Vehicle suspension adjusting method and device, vehicle and storage medium

Abstract

The invention discloses a vehicle suspension adjusting method and device, a vehicle and a storage medium, and the method comprises the steps: obtaining pavement image data and vehicle state data in a vehicle driving process; determining obstacle features and obstacle categories based on the road surface image data; determining a vehicle running track based on the vehicle state data; determining a wheel rolling state in combination with the obstacle characteristics and the vehicle running track; determining a suspension adjusting strategy according to the wheel rolling condition and the obstacle category; and controlling the vehicle to run based on the suspension adjusting strategy. According to the embodiment of the invention, the method can precisely recognize the features of the obstacle, gives a targeted presetting strategy, and achieves the closed-loop control from the recognition of the obstacle to the intelligent adjustment.

Description

Technical Field

[0001] This application belongs to the field of autonomous driving control technology, and in particular relates to a vehicle suspension adjustment method, a vehicle suspension adjustment device, a vehicle, and a computer-readable storage medium. Background Technology

[0002] With the development of intelligent driving and advanced chassis control technology, the requirements for vehicle ride smoothness, ride comfort and active safety are increasing. Traditional passive suspension can only respond to the current road excitation with lag, while active suspension relies more on feedback adjustment. Even if a pre-aiming sensor is introduced, its detection capability is limited and it has the problem of weak obstacle category recognition. It cannot accurately identify small and medium-sized obstacles and it is difficult to associate control strategies to achieve advance control. Summary of the Invention

[0003] In view of the above problems, embodiments of this application are proposed to provide a vehicle suspension adjustment method, a vehicle suspension adjustment device, a vehicle, and a computer-readable storage medium that overcome or at least partially solve the above problems.

[0004] To address the aforementioned problems, in a first aspect of this application, an embodiment discloses a vehicle suspension control method, comprising: During vehicle operation, road surface image data and vehicle status data are acquired; Based on the road surface image data, the characteristics and categories of obstacles are determined; Based on the vehicle status data, the vehicle's trajectory is determined; Based on the characteristics of the obstacle and the vehicle's trajectory, the wheel crushing status is determined; Based on the wheel crushing condition and the type of obstacle, determine the suspension adjustment strategy; The vehicle's movement is controlled based on the aforementioned suspension adjustment strategy.

[0005] Optionally, the step of determining obstacle features and obstacle categories based on the road surface image data includes: The road surface image data is classified, and obstacle categories are output. Based on the obstacle category, perform structural analysis on the obstacle to determine the bounding box data corresponding to the obstacle category; Based on preset transformation parameters, the point cloud position is transformed on the bounding box data to determine the obstacle region; Local point cloud reconstruction is performed on the obstacle region to determine the obstacle features.

[0006] Optionally, the preset transformation parameters include a camera parameter matrix and a rigid body transformation matrix. The step of performing point cloud transformation on the bounding box data based on the preset transformation parameters to determine the obstacle region includes: Obtain the corner pixel coordinates of the bounding box data in the image pixel coordinate system; Based on the camera parameter matrix, the corner pixel coordinates are back-projected onto the camera's three-dimensional space to determine the three-dimensional camera coordinates corresponding to the corner pixel coordinates; Based on the rigid body transformation matrix, the coordinates of the three-dimensional camera are transformed into the coordinate system of the lidar to determine the coordinates of the three-dimensional corner points; Connect the coordinates of the three-dimensional corner points to determine the obstacle area.

[0007] Optionally, based on the camera parameter matrix, the corner pixel coordinates are back-projected onto the camera's three-dimensional space to determine the three-dimensional camera coordinates corresponding to the corner pixel coordinates; The steps for making a point include: Obtain a set of laser projection points within a preset neighborhood of the corner pixel coordinates; the set of laser projection points includes multiple candidate laser projection points. Calculate the two-dimensional Euclidean distance between the candidate laser projection point and the corner pixel coordinates to determine the first distance difference; Based on the first distance difference, the candidate laser projection points are sorted in ascending order to generate a laser projection point sequence; The candidate laser projection point that is first in the laser projection point sequence is determined as the target projection point; Obtain the depth value of the target projection point; By combining the camera parameter matrix and the depth value, the corner pixel coordinates are transformed to determine the three-dimensional camera coordinates corresponding to the corner pixel coordinates.

[0008] Optionally, the step of determining the vehicle's trajectory based on the vehicle status data includes: The vehicle state data is integrated to output the set of centroid positions and the vehicle heading angle; Based on the vehicle heading angle, local coordinate transformation is performed on the centroid positions in the centroid position set to generate the vehicle's trajectory.

[0009] Optionally, the step of determining the wheel crushing state by combining the obstacle features and the vehicle trajectory includes: The vehicle's trajectory is mapped to the lidar coordinate system to generate a trajectory point cloud; Based on the obstacle features and the lidar coordinate system, obtain the obstacle point cloud corresponding to the obstacle; Determine the second distance difference between the trajectory point cloud and the obstacle point cloud. If the second distance difference is less than a preset threshold, the wheel crushing state is determined to be a crushing risk state.

[0010] Optionally, the step of determining the suspension adjustment strategy based on the wheel rolling condition and the obstacle type includes: If the wheel crushing condition is a crushing risk condition, obtain the wheel identification; Based on the wheel identification and the obstacle category, a suspension adjustment strategy is determined.

[0011] In a second aspect, this application discloses a vehicle suspension adjustment device, comprising: The acquisition module is used to acquire road surface image data and vehicle status data during vehicle operation. The determination module is used to determine obstacle features and obstacle categories based on the road surface image data; The trajectory module is used to determine the vehicle's running trajectory based on the vehicle status data; The module is used to determine the wheel crushing state by combining the obstacle features and the vehicle's trajectory; An adjustment module is used to determine a suspension adjustment strategy based on the wheel crushing condition and the type of obstacle; The control module is used to control the vehicle's movement based on the suspension adjustment strategy.

[0012] In a third aspect of this application, embodiments of this application disclose a vehicle including a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the computer program, when executed by the processor, implements the vehicle suspension adjustment method as described above.

[0013] In a fourth aspect, embodiments of this application disclose a computer-readable storage medium storing a computer program that, when executed by a processor, implements the vehicle suspension adjustment method as described above.

[0014] The embodiments of this application have the following advantages: This application embodiment acquires road surface image data and vehicle status data during vehicle operation; based on the road surface image data, it determines obstacle features and obstacle categories; based on the vehicle status data, it determines the vehicle's trajectory; combining the obstacle features and the vehicle's trajectory, it determines the wheel crushing state; based on the wheel crushing state and the obstacle category, it determines a suspension adjustment strategy; and based on the suspension adjustment strategy, it controls vehicle movement. This application embodiment can accurately identify obstacle features and provide targeted pre-adjustment strategies, achieving closed-loop control from obstacle identification to intelligent adjustment. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the steps of an embodiment of a vehicle suspension adjustment method according to this application; Figure 2 This is a flowchart illustrating the steps of another embodiment of the vehicle suspension adjustment method of this application; Figure 3 This is a flowchart illustrating the steps of a vehicle suspension adjustment method according to this application; Figure 4 This is a structural block diagram of an embodiment of a vehicle suspension adjustment device according to this application; Figure 5 This is a structural block diagram of a vehicle provided in an embodiment of this application; Figure 6 This is a structural block diagram of a storage medium provided in an embodiment of this application. Detailed Implementation

[0016] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0017] Reference Figure 1 The diagram illustrates a flowchart of an embodiment of a vehicle suspension adjustment method according to this application. The vehicle suspension adjustment may specifically include the following steps: Step 101: During vehicle operation, acquire road surface image data and vehicle status data; During vehicle operation, onboard forward / downward cameras acquire real-time road images, outputting raw image data streams (road image data) which are transmitted to the onboard computing unit. Vehicle dynamics data (vehicle status data), such as real-time speed and steering angle, can also be obtained via wheel speed sensors or the CAN bus. This data is used to calculate obstacle arrival times and predict trajectories. Specifically, the camera can be fixed below the front bumper or at the front of the chassis, 30-50cm above the ground, with the lens tilted downwards at 15°-25° to cover a 1-8m road surface area in front. This balances detection distance and image resolution. The camera type can be an automotive-grade global shutter CMOS camera (Complementary Metal-Oxide-Semiconductor) such as the ON Semiconductor AR0234 or Sony IMX390. The image resolution should be set to 1920×1080 (Full HD), with a frame rate ≥30FPS, ensuring that vehicle displacement between adjacent frames is ≤0.75 when the vehicle speed is ≤80 km / h. m, to meet continuous tracking requirements; enable automatic exposure (AE) and adaptive white balance (AWB) to adapt to the illumination range of 0.1–100000 lux (unit: lux), including uniformly scaling the input image to 640×640 pixels, maintaining the aspect ratio, padding the insufficient parts with zeros, and normalizing the pixel values ​​to the range of [0,1] or [-1,1] to adapt to the model input requirements.

[0018] Step 102: Based on the road surface image data, determine the obstacle features and obstacle categories; The system acquires real-time road image data from automotive-grade front-view or downward-view cameras and preprocesses the images, such as size normalization, pixel normalization, and data augmentation. The raw image data is transformed into a normalized input image tensor suitable for deep learning model processing. Utilizing the feature extraction and recognition capabilities of the deep learning model, the system understands and analyzes the image content, identifying obstacle features and categories. For example, a lightweight target detection model (MobileNetV3 Backbone + OBB prediction head) is deployed, supporting rotated bounding box output. The preprocessed image tensor is input into the model, which extracts multi-scale visual features and makes predictions. For six preset target categories, such as vibration strips, speed bumps, manhole covers, potholes, concrete road joints, and bridge joints, the system outputs probability distributions to determine the obstacle category and outputs a rotated bounding box (OBB) for each obstacle category, including center coordinates. ,width ,high and rotation angle To accurately describe directional, slender structures such as seams, it can be seen that by using a lightweight OBB detection model to perform end-to-end inference on preprocessed images, the directional obstacle category and the geometric features of the rotated box, i.e., obstacle features, are directly output. This achieves pixel-level classification and synchronous perception of spatial orientation, laying the foundation for subsequent accurate point cloud reconstruction and trajectory-obstacle matching.

[0019] Step 103: Determine the vehicle's trajectory based on the vehicle status data; The system collects multiple key parameters reflecting the instantaneous motion state of the vehicle, namely vehicle state data, in real time through in-vehicle networks (such as CAN bus) and various sensors. It uses a classic bicycle model as the vehicle's kinematic model, inputting the vehicle state data into this model and predicting the trajectory of the center points of the four tires within a fixed time window through integral calculations. The vehicle state data can include vehicle speed v(t), steering wheel angle δ(t) (which can be converted to front wheel steering angle), yaw rate ω(t), and vehicle heading angle. (t), wheelbase L (inherent vehicle parameter), and center of gravity sideslip angle (t) (which can be simplified to 0 in low-speed scenarios), using a bicycle model to describe vehicle motion, predict the center trajectories of the four tires within the next T milliseconds (e.g., 200–500 ms). ;in This represents the two-dimensional position of a tire center at time t in the vehicle's body coordinate system (or world coordinate system); T represents the prediction time window, for example, T=0.3s, corresponding to a look-ahead distance of approximately 3.3 meters when the vehicle is traveling at 40 km / h. Inputting this data into the bicycle model yields the tire center trajectory, which is then... By mapping the coordinates (such as from the vehicle coordinate system to the lidar coordinate system) to the point cloud space, a three-dimensional trajectory line is obtained. That is, the vehicle's trajectory.

[0020] Step 104: Determine the wheel crushing state by combining the obstacle characteristics and the vehicle's trajectory; Aligning obstacle features with vehicle trajectories within the same three-dimensional spatial reference system involves: reconstructing obstacles using visually guided LiDAR based on their features to obtain a high-precision 3D point cloud set of obstacles in the LiDAR coordinate system, where each point contains precise geometric coordinates representing the spatial contour of the obstacle (e.g., height, width, slope); predicting the 3D trajectory sequence of each tire within a future time window T by integrating kinematic models based on vehicle state data (vehicle speed, steering angle, etc.); expanding the predicted 2D tire trajectories into 3D and mapping them to the LiDAR coordinate system to ensure that the obstacle point cloud and trajectory data are in the same spatial reference system; for each tire trajectory, traversing all its trajectory points and calculating its minimum Euclidean distance to the obstacle point cloud; comparing the calculated minimum distance with a pre-set distance threshold to determine the tire's crushing state; and accurately locating which wheels are affected, such as only the left front wheel crushing a manhole cover, thus determining the wheel crushing state.

[0021] Step 105: Determine the suspension adjustment strategy based on the wheel crushing condition and the type of obstacle; The system pre-establishes a mapping relationship between obstacle categories and suspension control strategies, i.e., a category-suspension parameter mapping table (obtained through real vehicle calibration or simulation learning), as follows:

[0022] Table 1 When a system determines that a tire will run over an obstacle, it immediately retrieves the corresponding baseline control parameters based on the obstacle category label. Specifically, it can adaptively adjust the baseline parameters based on the obstacle's 3D geometric information reconstructed by the LiDAR, such as height correction: the greater the actual height of the obstacle, the greater the vehicle's height adjustment. Slope correction: The steeper the slope, the more significant the damping softening / hardening effect, etc.

[0023] Step 106: Control the vehicle's movement based on the suspension adjustment strategy.

[0024] Based on the suspension adjustment strategy, adjustments are only made to the affected wheels. For example, if only the left front wheel runs over a manhole cover, a fine-tuning command is sent to the left front wheel, while the other wheels remain in their original state. If both front wheels run over a speed bump at the same time, the left and right front wheels are adjusted simultaneously, while the rear wheels remain stationary. It supports concurrent handling of multiple obstacles: for different obstacles predicted by different tires, independent control commands are generated to control the vehicle to pass smoothly.

[0025] This application embodiment acquires road surface image data and vehicle status data during vehicle operation; determines obstacle features and categories based on the road surface image data; determines the vehicle's trajectory based on the vehicle status data; determines wheel rolling status by combining the obstacle features and the vehicle's trajectory; determines a suspension adjustment strategy based on the wheel rolling status and the obstacle category; and controls vehicle operation based on the suspension adjustment strategy. This achieves a semantic closed loop from recognition to intelligent adjustment, overcoming the limitations of existing systems that only recognize obstacles but do not adjust them or provide uniform adjustment. It significantly improves smoothness and energy efficiency, and supports concurrent processing of multiple obstacles: reconstructing, judging, and controlling multiple OBBs separately, enabling multi-objective responses in complex scenarios.

[0026] Reference Figure 2 The diagram illustrates a flowchart of another embodiment of a vehicle suspension adjustment method according to this application. The vehicle suspension adjustment may specifically include the following steps: Step 201: During the vehicle's operation, acquire road surface image data and vehicle status data; During vehicle operation, the camera operates at full HD resolution and a frame rate of at least 30 FPS, with automatic exposure and white balance enabled to adapt to complex lighting conditions. It continuously captures images of the road surface area ahead, and the raw image data stream, i.e., the road image data, is transmitted in real time to the onboard computing unit for preprocessing such as scale normalization (e.g., scaling to 640×640) and pixel value normalization to adapt to subsequent perception models. Simultaneously, the system can integrate real-time vehicle speed, steering angle, and other vehicle status data from wheel speed sensors or the CAN bus to support more accurate obstacle trajectory prediction and arrival time estimation.

[0027] Step 202: Classify the road surface image data and output the obstacle categories; The system first performs preprocessing on the acquired images, such as size standardization and pixel normalization, to adapt them to the model input. The preprocessed road image is then input into the trained model, which automatically analyzes the image content and identifies whether it contains preset typical obstacles. This can be achieved by deploying a lightweight target detection model with MobileNetV3 as the backbone network and designing a prediction head that supports rotating bounding box output. After inference on the input image, the system determines the category of the target from six preset target categories (vibration strips, speed bumps, manhole covers, potholes, cement road joints, and bridge joints) and directly outputs the obstacle category.

[0028] Step 203: Perform structural analysis on the obstacles based on the obstacle categories to determine the bounding box data corresponding to the obstacle categories; A lightweight detector supporting rotated bounding boxes (OBB) is used for end-to-end inference on the input image. The obstacle category is determined by selecting the one with the highest probability from multiple typical obstacle types (vibration strips, speed bumps, manhole covers, potholes, concrete road joints, and bridge joints). Then, structural analysis is performed on the obstacle based on its category. Considering the inherent geometric features of different obstacle categories, the corresponding bounding box data, such as center coordinates, can be determined. Width w, Height and rotation angle During training, the model automatically learns the typical structural priors of different types of obstacles, such as manhole covers being approximately circular with a width-to-height ratio of about 1, and speed bumps being horizontally elongated strips with a width much greater than their height and a rotation angle of about 0. This makes the OBB output more consistent with physical laws. Compared with traditional axis-aligned bounding boxes (AABB), OBB can closely fit directional, slender obstacles (such as bridge joints and vibration strips), improving the average IoU (Intersection over Union ratio, the area of ​​the intersection of two boxes divided by the area of ​​the union) by 15–25%. OBB corner points can be accurately mapped to the LiDAR coordinate system, guiding local point cloud reconstruction and significantly improving the positioning accuracy of ROI (Region of Interest).

[0029] Step 204: Based on preset transformation parameters, perform point cloud position transformation on the bounding box data to determine the obstacle region; During the vehicle offline calibration phase, preset transformation parameters can be determined through offline calibration. Using these preset transformation parameters (mainly camera parameter matrices and rigid body transformation matrices), the key points of each OBB (such as the center point or four corner points) are transformed from the image pixel coordinate system to the vehicle or world 3D coordinate system. Typically, this is combined with depth information to construct a 3D cubic or polygonal spatial region with a specific orientation and scale, i.e., the obstacle region. The 2D visual detection results are then correlated and matched with 3D environmental perception data (such as LiDAR point clouds) to provide accurate and directional 3D obstacle region information for subsequent precise obstacle avoidance planning.

[0030] In an optional embodiment of this application, the preset transformation parameters include a camera parameter matrix and a rigid body transformation matrix. The step of performing point cloud transformation on the bounding box data based on the preset transformation parameters to determine the obstacle region includes: Sub-step S2041: Obtain the corner pixel coordinates of the bounding box data in the image pixel coordinate system; Sub-step S2042: Based on the camera parameter matrix, back-project the corner pixel coordinates to the camera 3D space to determine the 3D camera coordinates corresponding to the corner pixel coordinates; Sub-step S2043: Based on the rigid body transformation matrix, transform the three-dimensional camera coordinates to the lidar coordinate system to determine the three-dimensional corner coordinates; Sub-step S2044: Connect the coordinates of the three-dimensional corner points to determine the obstacle area.

[0031] Camera-LiDAR calibration extrinsic parameter matrix obtained during vehicle calibration phase By back-projecting the OBB from the image coordinate system to the LiDAR 3D coordinate system, the approximate region of the obstacle in 3D space is obtained. Specifically, the rigid body transformation matrix from the camera coordinate system to the LiDAR coordinate system is obtained through offline calibration (e.g., using a calibration board).

[0032] in for The rotation matrix, t represents The translation vector, i.e. This represents a 3×3 rotation matrix used to describe the alignment of the coordinate axes from the lidar coordinate system to the camera coordinate system. This represents a 3×1 translation vector, used to describe the positional offset of the camera coordinate system origin relative to the lidar coordinate system origin. Representing a row vector [0,0,0], it is a fixed format of homogeneous coordinates, where 1 is a fixed element of the homogeneous coordinates, used to unify matrix operations for rotation and translation. It is a set of 4×4 rotation matrices; Obtain the known camera parameter matrix , It is a set of 3×3 rotation matrices; Obtain the representation of the detected OBB in the image: typically consisting of the center (u,v), width... Define the height h and angle θ. Convert the OBB into the four corner points in the image. And for each OBB corner point Find the nearest neighbor of the LiDAR projection point within the neighborhood and obtain its depth Z. Use the camera parameters for backprojection to obtain the 3D point coordinates.

[0033] The camera parameters are back-projected to obtain 3D point coordinates, which are then aligned to the LiDAR coordinate system. The coordinates of the obstacle in the LiDAR coordinate system are then obtained through parameter transformation. Represents the coordinates of a point in the camera coordinate system:

[0034] The approximate area of ​​the obstacle can be determined based on its coordinates. This represents the coordinate point in the lidar coordinate system.

[0035] In an optional embodiment of this application, the step of back-projecting the corner pixel coordinates to the camera's three-dimensional space based on the camera parameter matrix to determine the three-dimensional camera coordinates corresponding to the corner pixel coordinates includes: Sub-step S1: Obtain a set of laser projection points within a preset neighborhood of the corner pixel coordinates; the set of laser projection points includes multiple candidate laser projection points; On the image plane, the pixel coordinates of the four target corner points are respectively ( ) Centered on the point cloud, a local search window (such as a 5×5 pixel or a circular neighborhood with radius r) is defined. The original point cloud of the LiDAR is traversed, and all laser projection points that fall within the window after projection are found through camera parameter projection. These are the laser projection points within the preset neighborhood of the corner pixel coordinates, forming a set of laser projection points. Considering that the density of the LiDAR point cloud is much lower than that of the image pixels, neighborhood search is used to ensure that the associated depth information can be captured.

[0036] Sub-step S2: Calculate the two-dimensional Euclidean distance between the candidate laser projection point and the corner pixel coordinates, and determine the first distance difference; For each candidate laser projection point in the set of laser projection points, calculate the two-dimensional Euclidean distance between its pixel coordinates on the image plane and the target corner point to obtain the first distance difference value, which quantifies the spatial correlation between the laser point and the corner point. The smaller the distance, the more likely the laser point is to correspond to the actual object surface pointed to by the corner point.

[0037] Sub-step S3: Arrange the candidate laser projection points in ascending order based on the first distance difference to generate a laser projection point sequence; All candidate laser projection points are sorted in ascending order according to the first distance difference, generating a laser projection point sequence and constructing a priority queue to provide a basis for selecting the most relevant point in the future. At the same time, the second-best candidates are retained to support the backtracking mechanism for handling anomalies such as when the first point is noise.

[0038] Sub-step S4: Determine the first candidate laser projection point in the laser projection point sequence as the target projection point; The first element of the sorted sequence is directly selected as the target projection point. If the first distance difference exceeds the preset threshold (e.g., >3 pixels), it is determined that the matching has failed and an exception is triggered.

[0039] Sub-step S5: Obtain the depth value of the target projection point; Sub-step S6: Combining the camera parameter matrix and the depth value, the corner pixel coordinates are transformed to determine the three-dimensional camera coordinates corresponding to the corner pixel coordinates.

[0040] The Z coordinate value (i.e., the radial depth from the point p to the LiDAR) of the target projection point p is extracted from the raw LiDAR data. This depth value is a measurement in the LiDAR coordinate system. It needs to be aligned with the camera coordinate system through coordinate transformation. Using the camera parameter matrix and the obtained depth value Z, a back-projection transformation is performed, where... The homogeneous pixel coordinates are transformed to the normalized plane and then multiplied by the depth Z to recover the three-dimensional scale, thus obtaining the accurate three-dimensional coordinates of the corner pixel coordinates in the camera coordinate system, i.e., the three-dimensional camera coordinates. This completes the dimensional reconstruction from 2D to 3D. Through local neighborhood search, nearest neighbor matching, and back projection transformation, the problem of sparse and heterogeneous LiDAR-camera data is effectively solved, and pixel-level precision depth correlation is achieved, providing reliable three-dimensional anchor points for subsequent coordinate system transformation and ROI generation.

[0041] Step 205: Perform local point cloud reconstruction on the obstacle region to determine the obstacle features; Based on the 3D region of interest (ROI), i.e., the obstacle region, obtained from visual OBB transformation, spatial filtering is performed on the original LiDAR point cloud. This can involve traversing the entire point cloud acquired by the LiDAR in real time, retaining only points falling within the axial bounding box of the ROI. Local point cloud reconstruction (such as local voxelization, point cloud completion, or surface fitting) is then performed on the original LiDAR point cloud to generate a high-fidelity obstacle point cloud. ;in N represents the set of three-dimensional point clouds of obstacles in the lidar coordinate system. , The coordinates of the i-th point are represented by the three-dimensional coordinates, typically in meters. The raw point cloud acquired by the LiDAR is preprocessed, including denoising, ground point cloud segmentation, and local normal vector estimation. Based on this, a local 3D environment map is constructed using SLAM (Simultaneous Localization and Mapping) technology, and key geometric features such as height, depth, and slope of obstacles are accurately extracted from it—i.e., obstacle features. By focusing on visually guided regions of interest (ROIs) for local mapping, global reconstruction of the entire scene's point cloud is avoided, significantly reducing computational complexity and processing latency. This meets the stringent real-time requirements of active suspension systems (typically requiring end-to-end latency below 50 ms).

[0042] Step 206: Determine the vehicle's trajectory based on the vehicle status data; The vehicle's motion status is acquired in real time via onboard sensors (such as IMU and wheel speed sensors) and the CAN bus. Core vehicle status data includes vehicle speed, steering wheel angle (or front wheel steering angle), and yaw rate. This data is then input into a bicycle model. The model simplifies the vehicle to two wheels, front and rear, using the front wheel steering angle as the core control input. Combining the vehicle's heading and speed, it integrally calculates the two-dimensional motion trajectory of each tire's center point in the vehicle coordinate system within a short future window (e.g., 300ms). The predicted two-dimensional tire trajectory is then transformed into a three-dimensional point cloud space using a known coordinate transformation matrix (describing the fixed relationship between the vehicle coordinate system and the LiDAR / world coordinate system). This generates a three-dimensional trajectory line, i.e., the vehicle's running trajectory, that can be used for spatiotemporal correlation with obstacle positions.

[0043] In an optional embodiment of this application, the step of determining the vehicle trajectory based on the vehicle status data includes: Sub-step 2061: Perform an integral operation on the vehicle state data to output the centroid position set and the vehicle heading angle; Sub-step 2062: Based on the vehicle heading angle, perform local coordinate transformation on the centroid positions in the centroid position set to generate the vehicle trajectory.

[0044] Based on the vehicle's current state (vehicle speed v, steering wheel angle δ, yaw rate ω) and kinematic model, predict the center trajectories of the four tires within the next T milliseconds (e.g., 200-500ms). ,in, Indicates the position at time t. Indicates position coordinates, These represent the lateral and longitudinal coordinates of the tire's center point at time t, respectively. In actual implementation, It can be obtained by integration from a vehicle kinematics model (such as a bicycle model):

[0045]

[0046]

[0047] in For vehicle speed, For the vehicle's heading angle, This refers to the front wheel steering angle. Wheelbase Side slip angle of the center of mass, , , These are the first derivatives with respect to time t, representing how the state evolves with time. Therefore, to obtain... It is necessary to , Integrate points:

[0048] Where x(0) and y(0) represent the two-dimensional coordinates of the vehicle's center of mass at the initial moment. This represents the vehicle's heading angle at the initial moment. This represents the definite integral operation from time 0 to time t, i.e., numerical integration of the kinematic equations over the time interval [0,T] (such as the Euler method or the Runge-Kutta method), yielding a discrete trajectory sequence of the four tire centers. The coordinates of each time step are obtained through the accumulated velocity vector, forming a smooth predicted path curve. Then, the trajectory... Through coordinate transformation, the predicted two-dimensional trajectory The three-dimensional trajectory line is obtained by transforming from the vehicle coordinate system to the lidar coordinate system. This refers to the vehicle's trajectory. This step can be achieved through preset transformation parameters, ensuring that the trajectory and point cloud data are in a unified coordinate system for subsequent spatial intersection determination. This predicted trajectory mapped into the point cloud provides the system with a geometric reference of the "area the vehicle is about to enter," enabling previously detected 3D obstacles to be more accurately assessed for collision risk and tracked based on their relative position and distance from this trajectory.

[0049] Step 207: Determine the wheel crushing state by combining the obstacle characteristics and the vehicle's trajectory; Ensure that the obstacle point cloud (from the lidar coordinate system) and the predicted 3D trajectories of the four tires (generated by the kinematic model and mapped to the same coordinate system) are in a unified 3D spatial reference system. For each tire trajectory, calculate the distance relationship between each trajectory point and all points in the obstacle point cloud. Based on the distance relationship, determine the wheel crushing state.

[0050] In an optional embodiment of this application, the step of determining the wheel crushing state by combining the obstacle features and the vehicle trajectory includes: Sub-step 2071: Map the vehicle's trajectory to the lidar coordinate system to generate a trajectory point cloud; Sub-step 2072: Based on the obstacle features and the lidar coordinate system, obtain the obstacle point cloud corresponding to the obstacle; Sub-step 2073: Determine the second distance difference between the trajectory point cloud and the obstacle point cloud. Sub-step 2074: If the second distance difference is less than a preset threshold, determine that the wheel crushing state is a crushing risk state.

[0051] The future trajectories of the four wheel centers of mass predicted based on the vehicle kinematics model (usually a two-dimensional sequence) are transformed to the lidar coordinate system through coordinate transformation and expanded into a three-dimensional set of points, i.e., a trajectory point cloud. Based on the obstacle category, OBB orientation angle, and 3D ROI region reconstructed by LiDAR, a subset of point clouds belonging to the obstacle is accurately extracted from the overall view cloud, i.e., the obstacle point cloud. For each tire track Determine whether it is related to the obstacle point cloud. Spatially, there is an intersection or proximity: calculate the Euclidean distance between the trajectory point and the nearest point cloud point. Where p is the coordinate of the point cloud closest to the trajectory in the obstacle point cloud; if there exists Make If the distance is 10 cm, then it is determined that the tire will pass through the obstacle, and the trajectory point in the trajectory point cloud is calculated. To obstacle point cloud Find the minimum Euclidean distance for each tire track (4 tracks in total), iterate through all time points t, and output the second distance difference. Quantify the spatial intersection risk between the tire and the obstacle; the smaller the distance, the higher the risk. Calculate the minimum distance. With preset safety threshold By comparison, it is determined whether there is a risk of crushing. If the second distance difference is less than a preset threshold, the wheel crushing state is determined to be a crushing risk state. It can generate high-precision and executable control trigger signals to provide a decision basis for subsequent categorized suspension adjustment.

[0052] Step 208: Determine the suspension adjustment strategy based on the wheel crushing condition and the type of obstacle; Based on the determined obstacle category (such as speed bumps, potholes, manhole covers, etc.), the corresponding benchmark adjustment parameters are retrieved from the preset prior control strategy table. Based on the actual height, slope, direction angle and other geometric features of the obstacle reconstructed by the lidar, the benchmark parameters are adaptively weighted and corrected. Based on the rolling state determination result, the adjustment is only performed on the affected wheels.

[0053] In an optional embodiment of this application, the step of determining the suspension adjustment strategy based on the wheel rolling condition and the obstacle type includes: Sub-step 2081: If the wheel crushing state is a crushing risk state, obtain the wheel identification. Sub-step 2082: Combine the wheel identification and the obstacle category to determine the suspension adjustment strategy.

[0054] When the second distance difference of a certain wheel is determined to be less than a preset threshold, the tire is confirmed to be in a state of risk of being crushed. The unique identifier ID of the tire is extracted (e.g., k=1: left front wheel, k=2: right front wheel, k=3: left rear wheel, k=4: right rear wheel). According to the obstacle type (e.g., "speed bump"), the baseline parameters are retrieved from the prior control strategy table. Combined with the actual height, slope, direction angle and other features reconstructed by the lidar, the baseline parameters are adaptively weighted (e.g., the steeper the slope, the greater the damping adjustment). The corrected parameters are only sent to the identified risky wheels (e.g., only the left front wheel), while the other wheels maintain their original state or perform compensation adjustments. This achieves independent and differentiated control of the four wheels, upgrading the suspension adjustment from vehicle-wide average to wheel-level semantic drive, improving response accuracy and energy efficiency by more than 40%.

[0055] Step 209: Control the vehicle's movement based on the suspension adjustment strategy.

[0056] The chassis domain controller (suspension ECU) converts strategy parameters (vehicle height, damping stiffness, etc.) into electronic control commands, which are then sent in real time to the four-wheel independent actuators (air springs, CDC solenoid valves, or active dampers). The intervention timing is dynamically calculated based on vehicle speed and obstacle distance, ensuring that the adjustment is completed 0.3 seconds before the tires contact the obstacle, achieving true feedforward control rather than reactive response. If the point cloud is missing or the trajectory prediction fails, the system automatically reverts to traditional IMU feedback control. It supports differentiated adjustments for different wheels in scenarios with multiple obstacles. The entire process runs on the onboard computing platform and achieves high-speed synchronization of perception, decision-making, and execution modules through the ROS / DDS (Robot Operating System / Data Distribution Service, referring to the distributed real-time communication mechanism based on the data distribution service standard in the robot operating system) communication mechanism, with an end-to-end latency of <100ms, significantly improving driving smoothness and safety.

[0057] This application embodiment acquires road surface image data and vehicle state data during vehicle operation. The road surface image data is classified to output obstacle categories. Structural analysis is performed on the obstacles based on these categories to determine the corresponding bounding box data. Point cloud position transformation is applied to the bounding box data based on preset transformation parameters to determine the obstacle region. Local point cloud reconstruction is performed on the obstacle region to determine obstacle features. The vehicle trajectory is determined based on the vehicle state data. The wheel rolling state is determined by combining the obstacle features and the vehicle trajectory. A suspension adjustment strategy is determined based on the wheel rolling state and the obstacle category. The vehicle is then controlled based on this suspension adjustment strategy. The application employs a Rotating Bounding Box (OBB) detection of directional obstacles and a vision-guided LiDAR local reconstruction strategy. Furthermore, it performs precise intersection calculations between the predicted tire trajectory and the obstacle point cloud in three-dimensional space, achieving millimeter-level geometric perception accuracy. This solves the false triggering / missed triggering problem caused by traditional 2D projection, providing a high-confidence decision basis for suspension control and significantly improving ride comfort.

[0058] To enable those skilled in the art to clearly understand the implementation process of the embodiments of this application, an example is provided below: (Refer to...) Figure 3 The system first acquires images using a monocular camera. After data preprocessing, the obstacle OBB detection model identifies the type of road obstacles and their oriented bounding boxes. Simultaneously, LiDAR acquires 3D point cloud data. The visually detected OBBs are then projected onto 3D coordinates to obtain the 3D spatial region of the obstacle. This point cloud is then preprocessed in conjunction with the LiDAR point cloud to form a precise obstacle region point cloud. Based on this, the system uses a SLAM mapping and localization module to obtain the vehicle's own position and predicts the four-wheel trajectory based on the vehicle's status. Finally, the system determines whether there is an intersection between the predicted tire trajectory and the obstacle region point cloud. If there is an intersection, suspension control is triggered for active adjustment; otherwise, the process ends. The entire process achieves a closed loop from environmental perception, obstacle localization, trajectory prediction to active control.

[0059] It should be noted that, for the sake of simplicity, the method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments of this application are not limited to the described order of actions, because according to the embodiments of this application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also understand that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of this application.

[0060] Reference Figure 4 The diagram shows a structural block diagram of an embodiment of a vehicle suspension adjustment device according to this application. The vehicle suspension adjustment device may specifically include the following modules: The acquisition module 401 is used to acquire road image data and vehicle status data during vehicle operation. The determination module 402 is used to determine obstacle features and obstacle categories based on the road surface image data; The trajectory module 403 is used to determine the vehicle's running trajectory based on the vehicle status data; The module 404 is used to determine the wheel crushing state by combining the obstacle features and the vehicle's running trajectory; The adjustment module 405 is used to determine the suspension adjustment strategy based on the wheel crushing condition and the type of obstacle; The control module 406 is used to control the vehicle's movement based on the suspension adjustment strategy.

[0061] In an optional embodiment of this application, the determining module 402 includes: The output submodule is used to classify the road surface image data and output the obstacle categories; The structural analysis submodule is used to perform structural analysis on obstacles based on the obstacle category and determine the bounding box data corresponding to the obstacle category. The transformation submodule is used to transform the point cloud position of the bounding box data based on preset transformation parameters to determine the obstacle region; The reconstruction submodule is used to perform local point cloud reconstruction of the obstacle region and determine the obstacle features.

[0062] In an optional embodiment of this application, the preset transformation parameters include a camera parameter matrix and a rigid body transformation matrix, and the transformation submodule includes: The acquisition unit is used to acquire the corner pixel coordinates of the bounding box data in the image pixel coordinate system; The back projection unit is used to back project the corner pixel coordinates to the camera three-dimensional space based on the camera parameter matrix, and determine the three-dimensional camera coordinates corresponding to the corner pixel coordinates; The transformation unit is used to transform the three-dimensional camera coordinates to the lidar coordinate system based on the rigid body transformation matrix, and to determine the three-dimensional corner coordinates. A connection unit is used to connect the three-dimensional corner point coordinates to determine the obstacle area.

[0063] In an optional embodiment of this application, the back-projection unit includes: The first acquisition component is used to acquire a set of laser projection points within a preset neighborhood of the corner pixel coordinates; the set of laser projection points includes multiple candidate laser projection points. A calculation component is used to calculate the two-dimensional Euclidean distance between the candidate laser projection point and the corner pixel coordinates, and to determine the first distance difference. A sorting component is used to sort the candidate laser projection points in ascending order based on the first distance difference to generate a laser projection point sequence. A determining component is used to determine the first candidate laser projection point in the laser projection point sequence as the target projection point; The second acquisition component is used to acquire the depth value of the target projection point; The component is used to combine the camera parameter matrix and the depth value to transform the corner pixel coordinates and determine the three-dimensional camera coordinates corresponding to the corner pixel coordinates.

[0064] In an optional embodiment of this application, the trajectory module 403 includes: The integration submodule is used to perform integration on the vehicle state data and output the centroid position set and the vehicle heading angle; The generation submodule is used to perform local coordinate transformation on the centroid positions in the centroid position set based on the vehicle heading angle, and generate the vehicle's running trajectory.

[0065] In an optional embodiment of this application, the bonding module 404 includes: The mapping submodule is used to map the vehicle's trajectory to the lidar coordinate system and generate a trajectory point cloud; The point cloud submodule is used to obtain the obstacle point cloud corresponding to the obstacle based on the obstacle features and the lidar coordinate system; The difference determination submodule is used to determine the second distance difference between the trajectory point cloud and the obstacle point cloud. The comparison submodule is used to determine that the wheel crushing state belongs to the crushing risk state when the second distance difference is less than a preset threshold.

[0066] In an optional embodiment of this application, the adjustment module 405 includes: The identification submodule is used to obtain wheel identification when the wheel crushing state is a crushing risk state; The strategy submodule is used to determine the suspension adjustment strategy by combining the wheel identifier and the obstacle category.

[0067] As the device embodiment is basically similar to the method embodiment, the description is relatively simple, and relevant parts can be found in the description of the method embodiment.

[0068] Reference Figure 5 This application also provides a vehicle, including: The processor 501 and the storage medium 502 store a computer program executable by the processor 501. When the vehicle is running, the processor 501 executes the computer program to implement the vehicle suspension adjustment method as described in any of the embodiments of this application.

[0069] The memory may include random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0070] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0071] Reference Figure 6 This application also provides a computer-readable storage medium 601, on which a computer program is stored. When the computer program is run by a processor, it executes the vehicle suspension adjustment method as described in any one of the embodiments of this application.

[0072] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0073] Those skilled in the art will understand that embodiments of this application can be provided as methods, apparatus, or computer program products. Therefore, embodiments of this application can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of this application can take the form of computer program products implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0074] This application describes embodiments with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0075] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0076] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0077] Although preferred embodiments of the present application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present application.

[0078] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0079] The foregoing has provided a detailed description of a vehicle suspension adjustment method, a vehicle suspension adjustment device, a vehicle, and a computer-readable storage medium provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for adjusting vehicle suspension, characterized in that, include: During vehicle operation, road surface image data and vehicle status data are acquired. Based on the road surface image data, the characteristics and categories of obstacles are determined; Based on the vehicle status data, the vehicle's trajectory is determined; Based on the characteristics of the obstacle and the trajectory of the vehicle, the wheel crushing state is determined; Based on the wheel crushing condition and the type of obstacle, determine the suspension adjustment strategy; The vehicle's movement is controlled based on the aforementioned suspension adjustment strategy.

2. The method according to claim 1, characterized in that, The step of determining obstacle features and obstacle categories based on the road surface image data includes: The road surface image data is classified, and obstacle categories are output. Based on the obstacle category, perform structural analysis on the obstacle to determine the bounding box data corresponding to the obstacle category; Based on preset transformation parameters, the point cloud position is transformed on the bounding box data to determine the obstacle region; Local point cloud reconstruction is performed on the obstacle region to determine the obstacle features.

3. The method according to claim 2, characterized in that, The preset transformation parameters include a camera parameter matrix and a rigid body transformation matrix. The step of performing point cloud transformation on the bounding box data based on the preset transformation parameters to determine the obstacle region includes: Obtain the corner pixel coordinates of the bounding box data in the image pixel coordinate system; Based on the camera parameter matrix, the corner pixel coordinates are back-projected onto the camera's three-dimensional space to determine the three-dimensional camera coordinates corresponding to the corner pixel coordinates; Based on the rigid body transformation matrix, the coordinates of the three-dimensional camera are transformed into the coordinate system of the lidar to determine the coordinates of the three-dimensional corner points; Connect the coordinates of the three-dimensional corner points to determine the obstacle area.

4. The method according to claim 3, characterized in that, The step of back-projecting the corner pixel coordinates to the camera's 3D space based on the camera parameter matrix to determine the corresponding 3D camera coordinates includes: Obtain a set of laser projection points within a preset neighborhood of the corner pixel coordinates; the set of laser projection points includes multiple candidate laser projection points. Calculate the two-dimensional Euclidean distance between the candidate laser projection point and the corner pixel coordinates to determine the first distance difference; Based on the first distance difference, the candidate laser projection points are sorted in ascending order to generate a laser projection point sequence; The candidate laser projection point that is first in the laser projection point sequence is determined as the target projection point; Obtain the depth value of the target projection point; By combining the camera parameter matrix and the depth value, the corner pixel coordinates are transformed to determine the three-dimensional camera coordinates corresponding to the corner pixel coordinates.

5. The method according to claim 1, characterized in that, The step of determining the vehicle's trajectory based on the vehicle status data includes: The vehicle state data is integrated to output the set of centroid positions and the vehicle heading angle; Based on the vehicle heading angle, local coordinate transformation is performed on the centroid positions in the centroid position set to generate the vehicle's trajectory.

6. The method according to claim 1, characterized in that, The step of determining the wheel crushing state by combining the obstacle features and the vehicle trajectory includes: The vehicle's trajectory is mapped to the lidar coordinate system to generate a trajectory point cloud; Based on the obstacle features and the lidar coordinate system, obtain the obstacle point cloud corresponding to the obstacle; Determine the second distance difference between the trajectory point cloud and the obstacle point cloud. If the second distance difference is less than a preset threshold, the wheel crushing state is determined to be a crushing risk state.

7. The method according to claim 1, characterized in that, The step of determining the suspension adjustment strategy based on the wheel rolling condition and the type of obstacle includes: If the wheel crushing condition is a crushing risk condition, obtain the wheel identification; Based on the wheel identification and the obstacle category, a suspension adjustment strategy is determined.

8. A vehicle suspension adjustment device, characterized in that, include: The acquisition module is used to acquire road surface image data and vehicle status data during vehicle operation. The determination module is used to determine obstacle features and obstacle categories based on the road surface image data; The trajectory module is used to determine the vehicle's running trajectory based on the vehicle status data; The module is used to determine the wheel crushing state by combining the obstacle features and the vehicle's trajectory; An adjustment module is used to determine a suspension adjustment strategy based on the wheel crushing condition and the type of obstacle; The control module is used to control the vehicle's movement based on the suspension adjustment strategy.

9. A vehicle, characterized in that, It includes a processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein the program or instructions, when executed by the processor, implement the steps of the vehicle suspension adjustment method as described in claims 1-7.

10. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions that, when executed by a processor, implement the steps of the vehicle suspension adjustment method as described in claims 1-7.

Images