A method and system for controlling a vehicle's suspension system, and a storage medium.

Abstract

This invention relates to a vehicle suspension system control method and system, and a storage medium. The invention first fuses current road surface images, road surface point cloud information, and vehicle status information, and after considering historical road surface roughness levels, obtains the target road surface roughness level that best reflects the target road surface condition. Simultaneously, based on the target road surface roughness level, it predicts the feedforward damping force of the suspension system when the vehicle is about to travel on the target road surface. Furthermore, while the vehicle is traveling on the target road surface based on this feedforward damping force, it can calculate in real time, based on the vehicle's body parameters, the feedback damping force used to compensate for the feedforward damping force, enabling the vehicle to maintain stable driving on the target road surface and further improving the vehicle's ride comfort performance under uneven road conditions based on this control strategy.

Description

Technical Field

[0001] This invention relates to the automotive field, and more particularly to a method and system for controlling a vehicle's suspension system, as well as a storage medium. Background Technology

[0002] As the performance and user experience of new energy vehicles continue to improve, users have increasingly higher requirements for vehicle ride comfort, handling stability, and safety. Traditional suspension control mainly relies on vibration state feedback control, but this method suffers from significant latency, resulting in large initial impacts on uneven roads and a poor ride experience. Furthermore, semi-active suspension damping control primarily uses classic control methods such as ceiling damping, floor damping, and acceleration damping, requiring parameter calibration based on engineers' experience, and these control parameters remain fixed after calibration under typical operating conditions. In addition, traditional control methods do not consider the impact of unknown external disturbances, vehicle mass, and the nonlinear characteristics of the shock absorbers on ride comfort. Under complex road conditions and driving conditions, the vehicle's ride comfort and handling performance need further improvement. Therefore, it is necessary to design an adjustable damping force control method with low control latency and good adaptive robustness. Summary of the Invention

[0003] One objective of this invention is to provide a vehicle suspension system control method, system, and storage medium. By predicting the unevenness level of the target road surface that the vehicle will traverse, a feedforward control strategy is designed to control the vehicle's suspension system in advance. Furthermore, based on the vehicle's real-time state, an anti-disturbance adaptive feedback control strategy is designed to compensate for the feedforward control input, thereby improving the ride comfort when the vehicle traverses uneven road surfaces.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0005] A method for controlling a vehicle's suspension system, comprising:

[0006] Acquire the current road surface image and road surface point cloud information of the target road surface in a first direction, as well as the vehicle's status information; the first direction is the vehicle's driving direction;

[0007] The current road surface image, the road surface point cloud information, and the status information are input into a preset road surface recognition model to obtain the target road surface roughness level of the target road surface;

[0008] Based on the target road surface roughness level, predict the feedforward damping force required by the vehicle's suspension system when the vehicle is traveling on the target road surface;

[0009] When the vehicle travels to the target road surface based on the feedforward damping force, the feedback damping force is calculated in real time based on the vehicle's body parameters, and the feedforward damping force is compensated in real time based on the feedback damping force so that the vehicle can travel smoothly on the target road surface.

[0010] In the aforementioned vehicle suspension system control method, the step of calculating the feedback damping force in real time based on the vehicle's body parameters includes:

[0011] Construct alternative Lyapunov functions;

[0012] While maintaining the first derivative of the alternative Lyapunov function less than or equal to 0, determine the first formula for calculating the feedback damping force;

[0013] The feedback damping force is calculated according to the first calculation formula.

[0014] In the aforementioned vehicle suspension system control method, the construction of alternative Lyapunov functions includes:

[0015] The suspension system is decoupled into multiple sub-suspension systems in different parts, and a two-degree-of-freedom vibration model is established based on the sub-suspension systems and the vehicle body parameters.

[0016] The alternative Lyapunov function is constructed based on the sliding modal variables determined by the two-degree-of-freedom vibration model, the estimation error of the radial basis function weight coefficients, the estimation error of the radial basis function, the estimation error of the radial basis center, and the uncertainty of the disturbance force estimation error.

[0017] In the above-mentioned vehicle suspension system control method, the method further includes:

[0018] The relative displacement and relative velocity of the lower mass of the suspension system are determined based on the two-degree-of-freedom vibration model.

[0019] The relative displacement of the lower mass is multiplied by the sliding mode gain coefficient, and then the relative velocity of the lower mass is added to obtain the sliding mode variable.

[0020] In the above-mentioned vehicle suspension system control method, the method further includes:

[0021] The nominal disturbance force and the estimated disturbance force are determined based on the sliding mode variables.

[0022] Subtracting the estimated disturbance force from the nominal disturbance force yields the disturbance force estimation error; the disturbance force estimation error is composed of the product of the weighting coefficient and the radial basis function estimation error, the product of the weighting coefficient error and the radial basis function estimator, and the sum of the optimal approximation error;

[0023] The radial basis function estimation error is decomposed into the product of the partial derivative of the radial basis function with respect to the center of the radial basis and the estimation error of the center of the radial basis, the product of the partial derivative of the radial basis function with respect to the width of the radial basis and the estimation error of the width of the radial basis, and the sum of the higher-order terms after the Taylor expansion of the radial basis function.

[0024] Based on the product of the sliding mode variable, the uncertainty gain coefficient, and the first matrix, and the product of the sliding mode variable and the adjustable gain coefficient, the uncertainty of the disturbance force estimation error is obtained.

[0025] The first matrix consists of a first constant, a radial basis center estimate, a radial basis width estimate, and a weight coefficient estimate.

[0026] In the above-mentioned vehicle suspension system control method, the method further includes:

[0027] The radial basis function is obtained by subtracting the norm of the radial basis center from the sliding modal variable and then dividing by the radial basis width.

[0028] The nominal perturbation force is determined by the sum of the product of the radial basis function and the weighting coefficient and the optimal approximation error.

[0029] The estimated disturbance force is obtained by replacing the weight coefficients, radial basis center, and radial basis width in the nominal disturbance force with the corresponding estimates and discarding the optimal disturbance approximation error term.

[0030] In the above-mentioned vehicle suspension system control method, the first variable in the first calculation formula includes the radial basis function neural network weight coefficient, the radial basis center, the radial basis width, and the uncertainty of the disturbance force estimation error; the calculation of the feedback damping force according to the first calculation formula includes:

[0031] The radial basis function neural network weight coefficients, radial basis center, radial basis width, and disturbance force estimation error uncertainty are calculated in real time based on the real-time vehicle body parameters.

[0032] Substituting the radial basis function neural network weight coefficients, the radial basis center, the radial basis width, and the uncertainty of the disturbance force estimation error into the first calculation formula, the feedback damping force is obtained.

[0033] In the aforementioned vehicle suspension system control method, the preset road surface recognition model includes at least a first model, a second model, and a third model; the step of inputting the current road surface image, the road surface point cloud information, and the state information into the preset road surface recognition model to obtain the target road surface roughness level includes:

[0034] The current road surface image is input into the first model, and the first road surface feature in the current road surface image is extracted using the first model. The first road surface feature is then converted into a first road surface roughness level.

[0035] The road surface point cloud information is input into the second model, and the second model is used to extract the second road surface features from the road surface point cloud information. The second road surface features are then converted into a second road surface roughness level.

[0036] The state information is input into the third model to obtain the third road surface roughness level;

[0037] The first road surface roughness level, the second road surface roughness level, the third road surface roughness level, and the historical road surface roughness level of the target road surface are fused to obtain the target road surface roughness level.

[0038] In the above-mentioned vehicle suspension system control method, the method further includes:

[0039] Obtain historical road marking information corresponding to the target road surface; the historical road marking information is obtained after the target road surface was marked when the vehicle previously traveled to the target road surface.

[0040] The historical road marking information includes the historical road unevenness level, historical road coordinates, historical road length, and historical path curvature corresponding to the target road surface.

[0041] In the above-mentioned vehicle suspension system control method, the state information includes the vehicle's current first driving speed, and the method further includes:

[0042] Determine the current first position coordinates of the vehicle;

[0043] Based on the first position coordinates and the historical road surface coordinates, a first distance is determined between the first position of the vehicle and the starting position of the target road surface; the first position is the center position of the upper front axle of the vehicle.

[0044] Based on the first distance and the first driving speed, determine the first time required for the vehicle to travel to the starting position;

[0045] The control strategy is activated when the first duration is less than the first threshold and / or the first distance is less than the second threshold.

[0046] In the above-mentioned vehicle suspension system control method, the method further includes:

[0047] Based on the first driving speed and the historical road length, determine the second time required for the vehicle to travel the target road.

[0048] If the control policy is enabled and the second time period has elapsed, then the control policy is disabled.

[0049] In the above-mentioned vehicle suspension system control method, after real-time compensation of the feedforward damping force based on the feedback damping force, the method further includes:

[0050] Adjust the height of the suspension system to the target height, and / or adjust the stiffness of the air springs in the suspension system to the target stiffness.

[0051] In the above-mentioned vehicle suspension system control method, the method further includes:

[0052] Determine the current air suspension mode set for the vehicle;

[0053] The target height and the target stiffness are determined based on the air suspension mode, historical path curvature, the state information, and the target road surface roughness level.

[0054] In the above-mentioned vehicle suspension system control method, the method further includes:

[0055] If the vehicle is detected to be accelerating or braking while traveling on the target road surface, the target height is adjusted according to the vehicle's current second driving speed and the vehicle's first information;

[0056] The first information refers to the current position of the accelerator pedal and / or the current pressure of the brake cylinder in the vehicle.

[0057] In the above-mentioned vehicle suspension system control method, the method further includes:

[0058] If the vertical acceleration of the front axle wheels of the vehicle is greater than the third threshold value, it is determined that the vehicle has reached the starting point of the target road surface, and the current first position coordinates and the current first time point of the vehicle are determined.

[0059] If the vertical acceleration of the front axle wheels of the vehicle is less than the fourth threshold value, it is determined that the vehicle has left the target road surface, and the current second position coordinates and the current second time point of the vehicle are determined.

[0060] The first road length of the target road surface is determined based on the first position coordinates and the second position coordinates, and / or the first time point, the second time point, and the current third driving speed of the vehicle;

[0061] The target path curvature is determined based on the first path curvature and the historical path curvature, and the historical road surface marking information is updated based on the target path curvature, the target road surface unevenness level, the first road surface length, and the first location coordinates.

[0062] A vehicle suspension control system, the vehicle suspension control system comprising:

[0063] A binocular camera is used to acquire a current image of the target road surface in the first direction;

[0064] Radar is used to acquire road point cloud information of the target road surface in the first direction; the first direction is the vehicle's driving direction.

[0065] Vehicle body sensors are used to acquire vehicle status information;

[0066] The suspension controller is used to input the current road surface image, the road surface point cloud information, and the state information into a preset road surface recognition model to obtain the target road surface roughness level; predict the feedforward damping force required by the vehicle's suspension system when the vehicle is traveling on the target road surface based on the target road surface roughness level; when the vehicle travels to the target road surface based on the feedforward damping force, calculate the feedback damping force in real time based on the vehicle's body parameters, and compensate the feedforward damping force in real time based on the feedback damping force, so as to make the vehicle travel smoothly on the target road surface.

[0067] This invention first fuses current road surface images, road surface point cloud information, and vehicle state information, and then considers historical road surface roughness levels to obtain the target road surface roughness level that best reflects the target road surface condition. Therefore, based on this target road surface roughness level, the most accurate control strategy can be obtained. Furthermore, the damping force of the shock absorber determined in the control strategy is based on feedback control damping force and feedforward damping force, which can reduce the impact and vibration acceleration of the vehicle when entering the target road surface, further improving the vehicle's ride comfort performance under uneven road conditions. Simultaneously, the feedback control damping force can utilize the nonlinear characteristics of radial basis function neural networks to design a self-learning control strategy to adaptively compensate for external disturbances. The feedback control damping force can quickly eliminate the vibration acceleration of the vehicle body and wheels after entering the target road surface. In addition, the real-time update law of the rate of change of the radial basis function neural network weight coefficients, the radial basis center, the radial basis width, and the uncertainty of the disturbance force estimation error proposed in this invention greatly reduces the parameter calibration workload of the feedback control algorithm. Attached Figure Description

[0068] Figure 1A flowchart illustrating a vehicle suspension system control method provided in an embodiment of the present invention;

[0069] Figure 2 A schematic diagram of the composition structure of a vehicle suspension control system provided in an embodiment of the present invention;

[0070] Figure 3 A schematic diagram of an exemplary two-degree-of-freedom model provided for an embodiment of the present invention;

[0071] Figure 4 This is an exemplary flowchart for determining the roughness level of a target road surface, provided as an embodiment of the present invention.

[0072] Figure 5 This is an exemplary flowchart illustrating the determination of target height and target stiffness, provided for an embodiment of the present invention.

[0073] Figure 6 A schematic flowchart illustrating an exemplary process for determining the start time of a control strategy, provided as an embodiment of the present invention;

[0074] Figure 7 This is a schematic diagram illustrating an exemplary process for updating historical road marking information, provided as an embodiment of the present invention.

[0075] Figure 8 A schematic diagram illustrating an exemplary process for determining the optimal special road surface roughness level, provided as an embodiment of the present invention;

[0076] Figure 9 An exemplary signal processing flow provided for embodiments of the present invention. Figure 1 ;

[0077] Figure 10 An exemplary signal processing flow provided for embodiments of the present invention. Figure 2 ;

[0078] Figure 11 This is a schematic diagram of the composition of a vehicle suspension control device provided in an embodiment of the present invention. Detailed Implementation

[0079] The embodiments of the present invention will be described below with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention and not for limiting the scope of protection of the present invention.

[0080] As the performance and user experience of new energy vehicles continue to improve, users have increasingly higher requirements for vehicle ride comfort, handling stability, and safety. The gradual maturation of intelligent sensing, intelligent networking, and electronically controlled suspension technologies provides new technical pathways for improving overall suspension performance. The suspension controller adjusts the height, damping, and stiffness of the suspension system based on vehicle driving status and road conditions to adapt to changes in road conditions, effectively improving vehicle comfort. Currently, semi-active suspension damping control mainly relies on classic control methods such as ceiling, floor, and acceleration damping. However, after parameter calibration, these methods struggle to adapt to complex road conditions and changes in vehicle driving status, failing to guarantee smooth performance under multiple operating conditions. Furthermore, they address issues such as large impact vibrations and untimely control when vehicles pass over speed bumps, manhole covers, and other special road surfaces. While pure machine vision-based road condition prediction control is emerging, it still has significant room for improvement in terms of road surface recognition accuracy, scene coverage, and reliability. These issues are key factors limiting the performance improvement of semi-active suspension systems. Pure machine vision-based road condition prediction control mainly includes the following methods:

[0081] 1. By utilizing onboard driving data such as vehicle speed and acceleration, and map navigation route information, the boundary values ​​of the semi-active suspension damping coefficient are determined based on information such as the road surface roughness level, functional level, and category of the road segment where the vehicle is located. The suspension damping coefficient value is then calculated using a control algorithm. This approach addresses the impact of localized special road surfaces on ride comfort. However, this method only allows for suspension control through feedback control, which results in a lag.

[0082] 2. Two road surface information is calculated by using wheel vibration signals and road surface image recognition and uploaded to the cloud server. The suspension is controlled by receiving adjustment instructions issued by the cloud server through the receiving module. However, this method does not consider the fusion of multimodal road surface information and does not explain how to specifically control the suspension system.

[0083] 3. The received binocular images are used to reconstruct and extract three-dimensional road surface point clouds, and then the road surface unevenness level is calculated. Finally, the driver is reminded by the indicator light according to the road surface unevenness level. However, this method identifies road surface unevenness by pure road surface images, and its accuracy and reliability need to be improved.

[0084] 4. High-precision map acquisition vehicles and positioning base station equipment are used to scan road slopes and elevations. Based on the acquired three-dimensional point cloud icon information, local and global maps are built. Then, the elevation information of the vehicle's driving area is estimated, and the height and damping of the vehicle's active suspension are controlled. However, this method is characterized by high cost and long cycle in producing high-precision maps. Moreover, it is difficult to update the high-precision map in real time when the road conditions change, and the maintenance cost of the high-precision map is very high.

[0085] Electronically controlled suspension systems offer advantages such as low energy consumption, compact size, low cost, and continuous damping control, and currently have promising market prospects in mid-to-high-end vehicles. However, the control delays caused by signal transmission, feedback control calculations, and actuator response result in significant lag in electronically controlled suspension systems, hindering further improvements in ride comfort and handling stability. For example, air spring control also suffers from inherent issues with insufficient adjustment speed. Therefore, a systematic solution to the lag problem in semi-active suspension control is needed.

[0086] In view of the above-mentioned technical problems, the present invention provides a vehicle suspension system control method, which is implemented by a vehicle suspension control system. It can adjust the vehicle suspension system before the vehicle reaches a special road surface, thereby improving the vehicle's ride comfort, handling stability and safety under comprehensive operating conditions.

[0087] Figure 1 This is a flowchart illustrating a vehicle suspension system control method according to an embodiment of the present invention, as shown below. Figure 1 As shown, the process includes the following steps S101 to S103:

[0088] Step S101: Obtain the current road surface image and road surface point cloud information of the target road surface in the first direction, as well as the vehicle status information; the first direction is the vehicle's driving direction.

[0089] In an embodiment of the present invention, reference is made to Figure 2 The vehicle's suspension control system 200 includes at least a binocular camera 2001, a radar 2002, a body sensor 2003, an acceleration sensor 2004, a suspension height sensor 2005, a communication module 2006, a cloud server 2007, a vehicle networking module 2008, a suspension controller 2009, an air spring 2010, and a shock absorber 2011.

[0090] In some embodiments, the vehicle's suspension control system 200 acquires the current road surface image of the target road surface in the vehicle's driving direction through the binocular camera 2001 in the image acquisition device, acquires the road surface point cloud data of the target road surface through the radar 2002 in the image acquisition device, and acquires the vehicle's status information through the vehicle body sensor 2003.

[0091] In some embodiments, the vehicle's state information may be the vehicle's speed or vibration acceleration, etc.; the specific state information can be determined according to the actual situation, and the embodiments of the present invention do not impose specific limitations here.

[0092] Step S102: Input the current road surface image, road surface point cloud information and status information into the preset road surface recognition model to obtain the target road surface unevenness level.

[0093] In this embodiment of the invention, after acquiring the current road surface image, road surface point cloud data and vehicle status information, the vehicle suspension control system 200 inputs the current road surface image, road surface point cloud information and status information into a preset road surface recognition model through the suspension controller 2009 to obtain the target road surface roughness level of the target road surface.

[0094] In some embodiments, after acquiring the current road surface image of the target road surface in the vehicle's driving direction, the binocular camera 2001 uploads the current road surface image to the cloud server 2007 via the vehicle networking module 2008. Simultaneously, after acquiring the road surface point cloud data of the target road surface, the radar 2002 also uploads the current road surface image to the cloud server 2007 via the vehicle networking module 2008. Furthermore, after acquiring the vehicle's status information, the vehicle body sensor 2003 also uploads the current road surface image to the cloud server 2007 via the vehicle networking module 2008. Based on this, the suspension controller 2009 needs to download the current road surface image, road surface point cloud data, and status information from the cloud server 2007 to the suspension controller 2009. Then, the current road surface image, road surface point cloud data, and status information are respectively input into the pre-trained preset road surface recognition model in the suspension controller 2009 to obtain the target road surface roughness level of the target road surface.

[0095] In some embodiments, the target road surface roughness level refers to the degree of variation of the target road surface at the ideal road surface elevation; the specific target road surface roughness level can be determined according to the actual situation, and the embodiments of the present invention do not make specific limitations here.

[0096] Step S103: Predict the feedforward damping force required by the vehicle's suspension system when the vehicle is driving on the target road surface based on the target road surface roughness level.

[0097] In this embodiment of the invention, after obtaining the target road surface roughness level, the suspension controller 2009 in the vehicle suspension control system 200 predicts the feedforward damping force required by the vehicle's suspension system when the vehicle is driving on the target road surface based on the target road surface roughness level.

[0098] In some embodiments, the suspension controller 2009 can predict the feedforward damping force required by the vehicle's suspension system when the vehicle is driving on the target road surface based on the predicted target road surface roughness level, thereby enabling the suspension system of the vehicle to be controlled in advance to pass through the target road surface, thereby improving the driving experience of the vehicle.

[0099] It is understood that this invention mainly targets the suspension system control strategy when a vehicle is driving on special road surfaces. Special road surfaces mainly include speed bumps, manhole covers, damaged road surfaces, jointed road surfaces, raised road surfaces, settled road surfaces, and long-wave road surfaces, etc., which are discontinuous road surfaces. Special road surfaces can bring an uncomfortable experience to the driver or passengers of the vehicle. Based on the technical solution of this invention, the vehicle can adjust its suspension system before driving on special road surfaces and restore the suspension system after driving on them, thereby improving the driving experience on special road surfaces.

[0100] Step S104: When the vehicle travels to the target road surface based on the feedforward damping force, the feedback damping force is calculated in real time based on the vehicle body parameters, and the feedforward damping force is compensated in real time based on the feedback damping force so that the vehicle can travel smoothly on the target road surface.

[0101] In some embodiments, after the suspension controller 2009 in the vehicle's suspension control system 200 has traveled on the target road surface based on the feedforward damping force, since there are uncertain disturbance forces on the target road surface, the suspension controller 2009 can also calculate the feedback damping force in real time based on the vehicle's body parameters, and compensate the feedforward damping force in real time based on the feedback damping force, so as to make the vehicle travel smoothly on the target road surface.

[0102] In this embodiment of the invention, the target damping force of the shock absorber 2011 is determined according to one or more of the following: feedback control damping force, which is the damping force required for the vehicle to maintain stability while driving on the target road surface; and feedforward damping force, which is the predicted damping force required for the vehicle to drive on the target road surface.

[0103] Understandably, the feedback control damping force changes in real time and needs to be continuously maintained to ensure the stability of the vehicle when driving on the target road surface. Therefore, the feedback control damping force is constantly adjusted according to the vehicle's driving process on the target road surface in order to maintain the vehicle's stability.

[0104] In this embodiment of the invention, the method for calculating the feedback damping force in real time based on the vehicle's body parameters includes: constructing a candidate Lyapunov function; determining a first formula for calculating the feedback damping force while maintaining the first derivative of the candidate Lyapunov function less than or equal to 0; and calculating the feedback damping force according to the first formula.

[0105] In this embodiment of the invention, constructing an alternative Lyapunov function includes: decoupling the suspension system into multiple sub-suspension systems in different parts; establishing a two-degree-of-freedom vibration model based on the sub-suspension systems and vehicle body parameters; and constructing an alternative Lyapunov function based on the uncertainties in the sliding mode variables, radial basis function neural network weight coefficient estimation errors, radial basis function estimation errors, radial basis center estimation errors, and disturbance force estimation errors determined by the two-degree-of-freedom vibration model.

[0106] Specifically, the suspension system is decoupled into four relatively independent suspension subsystems, and a two-degree-of-freedom vibration model of the suspension subsystems is established. The model consists of inertial force, spring force, basic damping force of the shock absorber, adjustable damping force of the shock absorber, and unknown disturbance forces from the random road surface on the suspension system. Model parameters include the suspension system's lower spring mass, wheelbase, track width, spring stiffness coefficient, basic damping coefficient of the shock absorber, and the position and velocity relative to the lower spring mass. The displacement and velocity relative to the lower spring mass are directly measured by suspension height sensors and wheel acceleration sensors, respectively, and then calculated through differentiation and integration.

[0107] In this embodiment of the invention, the relative displacement and relative velocity of the lower mass of the suspension system are determined according to the two-degree-of-freedom vibration model; the relative displacement of the lower mass is multiplied by the sliding mode gain coefficient and then added to the relative velocity of the lower mass to obtain the sliding mode variables.

[0108] In this embodiment of the invention, the radial basis function is obtained by subtracting the norm of the radial basis center from the sliding modal variable and then dividing it by the radial basis width; the nominal perturbation force is determined by the product of the radial basis function and the weighting coefficients and the sum of the optimal approximation error; the weighting coefficients, the radial basis center, and the radial basis width in the nominal perturbation force are replaced with the corresponding estimates, and the optimal perturbation approximation error term is discarded.

[0109] Specifically, the relative displacement with respect to the lower mass is multiplied by the sliding mode gain coefficient, and then the relative velocity with respect to the lower mass is added to obtain the sliding mode variable. The sliding mode variable is then subtracted from the norm of the radial basis center and divided by the radial basis width to obtain the radial basis function. The radial basis function is multiplied by the weighting coefficients and then added to the optimal approximation error to obtain the nominal perturbation force. The weighting coefficients, radial basis center, and radial basis width in the nominal perturbation force are replaced with their corresponding estimators, and the optimal perturbation approximation error term is discarded to obtain the estimated perturbation force.

[0110] In this embodiment of the invention, the nominal disturbance force and the estimated disturbance force are determined based on the sliding mode variables; the estimated disturbance force is subtracted from the nominal disturbance force to obtain the disturbance force estimation error; the disturbance force estimation error is composed of the product of the weight coefficient and the radial basis function estimation error, the product of the weight coefficient error and the radial basis function estimate, and the sum of the optimal approximation error; the radial basis function estimation error is decomposed into the product of the partial derivative of the radial basis function with respect to the radial basis center and the estimation error of the radial basis center, the product of the partial derivative of the radial basis function with respect to the radial basis width and the estimation error of the radial basis width, and the sum of the higher-order terms after the Taylor expansion of the radial basis function; the uncertainty of the disturbance force estimation error is obtained based on the product of the sliding mode variables, the uncertainty gain coefficient, and the first matrix, and the product of the sliding mode variables and the adjustable gain coefficient; wherein, the first matrix is ​​composed of the first constant, the radial basis center estimate, the radial basis width estimate, and the weight coefficient estimate.

[0111] Specifically, the disturbance force estimation error is obtained by subtracting the estimated disturbance force from the nominal disturbance force. The disturbance force estimation error is then transformed into a linear sum of three terms: the first term is the weight coefficient multiplied by the radial basis function estimation error, the second term is the weight coefficient error multiplied by the radial basis function estimator, and the third term is the optimal approximation error.

[0112] Furthermore, the radial basis function estimation error is linearized into a sum of three terms: the first term is the partial derivative of the radial basis function with respect to the center of the radial basis multiplied by the estimation error of the center of the radial basis; the second term is the partial derivative of the radial basis function with respect to the width of the radial basis multiplied by the estimation error of the width of the radial basis; and the third term is all the higher-order terms after the Tyler series expansion of the radial basis function.

[0113] Furthermore, the uncertainty of the disturbance estimation error is obtained by superimposing two terms. The first term is the sliding mode variable multiplied by the uncertainty gain coefficient and then multiplied by a column matrix composed of constant, radial basis center estimate, radial basis width estimate and weight coefficient estimate. The second term is the sliding mode variable multiplied by the adjustable gain coefficient.

[0114] In this embodiment of the invention, the first variable in the first calculation formula includes the radial basis function (RBF) neural network weight coefficient, the radial basis center, the radial basis width, and the uncertainty of the disturbance force estimation error. Calculating the feedback damping force according to the first calculation formula includes: calculating the RBF neural network weight coefficient, the radial basis center, the radial basis width, and the uncertainty of the disturbance force estimation error in real time based on real-time vehicle body parameters; and substituting the RBF neural network weight coefficient, the radial basis center, the radial basis width, and the uncertainty of the disturbance force estimation error into the first calculation formula to obtain the feedback damping force.

[0115] Specifically, the uncertainties in the estimation of the sliding modal variables, radial basis function (RBF) neural network weight coefficients, RBF function, RBF center, and disturbance force are squared, then divided by the adjustable weight coefficients and multiplied by one-half to obtain the candidate Lyapunov function. The first derivative of the candidate Lyapunov function with respect to time is calculated. Then, the first derivatives of the RBF neural network weight coefficient estimation errors, RBF function estimation errors, RBF center estimation errors, and lumped disturbance estimation errors are substituted into the first derivative of the candidate Lyapunov function, ensuring that the first derivative of the candidate Lyapunov function remains constantly less than zero. In the process of proving the stability of the suspension control system, the formula for calculating the adjustable damping force of the suspension system, as well as the real-time update laws for the rates of change of the RBF neural network weight coefficients, the RBF center, the RBF width, and the disturbance force estimation error are derived and constructed.

[0116] Furthermore, the rate of change of the real-time updated radial basis function (RBF) neural network weight coefficients, the rate of change of the radial basis center, the rate of change of the radial basis width, and the rate of change of the uncertainty of the lumped disturbance estimation error are integrated to obtain the real-time updated RBF neural network weight coefficients, radial basis center, radial basis width, and uncertainty of the lumped disturbance estimation error.

[0117] Furthermore, the updated radial basis function neural network weight coefficients, radial basis center, radial basis width, and uncertainty of disturbance force estimation error are substituted into the adjustable damping force calculation formula to obtain the feedback damping force based on the suspension system state calculation.

[0118] The following is an exemplary calculation process for the feedback damping force provided by this invention:

[0119] 1. Reference Figure 3 The vehicle's overall suspension system is decoupled into a quarter-suspension system consisting of left front, right front, left rear, and right rear sections. This quarter-suspension system is then simplified into a two-degree-of-freedom vibration model. According to Newton's second law of motion, the dynamic equation for the upper mass of the spring is:

[0120]

[0121] In the above formula, m s Let z be the mass of the upper part of the spring. s Let be the vertical displacement of the upper mass of the spring. Let be the vertical velocity of the mass at the top of the spring. Let z be the vertical acceleration of the mass at the top of the spring. u This represents the vertical displacement of the mass at the bottom of the spring. Let c be the vertical velocity of the mass at the bottom of the spring. s k is the basic damping coefficient of the shock absorber in the sub-suspension system.s Let f be the helical spring stiffness of the shock absorber in the sub-suspension system, and f be the feedback damping force.

[0122] According to Newton's second law of motion, the second equation of motion for the lower mass of the spring is:

[0123]

[0124] In the above formula, m u The mass of the lower part of the spring. The vertical accelerations of the lower mass of the spring are k and k, respectively. u For the vertical stiffness of the tire corresponding to the sub-suspension system, z r The target road surface is given random road surface excitation.

[0125] 2. Establish a matrix-based semi-active suspension control model. Rearrange formulas (1) and (2) into a matrix-vector form, which can be expressed as:

[0126]

[0127] In the above formula, d is an unknown concentrated disturbance vector, which includes parameter uncertainty, unmodeled dynamics, random road surface excitation, and noise.

[0128] 3. Calculate the sliding modal variable s and its derivative.

[0129]

[0130] In the above formula, λ is the adjustable gain coefficient matrix, defined as λ = diag(λ1, λ2).

[0131] 4. Calculate the nominal disturbance force using a radial basis function neural network. To overcome the influence of the unknown disturbance force on the suspension control system, a radial basis function neural network is used to estimate the external disturbance force. The nominal disturbance force d is defined as:

[0132] d = W T h(s,c,b)+ε a (6);

[0133] In the above formula, W is the optimal weight matrix, and ε a The optimal approximation error is bounded and satisfies ||ε a ||≤b ε b ε It is a positive definite constant. It is a radial basis function vector. The center vector of the radial basis functions. denoted as the center vector component of the radial basis function, m is the number of hidden layers in the radial basis neural network, and N is the number of radial basis centers. is the width vector of the radial basis functions.

[0134] 5. Define the estimated disturbance force d of the nominal disturbance force. for:

[0135]

[0136] In the above formula, This is the estimated value of the weight matrix. The center estimation vector of the radial basis functions. Let be the width estimation vector of the radial basis functions, and s be the input of the neural network. This is the output of the neural network.

[0137] 6. Calculate the estimation error of the disturbance force. According to formulas (6) and (7), we can obtain:

[0138]

[0139] In the above formula, The estimation bias of the weight matrix W, Estimate the bias vector for the radial basis function.

[0140] 7. The radial basis function vector Perform a Taylor series expansion:

[0141]

[0142] In the above formula, These are the higher-order terms after the Taylor series expansion. Furthermore, matrices A and B can be represented as:

[0143]

[0144]

[0145] Substituting equation (9) into equation (8), we obtain the perturbation estimation error of the radial basis neural network. Third-party programs:

[0146]

[0147] The uncertainty δ of the disturbance estimation error can be expressed as:

[0148]

[0149] By calculating the boundary associated with the disturbance vector, we can obtain:

[0150]

[0151] According to the rules of matrix inequality operations, the norm of δ can be expressed by formula (14):

[0152]

[0153] Add to both sides of the inequality (15) It can be further converted into:

[0154]

[0155] The perturbation boundary gain matrix k in equation (16) δ The neural network estimation vector ν is defined as follows:

[0156]

[0157] Furthermore, the estimated value of the uncertainty δ of the disturbance estimation error is... Defined as:

[0158]

[0159] 8. Prove the stability of the feedback damping force control algorithm, and derive the calculated value of the adjustable damping force and the adaptive law. Define the alternative Lyapunov function as:

[0160]

[0161] In the above formula, Γ w ,Γ c ,Γ b and Γ δ Let be an adjustable positive definite matrix, and tr(·) be the trace of the matrix.

[0162] Taking the first derivative of the candidate Lyapunov function:

[0163]

[0164] 9. In the process of proving the stability of the control system, the following formula is derived, in which the feedback control damping force f can be expressed as:

[0165]

[0166] Meanwhile, the real-time update law for the rate of change of weight coefficients in a radial basis function neural network is defined as follows:

[0167]

[0168] The real-time update law for the radial basis center rate of change is defined as follows:

[0169]

[0170] Define the real-time update law for the radial basis width change rate as follows:

[0171]

[0172] The real-time update law for the rate of change of the uncertainty in the disturbance estimation error is:

[0173]

[0174] Substituting equations (22) to (26) into equation (21), we get:

[0175]

[0176] According to Lyapunov's stability theorem, the suspension damping control system is determined to be asymptotically stable, and the disturbance estimation error will converge to near zero within a finite time. The real-time updated rates of change of the radial basis function (RBF) neural network weight coefficients, the RBF center, the RBF width, and the lumped disturbance estimation error uncertainty are integrated to obtain the real-time updated RBF neural network weight coefficients, RBF center, RBF width, and lumped disturbance estimation error uncertainty. Finally, these values ​​are substituted into the formula for calculating the feedback control damping force f to obtain the feedback damping force of the suspension subsystem.

[0177] In some embodiments, the damping force of the vehicle's suspension system may also include one or more of the following: handling stability control damping force and safe driving control damping force.

[0178] For example, the Skyhook Control-Ground Hook Control (SH-GH) algorithm is used to calculate the control damping force for maneuvering stability. The SH-GH algorithm is calculated based on the first velocity direction of the damper 2011, the second velocity direction of the upper mass of the air spring 2010, and the acceleration direction of the upper mass of the air spring 2010.

[0179] For example, when calculating the safe driving control damping force, it is necessary to first determine the current driving state of the vehicle based on the vehicle's state information, that is, whether the vehicle is currently in a stable driving state or not. If the vehicle is currently in a stable driving state based on the vehicle's state information, the third damping force is defaulted to a preset value, such as 0. If the vehicle is currently in an unstable driving state based on the vehicle's state information, the current driving speed of the vehicle is determined based on the state information, and the safe driving control damping force is calculated based on the current driving speed and the pre-calibrated damping force.

[0180] In this embodiment of the invention, the preset road surface recognition model includes at least a first model, a second model, and a third model; the suspension controller 2009 in the vehicle's suspension control system 200 inputs the current road surface image, road surface point cloud information, and status information into the preset road surface recognition model to obtain the target road surface roughness level, and refers to... Figure 4 This includes the following steps S401 to S404:

[0181] Step S401: Input the current road surface image into the first model, use the first model to extract the first road surface features from the current road surface image, and convert the first road surface features into the first road surface roughness level.

[0182] In some embodiments, the first model is pre-trained using sample road surface images. During daily driving, the vehicle can acquire sample road surface images through a binocular camera 2001 and upload them to a cloud server 2007 via a vehicle networking module 2008. Simultaneously, the sample road surface images can be labeled with features (manual or automatic) and output as first data samples for training the first model. The first data samples include sample road surface images and sample road surface types. First, the sample road surface images are input into the first model, which can extract the sample road surface image features. Then, the sample road surface image features are converted into the first sample road surface roughness level and sample path curvature corresponding to the target road surface. Finally, the first model is trained based on the first sample road surface roughness level and the labeled sample road surface type.

[0183] In some embodiments, the vehicle can continuously acquire road surface images and train the first model. The vehicle networking module 2008 can periodically send the trained first model to the suspension controller 2009. At this time, after downloading the current road surface image from the cloud server 2007, the suspension controller 2009 can input the current road surface image into the trained first model, use the first model to extract the first road surface features in the current road surface image, and convert the first road surface features into a first road surface roughness level and a first path curvature.

[0184] Step S402: Input the road surface point cloud information into the second model, use the second model to extract the second road surface features from the road surface point cloud information, and convert the second road surface features into the second road surface roughness level.

[0185] In some embodiments, the second model is pre-trained using sample road surface point cloud information. During daily driving, the vehicle can scan the road surface in the direction of travel using radar 2002 to obtain three-dimensional road surface point cloud coordinates, which is the sample road surface point cloud information. The sample road surface point cloud information is uploaded to the cloud server 2007 through the vehicle networking module 2008. At the same time, the sample road surface image can be labeled with features (manual or automatic) to establish a second data sample for deep learning. The second model is trained using the second data sample, that is, the second model is iteratively trained using the second data sample, and finally a second model that meets the recognition accuracy is obtained.

[0186] In some embodiments, the vehicle can continuously collect road surface point cloud information and train the second model. The vehicle networking module 2008 can periodically send the trained second model to the suspension controller 2009. At this time, after downloading the current road surface point cloud information from the cloud server 2007, the suspension controller 2009 can input the road surface point cloud information into the trained second model, use the neural network algorithm in the second model to extract the second road surface features from the road surface point cloud information, and convert the second road surface features into the second road surface roughness level.

[0187] Step S403: Input the status information into the third model to obtain the third road surface roughness level.

[0188] In some embodiments, the third model is pre-trained using sample state information. During daily driving, the vehicle can obtain vehicle state information, such as vehicle speed and vibration acceleration, through the vehicle body sensor 2003 to obtain sample state information. The sample state information is then uploaded to the cloud server 2007 through the vehicle networking module 2008. At the same time, the sample state information can be used to establish a third data sample through feature annotation (manual annotation or automatic annotation). The third model is trained using the third data sample, that is, the third model is iteratively trained using the third data sample, and finally a third model that meets the recognition accuracy is obtained.

[0189] In some embodiments, the vehicle can continuously acquire the vehicle's state information and train the third model. The vehicle networking module 2008 can periodically send the trained third model to the suspension controller 2009. At this time, after downloading the current state information from the cloud server 2007, the suspension controller 2009 can input the state information into the trained third model, output the third road surface unevenness level, and also output the first road surface length of the target road surface.

[0190] Step S404: Integrate the information of the first road surface roughness level, the second road surface roughness level, the third road surface roughness level, and the historical road surface roughness level of the target road surface to obtain the target road surface roughness level.

[0191] In some embodiments, after inputting the current road surface image, road surface point cloud information, and status information into the first model, the second model, and the third model respectively to obtain the first road surface roughness level, the second road surface roughness level, and the third road surface roughness level for the target road surface under different evaluation dimensions, the suspension controller 2009 can comprehensively determine the target road surface roughness level together with the historical road surface roughness level in the historical road surface marking information for the target road surface issued by the vehicle network module 2008.

[0192] In this embodiment of the invention, the suspension controller 2009 acquires historical road marking information corresponding to the target road surface; the historical road marking information is obtained after the vehicle has marked the target road surface when it has traveled to it in the past; wherein, the historical road marking information includes the historical road surface roughness level, historical road surface coordinates, historical road surface length and historical path curvature corresponding to the target road surface.

[0193] In some embodiments, during vehicle operation, the vehicle networking module 2008 can send historical road surface marking information of the target road surface to the suspension controller 2009 based on the real-time positioning of the vehicle. The suspension controller 2009 can perform multimodal information fusion of the first road surface roughness level, the second road surface roughness level, the third road surface roughness level, and the historical road surface roughness level to obtain the most accurate target road surface roughness level.

[0194] In this embodiment of the invention, after the suspension controller 2009 performs real-time compensation of the feedforward damping force based on the feedback damping force, it further includes: adjusting the height of the suspension system to the target height, and / or adjusting the stiffness of the air spring in the suspension system to the target stiffness.

[0195] In some embodiments, the suspension controller 2009 can pre-calculate the control strategy of the suspension system when passing through a target road surface, such as the height at which the suspension system should pass through the target road surface, the target stiffness of the air spring 2010 in the suspension system, or the feedforward damping force of the shock absorber 2011 in the suspension system, etc. Then, the target height of the suspension system, the target stiffness of the air spring 2010, or the feedforward damping force of the shock absorber 2011 are respectively converted into the target control current of the solenoid valve, so as to adjust the suspension system at an appropriate time through the target control current, such as adjusting the suspension system when the vehicle is about to enter the target road surface.

[0196] In this embodiment of the invention, the suspension controller 2009 determines the target height of the suspension system and the target stiffness of the air spring 2010, with reference to... Figure 5 This includes the following steps S501 to S502:

[0197] Step S501: Determine the current air suspension mode of the vehicle.

[0198] In some embodiments, the suspension controller 2009 can determine the currently selected air suspension mode of the vehicle, and then determine the current height of the vehicle's suspension system based on the air suspension mode.

[0199] Step S502: Determine the target height and target stiffness based on the air suspension mode, historical path curvature, status information, and target road surface roughness level.

[0200] In some embodiments, after determining the current height of the vehicle's suspension system, the suspension controller 2009 can input the signals corresponding to the currently set air suspension mode, historical path curvature, status information, and target road surface roughness level into the signal processing module for signal analysis and validity judgment. Based on this, the signal processing module can input the processed signals into the air spring mode management module. The air spring mode management module can calculate the current mode of the air spring 2010 and the target stiffness of the air spring 2010. The suspension target height calculation module can calculate the target height of the suspension system. The air spring safety control module can calculate the safety control flag and safety height of the air spring 2010. Finally, the air spring control module converts the target height of the suspension system and the target stiffness of the air spring 2010 into the target current of the corresponding solenoid valve.

[0201] In this embodiment of the invention, after determining the control strategy, the suspension controller 2009 can also determine at what time the suspension system should be adjusted, referring to... Figure 6 This includes the following steps S601 to S604:

[0202] Step S601: Determine the current first position coordinates of the vehicle.

[0203] In some embodiments, the suspension controller 2009 can continuously collect the vehicle's current position coordinates, i.e., the first position coordinates, during driving, and thus determine the activation time of the control strategy based on the first position coordinates.

[0204] Step S602: Determine the first distance between the vehicle's first position and the starting position of the target road surface based on the first position coordinates and the historical road surface coordinates; the first position is the center position of the vehicle's upper front axle.

[0205] In some embodiments, since the vehicle networking module 2008 sends historical road marking information corresponding to the target road surface to the suspension controller 2009, the first distance, that is, the lateral distance, between the center position of the upper front axle of the vehicle and the starting position of the target road surface can be determined based on the historical road surface coordinates included in the historical road surface marking information and the first position coordinates.

[0206] Step S603: Determine the first time required for the vehicle to travel to the starting position based on the first distance and the first driving speed.

[0207] In some embodiments, after determining the first distance between the center position of the upper front axle of the vehicle and the starting position of the target road surface, the suspension controller 2009 calculates the travel time required for the center position of the upper front axle of the vehicle to travel to the starting position of the target road surface, i.e., the first duration, based on the first travel speed of the vehicle.

[0208] Step S604: If the first duration is less than the first threshold and / or the first distance is less than the second threshold, adjust the suspension system.

[0209] In some embodiments, if the first duration is short or the first distance is short, it indicates that the vehicle is about to reach the target road surface and the suspension system needs to be adjusted.

[0210] In this embodiment of the invention, after adjusting the suspension system, the suspension controller 2009 also needs to determine at what time to restore the suspension system. The specific steps include: determining the second time required for the vehicle to travel the target road surface based on the first driving speed and the historical road length; and restoring the suspension system after the control strategy is activated and the second time has elapsed.

[0211] In some embodiments, since the historical road marking information includes the historical road length, the suspension controller 2009 can calculate how long it will take to leave the target road based on the first driving speed and the historical road length. Based on this, the suspension system can be restored after adjustments are made to the suspension system.

[0212] In this embodiment of the invention, during the vehicle activation control strategy while driving on the target road surface, the target height of the suspension system can be adjusted under special circumstances. The specific steps include: when the vehicle is detected to be accelerating or braking while driving on the target road surface, the target height is determined and adjusted according to the vehicle's current second driving speed and the vehicle's first information; wherein, the first information is the current position of the accelerator pedal and / or the current pressure of the brake cylinder in the vehicle.

[0213] In some embodiments, when the suspension controller 2009 detects that the vehicle is accelerating or braking during driving, it calculates and determines a control amount to adjust the height of the vehicle suspension system based on the vehicle's second driving speed, the position of the accelerator pedal, and / or the current pressure of the brake cylinder, thereby suppressing the vehicle's pitch attitude.

[0214] In this embodiment of the invention, the suspension controller 2009 also needs to update the historical road marking information for the target road surface in the cloud server 2007, referring to... Figure 7 This includes the following steps S701 to S704:

[0215] Step S701: When the vertical acceleration of the front axle wheels of the vehicle is greater than the third threshold value, determine that the vehicle has reached the starting point of the target road surface, and determine the current first position coordinates and the current first time point of the vehicle.

[0216] In some embodiments, when a vehicle is about to enter the target road surface, since there is an elevation difference between the target road surface and the road surface that the vehicle is currently traveling on, it can be determined that the vehicle has reached the starting point of the target road surface when the vertical acceleration of the front axle wheels of the vehicle is detected to be greater than the third threshold value. The starting coordinate point of the target road surface, i.e. the first position coordinate, is calculated based on the vehicle's size parameters, and the current first time point is recorded as the starting time point of the vehicle entering the target road surface.

[0217] Step S702: If the vertical acceleration of the front axle wheels of the vehicle is less than the fourth threshold value, determine that the vehicle has left the target road surface, and determine the current second position coordinates and the current second time point of the vehicle.

[0218] In some embodiments, when a vehicle is about to leave the target road surface, since there is an elevation difference between the target road surface and the ordinary road surface that the vehicle is about to travel on, it can be determined that the vehicle has left the target road surface if the vertical acceleration of the front axle wheels of the vehicle is less than the fourth threshold value. The endpoint coordinates of the target road surface, i.e., the second position coordinates, are calculated based on the vehicle's size parameters, and the current second time point is recorded as the final time point when the vehicle leaves the target road surface.

[0219] Step S703: Determine the first road length of the target road surface based on the first position coordinates, the second position coordinates, and / or the first time point, the second time point, and the vehicle's current third driving speed.

[0220] In some embodiments, after determining the starting coordinate point of the target road surface, i.e., the first position coordinate, and the ending coordinate point of the target road surface, i.e., the second position coordinate, the first road surface length of the target road surface can be calculated based on the first position coordinate and the second position coordinate. Alternatively, after determining the starting time point when the vehicle enters the target road surface, i.e., the first time point, and the final time point when the vehicle leaves the target road surface, i.e., the second time point, the first road surface length of the target road surface can be calculated based on the first time point, the second time point, and the vehicle's driving speed.

[0221] Step S704: Determine the target path curvature based on the first path curvature and the historical path curvature, and update the historical road marking information based on the target path curvature, the target road surface unevenness level, the first road surface length, and the first position coordinates.

[0222] In some embodiments, after determining the first path curvature of the target road surface according to the first model, the suspension controller 2009 can also calculate the actual path curvature of the target road segment based on the first path curvature and the original path curvature.

[0223] In some embodiments, the historical road surface coordinates corresponding to the target road surface in the historical road surface marking information can be updated according to the first coordinate position, the historical road surface roughness level corresponding to the target road surface in the historical road surface marking information can be updated according to the target road surface roughness level, the historical road surface length corresponding to the target road surface in the historical road surface marking information can be updated according to the first road surface length, and the historical path curvature corresponding to the target road surface in the historical road surface marking information can be updated according to the actual path curvature to obtain the updated road surface marking information. Based on this, before the vehicle enters the target road surface again, the corresponding suspension system control strategy can be calculated according to the updated road surface marking information.

[0224] This invention provides a method for controlling a vehicle's suspension system. The method includes: acquiring a current road surface image and road surface point cloud information of a target road surface in a first direction, as well as vehicle state information; the first direction being the vehicle's driving direction; inputting the current road surface image, road surface point cloud information, and state information into a preset road surface recognition model to obtain the target road surface roughness level; determining the control strategy of the suspension system when the vehicle is driving on the target road surface based on the target road surface roughness level; using the above implementation scheme, firstly, the current road surface image, road surface point cloud information, and vehicle state information are fused, and then, considering historical road surface roughness levels, the method obtains the control strategy that best reflects the target road surface roughness level. The target road surface roughness level is used to determine the road surface roughness. Therefore, the most accurate control strategy can be obtained based on this target road surface roughness level. The damping force of the shock absorber determined in the control strategy is based on the feedback control damping force and the feedforward damping force, which can reduce the impact and vibration acceleration of the vehicle when it enters the target road surface. This further improves the driving comfort performance of the vehicle under rough road conditions based on this control strategy. At the same time, the feedback control damping force can immediately design a self-learning control strategy for the nonlinear characteristics of the radial basis neural network to adaptively compensate for external disturbances. Therefore, the feedback control damping force can quickly eliminate the vibration acceleration of the vehicle body and wheels after the vehicle enters the target road surface.

[0225] Based on the above embodiments, the present invention provides an exemplary vehicle suspension system control method, applied to a vehicle suspension control system, with reference to... Figure 2 The specific steps are as follows:

[0226] Step S1: The vehicle network module 2008 receives road surface image information (sample road surface image in the above embodiment), three-dimensional road surface point cloud (sample road surface point cloud information in the above embodiment), and vehicle status parameters (sample status information in the above embodiment) collected by the vehicle, and uploads them to the cloud server 2007. The special road surface recognition model (preset road surface recognition model in the above embodiment) is trained using the road surface image information, three-dimensional road surface point cloud, and vehicle status parameters.

[0227] Step S1.1: Obtain road surface image information in the direction of vehicle travel through the binocular camera 2001. The vehicle networking module 2008 uploads the road surface image information to the cloud server 2007, and then stores the road surface image information. After feature annotation, a road surface image sample dataset is formed. Using the road surface image sample dataset, the first road surface recognition model (the first model in the above embodiment) is trained, and the first road surface recognition model that meets the recognition accuracy requirements is output.

[0228] Step S1.2: The radar 2002 acquires a three-dimensional road surface point cloud along the vehicle's driving direction. The vehicle networking module 2008 uploads the three-dimensional road surface point cloud to the cloud server 2007 and stores it. A road surface point cloud training sample is established through feature annotation. The road surface point cloud training sample is used to iteratively train the second road surface recognition model (the second model in the above embodiment) to obtain a second road surface recognition model that meets the recognition accuracy.

[0229] Step S1.3: Obtain vehicle status parameters through vehicle body sensor 2003, vehicle network module 2008 uploads vehicle status parameters to cloud server 2007 and stores them to establish vehicle status dataset; use vehicle status dataset to iteratively train the third road surface recognition model based on vehicle status (the third model in the above embodiment) to obtain a third road surface recognition model that meets the recognition accuracy requirements.

[0230] Step S2: The vehicle networking module 2008 downloads the pre-trained special road surface recognition model and the special road surface information marked on the map (historical road surface marking information in the above embodiment) to the suspension controller 2009. The pre-trained special road surface recognition model includes a first road surface recognition model, a second road surface recognition model, and a third road surface recognition model. The special road surface information includes the special road surface roughness level (historical road surface roughness level in the above embodiment), the special road surface length (historical road surface length in the above embodiment), and the path curvature (historical path curvature in the above embodiment).

[0231] Step S3: The vehicle network module 2008 receives the current road surface image information (current road surface image in the above embodiment), three-dimensional road surface point cloud (road surface point cloud information in the above embodiment), and vehicle status parameters (status information in the above embodiment) collected by the vehicle and uploads them to the cloud server 2007. The suspension controller 2009 downloads them from the cloud server 2007, and then inputs the downloaded current road surface image information, three-dimensional road surface point cloud, and vehicle status parameters into the first road surface recognition model, the second road surface recognition model, and the third road surface recognition model, respectively, to obtain the first special road surface roughness level and path curvature, the second special road surface roughness level, and the third special road surface roughness level.

[0232] Step S3.1, Reference Figure 8 The current road surface image information 801 is input into the first road surface recognition model 802. The first road surface recognition model 802 is used to extract the first road surface features from the current road surface image information 801 and convert the first road surface features into the first special road surface unevenness level and path curvature.

[0233] Step S3.2, Reference Figure 8The three-dimensional road surface point cloud 803 is input into the second road surface recognition model 804. The second road surface recognition model 804 is used to extract the second road surface features from the three-dimensional road surface point cloud 803 and convert the second road surface features into a second special road surface unevenness level.

[0234] Step S3.3, Reference Figure 8 The vehicle state parameter 805 is input into the third road surface recognition model 806. The third road surface recognition model 806 is used to extract the third road surface features from the vehicle state parameter 805 and convert the third road surface features into the third special road surface unevenness level.

[0235] Step S4, Reference Figure 8 Multimodal road information fusion 808 is performed using the special road surface information 807 marked on the map, the first special road surface roughness level, the second special road surface roughness level, and the third special road surface roughness level to obtain the optimal special road surface roughness level 809 (the target road surface roughness level in the above embodiment).

[0236] Step S4.1: The vehicle networking module 2008 sends special road surface information marked on the map to the suspension controller 2009 based on the real-time positioning of the vehicle. The special road surface information includes the original unevenness level, the original length and curvature of the special road surface. After obtaining the original unevenness level of the special road surface, the suspension controller 2009 can use multimodal road surface information fusion to fuse the first special road surface unevenness level, the second special road surface unevenness level, the third special road surface unevenness level and the special road surface inhomogeneity level to calculate the optimal special road surface unevenness level.

[0237] Step S4.2: When the front axle wheels are subjected to vertical excitation from the special road surface, the accelerometer 2004 identifies the characteristics of the vertical acceleration signal of the front axle wheels and determines the starting point of the special road surface. The current vehicle position information and timestamp are recorded, and the coordinates of the special road surface are calculated based on the vehicle's dimensional parameters. Similarly, when the accelerometer 2004 identifies the characteristics of the front axle wheel acceleration signal and determines that the vehicle has exited the special road surface, the current vehicle position information and timestamp are recorded. Based on the vehicle position information, timestamps, and speed when entering and exiting the special road surface, as described above, the length of the special road surface is calculated.

[0238] Step S5: Calculate the target damping force for ride comfort control, handling stability control, driving safety control, and special road surface feed-out control.

[0239] Specific, for reference Figure 9The CAN signal output from the communication module, the sensor signals output from the body sensor 2003 and the suspension height sensor 2005 are input to the signal processing module 901 for analysis and effective judgment. Then, the damper mode management module 902 outputs the current mode of the damper 2011. Then, the self-learning anti-disturbance damping control module 903, the handling stability control module 904, the driving safety control module 905, and the special road surface feedforward control module 906 respectively calculate and determine the self-learning anti-disturbance control damping force (feedback control damping force in the above embodiment), the handling stability target damping force, the safe driving target damping force, and the special road surface feedforward target damping force (feedforward damping force in the above embodiment) of the damper 2011. The optimal damping force arbitration module 907 arbitrates the damping force calculated as described above to obtain the optimal damping force of the damper 2011, that is, the target damping force. Then, the target current of the solenoid valve of the damper 2011 is calculated by the solenoid valve target current module 1008.

[0240] Step S5.2: Calculate and adjust the control values ​​of the air spring 2010 stiffness and shock absorber 2011 damping force based on the original path curvature, driver input, and vehicle driving status to suppress vehicle roll. When vehicle acceleration or braking is detected, calculate the target damping force of the shock absorber to suppress vehicle pitch based on vehicle speed and accelerator pedal position or brake cylinder pressure.

[0241] Step S6: Calculate the suspension height and stiffness control of the air spring 2010 based on signals such as the user's selected air suspension mode, driver input, vehicle speed, and special road surface unevenness level.

[0242] Step S6.1, Reference Figure 10 The CAN signal and sensor signal are input to the signal processing module 901 for signal parsing and validity judgment. The air spring mode management module 1001 calculates the current mode of the air spring 2010, the suspension target height and stiffness. The suspension target height calculation module 1002 calculates the suspension target height based on the air spring mode and vehicle status signal. The air spring safety control module 1003 outputs the air spring safety control flag and target height based on the vehicle status signal. Finally, the air spring control module 1004 calculates the target current of the solenoid valve.

[0243] Step S6.2: If the lateral distance between the vehicle's front axle center and the special road surface exceeds a critical value or the driving time exceeds a critical value, the suspension adjustment function will be suppressed. Conversely, the damping force of the shock absorber 2011 will be adjusted according to the unevenness level of the special road surface.

[0244] Step S6.3: Convert the target height of the suspension system, the target stiffness of the air spring, and the target damping force of the shock absorber into the target control current of the corresponding solenoid valve.

[0245] Step S7: Calculate the actual path curvature of the target road segment based on the first path curvature and the original path curvature, upload the optimal special road surface unevenness level, special road surface length, actual path curvature and vehicle location information to the cloud server 2007, and mark it on the map.

[0246] Understandably, the vehicle suspension control method proposed in this invention first utilizes binocular cameras, radar, and vehicle status signals to determine the special road surface roughness level, special road surface length, and path curvature of the target road surface through multimodal information fusion. Simultaneously, it combines special road surface information from a cloud server to calculate the optimal special road surface information through secondary information fusion. Then, leveraging the nonlinear approximation characteristics of radial basis function neural networks (RBNs) and suspension vibration state feedback, a self-learning, disturbance-resistant semi-active suspension damping control strategy is proposed. The weight matrix, center vector, and width vector of the RBN can achieve online self-learning to adapt to changes in road conditions, providing adaptive robustness for the semi-active suspension damping control. Under the controller's action, the suspension height, shock absorber damping, and air spring stiffness are adjusted in advance beyond visual range. Based on suspension state feedback, the adjustable damping force of the semi-active suspension is self-learned to resist disturbances, thereby improving the vehicle's ride comfort.

[0247] Based on the above embodiments, in another embodiment of the present invention, a vehicle suspension control system is provided. Figure 2 A schematic diagram of the composition structure of a vehicle suspension control system provided by the present invention is shown below. Figure 2 As shown, the vehicle's suspension control system 200 includes:

[0248] The binocular camera 2001 is used to acquire the current road surface image of the target road surface in the first direction;

[0249] Radar 2002 is used to acquire road surface point cloud information of the target road surface in the first direction; the first direction is the vehicle's driving direction;

[0250] The vehicle body sensor 2003 is used to acquire the vehicle's status information;

[0251] The suspension controller 2009 is used to input the current road surface image, the road surface point cloud information, and the state information into a preset road surface recognition model to obtain the target road surface roughness level; predict the feedforward damping force required by the vehicle's suspension system when the vehicle is traveling on the target road surface based on the target road surface roughness level; when the vehicle travels to the target road surface based on the feedforward damping force, calculate the feedback damping force in real time based on the vehicle's body parameters, and compensate the feedforward damping force in real time based on the feedback damping force, so as to make the vehicle travel smoothly on the target road surface.

[0252] In some embodiments, the vehicle suspension control system 200 further includes: a model building module and a determination module;

[0253] The model building module is used to construct alternative Lyapunov functions;

[0254] The determining module is used to determine the first calculation formula of the feedback damping force while maintaining the first derivative of the candidate Lyapunov function less than or equal to 0.

[0255] The data processing module is also used to calculate the feedback damping force according to the first calculation formula.

[0256] In some embodiments, the model building module is further configured to decouple the suspension system into multiple sub-suspension systems in different parts, establish a two-degree-of-freedom vibration model based on the sub-suspension systems and the vehicle body parameters, and construct the alternative Lyapunov function based on the sliding modal variables, radial basis function neural network weight coefficient estimation error, radial basis function estimation error, radial basis center estimation error, and disturbance force estimation error uncertainty determined by the two-degree-of-freedom vibration model.

[0257] In some embodiments, the vehicle suspension control system 200 further includes: a computing module;

[0258] The determining module is also used to determine the relative displacement of the lower mass of the suspension system and the relative velocity of the lower mass based on the two-degree-of-freedom vibration model.

[0259] The calculation module is used to multiply the relative displacement of the lower mass by the sliding mode gain coefficient and add the relative velocity of the lower mass to obtain the sliding mode variable.

[0260] In some embodiments, the determining module is further configured to determine the nominal disturbance force and estimate the disturbance force based on the sliding mode variables;

[0261] The calculation module is further configured to subtract the estimated disturbance force from the nominal disturbance force to obtain the disturbance force estimation error; the disturbance force estimation error is composed of the product of the weighting coefficient and the radial basis function estimation error, the product of the weighting coefficient error and the radial basis function estimator, and the sum of the optimal approximation error;

[0262] The calculation module is further configured to decompose the radial basis function estimation error into the product of the partial derivative of the radial basis function with respect to the radial basis center and the estimation error of the radial basis center, the product of the partial derivative of the radial basis function with respect to the radial basis width and the estimation error of the radial basis width, and the sum of the higher-order terms after the Taylor expansion of the radial basis function.

[0263] The calculation module is further configured to obtain the uncertainty of the disturbance estimation error based on the product of the sliding mode variable, the uncertainty gain coefficient, and the first matrix, and the product of the sliding mode variable and the adjustable gain coefficient; wherein the first matrix consists of a first constant, a radial basis center estimate, a radial basis width estimate, and a weight coefficient estimate.

[0264] In some embodiments, the calculation module is further configured to subtract the norm of the radial basis center from the sliding mode variable and then divide by the radial basis width to obtain the radial basis function;

[0265] The determining module is further configured to determine the nominal perturbation force by summing the product of the radial basis function and the weighting coefficient with the optimal approximation error;

[0266] The calculation module is further configured to replace the weight coefficients, radial basis center, and radial basis width in the nominal perturbation force with the corresponding estimates, and discard the optimal perturbation approximation error term to obtain the estimated perturbation force.

[0267] In some embodiments, the calculation module is further configured to calculate the radial basis function neural network weight coefficients, the radial basis center, the radial basis width, and the uncertainty of the disturbance force estimation error in real time based on the real-time vehicle body parameters; and to substitute the radial basis function neural network weight coefficients, the radial basis center, the radial basis width, and the uncertainty of the disturbance force estimation error into the first calculation formula to obtain the feedback damping force.

[0268] In some embodiments, the vehicle suspension control system 200 further includes: an input module;

[0269] The input module is used to input the current road surface image into the first model, extract a first road surface feature from the current road surface image using the first model, and convert the first road surface feature into a first road surface roughness level; input the road surface point cloud information into the second model, extract a second road surface feature from the road surface point cloud information using the second model, and convert the second road surface feature into a second road surface roughness level; input the state information into the third model to obtain a third road surface roughness level; and perform information fusion on the first road surface roughness level, the second road surface roughness level, the third road surface roughness level, and the historical road surface roughness level of the target road surface to obtain the target road surface roughness level.

[0270] In some embodiments, the vehicle suspension control system 200 further includes: an acquisition module;

[0271] The acquisition module is used to acquire historical road marking information corresponding to the target road surface; the historical road marking information is obtained after the vehicle has historically traveled to the target road surface and marked it; wherein, the historical road marking information includes the historical road surface roughness level, historical road surface coordinates, historical road surface length and historical path curvature corresponding to the target road surface.

[0272] In some embodiments, the determining module is further configured to determine the current first position coordinates of the vehicle; determine a first distance between the first position of the vehicle and the starting position of the target road surface based on the first position coordinates and the historical road surface coordinates; the first position is the center position of the upper front axle of the vehicle; determine a first time required for the vehicle to travel to the starting position based on the first distance and the first driving speed; and activate the control strategy if the first time is less than a first threshold and / or the first distance is less than a second threshold.

[0273] In some embodiments, the determining module is further configured to determine a second time required for the vehicle to travel the target road surface based on the first driving speed and the historical road surface length; and to deactivate the control strategy if the control strategy is activated and the second time has elapsed.

[0274] In some embodiments, the vehicle suspension control system 200 further includes: an adjustment module;

[0275] The adjustment module is used to adjust the height of the suspension system to a target height and / or adjust the stiffness of the air spring in the suspension system to a target stiffness.

[0276] In some embodiments, the determining module is further configured to determine the air suspension mode currently set by the vehicle; and to determine the target height and the target stiffness based on the air suspension mode, historical path curvature, the state information, and the target road surface roughness level.

[0277] In some embodiments, the determining module is further configured to determine the air suspension mode currently set by the vehicle; and to determine the target height and the target stiffness based on the air suspension mode, historical path curvature, the state information, and the target road surface roughness level.

[0278] In some embodiments, the determining module is further configured to adjust the target height based on the vehicle's current second driving speed and the vehicle's first information when the vehicle is detected to be accelerating or braking while driving on the target road surface; wherein the first information is the current position of the accelerator pedal and / or the current pressure of the brake cylinder in the vehicle.

[0279] In some embodiments, the determining module is further configured to: determine that the vehicle has reached the starting point of the target road surface when the vertical acceleration of the front axle wheels of the vehicle is greater than a third threshold value; determine the current first position coordinates and the current first time point of the vehicle; determine that the vehicle has left the target road surface when the vertical acceleration of the front axle wheels of the vehicle is less than a fourth threshold value; determine the current second position coordinates and the current second time point of the vehicle; determine the first road surface length of the target road surface based on the first position coordinates and the second position coordinates, and / or the first time point, the second time point, and the current third driving speed of the vehicle; determine the target path curvature based on the first path curvature and the historical path curvature; and update the historical road surface marking information based on the target path curvature, the target road surface unevenness level, the first road surface length, and the first position coordinates.

[0280] This invention provides a vehicle suspension control system, comprising: a binocular camera for acquiring a current road surface image of a target road surface in a first direction; a radar for acquiring road surface point cloud information of the target road surface in the first direction; the first direction being the vehicle's driving direction; a vehicle body sensor for acquiring vehicle status information; and a suspension controller for inputting the current road surface image, road surface point cloud information, and status information into a preset road surface recognition model to obtain the target road surface roughness level; predicting the feedforward damping force required by the vehicle's suspension system when the vehicle is driving on the target road surface based on the target road surface roughness level; and, when the vehicle travels onto the target road surface based on the feedforward damping force, calculating the feedback damping force in real time based on the vehicle's body parameters, and compensating the feedforward damping force in real time based on the feedback damping force to ensure smooth vehicle travel on the target road surface. Using the above implementation scheme, the current road surface image, road surface point cloud information, and vehicle status information are first fused together. After considering the historical road surface roughness level, the target road surface roughness level that best reflects the target road surface condition is obtained. Therefore, the most accurate control strategy can be obtained based on this target road surface roughness level. Furthermore, the damping force of the shock absorber determined in the control strategy is based on the feedback control damping force and the feedforward damping force, which can reduce the impact and vibration acceleration of the vehicle when it enters the target road surface. This further improves the driving comfort performance of the vehicle under uneven road surface conditions based on this control strategy. At the same time, the feedback control damping force can immediately design a self-learning control strategy for the nonlinear characteristics of the radial basis neural network to adaptively compensate for external disturbances. Therefore, the feedback control damping force can quickly eliminate the vibration acceleration of the vehicle body and wheels after the vehicle enters the target road surface.

[0281] Figure 11 This is a schematic diagram of the composition structure of a vehicle suspension control device provided in an embodiment of the present invention. In practical applications, based on the same disclosed concept of the above embodiments, such as... Figure 11 As shown, the vehicle suspension control device 11 in this embodiment includes a processor 110, a memory 111, and a communication bus 112.

[0282] In a specific embodiment, the aforementioned model building module, determination module, calculation module, input module, acquisition module, and adjustment module can be implemented by a processor 110 located on the vehicle's suspension control device 11. The processor 110 can be at least one of the following: Application Specific Integrated Circuit (ASIC), Digital Signal Processor (DSP), Digital Signal Processing Device (DSPD), Programmable Logic Device (PLD), Field Programmable Gate Array (FPGA), CPU, controller, microcontroller, or microprocessor. It is understood that for different vehicle suspension control devices 11, the electronic devices used to implement the above-mentioned processor functions can also be other types; this embodiment does not impose specific limitations.

[0283] In this embodiment of the invention, the communication bus 112 is used to realize the connection communication between the processor 110 and the memory 111; when the processor 110 executes the running program stored in the memory 111, it implements the following vehicle suspension system control method:

[0284] Acquire the current road surface image and road surface point cloud information of the target road surface in a first direction, as well as the vehicle's status information; the first direction is the vehicle's driving direction;

[0285] The current road surface image, the road surface point cloud information, and the status information are input into a preset road surface recognition model to obtain the target road surface roughness level of the target road surface;

[0286] Based on the target road surface roughness level, predict the feedforward damping force required by the vehicle's suspension system when the vehicle is traveling on the target road surface;

[0287] When the vehicle travels to the target road surface based on the feedforward damping force, the feedback damping force is calculated in real time based on the vehicle's body parameters, and the feedforward damping force is compensated in real time based on the feedback damping force so that the vehicle can travel smoothly on the target road surface.

[0288] In some embodiments, the processor 110 is further configured to construct alternative Lyapunov functions; determine a first formula for calculating the feedback damping force while maintaining the first derivative of the alternative Lyapunov functions less than or equal to 0; and calculate the feedback damping force according to the first formula.

[0289] In some embodiments, the processor 110 is further configured to decouple the suspension system into multiple sub-suspension systems in different parts, establish a two-degree-of-freedom vibration model based on the sub-suspension systems and the vehicle body parameters, and construct the alternative Lyapunov function based on the sliding modal variables, radial basis function neural network weight coefficient estimation error, radial basis function estimation error, radial basis center estimation error, and disturbance force estimation error uncertainty determined by the two-degree-of-freedom vibration model.

[0290] In some embodiments, the processor 110 is further configured to determine the relative displacement of the lower mass of the suspension system and the relative velocity of the lower mass according to the two-degree-of-freedom vibration model; multiply the relative displacement of the lower mass by the sliding mode gain coefficient and add the relative velocity of the lower mass to obtain the sliding mode variable.

[0291] In some embodiments, the processor 110 is further configured to determine the nominal disturbance force and estimate the disturbance force based on the sliding mode variables; subtract the estimated disturbance force from the nominal disturbance force to obtain the disturbance force estimation error; the disturbance force estimation error is composed of the product of the weighting coefficients and the radial basis function estimation error, the product of the weighting coefficient error and the radial basis function estimate, and the sum of the best approximation error; the radial basis function estimation error is decomposed into the product of the partial derivative of the radial basis function with respect to the radial basis center and the estimation error of the radial basis center, the product of the partial derivative of the radial basis function with respect to the radial basis width and the estimation error of the radial basis width, and the sum of the higher-order terms after Taylor expansion of the radial basis function; based on the product of the sliding mode variables, the uncertainty gain coefficient, and the first matrix, and the product of the sliding mode variables and the adjustable gain coefficient, the uncertainty of the disturbance force estimation error is obtained; wherein, the first matrix is ​​composed of a first constant, the radial basis center estimate, the radial basis width estimate, and the weighting coefficient estimate.

[0292] In some embodiments, the processor 110 is further configured to subtract the norm of the radial basis center from the sliding modal variable and then divide it by the radial basis width to obtain the radial basis function; to determine the nominal perturbation force by the product of the radial basis function and the weighting coefficients and the sum of the optimal approximation error; to replace the weighting coefficients, the radial basis center, and the radial basis width in the nominal perturbation force with the corresponding estimates, and to discard the optimal perturbation approximation error term to obtain the estimated perturbation force.

[0293] In some embodiments, the processor 110 is further configured to calculate in real time the radial basis function (RBF) neural network weight coefficients, the radial basis center, the radial basis width, and the uncertainty of the disturbance force estimation error based on the real-time vehicle body parameters; and to substitute the RBF neural network weight coefficients, the radial basis center, the radial basis width, and the uncertainty of the disturbance force estimation error into the first calculation formula to obtain the feedback damping force.

[0294] In some embodiments, the processor 110 is further configured to: input the current road surface image into the first model; extract a first road surface feature from the current road surface image using the first model; convert the first road surface feature into a first road surface roughness level; input the road surface point cloud information into the second model; extract a second road surface feature from the road surface point cloud information using the second model; convert the second road surface feature into a second road surface roughness level; input the state information into the third model to obtain a third road surface roughness level; and perform information fusion on the first road surface roughness level, the second road surface roughness level, the third road surface roughness level, and the historical road surface roughness level of the target road surface to obtain the target road surface roughness level.

[0295] In some embodiments, the processor 110 is further configured to obtain historical road marking information corresponding to the target road surface; the historical road marking information is obtained after the vehicle has historically traveled to the target road surface and marked it; wherein, the historical road marking information includes the historical road surface roughness level, historical road surface coordinates, historical road surface length, and historical path curvature corresponding to the target road surface.

[0296] In some embodiments, the processor 110 is further configured to determine the current first position coordinates of the vehicle; determine a first distance between the first position of the vehicle and the starting position of the target road surface based on the first position coordinates and the historical road surface coordinates; the first position is the center position of the upper front axle of the vehicle; determine a first time required for the vehicle to travel to the starting position based on the first distance and the first driving speed; and activate the control strategy if the first time is less than a first threshold and / or the first distance is less than a second threshold.

[0297] In some embodiments, the processor 110 is further configured to determine a second time required for the vehicle to travel the target road surface based on the first driving speed and the historical road surface length; and to disable the control strategy if the control strategy is enabled and the second time has elapsed.

[0298] In some embodiments, the processor 110 is further configured to adjust the height of the suspension system to a target height and / or adjust the stiffness of the air springs in the suspension system to a target stiffness.

[0299] In some embodiments, the processor 110 is further configured to determine the air suspension mode currently set for the vehicle; and to determine the target height and the target stiffness based on the air suspension mode, historical path curvature, the status information, and the target road surface roughness level.

[0300] In some embodiments, the processor 110 is further configured to adjust the target height based on the vehicle's current second driving speed and the vehicle's first information when the processor detects that the vehicle is accelerating or braking while driving on the target road surface; wherein the first information is the current position of the accelerator pedal and / or the current pressure of the brake cylinder in the vehicle.

[0301] In some embodiments, the processor 110 is further configured to: determine that the vehicle has reached the starting point of the target road surface when the vertical acceleration of the front axle wheels of the vehicle is greater than a third threshold value; determine the current first position coordinates and the current first time point of the vehicle; determine that the vehicle has left the target road surface when the vertical acceleration of the front axle wheels of the vehicle is less than a fourth threshold value; determine the current second position coordinates and the current second time point of the vehicle; determine the first road surface length of the target road surface based on the first position coordinates and the second position coordinates, and / or the first time point, the second time point, and the current third driving speed of the vehicle; determine the target path curvature based on the first path curvature and the historical path curvature; and update the historical road surface marking information based on the target path curvature, the target road surface unevenness level, the first road surface length, and the first position coordinates.

[0302] This invention provides a storage medium storing a computer program thereon. The computer-readable storage medium stores one or more programs, which can be executed by one or more processors and applied to a vehicle suspension control system. The computer program implements the vehicle suspension system control method described above.

[0303] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0304] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will 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 apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0305] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function 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.

[0306] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable 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.

[0307] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for controlling a vehicle's suspension system, characterized in that, The method includes: Acquire the current road surface image and road surface point cloud information of the target road surface in a first direction, as well as the vehicle's status information; the first direction is the vehicle's driving direction; The current road surface image, the road surface point cloud information, and the status information are input into a preset road surface recognition model to obtain the target road surface roughness level of the target road surface; Based on the target road surface roughness level, predict the feedforward damping force required by the vehicle's suspension system when the vehicle is traveling on the target road surface; When the vehicle travels to the target road surface based on the feedforward damping force, the feedback damping force is calculated in real time based on the vehicle's body parameters, and the feedforward damping force is compensated in real time based on the feedback damping force so that the vehicle can travel smoothly on the target road surface. The real-time calculation of feedback damping force based on the vehicle's body parameters includes: The suspension system is decoupled into multiple sub-suspension systems in different parts, and a two-degree-of-freedom vibration model is established based on the sub-suspension systems and the vehicle body parameters. Based on the uncertainties of the sliding modal variables, radial basis function neural network weight coefficient estimation errors, radial basis function estimation errors, radial basis center estimation errors, and disturbance force estimation errors determined by the two-degree-of-freedom vibration model, alternative Lyapunov functions are constructed. While maintaining the first derivative of the candidate Lyapunov function less than or equal to 0, a first formula for calculating the feedback damping force is determined; the first variable in the first formula includes the radial basis function weight coefficient, the radial basis center, the radial basis width, and the uncertainty of the disturbance force estimation error. The radial basis function neural network weight coefficients, radial basis center, radial basis width, and disturbance force estimation error uncertainty are calculated in real time based on the real-time vehicle body parameters. Substituting the radial basis function neural network weight coefficients, the radial basis center, the radial basis width, and the uncertainty of the disturbance force estimation error into the first calculation formula, the feedback damping force is obtained.

2. The vehicle suspension system control method according to claim 1, characterized in that, The method further includes: The relative displacement and relative velocity of the lower mass of the suspension system are determined based on the two-degree-of-freedom vibration model. The relative displacement of the lower mass is multiplied by the sliding mode gain coefficient, and then the relative velocity of the lower mass is added to obtain the sliding mode variable.

3. The vehicle suspension system control method according to claim 1, characterized in that, The method further includes: The nominal disturbance force and the estimated disturbance force are determined based on the sliding mode variables. Subtracting the estimated disturbance force from the nominal disturbance force yields the disturbance force estimation error; the disturbance force estimation error is composed of the product of the weighting coefficient and the radial basis function estimation error, the product of the weighting coefficient error and the radial basis function estimator, and the sum of the optimal approximation error; The radial basis function estimation error is decomposed into the product of the partial derivative of the radial basis function with respect to the center of the radial basis and the estimation error of the center of the radial basis, the product of the partial derivative of the radial basis function with respect to the width of the radial basis and the estimation error of the width of the radial basis, and the sum of the higher-order terms after the Taylor expansion of the radial basis function. Based on the product of the sliding mode variable, the uncertainty gain coefficient, and the first matrix, and the product of the sliding mode variable and the adjustable gain coefficient, the uncertainty of the disturbance force estimation error is obtained. The first matrix consists of a first constant, a radial basis center estimate, a radial basis width estimate, and a weight coefficient estimate.

4. The vehicle suspension system control method according to claim 3, characterized in that, The method further includes: The radial basis function is obtained by subtracting the norm of the radial basis center from the sliding modal variable and then dividing by the radial basis width. The nominal perturbation force is determined by the sum of the product of the radial basis function and the weighting coefficient and the optimal approximation error. The estimated disturbance force is obtained by replacing the weight coefficients, radial basis center, and radial basis width in the nominal disturbance force with the corresponding estimates and discarding the optimal disturbance approximation error term.

5. The vehicle suspension system control method according to any one of claims 1 to 4, characterized in that, The preset road surface recognition model includes at least a first model, a second model, and a third model; the step of inputting the current road surface image, the road surface point cloud information, and the state information into the preset road surface recognition model to obtain the target road surface roughness level includes: The current road surface image is input into the first model, and the first road surface feature in the current road surface image is extracted using the first model. The first road surface feature is then converted into a first road surface roughness level. The road surface point cloud information is input into the second model, and the second model is used to extract the second road surface features from the road surface point cloud information. The second road surface features are then converted into a second road surface roughness level. The state information is input into the third model to obtain the third road surface roughness level; The first road surface roughness level, the second road surface roughness level, the third road surface roughness level, and the historical road surface roughness level of the target road surface are fused to obtain the target road surface roughness level.

6. The vehicle suspension system control method according to claim 5, characterized in that, The method further includes: Obtain historical road marking information corresponding to the target road surface; the historical road marking information is obtained after the target road surface was marked when the vehicle previously traveled to the target road surface. The historical road marking information includes the historical road unevenness level, historical road coordinates, historical road length, and historical path curvature corresponding to the target road surface.

7. The vehicle suspension system control method according to claim 6, characterized in that, The status information includes the vehicle's current first driving speed, and the method further includes: Determine the current first position coordinates of the vehicle; Based on the first position coordinates and the historical road surface coordinates, a first distance is determined between the first position of the vehicle and the starting position of the target road surface; the first position is the center position of the upper front axle of the vehicle. Based on the first distance and the first driving speed, determine the first time required for the vehicle to travel to the starting position; The control strategy is activated if the first duration is less than the first threshold and / or the first distance is less than the second threshold.

8. The vehicle suspension system control method according to claim 7, characterized in that, The method further includes: Based on the first driving speed and the historical road length, determine the second time required for the vehicle to travel the target road. If the control policy is enabled and the second time period has elapsed, then the control policy is disabled.

9. The vehicle suspension system control method according to claim 1, characterized in that, After real-time compensation of the feedforward damping force based on the feedback damping force, the method further includes: Adjust the height of the suspension system to the target height, and / or adjust the stiffness of the air springs in the suspension system to the target stiffness.

10. The vehicle suspension system control method according to claim 9, characterized in that, The method further includes: Determine the current air suspension mode set for the vehicle; The target height and the target stiffness are determined based on the air suspension mode, historical path curvature, state information, and target road surface roughness level.

11. The vehicle suspension system control method according to claim 10, characterized in that, The method further includes: If the vehicle is detected to be accelerating or braking while traveling on the target road surface, the target height is adjusted according to the vehicle's current second driving speed and the vehicle's first information; The first information refers to the current position of the accelerator pedal and / or the current pressure of the brake cylinder in the vehicle.

12. The vehicle suspension system control method according to claim 11, characterized in that, The method further includes: If the vertical acceleration of the front axle wheels of the vehicle is greater than the third threshold value, it is determined that the vehicle has reached the starting point of the target road surface, and the current first position coordinates and the current first time point of the vehicle are determined. If the vertical acceleration of the front axle wheels of the vehicle is less than the fourth threshold value, it is determined that the vehicle has left the target road surface, and the current second position coordinates and the current second time point of the vehicle are determined. The first road length of the target road surface is determined based on the first position coordinates and the second position coordinates, and / or the first time point, the second time point, and the current third driving speed of the vehicle; The target path curvature is determined based on the first path curvature and the historical path curvature, and the historical road surface marking information is updated based on the target path curvature, the target road surface unevenness level, the first road surface length, and the first location coordinates.

13. A vehicle suspension control system, characterized in that, The system includes: A binocular camera is used to acquire a current image of the target road surface in the first direction; Radar is used to acquire road point cloud information of the target road surface in the first direction; the first direction is the vehicle's driving direction. Vehicle body sensors are used to acquire the vehicle's status information; A suspension controller is used to input the current road surface image, the road surface point cloud information, and the state information into a preset road surface recognition model to obtain the target road surface roughness level; predict the feedforward damping force required by the vehicle's suspension system when the vehicle travels on the target road surface based on the target road surface roughness level; when the vehicle travels on the target road surface based on the feedforward damping force, calculate the feedback damping force in real time based on the vehicle's body parameters, and compensate the feedforward damping force in real time based on the feedback damping force to enable the vehicle to travel smoothly on the target road surface; The suspension controller is further configured to decouple the suspension system into multiple sub-suspension systems in different parts, establish a two-degree-of-freedom vibration model based on the sub-suspension systems and the vehicle body parameters, construct a candidate Lyapunov function based on the sliding modal variables, radial basis function (RBF) neural network weight coefficient estimation error, RBF function estimation error, RBF center estimation error, and disturbance force estimation error uncertainty determined by the two-degree-of-freedom vibration model, and determine a first calculation formula for the feedback damping force while maintaining the first derivative of the candidate Lyapunov function less than or equal to 0. The first variables in the first calculation formula include the RBF neural network weight coefficient, RBF center, RBF width, and disturbance force estimation error uncertainty. The RBF neural network weight coefficient, RBF center, RBF width, and disturbance force estimation error uncertainty are calculated in real time based on the real-time vehicle body parameters. The RBF neural network weight coefficient, RBF center, RBF width, and disturbance force estimation error uncertainty are substituted into the first calculation formula to obtain the feedback damping force.

14. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the method as described in any one of claims 1-12.

Images