A vehicle semi-active suspension control method, device, medium and product

Abstract

The application discloses a vehicle semi-active suspension control method, device, medium and product, relates to the field of vehicle semi-active suspension control, and establishes an equivalent skyhook damping control force model of a target vehicle under random road surface excitation and a suspension preview control model under local feature excitation of a road surface; whether the wheel predicted track of the target vehicle passes through the local feature area of the digital elevation map is judged, the suspension preview control model and the equivalent skyhook damping control force model are switched and used, if yes, the state variables of the target vehicle are used to solve the suspension preview control model by using a model predictive control algorithm, and the damping control force of the adjustable shock absorber is determined; if not, the state variables of the target vehicle and the skyhook damping controller parameters are used to solve the equivalent skyhook damping control force model, and the damping control force of the adjustable shock absorber is determined, so that the semi-active suspension of the target vehicle is controlled. The application improves the control precision of the semi-active suspension of the vehicle.

Description

Technical Field

[0001] This application relates to the field of semi-active vehicle suspension control, and in particular to a semi-active vehicle suspension control method, device, medium, and product. Background Technology

[0002] As a crucial component of the automotive chassis system, the suspension system has a significant impact on vehicle ride comfort and handling stability. Based on current technology, suspensions can be categorized into three main types: passive, semi-active, and active suspensions. Semi-active suspensions improve vehicle ride comfort by adjusting shock absorber damping. This not only compensates for the limitations of non-adjustable parameters in passive suspensions but also offers advantages over active suspensions, such as lower energy consumption and higher cost-effectiveness. It holds immense application potential in the mid-to-low-end car market, where controllable suspensions are not yet widespread, and has gained favor among many automakers in recent years.

[0003] Traditional semi-active suspension control algorithms typically calculate the ideal damping force based on pre-calibrated control parameters using vibration signals such as the vehicle's vertical acceleration, and then control the shock absorbers to follow this damping force. In real-world scenarios, different road types often have different control requirements, thus requiring control parameters to dynamically change with road type. Furthermore, it is desirable to use onboard vision, LiDAR, and other sensors to detect road surface features such as speed bumps and manhole covers in advance, and to precisely adjust the damping force of the shock absorbers based on the elevation information of the road surface the wheels are about to pass over. Benefiting from the booming intelligent vehicle industry, advanced sensors such as binocular cameras and LiDAR are widely used in the development of autonomous driving or assisted driving functions, providing richer hardware resources for solving the problem of intelligent road surface recognition and achieving flexible control of semi-active suspension in various scenarios. However, due to the complexity and diversity of road types and local features, various control algorithms have their own advantages and disadvantages, and a unified solution has not yet been formed for this problem. In addition, most existing solutions do not predict the vehicle's driving trajectory, making it difficult to accurately obtain the road surface elevation input for each wheel in the future time domain.

[0004] Therefore, how to accurately and efficiently identify and detect road types and local road surface features using vehicle-mounted sensors, accurately obtain the road surface elevation input of each wheel, and simultaneously use appropriate control algorithms to control the damping force of the shock absorber in real time in order to achieve the optimal control effect under complex and ever-changing working conditions is an urgent problem to be solved in the field of semi-active suspension technology. Summary of the Invention

[0005] The purpose of this application is to provide a method, device, medium, and product for controlling a vehicle's semi-active suspension, so as to improve the accuracy of vehicle semi-active suspension control.

[0006] To achieve the above objectives, this application provides the following solution:

[0007] In a first aspect, this application provides a semi-active suspension control method for a vehicle, comprising:

[0008] Establish an equivalent canopy damping control force model for the target vehicle under random road surface excitation;

[0009] Establish a suspension anti-aiming control model for the target vehicle under local road surface feature excitation;

[0010] It is determined whether the predicted wheel trajectory of the target vehicle passes through the local feature area of ​​the road surface in the digital elevation map; the predicted wheel trajectory is determined based on the wheel turning angle, longitudinal velocity at the vehicle's center of gravity, lateral velocity, vehicle yaw angle, yaw rate, distance from the vehicle's center of gravity to the front and rear axles, and the wheel track of the front and rear axles; the digital elevation map is determined based on the road image in front of the target vehicle, using target detection algorithm and semi-local stereo matching algorithm;

[0011] If so, switch to the aiming controller; the aiming controller is a suspension aiming control model;

[0012] Based on the predicted wheel trajectory and the digital elevation map, the road surface elevation of multiple predicted trajectory points is determined and input into the suspension pre-aiming control model;

[0013] Based on the state variables of the target vehicle, the suspension preview control model is solved using a model predictive control algorithm to determine the damping control force of the adjustable shock absorber in order to control the semi-active suspension of the target vehicle; the state variables include longitudinal velocity, lateral velocity, yaw angle, yaw rate and wheel rotation angle.

[0014] If not, switch to the ceiling damping controller; the ceiling damping controller is an equivalent ceiling damping control force model;

[0015] Based on the current road surface roughness level, the parameters of the roof damping controller are determined; the roof damping controller parameters include passive damping control parameters and roof damping control parameters; the current road surface roughness level is determined based on the current vibration response signal of the target vehicle using a road surface roughness level identification model; the road surface roughness level identification model is determined by training a self-attention mechanism model using a first training dataset; the first training dataset includes the vibration response signal of the training vehicle and the corresponding roughness level label; the self-attention mechanism model includes a self-attention mechanism layer, a linear layer, and a Softmax layer connected in sequence;

[0016] Based on the state variables of the target vehicle and the parameters of the roof damping controller, the equivalent roof damping control force model is solved to determine the damping control force of the adjustable shock absorber, so as to control the semi-active suspension of the target vehicle.

[0017] Optionally, the equivalent ceiling damping control force model is:

[0018]

[0019] Among them, F sky For equivalent ceiling damping control force; c p For passive damping control parameters; c sky These are the control parameters for ceiling damping; The vertical velocity of the vehicle body; c is the vertical velocity of the wheel. min F is the minimum damping coefficient of the adjustable vibration damper. out The output force of the adjustable vibration damper; F max F is the maximum output force of the adjustable vibration damper. min This is the minimum output force of the adjustable vibration damper.

[0020] Optionally, the suspension anti-aiming control model is:

[0021]

[0022] Where, N p N represents the prediction time domain; c Represents the control time domain; Q and R are weight matrices; y(k+i|k) is the vehicle output at time k+i predicted at time k; u(k+j|k) represents the vehicle control input at time k+j predicted at time k; T is the transpose.

[0023] Optionally, based on road images ahead of the target vehicle, a digital elevation map is determined using target detection algorithms and semi-local stereo matching algorithms, specifically including:

[0024] Acquire images of the road ahead of the target vehicle and images from a binocular camera;

[0025] Based on the road image, a target detection model is used to determine the region of interest (ROI) of the road image. The target detection model is obtained by training a YOLO v3 target detection model using a second training dataset. The second training dataset is obtained by using a binocular stereo camera mounted on the windshield of a vehicle to capture road images in real time ahead of the vehicle, filtering out images containing two types of road surface features (speed bumps and manhole covers), and then labeling the types and locations of these features using image annotation software. The local features include speed bumps and manhole covers.

[0026] Based on the binocular camera images, a disparity map is determined using a semi-local stereo matching algorithm;

[0027] Based on the disparity map and the intrinsic and extrinsic parameters of the binocular camera, three-dimensional reconstruction is performed on the spatial points corresponding to each pixel in the region of interest to obtain the three-dimensional point cloud of local features in the road image in the camera coordinate system.

[0028] The local features in the 3D point cloud in the camera coordinate system are transformed into the world coordinate system to obtain a digital elevation map.

[0029] Optionally, the 3D point cloud of local features in the camera coordinate system is transformed to the world coordinate system to obtain a digital elevation map, specifically including:

[0030] Based on the 3D point cloud of the local features in the camera coordinate system, a feature-point-based inter-frame registration algorithm is used to determine the spatial transformation relationship between the current frame's camera coordinate system and the world coordinate system; the spatial transformation relationship includes rotation and translation matrices.

[0031] Based on the aforementioned spatial transformation relationship, the 3D point cloud of the local features in the camera coordinate system is transformed into a 3D point cloud of the local features in the world coordinate system.

[0032] Based on the three-dimensional point cloud of the local features in the world coordinate system, a local digital elevation map is established for the current moment.

[0033] The local digital elevation map at the current moment and the global digital elevation map at the previous moment are merged to obtain the digital elevation map at the current moment.

[0034] Optionally, the process of determining the predicted wheel trajectory specifically includes:

[0035] Establish a target vehicle state equation with wheel rotation angle as input and longitudinal velocity, lateral velocity, yaw angle, and yaw rate at the vehicle's center of gravity as state variables.

[0036] The state equation of the target vehicle is discretized using the zero-order hold discretization method to obtain the recursive expression of the state variables of the target vehicle;

[0037] Collect the wheel rotation angle and state variables at the current moment; the state variables include the longitudinal velocity, lateral velocity, yaw angle, and yaw rate at the vehicle's center of gravity;

[0038] Based on the wheel angle and the state variable at the current moment, the state variables at the next K sampling moments are determined using the recursive expression of the target vehicle's state variables;

[0039] Transform the longitudinal and lateral velocities at time k into longitudinal and lateral velocities in the world coordinate system; k = 0, 1, 2, ..., K-1; k = 0 represents the current time.

[0040] Based on the longitudinal and lateral velocities in the world coordinate system, the position information of the target vehicle's centroid in the prediction time domain of M predicted trajectory points is determined, and the predicted trajectory of the target vehicle's centroid in the prediction time domain is obtained.

[0041] Based on the distance from the vehicle's center of gravity to the front and rear axles and the wheelbase parameters of the front and rear axles, the predicted trajectory of the target vehicle's center of gravity in the prediction time domain is converted into the predicted wheel trajectories of the four wheels of the target vehicle.

[0042] Optionally, the longitudinal and lateral velocities at time k are transformed into longitudinal and lateral velocities in the world coordinate system, specifically including:

[0043] Using formula Transform the longitudinal and lateral velocities at time k into longitudinal and lateral velocities in the world coordinate system; where v x k v is the lateral velocity at time k; y k V is the longitudinal velocity at time k; X k V represents the lateral velocity in the world coordinate system. Y k The longitudinal velocity in the world coordinate system; This refers to the vehicle's yaw angle.

[0044] In a second aspect, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the vehicle semi-active suspension control method described above.

[0045] Thirdly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the vehicle semi-active suspension control method described above.

[0046] Fourthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the vehicle semi-active suspension control method described above.

[0047] According to the specific embodiments provided in this application, the following technical effects are disclosed:

[0048] This application provides a semi-active suspension control method, device, medium, and product for vehicles. It establishes an equivalent roof damping control force model for the target vehicle under random road surface excitation and a suspension anti-misalignment control model under local road surface feature excitation. By determining whether the predicted wheel trajectory of the target vehicle passes through a local road surface feature area in a digital elevation map, the method switches between the suspension anti-misalignment control model and the equivalent roof damping control force model. When switching to the suspension anti-misalignment control model, the road surface elevation of multiple predicted trajectory points is determined based on the predicted wheel trajectory and the digital elevation map, and input into the suspension anti-misalignment control model. Based on the state variables of the target vehicle, a model predictive control algorithm is used to solve the suspension anti-misalignment control model to determine the damping control force of the adjustable shock absorber. When switching to the equivalent roof damping control force model, the roof damping controller parameters are determined based on the current road surface roughness level. Based on the state variables of the target vehicle and the roof damping controller parameters, the equivalent roof damping control force model is solved to determine the damping control force of the adjustable shock absorber, thereby controlling the semi-active suspension of the target vehicle. This application improves the control precision of the vehicle's semi-active suspension. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 A schematic flowchart of a semi-active suspension control method for a vehicle provided in an embodiment of this application;

[0051] Figure 2 A flowchart of a semi-active suspension control method combining intelligent road surface recognition and vehicle trajectory prediction provided in an embodiment of this application;

[0052] Figure 3 This is a schematic diagram of the equivalent ceiling damping model structure;

[0053] Figure 4 This is a schematic diagram of semi-active suspension control based on the ceiling damping algorithm;

[0054] Figure 5 This is a network structure diagram of the road surface roughness level identification method based on self-attention mechanism proposed in this application;

[0055] Figure 6 This is a schematic diagram of semi-active suspension anti-aiming control based on MPC.

[0056] Figure 7This is an application environment diagram of a vehicle semi-active suspension control method according to an embodiment of this application;

[0057] Figure 8 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0058] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0059] 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.

[0060] In an exemplary embodiment, a two-axle vehicle with independent suspensions equipped with continuously adjustable dampers on both the front and rear axles is used as the object. The damper damping control force models under random road excitation and local road feature excitation are built based on the ceiling damping algorithm and the model predictive control algorithm (MPC), respectively. The control algorithm is switched according to the results of intelligent road recognition and wheel trajectory prediction.

[0061] like Figure 1 and Figure 2 As shown, a semi-active suspension control method for a vehicle is provided, including the following S1 to S9. Wherein:

[0062] S1: Establish an equivalent ceiling damping control force model for the target vehicle under random road surface excitation.

[0063] In practical applications, an equivalent ceiling damping control force model is built under random road surface excitation.

[0064] like Figure 3 The ceiling damping and equivalent ceiling damping models shown are used to build an equivalent ceiling damping simulation model in simulation software based on their vertical dynamic equations. Figure 3 In the middle, c p c sky These represent the passive damping control parameters and the roof damping control parameters, respectively. In this embodiment, considering various ride comfort indicators of the suspension under different road surface grades i (i = A ~ E), a pair of control parameters (c) are selected based on simulation results to achieve the optimal overall performance of the suspension under random road surface excitation at the current grade. p c sky ) i .

[0065] like Figure 3 As shown, based on the principle of equivalent damping forces in the two systems, the equivalent ceiling damping control force F can be obtained. sky The expression for c is shown in formula (1); at the same time, considering the maximum and minimum output force constraints of the damper, the actual output damping force should also satisfy formula (2). Where, c min F is the minimum damping coefficient of the adjustable vibration damper. max F min These are the maximum and minimum output forces of the vibration damper, respectively.

[0066]

[0067] Among them, F sky For equivalent ceiling damping control force; The vertical velocity of the vehicle body; F is the vertical velocity of the wheel. out This refers to the output force of the adjustable vibration damper.

[0068] S2: Establish a suspension anti-aiming control model for the target vehicle under the excitation of local road surface features.

[0069] In practical applications, a suspension aiming control model based on model predictive control and local road feature excitation is built.

[0070] like Figure 3 The equivalent ceiling damping model shown is used to define the damper damping control force as F. μ Choose the state vector Output vector Input quantity u = F μ The disturbance quantity w = x r We can establish the state-space equations of a continuous system, and then discretize them using the zero-order hold discretization method, to obtain:

[0071]

[0072] Where x(k|k) is the actual system state at time k; x(k+1|k) is the system state predicted at time k+1; y(k|k) and u(k|k) represent the system output and control input predicted at time k, respectively; and w(k|k) represents the road surface excitation at time k.

[0073] Model predictive control (MPC) is essentially solving an optimal control problem. For the semi-active suspension control problem, in order to reduce the energy consumption of the adjustable dampers while ensuring vehicle ride comfort, the objective function of the optimization problem can be set as follows:

[0074]

[0075] In the formula, Np N c Let Q and R represent the prediction time domain and the control time domain, respectively; Q and R are weight matrices; y(k+i|k) is the vehicle output at time k+i predicted at time k; u(k+j|k) represents the vehicle control input at time k+j predicted at time k; and T is the transpose.

[0076] Based on the recursive relationship between the state variables described by formula (3), the optimization problem in formula (4) can be transformed into a quadratic programming problem about the control input sequence U(k) = [u(k|k), u(k+1|k), u(k+2|k), ..., u(k+Nc|k)]. Then, the optimal control input sequence U(k)* can be solved using the effective set method or other optimization methods, and u(k|k) in U(k)* is selected as the ideal damping force of the damper at time k. At time k+1, the optimization solution is performed again. This process is repeated to form a complete model predictive control process.

[0077] S3: Determine whether the predicted wheel trajectory of the target vehicle passes through the local feature area of ​​the road surface in the digital elevation map; the predicted wheel trajectory is determined based on the wheel turning angle, longitudinal velocity at the vehicle's center of gravity, lateral velocity, vehicle yaw angle, yaw rate, distance from the vehicle's center of gravity to the front and rear axles, and the wheel track of the front and rear axles; the digital elevation map is determined based on the road image in front of the target vehicle, using a target detection algorithm and a semi-local stereo matching algorithm.

[0078] The switching rule for the controller based on the predicted wheel trajectory and the digital elevation map is as follows: Based on the trajectories of the four wheels in the trajectory prediction time domain obtained in the second embodiment and the digital elevation map obtained in the first embodiment, the predicted trajectory is projected onto the digital elevation map, and it is determined whether the predicted wheel trajectory passes through the local feature area of ​​the road surface in the digital elevation map. For example... Figure 4 As shown, if the predicted wheel trajectory will not pass through local road surface features, a roof damping controller is selected. The roof damping controller parameters are chosen based on the road surface roughness level obtained from the first embodiment, and then the damping control force of the adjustable shock absorber is calculated in conjunction with the current suspension state parameters. For example... Figure 6 As shown, if the predicted wheel trajectory passes through local road features, the system switches to the aiming controller and inputs the road elevation on the predicted trajectory to the aiming controller. The aiming controller then calculates the damping control force of the adjustable shock absorber based on the current suspension state and the principle of MPC rolling optimization.

[0079] In practical applications, the controller is switched based on the predicted wheel trajectory and the digital elevation map. When switching to the aiming controller, the road surface elevation extracted from the predicted wheel trajectory in the digital elevation map is input to the aiming controller. When switching to the ceiling damping controller, the ceiling damping control parameters are selected based on the results of the road surface roughness level identification.

[0080] The process of determining the predicted wheel trajectory is as follows:

[0081] Step 1: Establish the target vehicle state equation with wheel rotation angle as input and longitudinal velocity, lateral velocity, yaw angle, and yaw rate at the vehicle's center of gravity as state variables.

[0082] Step 2: Discretize the state equation of the target vehicle using the zero-order hold discretization method to obtain the recursive expression of the state variables of the target vehicle.

[0083] In practical applications, a two-degree-of-freedom vehicle model is used to describe the dynamic characteristics of the vehicle during steering. A model is established with the wheel steering angle δ as input and the lateral velocity v at the vehicle's center of mass as input. y Vehicle yaw angle φ, yaw rate The state equation for the state variables is shown in equation (5); a suitable sampling period T is selected. T and trajectory prediction time domain length T p The state equations are discretized using the zero-order hold discretization method to obtain the recursive expressions for the state variables, as shown in formula (7). Here, m represents the vehicle mass; l f l r These represent the distances from the center of mass to the front and rear axes, respectively; C αf C αr Represents the lateral stiffness of the front and rear wheels; I z This represents the moment of inertia of the car about the z-axis.

[0084]

[0085] Step 3: Collect the wheel angle and state variables at the current moment; the state variables include the longitudinal velocity, lateral velocity, yaw angle and yaw rate at the vehicle's center of gravity.

[0086] In practical applications, vehicle speed sensors, yaw rate sensors, angular velocity sensors, and wheel rotation angle sensors are used to collect the longitudinal velocity v at the current time t0. x 0 Lateral velocity v y 0 Horizontal sway angle φ 0 yaw rate ω r 0 and wheel rotation angle δ 0And it is believed that in the trajectory prediction time domain T p Inside, the longitudinal velocity, yaw rate, and wheel rotation angle remain constant.

[0087] Step 4: Based on the wheel angle and the state variable at the current moment, use the recursive expression of the target vehicle's state variable to determine the state variables at the next K sampling moments after the current moment.

[0088] In practical applications, the collected signal values ​​are substituted into formula (7) to obtain the values ​​of the state variables at the 1st, 2nd, ..., Kth sampling times (K = Tp / T).

[0089] Step 5: Convert the longitudinal and lateral velocities at time k into longitudinal and lateral velocities in the world coordinate system; k = 0, 1, 2, ..., K-1; k = 0 represents the current time.

[0090] In practical applications, the longitudinal and lateral vehicle speeds v in the vehicle coordinate system at time k (k = 0, 1, 2, ..., K-1) are calculated according to equation (8). x k v y k The velocity components V along the X and Y axes relative to the world coordinate system XOY are transformed into these components. X k V Y k .

[0091]

[0092] Step 6: Based on the longitudinal and lateral velocities in the world coordinate system, determine the position information of the target vehicle's center of mass at M predicted trajectory points in the prediction time domain, and obtain the predicted trajectory of the target vehicle's center of mass in the prediction time domain.

[0093] In practical applications, the coordinates (X, X) of the vehicle's center of mass in the world coordinate system at time k+1 are calculated according to equation (9). k+1 Y k+1 Thus, the vehicle's centroid in the prediction time domain T is obtained. p The location information of the K predicted trajectory points.

[0094]

[0095] Step 7: Based on the distance from the vehicle's center of gravity to the front and rear axles and the wheelbase parameters of the front and rear axles, convert the predicted trajectory of the target vehicle's center of gravity in the prediction time domain into the predicted wheel trajectories of the four wheels of the target vehicle.

[0096] In practical applications, based on the distance from the vehicle's center of gravity to the front and rear axles and the wheelbase parameters of the front and rear axles, the predicted trajectory of the vehicle's center of gravity in the prediction time domain is converted into the predicted trajectories of the four wheels in the prediction time domain.

[0097] In practical applications, digital elevation maps containing local road surface elevation information are established based on target detection and binocular vision methods.

[0098] The specific process of creating a digital elevation map is as follows:

[0099] 1. Acquire images of the road ahead of the target vehicle and images from the binocular camera;

[0100] 2. Based on the road image, a target detection model is used to determine the region of interest (ROI) of the road image; wherein, the target detection model is obtained by training a YOLO v3 target detection model using a second training dataset. The second training dataset is obtained by using a binocular stereo camera mounted on the windshield of a vehicle to acquire road images in real time ahead of the vehicle, filtering out images containing two types of road surface features: speed bumps and manhole covers, and labeling the types and locations of these local features using image annotation software; the local features include speed bumps and manhole covers.

[0101] In practical applications, a binocular stereo camera mounted on the windshield of a vehicle is used to acquire road images ahead of the vehicle under different weather, lighting conditions and locations. Images containing two types of road surface features, namely speed bumps and manhole covers, are then selected. Labeling software is used to label the categories and locations of the features contained therein, thereby forming a second training dataset. The YOLO v3 object detection model is then trained based on this second training dataset.

[0102] The target detection model is deployed on an in-vehicle edge computing platform to identify road images collected in real time during vehicle operation and determine whether there are two types of local road features: speed bumps and manhole covers. If they exist, the model will return the position and size information of a rectangular bounding box containing the speed bump or manhole cover features. The region within the rectangular bounding box is then taken as the region of interest.

[0103] 3. Based on the binocular camera images, a disparity map is determined using a semi-local stereo matching algorithm.

[0104] For the acquired binocular camera images, the disparity map is obtained using the semi-local stereo matching algorithm SGBM.

[0105] 4. Based on the disparity map and the intrinsic and extrinsic parameters of the binocular camera, perform three-dimensional reconstruction of the spatial points corresponding to each pixel in the region of interest to obtain a three-dimensional point cloud of local features in the road image in the camera coordinate system.

[0106] 5. Transform the 3D point cloud of local features in the camera coordinate system to the world coordinate system to obtain a digital elevation map.

[0107] Based on the principle of binocular ranging and the pre-calibrated intrinsic and extrinsic parameters of the binocular camera, three-dimensional reconstruction is performed on the spatial points corresponding to each pixel in the region of interest to obtain the three-dimensional point cloud of local features in the road image in the camera coordinate system; the three-dimensional point cloud in the camera coordinate system is then transformed into the world coordinate system to establish a digital elevation map.

[0108] (1) Perform three-dimensional reconstruction on each spatial point. The specific method is as follows: based on the relationship between disparity and depth in binocular vision theory, convert the disparity value of the spatial point into a depth value; further, calculate the coordinate value of the spatial point in the left camera coordinate system, thereby obtaining the three-dimensional point cloud in the camera coordinate system.

[0109] (2) Based on the three-dimensional point cloud of the local features in the camera coordinate system, the spatial transformation relationship of the current frame camera coordinate system relative to the world coordinate system is determined by using the feature point-based inter-frame registration algorithm; the spatial transformation relationship includes rotation matrix and translation matrix.

[0110] In practical applications, based on the 3D point cloud in the camera coordinate system, a feature-point-based inter-frame registration algorithm is used to estimate the spatial transformation relationship between the current frame camera coordinate system and the world coordinate system, thereby converting the 3D point cloud in the current frame camera coordinate system into a 3D point cloud in the world coordinate system.

[0111] It should be noted that each time the stereo camera inputs a frame captured by each of the left and right cameras into the edge computing platform, a 3D point cloud in the camera coordinate system at the corresponding moment of the current frame is obtained. However, in order to convert this into a 3D point cloud in the world coordinate system, it is necessary to estimate the spatial transformation relationship between the two frames by using the correspondence between feature points in the previous and next frames. That is, to calculate the rotation and translation matrices of the current frame's camera coordinate system relative to the previous frame's camera coordinate system, and then obtain the rotation and translation transformation matrices of the current frame's camera coordinate system relative to the world coordinate system according to the recursive relationship. The feature point-based inter-frame registration method uses the Speeded Up Robust Feature (SURF) feature extraction method to extract feature points from the previous and next frames, and uses the Random Sample Consensus algorithm to eliminate erroneous matching points, thereby estimating the inter-frame spatial transformation relationship.

[0112] (3) Based on the spatial transformation relationship, the three-dimensional point cloud of the local features in the camera coordinate system is converted into a three-dimensional point cloud of the local features in the world coordinate system.

[0113] (4) Based on the three-dimensional point cloud of the local features in the world coordinate system, establish a local digital elevation map for the current moment.

[0114] In practical applications, a local digital elevation map is created for the current time t. For example... Figure 4 As shown, the XOY plane in the world coordinate system is divided into a grid map according to a certain grid size, and the three-dimensional point cloud obtained above is vertically projected downwards into the grid map. The average height of the three-dimensional point cloud projected by each grid is counted as the fitted value of the local feature height of the road surface at that grid location.

[0115] (5) The local digital elevation map at the current moment and the global digital elevation map at the previous moment are merged to obtain the digital elevation map at the current moment.

[0116] To create a global digital elevation map at the current time t, we need to use the averaging method to create a local digital elevation map M at the current time t. local The global digital elevation maps established at time (t) and t-1 are fused to obtain the global digital elevation map M at time t. global (t). Specifically, the method of taking the average is a weighted average method for fusion calculation, as shown in the following formula:

[0117] H global (t)=αH local (t)+(1-α)H global (t-1) (10)

[0118] Where α is the weighting coefficient, set to 0.5; H global (t), H local (t), H global (t-1) represents the fitted values ​​of local feature heights at the same grid location in the global digital elevation map at time t, the local digital elevation map at time t, and the global digital elevation map at time t-1, respectively.

[0119] S4: If yes, switch to the pre-aiming controller and execute S5-S6; the pre-aiming controller is the suspension pre-aiming control model.

[0120] S5: Based on the predicted wheel trajectory and the digital elevation map, determine the road surface elevation of multiple predicted trajectory points and input them into the suspension pre-aiming control model.

[0121] S6: Based on the state variables of the target vehicle, the suspension preview control model is solved using a model predictive control algorithm to determine the damping control force of the adjustable shock absorber, so as to control the semi-active suspension of the target vehicle; the state variables include longitudinal velocity, lateral velocity, yaw angle, yaw rate and wheel rotation angle.

[0122] S7: If not, switch to the ceiling damping controller and execute S8-S9; the ceiling damping controller is an equivalent ceiling damping control force model.

[0123] S8: Based on the current road surface roughness level, determine the ceiling damping controller parameters; the ceiling damping controller parameters include passive damping control parameters and ceiling damping control parameters; the current road surface roughness level is determined based on the current vibration response signal of the target vehicle using a road surface roughness level identification model; the road surface roughness level identification model is determined by training a self-attention mechanism model using a first training dataset; the first training dataset includes the vibration response signal of the training vehicle and the corresponding roughness level label; the self-attention mechanism model includes a self-attention mechanism layer, a linear layer, and a Softmax layer connected in sequence.

[0124] In practical applications, a stochastic road surface roughness level identification model based on a self-attention mechanism model and vehicle vibration signals is established. The specific process is as follows:

[0125] Step 1.1: In the simulation software, use a filtered white noise model to establish the road unevenness displacement input signals of the four wheels of a vehicle under five road surface levels: A, B, C, D, and E.

[0126] Step 1.2: Select the road surface unevenness displacement of the four wheels of the vehicle as the input to the whole vehicle simulation model. Considering the vehicle vibration response quantity related to the road surface unevenness level, select the vehicle body vertical acceleration. Left front suspension dynamic deflection f L1 and right rear suspension dynamic deflection f R2 The output frequency of the simulation model was set to 100Hz, and the simulation was performed for 300 seconds under each road surface condition to obtain the raw data of vehicle vibration response under five road surface grades. The raw data obtained from the simulation under each road surface grade was then sliced ​​into 300 segments with a time step of 1 second. For example... Figure 5 As shown, each segment is a sequence of length 100 and dimension 3 {x1, x2, ..., x...} 98 x 99 x 100 Each segment represents a sample, and the collection of all segments constitutes the dataset. The dataset is divided into three parts: training set, validation set, and test set, in a 6:2:2 ratio.

[0127] Step 1.3: As Figure 5As shown, the self-attention mechanism in deep learning theory is used to extract the contextual relationship of the vehicle vibration signal time series, thereby constructing a road surface roughness classification network based on the self-attention mechanism. The samples in the training dataset are used as the input of the self-attention mechanism layer, and a linear layer is connected after the self-attention mechanism layer. Since the road surface roughness classification problem is essentially a multi-class problem, a Softmax layer is connected at the end to transform the features output by the linear layer into the probability distribution of the sample belonging to the five categories.

[0128] The above describes the forward propagation process of the network model. To obtain the optimal solutions for each parameter in the network structure, it is also necessary to set a loss function and perform backpropagation iterations on the parameters. The cross-entropy loss function and the Adam optimization algorithm are selected, and hyperparameters such as learning rate and batch size are set to iteratively train the parameters in the network model. The hyperparameter values ​​are adjusted based on the model's performance on the validation dataset, and multiple training iterations are performed to obtain the model with the best performance on the validation dataset. The model's performance on the test dataset is then tested to obtain the road surface roughness classification accuracy, thus obtaining the road surface roughness classification model.

[0129] Step 1.4: Deploy the obtained self-attention mechanism model to the vehicle edge computing platform. During the actual vehicle driving process, the vehicle sensors collect the vibration response signals required by the model in real time. The data segments are still divided with a time step of 1 second and input into the network model to identify the road surface unevenness level in real time.

[0130] S9: Based on the state variables of the target vehicle and the parameters of the roof damping controller, solve the equivalent roof damping control force model to determine the damping control force of the adjustable shock absorber, so as to control the semi-active suspension of the target vehicle.

[0131] The beneficial effects of this application are reflected in:

[0132] 1. This application uses a self-attention mechanism model and vehicle vibration signals to identify road surface roughness levels. Vehicle vibration signals are easy to obtain during vehicle operation, and the self-attention mechanism model can eliminate the cumbersome feature extraction steps in traditional road surface roughness identification methods. It leverages the advantages of edge computing platforms in parallel computing, and combining the two can quickly, accurately, and robustly identify road surface roughness levels during actual vehicle operation, providing real-time and accurate road roughness information for semi-active suspension control systems.

[0133] 2. This application utilizes object detection and binocular vision methods, employing a binocular camera to identify two types of road surface features: speed bumps and manhole covers. After 3D reconstruction, a digital elevation map is created. This method accurately identifies the position and elevation information of local road surface features relative to the vehicle, providing a basis for subsequent switching control strategies of the semi-active suspension and offering road elevation prediction information to the suspension anti-suspension controller.

[0134] 3. This application combines the results of real-time vehicle trajectory prediction to extract the road surface elevation of each wheel in the trajectory prediction time domain from the digital elevation map. Real-time trajectory prediction based on the vehicle dynamics model can not only accurately determine the trajectory of each wheel in the prediction time domain and obtain a more realistic road surface elevation input from the digital elevation map, but also ensure high prediction accuracy even when the vehicle is traveling at medium and high speeds, making up for the shortcomings of previous studies in considering wheel trajectories.

[0135] 4. This application combines the results of intelligent road surface recognition and vehicle trajectory prediction to implement switching control for the semi-active suspension. Based on the equivalent ceiling damping principle and the MPC principle, an equivalent ceiling damping controller under random road surface excitation and a suspension pre-aiming controller under local road surface feature excitation are built respectively. Switching rules for the controller are set, enabling dynamic adjustment of control parameters according to the road surface unevenness level under random road surface excitation, meeting the suspension performance requirements of different road surface levels; under local road surface feature excitation, the damping force of the adjustable shock absorber can be accurately calculated based on the road surface elevation obtained from pre-aiming, greatly improving vehicle ride comfort.

[0136] The semi-active suspension control method for vehicles provided in this application can be applied to, for example... Figure 7In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be set up independently, integrated into server 104, or placed in the cloud or on other servers. Terminal 102 can send the state variables of the target vehicle to server 104. After receiving the state variables of the target vehicle, server 104 establishes an equivalent ceiling damping control force model for the target vehicle under random road surface excitation and a suspension anti-aiming control model under local road surface feature excitation. It switches between the suspension anti-aiming control model and the equivalent ceiling damping control force model based on whether the predicted wheel trajectory of the target vehicle passes through the local road surface feature area in the digital elevation map. If yes, based on the state variables of the target vehicle, the model predictive control algorithm is used to solve the suspension anti-aiming control model to determine the damping control force of the adjustable shock absorber; otherwise, based on the state variables of the target vehicle and the ceiling damping controller parameters, the equivalent ceiling damping control force model is solved to determine the damping control force of the adjustable shock absorber. Server 104 can feed back the obtained damping control force of the adjustable shock absorber to terminal 102. In addition, in some embodiments, the vehicle semi-active suspension control method can also be implemented by server 104 or terminal 102 alone. For example, terminal 102 can directly process the state variables of the target vehicle, or server 104 can obtain the state variables of the target vehicle from the data storage system and process them.

[0137] The terminal 102 can be, but is not limited to, various desktop computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, and smart in-vehicle devices. Portable wearable devices can include smartwatches, smart bracelets, and head-mounted devices. The server 104 can be implemented using a standalone server or a server cluster composed of multiple servers, or it can be a cloud server.

[0138] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 8As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores training data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a semi-active suspension control method for vehicles.

[0139] Those skilled in the art will understand that Figure 8 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0140] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0141] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0142] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0143] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0144] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0145] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0146] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0147] This document uses specific examples 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 methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A semi-active suspension control method for a vehicle, characterized in that, include: Establish an equivalent canopy damping control force model for the target vehicle under random road surface excitation; Establish a suspension anti-aiming control model for the target vehicle under local road surface feature excitation; It is determined whether the predicted wheel trajectory of the target vehicle passes through the local feature area of ​​the road surface in the digital elevation map; the predicted wheel trajectory is determined based on the wheel turning angle, longitudinal velocity at the vehicle's center of gravity, lateral velocity, vehicle yaw angle, yaw rate, distance from the vehicle's center of gravity to the front and rear axles, and the wheel track of the front and rear axles; the digital elevation map is determined based on the road image in front of the target vehicle, using target detection algorithm and semi-local stereo matching algorithm; If so, switch to the aiming controller; the aiming controller is a suspension aiming control model; Based on the predicted wheel trajectory and the digital elevation map, the road surface elevation of multiple predicted trajectory points is determined and input into the suspension pre-aiming control model; Based on the state variables of the target vehicle, the suspension preview control model is solved using a model predictive control algorithm to determine the damping control force of the adjustable shock absorber in order to control the semi-active suspension of the target vehicle; the state variables include longitudinal velocity, lateral velocity, yaw angle, yaw rate and wheel rotation angle. If not, switch to the ceiling damping controller; the ceiling damping controller is an equivalent ceiling damping control force model; Based on the current road surface roughness level, the parameters of the roof damping controller are determined; the roof damping controller parameters include passive damping control parameters and roof damping control parameters; the current road surface roughness level is determined based on the current vibration response signal of the target vehicle using a road surface roughness level identification model; the road surface roughness level identification model is determined by training a self-attention mechanism model using a first training dataset; the first training dataset includes the vibration response signal of the training vehicle and the corresponding roughness level label; the self-attention mechanism model includes a self-attention mechanism layer, a linear layer, and a Softmax layer connected in sequence; Based on the state variables of the target vehicle and the parameters of the roof damping controller, the equivalent roof damping control force model is solved to determine the damping control force of the adjustable shock absorber, so as to control the semi-active suspension of the target vehicle.

2. The vehicle semi-active suspension control method according to claim 1, characterized in that, The equivalent ceiling damping control force model is as follows: Among them, F sky For equivalent ceiling damping control force; c p For passive damping control parameters; c sky These are the control parameters for ceiling damping; The vertical velocity of the vehicle body; c is the vertical velocity of the wheel. min F is the minimum damping coefficient of the adjustable vibration damper. out The output force of the adjustable vibration damper; F max F is the maximum output force of the adjustable vibration damper. min This is the minimum output force of the adjustable vibration damper.

3. The vehicle semi-active suspension control method according to claim 1, characterized in that, Based on road images ahead of the target vehicle, a digital elevation map is determined using target detection algorithms and semi-local stereo matching algorithms, specifically including: Acquire images of the road ahead of the target vehicle and images from a binocular camera; Based on the road image, a target detection model is used to determine the region of interest (ROI) of the road image. The target detection model is obtained by training a YOLO v3 target detection model using a second training dataset. The second training dataset is obtained by using a stereo camera mounted on the windshield of a vehicle to capture road images in real time ahead of the vehicle, filtering out images containing two types of road surface features: speed bumps and manhole covers, and then using image annotation software to label the types and locations of these local features. The local features include speed bumps and manhole covers. Based on the binocular camera images, a disparity map is determined using a semi-local stereo matching algorithm; Based on the disparity map and the intrinsic and extrinsic parameters of the binocular camera, three-dimensional reconstruction is performed on the spatial points corresponding to each pixel in the region of interest to obtain the three-dimensional point cloud of local features in the road image in the camera coordinate system. The local features in the 3D point cloud in the camera coordinate system are transformed into the world coordinate system to obtain a digital elevation map.

4. The vehicle semi-active suspension control method according to claim 3, characterized in that, Transforming the 3D point cloud of local features in the camera coordinate system to the world coordinate system yields a digital elevation map, specifically including: Based on the 3D point cloud of the local features in the camera coordinate system, a feature-point-based inter-frame registration algorithm is used to determine the spatial transformation relationship between the current frame's camera coordinate system and the world coordinate system; the spatial transformation relationship includes rotation and translation matrices. Based on the aforementioned spatial transformation relationship, the 3D point cloud of the local features in the camera coordinate system is transformed into a 3D point cloud of the local features in the world coordinate system. Based on the three-dimensional point cloud of the local features in the world coordinate system, a local digital elevation map is established for the current moment. The local digital elevation map at the current moment and the global digital elevation map at the previous moment are merged to obtain the digital elevation map at the current moment.

5. The vehicle semi-active suspension control method according to claim 1, characterized in that, The process of determining the predicted wheel trajectory specifically includes: Establish a target vehicle state equation with wheel rotation angle as input and longitudinal velocity, lateral velocity, yaw angle, and yaw rate at the vehicle's center of gravity as state variables. The state equation of the target vehicle is discretized using the zero-order hold discretization method to obtain the recursive expression of the state variables of the target vehicle; Collect the wheel rotation angle and state variables at the current moment; the state variables include the longitudinal velocity, lateral velocity, yaw angle, and yaw rate at the vehicle's center of gravity; Based on the wheel angle and the state variable at the current moment, the state variables at the next K sampling moments are determined using the recursive expression of the target vehicle's state variables; Transform the longitudinal and lateral velocities at time k into longitudinal and lateral velocities in the world coordinate system; k = 0, 1, 2, ..., K-1; k = 0 represents the current time. Based on the longitudinal and lateral velocities in the world coordinate system, the position information of the target vehicle's centroid in the prediction time domain of M predicted trajectory points is determined, and the predicted trajectory of the target vehicle's centroid in the prediction time domain is obtained. Based on the distance from the vehicle's center of gravity to the front and rear axles and the wheelbase parameters of the front and rear axles, the predicted trajectory of the target vehicle's center of gravity in the prediction time domain is converted into the predicted wheel trajectories of the four wheels of the target vehicle.

6. The vehicle semi-active suspension control method according to claim 5, characterized in that, The longitudinal and lateral velocities at time k are transformed into longitudinal and lateral velocities in the world coordinate system, specifically including: Using formula Transform the longitudinal and lateral velocities at time k into longitudinal and lateral velocities in the world coordinate system; where v x k v is the lateral velocity at time k; y k V is the longitudinal velocity at time k; X k V represents the lateral velocity in the world coordinate system. Y k The longitudinal velocity in the world coordinate system; This refers to the vehicle's yaw angle.

7. A computer device, comprising: The memory and processor contain a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the vehicle semi-active suspension control method according to any one of claims 1-6.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the vehicle semi-active suspension control method according to any one of claims 1-6.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the vehicle semi-active suspension control method according to any one of claims 1-6.

Images