresponding to multiple consecutive valid steering segments within a preset sliding window are fused to obtain a fused driving style result. The current driver's driving style is determined based on the results of the fused driving style analysis.

6. The vehicle control method based on driving style recognition as described in claim 1, characterized in that, The vehicle dynamics parameters are updated based on the driver's driving style to obtain updated dynamics parameters, including: Based on the current driver's driving style, a pre-established style-parameter mapping table is queried to obtain the corresponding vehicle target dynamic parameters; Physical constraints and safety boundary judgments are performed on the vehicle target dynamic parameters to obtain the verified vehicle target dynamic parameters; The difference between the verified vehicle target dynamics parameters and the currently effective vehicle dynamics parameters is calculated to obtain the parameter increment; The parameter increment is smoothed based on the rate of change limit and the smoothing coefficient to obtain the processed parameter increment. The updated vehicle dynamics parameters are obtained by summing the currently effective vehicle dynamics parameters with the processed parameter increments.

7. The vehicle control method based on driving style recognition as described in claim 1, characterized in that, The target drive torque for each wheel of the vehicle is calculated based on the updated dynamic parameters, including: The model parameters of the vehicle's desired yaw rate transfer function model are configured based on the vehicle's updated dynamic parameters. Input the steering wheel angle into the desired yaw rate transfer function model to obtain the desired yaw rate trajectory; Based on the updated vehicle dynamics parameters and the desired yaw rate trajectory, combined with the real-time driving data, the target drive torque distribution ratio of each drive wheel is calculated to obtain the target drive torque of each wheel.

8. A vehicle control device based on driving style recognition, characterized in that, The vehicle control device based on driving style recognition includes: The acquisition module is used to acquire real-time driving data during vehicle operation; The recognition module is used to input the real-time driving data into a pre-trained driving style recognition model to obtain the driver's driving style. The driving style recognition model is trained based on the original driving data of multiple drivers with different driving styles under different preset working conditions. An update module is used to update the vehicle dynamics parameters according to the driver's driving style to obtain updated vehicle dynamics parameters. The control module is used to calculate the target driving torque of each wheel of the vehicle based on the updated vehicle dynamics parameters, and to control the vehicle based on the target driving torque.

9. A vehicle control device based on driving style recognition, characterized in that, The vehicle control device based on driving style recognition includes a processor, a memory, and a vehicle control program based on driving style recognition stored in the memory and executable by the processor, wherein when the vehicle control program based on driving style recognition is executed by the processor, it implements the steps of the vehicle control method based on driving style recognition as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a vehicle control program based on driving style recognition, wherein when the vehicle control program based on driving style recognition is executed by a processor, it implements the steps of the vehicle control method based on driving style recognition as described in any one of claims 1 to 7.

Images