Vehicle control methods, systems and vehicles

Abstract

This application provides a vehicle control method, system, and vehicle. The vehicle control method includes: acquiring road image data of the road ahead corresponding to the vehicle's driving direction; constructing a road environment model based on a visual perception model and the road image data to characterize the road environment attributes of the road ahead; acquiring the vehicle's driving state information; obtaining a sequence of wheel control parameters based on the driving state information and the road environment model; and controlling the vehicle's wheel state based on the wheel control parameter sequence. Thus, the road environment model can intuitively characterize the road surface environment attributes, and the vehicle's driving state information can yield a sequence of wheel control parameters that satisfies both the road environment attributes and the vehicle's current driving state. This allows for better vehicle control to adapt to the road conditions the vehicle is about to travel on, enabling the vehicle's wheels to adapt to complex and changing real-world road conditions and improving the user's driving experience.

Description

Technical Field

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

[0002] When a vehicle travels on different road surfaces, control parameters such as the toe angle and camber angle of the wheels directly determine the vehicle's handling stability, steering sensitivity, and tire wear rate. Traditionally, vehicle wheel alignment parameters are set statically during manufacturing or repair, and these settings remain fixed. This means the vehicle can only achieve optimal performance under specific driving conditions and cannot adapt to complex and changing real-world road conditions. Therefore, existing fixed settings for parameters such as tire control cannot meet the driving needs of vehicles on different road surfaces. Summary of the Invention

[0003] This application provides a vehicle control method, system, and vehicle. Based on a visual perception model, a road environment model characterizing the road environment attributes of the road ahead is obtained from road image data. Furthermore, based on vehicle driving state information, a sequence of wheel control parameters for the vehicle on the road it is about to travel on is determined. The vehicle's wheel states are then controlled based on this sequence of wheel control parameters. This allows for real-time control of the vehicle's wheel states based on the wheel control parameter sequence corresponding to the road ahead, ensuring that the vehicle's wheels meet the driving requirements of the road and improving the user's driving experience.

[0004] To address the aforementioned technical problems, in a first aspect, embodiments of this application provide a vehicle control method. This method includes: acquiring road image data of the road ahead corresponding to the vehicle's driving direction; constructing a road environment model of the road ahead based on a visual perception model and the road image data; the visual perception model includes multiple sub-models, each sub-model being used to perform corresponding recognition processing on the road ahead based on the road image data to obtain a corresponding recognition result; and the road environment model being used to characterize the road environment attributes of the road ahead; acquiring vehicle driving state information; obtaining a sequence of wheel control parameters for the vehicle based on the driving state information and the road environment model; and controlling the wheel state of the vehicle based on the wheel control parameter sequence.

[0005] Using the above technical solution, road image data of the road ahead corresponding to the vehicle's driving direction is obtained; based on multiple sub-models included in the visual perception model, the road ahead is identified according to the road image data to obtain corresponding identification results; based on the corresponding identification results, the road ahead is constructed into a road environment model to characterize the road environment attributes of the road ahead; based on the road environment model and driving state information, the wheel control parameter sequence corresponding to the road ahead is obtained; and the vehicle's wheels are controlled according to the wheel control parameter sequence corresponding to the road ahead. Thus, based on the multiple sub-models included in the visual perception model, corresponding recognition results are obtained by performing corresponding recognition processing on the road image data of the road ahead of the vehicle. This is used to construct a road environment model to characterize the road environment attributes of the road ahead, enabling the road environment model to intuitively represent the road surface features of the road ahead. Furthermore, based on the vehicle's driving state information, the sequence of wheel control parameters for the vehicle on the road it is about to travel on is determined. This allows for the dynamic acquisition of wheel control parameters that better meet the driving needs of the road ahead. Based on the sequence of wheel control parameters, the vehicle's wheels are controlled so that they can adapt to complex and changing real road conditions, meeting the driving needs of the vehicle on complex and changing roads and improving the user's driving experience.

[0006] In one possible implementation of the first aspect described above, the visual perception model includes a road surface semantic segmentation sub-model, a road parameter extraction sub-model, and an obstacle detection sub-model. Based on the visual perception model, a road environment model of the road ahead is constructed according to road image data, including: based on the road surface semantic segmentation sub-model, performing road surface type identification processing on road image data to obtain the road surface type of the road ahead, and obtaining the road surface adhesion coefficient information corresponding to the road surface type; based on the road parameter extraction sub-model, performing road surface condition identification processing on road image data to obtain road surface condition information of the road ahead; based on the obstacle detection sub-model, performing obstacle identification processing on road image data to obtain road surface obstacle information of the road ahead; and obtaining the road environment model of the road ahead based on the road surface type and the corresponding road surface adhesion coefficient information, road surface condition information, and road surface obstacle information.

[0007] The above technical solution employs a road surface semantic segmentation sub-model to perform deep recognition processing on road image data, obtaining the road surface type in the captured road images and determining the corresponding road surface adhesion coefficient information. Furthermore, a road parameter extraction sub-model is used to obtain road surface condition information from the captured road surface images, and an obstacle detection sub-model is used to obtain road surface obstacle information. Based on the road surface type and the corresponding road surface adhesion coefficient, road surface condition, and obstacle information, a road environment model is constructed. This allows the road environment model to characterize road environment attributes, facilitating predictive control of the road surface ahead.

[0008] In one possible implementation of the first aspect above, determining the road adhesion coefficient information corresponding to the road surface type according to the road surface type includes: obtaining the road adhesion coefficient information corresponding to the road surface type based on the road surface type and the distribution map of road surface type and road adhesion coefficient, wherein the distribution map of road surface type and road adhesion coefficient is obtained based on the correspondence table between road surface type and road adhesion coefficient.

[0009] In one possible implementation of the first aspect described above, the road surface condition information includes longitudinal slope, lateral slope, radius of curvature, and rate of change of radius of curvature. Road surface condition recognition processing is performed based on road image data to obtain road surface condition information of the road ahead, including: performing depth estimation processing on the road image data to generate a two-dimensional depth map; mapping the two-dimensional depth map to three-dimensional space to generate a three-dimensional depth map; performing planar fitting processing on the three-dimensional road surface point cloud data in the three-dimensional depth map to obtain the longitudinal slope and lateral slope of the road ahead; and determining the lane line of the vehicle's lane and the vehicle's historical driving trajectory based on the three-dimensional depth map, performing trajectory fitting processing on the lane line of the vehicle's lane and the historical driving trajectory to obtain the vehicle's predicted driving trajectory and the radius of curvature and rate of change of radius of curvature corresponding to the predicted driving trajectory.

[0010] By adopting the above technical solution, considering the longitudinal slope and lateral slope of the road surface ahead, the radius of curvature and the rate of change of the radius of curvature of the vehicle's predicted driving trajectory, and taking into account multiple factors, wheel control based on wheel control parameters can better cope with the road conditions the vehicle is about to encounter, better achieve vehicle control, and improve the user's driving experience.

[0011] In one possible implementation of the first aspect above, the driving state information includes vehicle speed information. Based on the road environment model and the driving state information, a sequence of wheel control parameters for the vehicle is obtained, including: determining the look-ahead distance based on the vehicle speed information and a predefined look-ahead time; determining the sequence of key point road parameters for the vehicle within the look-ahead distance based on the road environment model; and determining the vehicle control parameters corresponding to each key point in the key point road parameter sequence before the vehicle reaches the corresponding key point, based on the vehicle dynamics model and the key point road parameter sequence information, to obtain the wheel control parameter sequence for the vehicle corresponding to the key point road parameter sequence.

[0012] By employing the above technical solution, the look-ahead distance is determined based on vehicle speed information and a predefined look-ahead time. Based on a road environment model, a sequence of key point road parameters within the look-ahead distance is determined, resulting in a sequence of key point road parameters corresponding to multiple key points within the look-ahead distance on the road the vehicle is about to travel on. This, combined with a vehicle dynamics model, yields vehicle control parameters for each key point before the vehicle reaches its corresponding key point, and consequently, a sequence of wheel control parameters within the look-ahead distance on the road ahead. Thus, based on the road parameter information of each key point included in the key point road parameter sequence, the control parameters of the vehicle's wheels before reaching each key point on the road can be predicted in advance. Furthermore, based on the wheel control parameter sequence, the vehicle's wheels can be controlled ahead of time of arrival at each key point. In other words, predictive determination and predictive control of the vehicle's wheel control parameters are achieved for different driving roads, improving the accuracy and efficiency of wheel control and enhancing the user's driving experience.

[0013] In one possible implementation of the first aspect above, the key point road parameter sequence information of the vehicle within the look-ahead distance is determined according to the road environment model, including: determining the road surface adhesion coefficient information, road surface condition information and road obstacle information of each key point corresponding to the predicted driving trajectory of the vehicle within the look-ahead distance according to the road environment model, obtaining the road parameter information corresponding to each key point, and obtaining the key point road parameter sequence information according to the correspondence between each key point and the corresponding road parameter information.

[0014] By employing the above technical solution, road parameter information of key points on the predicted driving trajectory of the vehicle within the look-ahead distance is extracted, thereby obtaining the key point road parameter sequence information within the look-ahead distance corresponding to the road ahead of the vehicle. In this way, by extracting the road parameter information of key points on the predicted driving trajectory of the vehicle within the look-ahead distance, the wheel control parameters of the vehicle before reaching each key point on the predicted driving trajectory path within the look-ahead distance can be accurately predicted, resulting in an accurate wheel control parameter sequence. Furthermore, the wheel control parameter sequence can be determined directly based on the key point road parameter sequence, which can accelerate the generation efficiency of the wheel control parameter sequence on the predicted driving trajectory path.

[0015] In one possible implementation of the first aspect above, the wheel control parameter sequence includes wheel control parameters corresponding to each key point. Controlling the wheel state of the vehicle according to the wheel control parameter sequence includes: controlling the wheel state of the vehicle according to the target wheel control parameters corresponding to the target key point in the wheel control parameter sequence before the vehicle reaches the target key point.

[0016] By adopting the above technical solution, the wheel state of the vehicle is controlled in advance based on the wheel control parameters corresponding to the key point before the vehicle reaches the corresponding key point. This allows the vehicle to prepare in advance for the road surface to be traveled. This forward-looking control can achieve precise synchronization between the vehicle chassis adjustment and future road conditions, so that the vehicle can always meet the upcoming road conditions with the optimal chassis posture, which greatly improves the vehicle's driving performance in complex road conditions.

[0017] In one possible implementation of the first aspect above, the wheel control parameter sequence of the vehicle is obtained based on the driving state information and the road environment model, including: updating the road environment model based on the driving state information to obtain an updated road environment model; and obtaining the wheel control parameter sequence of the vehicle based on the updated road environment model and the driving state information.

[0018] By adopting the above technical solution and updating the road environment model in conjunction with vehicle driving status information, the road environment model can be made to fit the current state of the vehicle and better represent the road surface environment information.

[0019] Secondly, this application also discloses a vehicle control system, including: a visual perception module, a decision control module, and an execution module. The visual perception module is used to acquire road image data of the road ahead corresponding to the vehicle's driving direction, and constructs a road environment model of the road ahead based on the visual perception model and the road environment model to characterize the road environment attributes of the road ahead. The decision control module is used to acquire the vehicle's driving state information, and obtain the vehicle's wheel control parameter sequence based on the driving state information and the road environment model. The execution module is used to control the wheel state of the vehicle according to the wheel control parameter sequence.

[0020] Thirdly, this application also discloses a vehicle for executing the vehicle control method provided by any of the implementations of the first aspect.

[0021] The technical effects of the second and third aspects mentioned above can be found in the technical effects of the first aspect mentioned above. Attached Figure Description

[0022] To more clearly illustrate the technical solution of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below.

[0023] Figure 1 A schematic diagram of a vehicle control system provided in an embodiment of this application;

[0024] Figure 2 Another structural schematic diagram of the vehicle control system provided in the embodiments of this application;

[0025] Figure 3 A schematic diagram of yet another structure of the vehicle control system provided in an embodiment of this application;

[0026] Figure 4 A schematic flowchart of a vehicle control method provided in an embodiment of this application;

[0027] Figure 5 This is a schematic diagram of a process for constructing a road environment model provided in an embodiment of this application;

[0028] Figure 6 This is a schematic flowchart illustrating the process of determining a sequence of wheel control parameters provided in an embodiment of this application.

[0029] Figure 7 A schematic diagram of a wheel control scenario provided in an embodiment of this application;

[0030] Figure 8 This is a schematic diagram illustrating the principle of vehicle control based on conventional methods and vehicle control based on this application for obstacles in front, as provided in the embodiments of this application. Detailed Implementation

[0031] The camber and toe angles of a vehicle's wheels directly determine its grip performance on curves and straightaways. Different angles result in different driving experiences. By pre-setting the camber and toe angles, some vehicles have strong cornering ability, while others have strong acceleration on straightaways. For each vehicle, a uniform camber and toe angle design will result in relatively good driving performance in certain road conditions, while performance will be sacrificed in other road conditions.

[0032] Therefore, the control parameters of existing vehicle wheels are all preset and cannot be used in complex and ever-changing real road conditions.

[0033] Furthermore, CN116142306A discloses a dynamic wheel adjustment system for automobiles. This system uses a driving detection module to detect the vehicle's turning radius and bump level, allowing the control module to calculate parameters such as toe angle based on these parameters and adjust the wheels accordingly. However, this method relies primarily on road geometry (e.g., turning radius) and simple physical signals (e.g., bump level) at the perception level, lacking a deep understanding of road environment attributes, leading to poor accuracy in wheel control parameters. Secondly, the control logic of this patent is reactive control based on the vehicle's current or past road conditions. This means its control can only respond to the road conditions currently perceived by the vehicle and cannot anticipate and act on changes in road conditions ahead, resulting in a lag in wheel adjustment. Moreover, its decision-making process is largely based on preset formulas, lacking flexibility and struggling to handle complex scenarios with coupled road conditions. Therefore, a more intelligent and forward-looking wheel control solution is urgently needed.

[0034] Based on this, this application proposes a vehicle control method and system that can deeply understand the road environment data of the road ahead of the vehicle through a visual perception model, obtain a road environment model, obtain more geometric and physical attributes of the road ahead, and obtain a vehicle-road control parameter sequence based on the vehicle driving state information and the road environment model, so as to realize the forward dynamic adjustment of the vehicle as it passes through the road ahead, thereby significantly improving the vehicle's adaptability, safety and handling performance under various road conditions.

[0035] In the implementation method of this application, such as Figure 1 As shown, the vehicle control system includes a vision perception module, a decision control module, and an execution module.

[0036] The visual perception module is used to acquire road image data of the road ahead corresponding to the vehicle's driving direction. Based on the visual perception model, a road environment model of the road ahead is constructed according to the road image data. The road environment model is used to characterize the road environment attributes of the road ahead.

[0037] For example, the visual perception module collects road image data of the road ahead corresponding to the vehicle's driving direction based on the vehicle's visual image acquisition device, such as a binocular camera, a front-view camera, a rear-view camera, a panoramic camera, etc.

[0038] Specifically, when the vehicle is moving forward, the road image data is the road image data of the road ahead corresponding to the vehicle's forward direction of travel. When the vehicle is reversing, the road image data is the road image data of the road ahead corresponding to the vehicle's reversing direction of travel.

[0039] In this implementation, the visual perception module uses binocular cameras positioned in front of or behind the vehicle to acquire road images, and the visual perception model runs at an accelerated speed on a graphics processing unit (GPU).

[0040] Furthermore, such as Figure 2 As shown, the visual image acquisition device is connected to the visual perception module to input visual image data into the visual perception model of the visual perception module. The visual perception model includes multiple sub-models such as road surface semantic segmentation sub-model, road parameter extraction sub-model, and obstacle detection sub-model. Each sub-model performs corresponding recognition processing on the road ahead based on the road image data to obtain the corresponding recognition result.

[0041] Based on the road surface semantic segmentation sub-model, road surface type recognition processing is performed on road image data to obtain the road surface type of the road ahead. Based on the road surface type, the road surface adhesion coefficient information corresponding to the road surface type is obtained. The road surface type and the road surface adhesion coefficient information corresponding to the road surface type are used as the corresponding recognition results of the road surface semantic segmentation sub-model.

[0042] For example, based on the road surface semantic segmentation sub-model, the road surface adhesion coefficient information corresponding to each road surface type is obtained from the road surface type and the distribution map of road surface type and road surface adhesion coefficient. The distribution map of road surface type and road surface adhesion coefficient is obtained from the correspondence table between road surface type and road surface adhesion coefficient.

[0043] Among them, the road surface type is also the road surface material, which specifically includes different materials such as dry asphalt, wet asphalt, snow, and ice.

[0044] Based on the road parameter extraction sub-model, road surface condition recognition processing is performed on road image data to obtain road surface condition information of the road ahead, and the road surface condition information is used as the corresponding recognition result of the road parameter extraction sub-model.

[0045] For example, the road parameter extraction sub-model is also known as the road geometry extraction sub-model, which is used to obtain road surface condition information of the road ahead by depth estimation and lane line fitting.

[0046] The road surface condition information includes longitudinal slope. Lateral slope radius of curvature and rate of change of curvature .

[0047] Specifically, the road image data is depth estimated based on the road parameter extraction sub-model to generate a two-dimensional depth map. The two-dimensional depth map is then mapped to three-dimensional space to generate a three-dimensional depth map. The three-dimensional road surface point cloud data in the three-dimensional depth map is subjected to planar fitting to obtain the longitudinal and lateral slopes of the road ahead. The lane line of the vehicle's lane is determined based on the three-dimensional depth map, as well as the vehicle's historical driving trajectory. The lane line of the vehicle's lane and the historical driving trajectory are subjected to trajectory fitting to obtain the vehicle's predicted driving trajectory and the radius of curvature and rate of change of the radius of curvature corresponding to the predicted driving trajectory.

[0048] Based on the obstacle detection sub-model, obstacle recognition processing is performed on road image data to obtain road surface obstacle information of the road ahead, and the road surface obstacle information is used as the corresponding recognition result.

[0049] For example, the obstacle detection sub-model is used to identify obstacle information on the road ahead, so as to obtain the road obstacle information as the corresponding identification result of the obstacle detection sub-model.

[0050] Obstacle information includes obstacle type, distance between the vehicle and the obstacle, and obstacle bounding box. Specific obstacle types include potholes and speed bumps.

[0051] Furthermore, the visual perception module connects to the decision control module. Based on the road surface type and corresponding road surface adhesion coefficient information output by the road surface semantic segmentation sub-model, the road surface condition information output by the road parameter extraction sub-model, and the road obstacle information output by the obstacle detection sub-model, the visual perception module obtains the road surface environment model of the road ahead. The visual perception module then sends the road environment model (i.e., the dynamic road environment model) to the decision control module.

[0052] The decision control module is used to acquire the vehicle's driving status information and, based on the driving status information and the road environment model, obtain the vehicle's wheel control parameter sequence.

[0053] For example, such as Figure 2 As shown, the visual perception module acquires real-time vehicle driving status information based on the vehicle's CAN bus, and constructs a road environment model of the road ahead based on the vehicle driving status information and road image data. The vehicle driving status information, including vehicle speed and yaw rate, is then sent to the decision control module.

[0054] Furthermore, the decision control module establishes a look-ahead window, determines the vehicle speed information v and the predefined look-ahead time T, and calculates the look-ahead distance D based on the vehicle speed information and the look-ahead time. In this application, the look-ahead distance is the road distance in the road ahead corresponding to the vehicle's direction of travel.

[0055] Based on the predictive control strategy, the key point road parameter sequence of the vehicle within the look-ahead distance is determined according to the road environment model. Based on the vehicle dynamics model, the wheel control parameters corresponding to each key point in the key point road parameter sequence before the vehicle reaches the corresponding key point are determined, thus obtaining the wheel control parameter sequence (i.e., generating the target angle sequence).

[0056] For example, based on the road environment model, the road surface adhesion coefficient information, road surface condition information, and road obstacle information of each key point corresponding to the driving trajectory are predicted within the forward look-ahead distance of the vehicle on the road ahead, and the road parameter information corresponding to each key point is obtained. The key point road parameter sequence is obtained according to the correspondence between each key point and the corresponding road parameter information.

[0057] Furthermore, based on the vehicle dynamics model, the wheel control parameters corresponding to each key point before the vehicle reaches the corresponding key point are determined according to the key point road parameter sequence, thus obtaining the wheel control parameter sequence (i.e., generating the target angle sequence).

[0058] For example, based on the key point road parameter sequence and vehicle dynamics model, a set of wheel control parameters that are ahead of the actual arrival of the wheels at the corresponding key point positions in time are calculated, and a wheel control parameter sequence is generated.

[0059] Among them, the vehicle's wheel control parameter sequence is the sequence of wheel control parameters before the vehicle passes through each key point at a future time point. The wheel control parameter sequence includes the target toe angle sequence and the target camber angle sequence.

[0060] Furthermore, the decision control module updates the road environment model based on the driving status information to obtain the updated road environment model. Based on the updated road environment model and the driving status information, the vehicle's wheel control parameter sequence is obtained.

[0061] For example, yaw rate information can be used to describe the key state of the vehicle's current dynamics. The decision control module updates the road environment model based on the yaw rate information. Based on the updated road environment model and vehicle speed information, the wheel control parameter sequence of the vehicle is obtained according to the aforementioned wheel control parameter sequence acquisition method.

[0062] Furthermore, the decision control module sends a target angle command to the execution module based on the target toe angle sequence and the target camber angle sequence at the target time before the vehicle arrives at the corresponding key point. The target angle command includes the target toe angle and the target camber angle of the wheel control parameters at the corresponding key point.

[0063] The execution module is used to control the wheel state of the vehicle according to the wheel control parameter sequence.

[0064] Specifically, based on the wheel control parameter sequence, before the vehicle reaches the target key point, the wheel state of the vehicle is controlled according to the target wheel control parameters corresponding to the target key point in the wheel control parameter sequence.

[0065] The execution module includes a toe angle adjustment unit and a camber angle adjustment unit. The toe angle adjustment unit adjusts the wheel toe angle to the target toe angle value by controlling the power source servo motor installed on the wheel. The camber angle adjustment unit adjusts the wheel camber angle to the target camber angle value based on the power source servo motor installed on the wheel.

[0066] In another implementation of this application, such as Figure 3 As shown, the decision control module includes a dynamic road environment model construction unit and a target angle sequence generation unit. The visual perception model directly sends the road surface type and corresponding road surface adhesion coefficient information output by the road surface semantic segmentation sub-model, the road surface condition information output by the road parameter extraction sub-model, and the road obstacle information output by the obstacle detection sub-model to the decision control module. The dynamic road environment model construction unit included in the decision control module directly receives the road surface type and corresponding road surface adhesion coefficient information output by the road surface semantic segmentation sub-model, the road surface condition information output by the road parameter extraction sub-model, and the road obstacle information output by the obstacle detection sub-model sent by the visual perception module. Based on the road surface type and the corresponding road surface adhesion coefficient information, road surface condition information, and road obstacle information, the road environment model of the road ahead is obtained.

[0067] Furthermore, the visual perception module sends the yaw rate information input from the vehicle's CAN bus to the dynamic road environment model building unit included in the decision control module, so that the dynamic road environment model building unit updates the road environment model according to the yaw rate information, thereby obtaining the updated road environment model.

[0068] Furthermore, the vision perception module sends the vehicle speed information input from the vehicle CAN bus to the target angle sequence generation unit included in the decision control module, so that the target angle sequence generation unit can determine the wheel control parameter sequence, including the target toe angle sequence and the target camber angle sequence, based on the updated road environment model and the vehicle speed information input from the vehicle CAN bus.

[0069] The decision control module sends a target angle command to the execution module based on the target toe angle sequence and the target camber angle sequence at the target time before the vehicle arrives at the corresponding key point. The target angle command includes the target toe angle and the target camber angle of the wheel control parameters at the corresponding key point.

[0070] The execution module includes a toe angle adjustment unit that controls a servo motor mounted on the wheel to drive the wheel toe angle to a target toe angle value, and a camber angle adjustment unit that drives the wheel camber angle to a target camber value based on a servo motor mounted on the wheel.

[0071] It should be noted that the visual perception module, decision control module, and execution module in this application can be corresponding modules under different domain controllers of the vehicle, modules under the same domain controller of the vehicle, or modules under the vehicle controller. Software programs for implementing the corresponding functions are deployed in the corresponding domain controller.

[0072] If the modules are under different domain controllers, the domain controller corresponding to the execution module can be installed on the wheel to directly control the servo motors of each wheel.

[0073] It should also be noted that in this application, the vehicle CAN bus can be connected to a vision perception module so that the vision perception module can obtain vehicle driving status information and send the vehicle driving status information to the decision control module. Alternatively, the vehicle CAN bus can be connected to the decision control module so that the decision control module can obtain vehicle driving status information. The vehicle CAN bus can also be connected to an execution module so that the execution module can control the wheel status of the vehicle based on the vehicle driving status information (such as vehicle speed, position, acceleration, steering direction, etc.).

[0074] like Figure 4 As shown, the vehicle control method provided in this application specifically includes the following steps.

[0075] S100: Obtain road image data of the road ahead corresponding to the vehicle's driving direction. Based on the visual perception model, construct a road environment model of the road ahead based on the road image data. The road environment model is used to characterize the road environment attributes of the road ahead.

[0076] The visual perception model includes multiple sub-models, each of which is used to perform corresponding recognition processing on the road ahead based on road image data to obtain the corresponding recognition result.

[0077] S200 acquires the vehicle's driving status information and, based on the driving status information and the road environment model, obtains the vehicle's wheel control parameter sequence.

[0078] S300 controls the wheel state of the vehicle based on the wheel control parameter sequence.

[0079] In the implementation of this application, for step S100, the visual perception module collects road image data of the road ahead corresponding to the vehicle's driving direction based on the visual image acquisition device, and constructs a road environment model of the road ahead based on the visual perception model and the road image data.

[0080] In one implementation of this application, the road environment model includes road parameter information for each key point on the road ahead. The road parameter information includes at least one of the following: the road surface type corresponding to each key point, the adhesion coefficient information corresponding to the road surface type, road surface condition information, and road obstacle information. The road surface condition information includes longitudinal slope, lateral slope, radius of curvature, and rate of change of radius of curvature.

[0081] In the implementation of this application, the visual perception model includes multiple sub-models such as a road surface semantic segmentation sub-model, a road parameter extraction sub-model, and an obstacle detection sub-model. Each sub-model is used to perform corresponding recognition processing on the road ahead based on road image data to obtain corresponding recognition results. For example... Figure 5 As shown, a road environment model of the road ahead is constructed based on road image data, including the following steps.

[0082] S110: Based on the road image data, perform road surface type recognition processing to obtain the road surface type of the road ahead, and determine the road surface adhesion coefficient information corresponding to the road surface type.

[0083] For example, based on the road surface semantic segmentation sub-model, road surface material recognition processing is performed on road image data (as an example of road surface type recognition processing) to obtain the road surface material of the road ahead (as an example of road surface type). Based on the road surface type, the road surface adhesion coefficient information corresponding to the road surface type is determined. The road surface type and the road surface adhesion coefficient information corresponding to the road surface type are used as the corresponding recognition results of the road surface semantic segmentation sub-model.

[0084] Specifically, the road surface semantic segmentation sub-model adopts the DeepLabV3+ architecture. It is used to perform pixel-level classification processing (i.e., pixel-level parsing processing) on ​​the input road surface image data, identify the road surface material, and map it to the estimated road surface adhesion coefficient information.

[0085] Furthermore, to achieve the conversion from road visual features to key vehicle dynamics parameters, the visual perception module has a built-in or accessible road surface material-adhesion coefficient mapping table (as an example of a table showing the correspondence between road surface types and road surface adhesion coefficients). This mapping table defines different road surface materials and their estimated adhesion coefficients (…). The visual perception module, by querying a mapping table, converts the road surface material classification map into a quantified estimated adhesion coefficient distribution map, obtaining a distribution map of road surface materials and their adhesion coefficients. Based on the road surface semantic segmentation sub-model, it obtains the corresponding road adhesion coefficient information for each road surface material, thereby mapping the road surface material to physical parameters that directly characterize the road's gripping ability. In other words, the distribution map of road surface types and their adhesion coefficients is obtained from the correspondence table between road surface types and their adhesion coefficients.

[0086] The road surface materials include different materials such as dry asphalt, wet asphalt, snow, and ice.

[0087] In one implementation, the pavement material-adhesion coefficient mapping table corresponds to the dry asphalt pavement. The value is 0.8-1.0, corresponding to wet and slippery asphalt pavement. The value is 0.4-0.7, corresponding to snow-covered road surfaces. The value is 0.2-0.3, corresponding to the ice surface. The value is 0.05-0.1.

[0088] S120 performs road condition recognition processing based on road image data to obtain road condition information of the road ahead.

[0089] For example, based on the road parameter extraction sub-model, road surface condition recognition processing is performed on road image data to obtain road surface condition information of the road ahead. This road surface condition information is then used as the corresponding recognition result of the road parameter extraction sub-model. The road surface condition information includes longitudinal slope, lateral slope, radius of curvature, and rate of change of radius of curvature.

[0090] Specifically, the road parameter extraction sub-model utilizes the parallax principle of binocular vision from a stereo camera to perform depth estimation on road image data, generating a two-dimensional depth map. This two-dimensional depth map is then mapped to three-dimensional space to generate a three-dimensional depth map. Subsequently, the three-dimensional road surface point cloud data within the three-dimensional depth map undergoes planar fitting processing to calculate the longitudinal slope of the road ahead. and lateral slope Furthermore, based on the 3D depth map, the lane line of the vehicle's current lane and the vehicle's historical driving trajectory are determined. Trajectory fitting processing is then performed on the lane line and historical driving trajectory to calculate the vehicle's predicted driving trajectory and the radius of curvature corresponding to the predicted trajectory in real time. and the rate of change of radius of curvature .

[0091] S130 performs obstacle recognition processing based on road image data to obtain road surface obstacle information of the road ahead.

[0092] For example, based on the obstacle detection sub-model, obstacle recognition processing is performed on road image data to obtain road surface obstacle information of the road ahead. This road surface obstacle information is then used as the corresponding recognition result of the obstacle detection sub-model. The road surface obstacle information includes the type of obstacle, bounding box information, and the distance between the vehicle and the obstacle.

[0093] Specifically, the obstacle detection sub-model adopts the YOLOv5 architecture to quickly and accurately select discrete obstacles such as potholes and speed bumps from complex image scenes, and outputs the type of each obstacle, the precise bounding box, and the distance of each obstacle relative to the vehicle, providing key input information for the predictive protection function of the decision control module.

[0094] S140: Based on the road surface type and the corresponding road surface adhesion coefficient information, road surface condition information, and road obstacle information, a road environment model of the road ahead is obtained.

[0095] For example, the corresponding recognition results output by the three sub-models are integrated by the visual perception module into a unified, structured dynamic road environment model. This road environment model is not a meaningless collection of pixels, but a "digital twin road" containing road surface parameter information (such as coordinates, road surface adhesion coefficient, radius of curvature, slope, and obstacle information) of a series of road condition feature points (as examples of key points) along the vehicle's predicted driving trajectory path. In other words, the road environment model includes road parameter information of each key point on the vehicle's predicted driving trajectory path on the road ahead, so as to provide a unique and reliable factual basis for subsequent predictive decisions.

[0096] In other words, the road environment model is used to characterize the road environment attributes of the road ahead. The road environment attributes include physical attributes (such as road parameter information of each key point along the predicted driving trajectory path of the vehicle on the road ahead) and geometric attributes (such as images of the road ahead).

[0097] For step S200, the decision control module obtains the vehicle's wheel control parameter sequence based on the driving status information and the road environment model.

[0098] In the implementation of this application, the road environment model is updated according to the driving status information to obtain the updated road environment model, and the wheel control parameter sequence of the vehicle is obtained according to the updated road environment model and the driving status information.

[0099] The driving status information includes vehicle speed and yaw rate.

[0100] In one implementation, the decision control module receives a road environment model sent by the visual perception module and obtains the vehicle's driving status information, including yaw rate information, based on the CAN bus. The road environment model is then fused based on the yaw rate information to finally construct a unified and structured road environment model. Based on the updated road environment model and driving status information, the sequence of vehicle wheel control parameters is obtained.

[0101] In another implementation, the visual perception module does not execute step S140, but directly outputs all the information from the road surface semantic segmentation sub-model, the road parameter extraction sub-model, and the obstacle detection sub-model (including road surface material and adhesion coefficient). Radius of curvature R, slope rate of change of curvature (Information on road obstacles) is sent to the decision control module.

[0102] The decision control module includes a road environment model building unit that receives all the output information from the road surface semantic segmentation sub-model, road parameter extraction sub-model, and obstacle detection sub-model input from the visual perception module, and integrates it with real-time status data (such as yaw rate) from the vehicle CAN bus to finally build a unified and structured dynamic road environment model.

[0103] In the implementation of this application, the vehicle's driving direction and driving trend can be obtained based on the yaw rate information, so as to update the vehicle's predictive control trajectory and obtain the road parameter information of each key point on the updated predictive control trajectory path.

[0104] Furthermore, based on the road environment model (e.g., the updated road environment model) and driving state information, the sequence of vehicle wheel control parameters is obtained.

[0105] For example, the target angle sequence generation unit of the decision control module executes a predictive control strategy based on the road environment model. It calculates the wheel control parameters for different key points ahead based on the vehicle's current speed information and the road parameter sequence information of key points corresponding to the road conditions ahead (i.e., future road conditions) included in the road environment module, so as to obtain the wheel control parameter sequence of the vehicle.

[0106] like Figure 6 As shown, the sequence of wheel control parameters for the vehicle is obtained based on the road environment model (e.g., the updated road environment model) and driving state information, including the following steps.

[0107] S210 determines the look-ahead distance based on vehicle speed information and a predefined look-ahead time.

[0108] For example, the target angle sequence generation unit determines the vehicle's look-ahead distance D based on the vehicle speed information v and the preview time T, where, .

[0109] S220, based on the road environment model, determines the sequence of key point road parameters for vehicles within the look-ahead distance.

[0110] For example, within a look-ahead distance D, road parameter information for each key point along the predicted vehicle trajectory, included in the road environment model, is determined to obtain a key point road parameter sequence. That is, in this application, the key point road parameter sequence includes road parameter information for each key point.

[0111] In the implementation of this application, the key point road parameter sequence information of the vehicle within the look-ahead distance is determined according to the road environment model, including: determining the road surface adhesion coefficient information, road surface condition information and road obstacle information of each key point corresponding to the predicted driving trajectory of the vehicle within the look-ahead distance according to the road environment model, obtaining the road parameter information corresponding to each key point, and obtaining the key point road parameter sequence information according to the correspondence between each key point and the corresponding road parameter information.

[0112] S230, based on the vehicle dynamics model, determines the wheel control parameters corresponding to each key point in the key point road parameter sequence before the vehicle reaches the corresponding key point, based on the key point road parameter sequence information, and obtains the wheel control parameter sequence of the vehicle corresponding to the key point road parameter sequence.

[0113] For example, based on a vehicle dynamics model, and according to the road parameter information of each key point included in the key point road parameter sequence, a set of wheel control parameters that precede the vehicle's arrival at each key point in time is determined to obtain a wheel control parameter sequence. This wheel control parameter sequence includes the wheel control parameters corresponding to each key point.

[0114] Wheel control parameters include target toe angle parameters and target camber angle parameters, that is, the wheel control parameter sequence includes the target toe angle sequence ( , , ) and target outward tilt sequence ( , , ).

[0115] Based on the vehicle dynamics model, the wheel control parameters corresponding to each key point are determined according to the road parameter information of each key point included in the key point road parameter sequence. This includes: establishing a force and motion model based on the vehicle dynamics model, and determining the wheel control parameters corresponding to each key point by combining the road parameter information of each key point.

[0116] Among them, the force and motion model is a model that determines the adjustment values ​​of the vehicle's toe angle and camber angle based on road parameter information.

[0117] In this application, the toe angle of the wheel corresponding to each key point is determined based on the curvature radius, longitudinal slope and road adhesion coefficient information of each key point, and the camber angle of each key point is determined based on the curvature radius, curvature radius change rate, lateral slope and road adhesion coefficient information of each key point.

[0118] For example, a first toe angle is determined based on the radius of curvature of a key point, a second toe angle is determined based on the longitudinal slope, and a third toe angle is determined based on the road surface adhesion coefficient information. A weighted average of the first, second, and third toe angles is then performed to obtain the target toe angle corresponding to the key point. Similarly, a first camber angle is determined based on the radius of curvature of a key point, a second camber angle is determined based on the rate of change of the radius of curvature of the key point, a third camber angle is determined based on the lateral slope of the key point, and a fourth camber angle is determined based on the road surface adhesion coefficient information of the key point. A weighted average of the first, second, third, and fourth camber angles is then performed to obtain the target camber angle corresponding to the key point.

[0119] In one implementation, the first toe-in angle corresponding to the keypoint is determined based on the radius of curvature of the keypoint, wherein... Where R is the radius of curvature and L is the vehicle wheelbase. This represents the value of the toe angle.

[0120] Furthermore, the first toe angle and the first camber angle of the key point are determined based on the radius of curvature of the key point. For example, the lateral acceleration information of the vehicle is determined based on the radius of curvature of the key point and the vehicle speed information, the lateral force of the vehicle is determined based on the lateral acceleration information and the vehicle mass, and the first toe angle and the first camber angle corresponding to the key point are determined based on the lateral force of the vehicle.

[0121] Furthermore, the second camber angle corresponding to the key point is determined based on the rate of change of the radius of curvature of the key point. When dR / dt>0 (the turning radius increases), the camber angle needs to be reduced to reduce lateral force; when dR / dt<0 (the turning radius decreases), the camber angle needs to be increased (such as negative camber in sports cars) to improve grip.

[0122] Furthermore, the third outward tilt angle corresponding to the key point is determined based on the lateral and longitudinal slopes.

[0123] Among them, determining the third outward tilt angle corresponding to the key point based on the lateral slope includes obtaining the outward tilt angle in the following ways:

[0124]

[0125] in, The third outward tilt angle, The current camber angle, This refers to the lateral slope.

[0126] Furthermore, determining the second toe angle based on the lateral slope can also involve reducing the toe angle when the lateral slope is positive (i.e., reducing the toe angle when going uphill to avoid excessive tire wear) and increasing the toe angle when the lateral slope is negative (i.e., increasing the toe angle when going downhill).

[0127] Furthermore, the second toe angle can be determined based on the longitudinal slope. This can be achieved by reducing the toe angle when the longitudinal slope is positive, i.e., reducing the toe angle when going uphill to avoid excessive tire wear, and increasing the toe angle when the longitudinal slope is negative, i.e., increasing the toe angle when going downhill.

[0128] Furthermore, the third toe angle and the fourth camber angle are determined based on the road surface adhesion coefficient. Specifically, for roads with low adhesion coefficients (such as icy or snowy roads), the camber angle should be reduced (approaching 0°) and the toe angle to reduce the tendency to lateral slip; for roads with high adhesion coefficients (such as dry asphalt roads), the camber angle can be appropriately increased (negative camber) to improve handling.

[0129] In one implementation, T=2.5s, D=50m, the decision control module analyzes the road environment model along the predicted vehicle trajectory over the next 50 meters to obtain a series of key points ( , ... ).For example, (At 20 meters) is a dry asphalt curve (R=100m). (At 35 meters) is a sharp bend (R=50m, <0), (At 45 meters) the road surface is slippery (low) ).

[0130] Then, the decision control module calculates the target toe angle based on the vehicle dynamics model to ensure the vehicle maintains optimal attitude when reaching each key point. , , ) and target outward tilt sequence ( , , This allows us to obtain the wheel control parameter sequence information.

[0131] Finally, starting from the current moment, the decision control module sends control commands to the execution module to ensure that the vehicle smoothly adjusts its wheels to the target state before reaching the corresponding key point, thereby achieving spatiotemporal synchronization of wheel chassis parameters with future road conditions. For example, requiring the vehicle to reach... Before clicking, smoothly adjust the angle to... and .

[0132] For step S300, controlling the wheel state of the vehicle according to the wheel control parameter sequence includes: controlling the wheel state of the vehicle according to the target wheel control parameters corresponding to the target key point in the wheel control parameter sequence before the vehicle reaches the target key point.

[0133] For example, before the decision control module determines that the vehicle has traveled to the corresponding key point, it sends the wheel control parameters of the corresponding key point in the wheel control sequence to the execution module. The execution module sends the target toe angle included in the wheel control parameters to the toe angle adjustment unit, so that the toe angle adjustment unit drives the servo motor to adjust the toe angle of the wheel as the target toe angle. The execution module sends the target camber angle included in the wheel control parameters to the camber angle adjustment unit, so that the camber angle adjustment unit drives the servo motor to adjust the camber angle of the wheel as the target camber angle.

[0134] The vehicle control method provided in this application will be illustrated with specific examples below.

[0135] In a specific application scenario, when the decision control module determines, based on road model data, that a key point is a curve with a lateral slope, i.e., the lateral slope... If the camber is not zero, the decision control module will execute a camber compensation strategy based on the lateral slope information to determine the corresponding target camber angle. Before the vehicle is about to reach the critical point, the calculated target camber angle will be sent to the execution module so that the camber angle execution unit included in the execution module will adjust the camber angle of the wheels to the determined target camber angle.

[0136] like Figure 7 As shown on the left, on a normal road surface, the wheel camber angle is... If the decision control module determines that the lateral slope is negative, it determines that the road ahead is downhill, calculates and applies a positive camber compensation value to the wheel on the lower side based on the execution module (e.g., (e.g., X=1.5) Figure 7 (For the outer wheel in the compensated state on the right), if the decision control module determines that the lateral slope is positive, it determines that the road ahead is uphill, calculates and applies a negative camber compensation value to the uphill wheel on the higher side (e.g., the outer wheel in the compensated state on the right). (X=1.5).

[0137] In this way, by compensating for the camber angle to offset the extra tire deformation caused by the vehicle's weight, the tire tread can maintain the fullest contact with the sloping road surface when cornering, thereby greatly improving cornering grip.

[0138] In another specific application scenario, the vehicle control method provided in this application can achieve predictive protection against obstacles ahead.

[0139] like Figure 8 As shown, when the visual perception module detects obstacles such as speed bumps and potholes on the road ahead (corresponding to...), Figure 8 Identification points in The obstacle information is sent to the decision control module, which calculates the starting point for protective adjustments based on the vehicle's current speed and the distance between the vehicle and the obstacle. Figure 8 Action points in (and the wheel control parameters at the start time).

[0140] At the action point The decision control module sends adjustment commands to the toe-in and camber adjustment units, along with target toe-in and camber adjustment values ​​(i.e., setting the look-ahead buffer attitude), so that the toe-in adjustment unit dynamically adjusts the wheel's toe-in and camber angles to a neutral attitude near zero degrees (corresponding to...). Figure 4 (Protective neutral stance). This allows the tire to maintain a protective neutral stance at the moment of contact with the obstacle (corresponding to...). Figure 8 Impact point in It can meet the impact in a posture that is roughly perpendicular to the road surface, thereby converting the impact force into mainly vertical motion, so as to directly absorb the vertical impact energy and effectively avoid the tire lateral scraping, shimmy and additional stress on the steering tie rod and suspension system caused by the presence of toe angle or camber angle.

[0141] After the vehicle has completely passed the obstacle (corresponding to) Figure 8 Recovery points The control module then issues another command, instructing the toe-in and camber adjustment units to smoothly restore the wheel's toe-in and camber angles to the target values ​​required for normal driving (the tires quickly return to their pre-adjustment settings). This predictive protection control based on spatiotemporal planning not only improves ride comfort when traversing obstacles but also helps protect the mechanical structure of the chassis and steering system, extending their service life and ensuring driving safety.

[0142] The traditional vehicle control strategy when encountering obstacles is: upon detecting an obstacle (corresponding to...) Figure 8 Identification points in ), controlling the tire to maintain a vertical camber angle without predictable movement, at the instant the tire contacts the obstacle (corresponding to Figure 8 Impact point in ), maintain a neutral posture to cope with the impact and continue neutral posture after the impact, control the tire tilt and deformation to respond to the impact, and after the vehicle has completely passed the obstacle (corresponding to Figure 8 Recovery points The vehicle continues to tilt and deform accordingly, with residual body sway slowly returning to its original state. This method, which only engages wheel control upon contact with an obstacle, generates a strong impact, causing the user to feel a strong sway when the vehicle passes over an obstacle, which is detrimental to the user experience.

[0143] In the implementation of this application, both the toe angle adjustment unit and the camber angle adjustment unit in the execution module use high-response-speed servo motors as power sources, and drive the corresponding linkage mechanisms to achieve precise and rapid adjustment of the wheel angle.

[0144] The vehicle control method provided in this application is a method for detecting, predicting, and dynamically controlling the camber and toe angles of the vehicle chassis and tires based on a visual perception model. By introducing an end-to-end visual perception model, it achieves a process from perception to cognition based on a large visual perception model. This not only identifies the geometric parameters of obstacles but also understands the physical properties of the road surface, such as road material and road adhesion coefficient, surpassing existing technologies that only identify simple road surface features. Furthermore, it employs predictive control capabilities. Based on scene reconstruction of the road ahead, it analyzes the future vehicle sequence through a "look-ahead window" and issues control commands in advance. The chassis parameters are adjusted before the wheels contact specific road conditions, ensuring precise synchronization between wheel adjustment and future road conditions, thus solving the lag problem of traditional reactive control. This allows the vehicle to always meet upcoming road conditions with optimal chassis posture, greatly improving cornering performance, stability on slippery roads, and passability in complex road conditions.

[0145] The visual perception model implemented in this application is the core technology for road surface perception analysis. It is based on multi-objective optimization and learning, and compared with the existing technology that obtains simple road surface parameters based on simple fixed formulas, it can obtain multi-faceted road surface environmental parameters.

[0146] This application also provides a vehicle, which includes a vehicle control system to implement the vehicle control method.

[0147] It should be noted that, in addition to the specific embodiments described above, those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. Although the description of this application is presented in conjunction with preferred embodiments, this does not mean that the features of this application are limited to this implementation. On the contrary, the purpose of describing the application in conjunction with the implementation is to cover other options or modifications that may be derived from this application. To provide a thorough understanding of this application, many specific details are included in the above description, and this application may also be implemented without using these details. Furthermore, to avoid confusion or obscuring the focus of this application, some specific details will be omitted in the description. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other.

[0148] It should be noted that in this specification, similar reference numerals and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0149] It should be noted that some structural or methodological features may be shown in the accompanying drawings in a specific arrangement and / or order. However, it should be understood that such a specific arrangement and / or order may not be necessary. Rather, in some embodiments, these features may be arranged in a manner and / or order different from that shown in the illustrative drawings. Furthermore, including structural or methodological features in a particular figure does not imply that such features are required in all embodiments, and in some embodiments, these features may be omitted or may be combined with other features.

[0150] Although this application has been illustrated and described with reference to certain preferred embodiments, those skilled in the art should understand that the above description is a further detailed explanation of the application in conjunction with specific implementations, and should not be construed as limiting the specific implementation of the application to these descriptions. Those skilled in the art can make various changes in form and detail, including some simple deductions or substitutions, without departing from the spirit and scope of this application.

Claims

1. A vehicle control method, characterized in that, The method includes: The process involves acquiring road image data of the road ahead corresponding to the vehicle's driving direction. Based on a visual perception model, a road environment model of the road ahead is constructed using the road image data. The visual perception model includes a road surface semantic segmentation sub-model, a road parameter extraction sub-model, and an obstacle detection sub-model. Each sub-model is used to perform corresponding recognition processing on the road ahead based on the road image data to obtain corresponding recognition results. The road environment model is used to characterize the road environment attributes of the road ahead. Specifically, constructing the road environment model based on the road image data using the visual perception model includes: based on the road surface semantic segmentation sub-model, performing road surface type recognition processing on the road image data to obtain the road surface type of the road ahead, and obtaining the road surface adhesion coefficient information corresponding to the road surface type; and based on the road parameter extraction sub-model, performing road surface condition recognition processing on the road image data to obtain road surface condition information of the road ahead, including longitudinal slope, lateral slope, radius of curvature, and changes in radius of curvature. The method involves performing road condition recognition processing on the road image data to obtain road condition information of the road ahead, including: performing depth estimation processing on the road image data to generate a two-dimensional depth map; mapping the two-dimensional depth map to three-dimensional space to generate a three-dimensional depth map; performing planar fitting processing on the three-dimensional road surface point cloud data in the three-dimensional depth map to obtain the longitudinal slope and the lateral slope of the road ahead; determining the lane line of the lane where the vehicle is located based on the three-dimensional depth map; determining the historical driving trajectory of the vehicle; performing trajectory fitting processing on the lane line of the lane where the vehicle is located and the historical driving trajectory to obtain the predicted driving trajectory of the vehicle and the radius of curvature and the rate of change of the radius of curvature of the vehicle corresponding to the predicted driving trajectory; performing obstacle recognition processing on the road image data based on the obstacle detection sub-model to obtain road obstacle information of the road ahead; and obtaining a road environment model of the road ahead based on the road surface type and the road surface adhesion coefficient information corresponding to the road surface type, the road condition information, and the road obstacle information. Obtain the vehicle's driving status information, and based on the driving status information and the road environment model, obtain the vehicle's wheel control parameter sequence; The wheel state of the vehicle is controlled according to the wheel control parameter sequence.

2. The vehicle control method according to claim 1, characterized in that, Based on the road surface type, determine the road surface adhesion coefficient information corresponding to the road surface type, including: Based on the road surface type and the distribution map of the road surface type and the road surface adhesion coefficient, the road surface adhesion coefficient information corresponding to the road surface type is obtained, wherein the distribution map of the road surface type and the road surface adhesion coefficient is obtained according to the correspondence table between the road surface type and the road surface adhesion coefficient.

3. The vehicle control method according to claim 2, characterized in that, The driving status information includes vehicle speed information. Based on the road environment model and the driving status information, the wheel control parameter sequence of the vehicle is obtained, including: The look-ahead distance is determined based on the vehicle speed information and the predefined look-ahead time. Based on the road environment model, the key point road parameter sequence of the vehicle within the look-ahead distance is determined, and the key point road parameter sequence includes road parameter information of each key point; Based on the vehicle dynamics model, according to the key point road parameter sequence, the wheel control parameters corresponding to each key point included in the key point road parameter sequence before the vehicle reaches the corresponding key point are determined, thus obtaining the wheel control parameter sequence of the vehicle corresponding to the key point road parameter sequence.

4. The vehicle control method according to claim 3, characterized in that, Based on the road environment model, the key point road parameter sequence information of the vehicle within the look-ahead distance is determined, including: Based on the road environment model, the road surface adhesion coefficient information, road surface condition information, and road obstacle information of each key point corresponding to the predicted driving trajectory of the vehicle within the look-ahead distance are determined, and the road parameter information corresponding to each key point is obtained. The key point road parameter sequence information is obtained according to the correspondence between each key point and the corresponding road parameter information.

5. The vehicle control method according to claim 4, characterized in that, The wheel control parameter sequence includes wheel control parameters corresponding to each of the key points. Based on the wheel control parameter sequence, the wheel state of the vehicle is controlled, including: Based on the wheel control parameter sequence, before the vehicle reaches the target key point, the wheel state of the vehicle is controlled according to the target wheel control parameters corresponding to the target key point in the wheel control parameter sequence.

6. The vehicle control method according to any one of claims 1-5, characterized in that, Based on the driving state information and the road environment model, the wheel control parameter sequence of the vehicle is obtained, including: The road environment model is updated based on the driving status information to obtain the updated road environment model; Based on the updated road environment model and the driving state information, the wheel control parameter sequence of the vehicle is obtained.

7. A vehicle control system, characterized in that, The system includes: a visual perception module, a decision control module, and an execution module, wherein... The visual perception module is used to acquire road image data of the road ahead corresponding to the vehicle's driving direction. Based on the visual perception model, a road environment model of the road ahead is constructed according to the road image data. The visual perception model includes a road surface semantic segmentation sub-model, a road parameter extraction sub-model, and an obstacle detection sub-model. Each sub-model is used to perform corresponding recognition processing on the road ahead based on the road image data to obtain corresponding recognition results. The road environment model is used to characterize the road environment attributes of the road ahead. Specifically, constructing the road environment model of the road ahead based on the visual perception model and the road image data includes: based on the road surface semantic segmentation sub-model, performing road surface type recognition processing on the road image data to obtain the road surface type of the road ahead, and obtaining the road surface adhesion coefficient information corresponding to the road surface type; based on the road parameter extraction sub-model, performing road surface condition recognition processing on the road image data to obtain the road surface condition information of the road ahead, including longitudinal slope, lateral slope, radius of curvature, and curvature. The method involves processing road surface conditions based on the road image data to obtain road surface condition information for the road ahead. This includes: performing depth estimation processing on the road image data to generate a two-dimensional depth map; mapping the two-dimensional depth map to three-dimensional space to generate a three-dimensional depth map; performing planar fitting processing on the three-dimensional road surface point cloud data in the three-dimensional depth map to obtain the longitudinal slope and the lateral slope of the road ahead; determining the lane line of the vehicle's lane based on the three-dimensional depth map; determining the vehicle's historical driving trajectory; performing trajectory fitting processing on the lane line of the vehicle's lane and the historical driving trajectory to obtain the vehicle's predicted driving trajectory and the radius of curvature and the rate of change of the radius of curvature corresponding to the predicted driving trajectory; performing obstacle recognition processing based on the obstacle detection sub-model and the road image data to obtain road surface obstacle information for the road ahead; and obtaining a road environment model for the road ahead based on the road surface type and the road surface adhesion coefficient information corresponding to the road surface type, the road surface condition information, and the road surface obstacle information. The decision control module is used to acquire the vehicle's driving status information and, based on the driving status information and the road environment model, obtain the vehicle's wheel control parameter sequence. The execution module is used to control the wheel state of the vehicle according to the wheel control parameter sequence.

8. A vehicle, characterized in that, The vehicle is used to perform the vehicle control method as described in any one of claims 1-6.

Images