Automobile suspension pre-control method and system based on lqr and artificial neural network

Abstract

This invention provides a pre-control method and system for automotive suspension based on LQR and artificial neural networks. The method acquires road surface undulation information in advance using a binocular camera, and combines this with vehicle attitude and motion parameters obtained from onboard sensors to construct a dynamic model of the suspension system, expressed using linear space equations. Subsequently, an artificial neural network is designed to calculate a weighted coefficient matrix based on vehicle parameters and road surface height. Optimal suspension parameters are solved using a linear quadratic regulator (LQR) to achieve pre-control of the suspension. The suspension's attitude after deformation is calculated using a precise discretization method, and the artificial neural network is trained online under supervised supervision. This method can pre-calculate suspension deformation, effectively absorbing road impacts and improving driving safety and comfort. Compared to traditional real-time feedback control, this invention has stronger environmental adaptability and reliability.

Description

Technical Field

[0001] This invention relates to the technical field of automobiles, and more specifically, to a pre-control method for automobile suspension based on LQR and artificial neural networks. Background Technology

[0002] As a crucial component of a vehicle, the suspension system's performance directly impacts driving performance and passenger comfort. Excellent suspension control effectively absorbs road impacts and reduces the effects of vibrations on the vehicle body, thereby improving driving safety and comfort. With the continuous advancement of automotive technology, the intelligentization and automation of suspension systems have gradually become development trends. Achieving dynamic adjustments to suspension characteristics based on complex road conditions to achieve optimal driving performance has become a key focus of current research.

[0003] However, most current suspension control methods are limited to real-time feedback control, meaning they determine the suspension control strategy solely based on data collected by sensors at the current moment. This method can lead to a significant decrease in ride comfort when encountering rough road surfaces because the suspension cannot deform in time. To overcome this limitation, the industry urgently needs a method that can calculate suspension deformation in advance and achieve pre-control of the suspension.

[0004] Although some related patents have proposed improved suspension control methods, these methods still have shortcomings. For example, the semi-active suspension control method for vehicles proposed in patent document CN114919365B (application number: CN202210741404.0), while able to determine whether a vehicle has entered a special operating condition by acquiring the vertical acceleration of the wheels and controlling the working state of the wheel shock absorbers accordingly, fails to obtain road information in advance and cannot pre-control the suspension attitude. Another active suspension control method proposed in patent document CN110901325B (application number: CN201911197599.1), while able to observe the road roughness curve ahead and estimate the desired vehicle body attitude, relies on a classic PID algorithm for subsequent control, resulting in significant differences in control performance under different road conditions.

[0005] Therefore, a new technical solution is needed to improve the above-mentioned technical problems. Summary of the Invention

[0006] In view of the deficiencies in the prior art, the purpose of this invention is to provide a method and system for pre-control of automobile suspension based on LQR and artificial neural networks.

[0007] According to the present invention, a pre-control method for vehicle suspension based on LQR and artificial neural networks is provided, the method comprising the following steps:

[0008] Step S1: Based on binocular depth estimation technology, the undulation information of the road surface ahead is obtained in advance, and combined with all attitude parameters and motion parameters of the vehicle and suspension system obtained in real time by the on-board sensors;

[0009] Step S2: Based on the obtained parameters, construct a dynamic model of the suspension system and describe it using state-space expressions;

[0010] Step S3: Design an artificial neural network that takes vehicle parameters and road surface height as input, and calculates and outputs a weighted coefficient matrix for LQR control through learning.

[0011] Step S4: Using the LQR method and the weighted coefficient matrix output by the artificial neural network, the feedback gain matrix is ​​obtained, thereby determining the magnitude of the active control external force. Based on the precise discretization method, the attitude of the suspension after deformation is predicted, which is used for online supervised training of the artificial neural network.

[0012] Step S5: Based on the LQR solution, dynamically adjust the air spring stiffness of the suspension, and repeat steps S1 to S5 for continuous pre-control.

[0013] Preferably, the road surface undulation information in step S1 is calculated based on the depth image acquired by the binocular camera, while taking into account the current vehicle speed to determine the time it takes for the tire to reach each road surface undulation.

[0014] Preferably, when constructing the dynamic model of the suspension system in step S2, the sprung mass, unsprung mass, equivalent stiffness of the air spring, damper, active control force, equivalent stiffness of the tire, and sprung mass displacement and unsprung mass displacement parameters are considered.

[0015] Preferably, the artificial neural network designed in step S3 has a multi-layer fully connected structure, including an input layer, at least three hidden layers and an output layer, wherein the hidden layers are activated by the ReLU function, and the output layer is designed with specific dimensions to generate the weighted coefficient matrix according to the requirements.

[0016] The output layer of the artificial neural network includes two sub-output layers: one for generating a symmetric positive semi-definite weighted sparse matrix, and the other for outputting a symmetric positive definite weighted coefficient matrix after activation by the Softplus function.

[0017] Preferably, in step S4, when solving the Riccati equation to obtain the matrix, the weighted coefficient matrix output by the artificial neural network is used as part of the equation parameters for learning-based LQR parameter adjustment.

[0018] When predicting the attitude of the suspension after deformation based on the precise discretization method, the matrix, state variables and optimal control force under the current state are considered, and Taylor series expansion is used for approximate calculation.

[0019] When training an artificial neural network in an online supervised manner, the smooth L1 loss function is used to minimize the change in the vehicle's attitude in the vertical direction between two consecutive time points.

[0020] Preferably, in step S5, when dynamically adjusting the air spring stiffness of the suspension, the changes in the air spring cavity volume and effective force-bearing area after the suspension attitude changes are considered, and a new equivalent stiffness is calculated.

[0021] The present invention also provides a vehicle suspension pre-control system based on LQR and artificial neural networks, the system comprising the following modules:

[0022] Module M1: Based on binocular depth estimation technology, it pre-acquires information on the undulation of the road surface ahead, and combines it with all attitude and motion parameters of the vehicle and suspension system acquired in real time by onboard sensors;

[0023] Module M2: Based on the acquired parameters, construct a dynamic model of the suspension system and describe it using state-space expressions;

[0024] Module M3: Design an artificial neural network that takes vehicle parameters and road surface height as input, and calculates and outputs a weighted coefficient matrix for LQR control through learning.

[0025] Module M4: Using the LQR system and the weighted coefficient matrix output by the artificial neural network, the feedback gain matrix is ​​obtained, thereby determining the magnitude of the active control force and predicting the attitude of the suspension after deformation based on the precise discretization system, which is used for online supervised training of the artificial neural network.

[0026] Module M5: Based on the LQR solution, dynamically adjust the air spring stiffness of the suspension and repeat the process from Module M1 to S5 for continuous pre-control.

[0027] Preferably, the road surface undulation information in module M1 is calculated based on the depth image acquired by the binocular camera, while taking into account the current vehicle speed to determine the time it takes for the tire to reach each road surface undulation position.

[0028] When constructing the dynamic model of the suspension system in module M2, the sprung mass, unsprung mass, equivalent stiffness of the air spring, damper, active control force, equivalent stiffness of the tire, and sprung mass displacement and unsprung mass displacement parameters are considered.

[0029] Preferably, the artificial neural network designed in module M3 has a multi-layer fully connected structure, including an input layer, at least three hidden layers and an output layer, wherein the hidden layers are activated by the ReLU function, and the output layer is designed with specific dimensions to generate the weighted coefficient matrix according to the requirements.

[0030] The output layer of the artificial neural network includes two sub-output layers: one for generating a symmetric positive semi-definite weighted sparse matrix, and the other for outputting a symmetric positive definite weighted coefficient matrix after activation by the Softplus function.

[0031] Preferably, when solving the Riccati equation in module M4 to obtain the matrix, the weighted coefficient matrix output by the artificial neural network is used as part of the equation parameters for learning-based LQR parameter adjustment.

[0032] When predicting the attitude of the suspension after deformation based on the precise discretization system, the matrix, state variables and optimal control external force under the current state are considered, and Taylor series expansion is used for approximate calculation.

[0033] When training an artificial neural network in an online supervised manner, the smooth L1 loss function is used to minimize the change in the vehicle's attitude in the vertical direction between two consecutive time points.

[0034] When dynamically adjusting the air spring stiffness of the suspension in module M5, the changes in the air spring cavity volume and effective force-bearing area after the suspension attitude changes are considered, and a new equivalent stiffness is calculated.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] 1. This invention proposes a novel pre-control method for automotive suspension. By obtaining information on road surface undulation in advance and combining it with current vehicle and suspension parameters, the optimal suspension posture for future moments can be calculated in advance, allowing the suspension to deform earlier. This more effectively absorbs road impacts, reduces the impact of vibrations on the vehicle body, and significantly improves driving safety and passenger comfort. The effect is particularly noticeable when facing complex and ever-changing road environments.

[0037] 2. This invention employs an LQR parameter matrix adjustment method based on artificial neural networks. Through learning, it can adaptively adjust the values ​​of the weighting coefficient matrix according to the current parameters of the vehicle and suspension, enabling the suspension system to better adapt to different road conditions and driving conditions, thereby improving the environmental adaptability and reliability of the suspension.

[0038] 3. Through experimental testing, the suspension pre-control method of the present invention has shown significant control effect. When using the pre-control method, the overall control effect of the suspension is significantly stronger than that without pre-control. This not only verifies the effectiveness and feasibility of the present invention, but also further proves its great potential in improving the overall performance of the automotive suspension system. Attached Figure Description

[0039] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0040] Figure 1 This is a flowchart illustrating the implementation of the method described in this embodiment of the invention;

[0041] Figure 2 This is an equivalent model diagram of the suspension according to an embodiment of the present invention;

[0042] Figure 3 This is a graph showing the change in road surface undulation during the control experiment of an embodiment of the present invention;

[0043] Figure 4 This is a comparison chart of suspension control results according to an embodiment of the present invention. Detailed Implementation

[0044] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0045] Example 1:

[0046] Reference Figure 1 According to the present invention, a pre-control method for automobile suspension based on LQR and artificial neural networks is provided, the method comprising the following steps:

[0047] Step S1: Based on binocular depth estimation technology, the undulation information of the road surface ahead is obtained in advance, and combined with all attitude parameters and motion parameters of the vehicle and suspension system obtained in real time by the on-board sensors; the road undulation information is calculated based on the depth image obtained by the binocular camera, while taking into account the current vehicle speed to determine the time it takes for the tire to reach each road undulation position.

[0048] Step S2: Based on the obtained parameters, construct a dynamic model of the suspension system and describe it using a state-space expression. When constructing the dynamic model of the suspension system, consider the sprung mass, unsprung mass, equivalent stiffness of the air spring, damper, active control force, equivalent stiffness of the tire, and sprung mass displacement and unsprung mass displacement parameters.

[0049] Step S3: Design an artificial neural network that takes vehicle parameters and road surface height as input, and calculates and outputs a weighted coefficient matrix for LQR control through learning. The designed artificial neural network has a multi-layer fully connected structure, including an input layer, at least three hidden layers and an output layer. The hidden layers are activated using the ReLU function, and the output layer is designed with specific dimensions to generate the weighted coefficient matrix according to requirements. The output layer of the artificial neural network includes two sub-output layers: one for generating a symmetric positive semi-definite weighted sparse matrix, and the other for outputting a symmetric positive definite weighted coefficient matrix after activation by the Softplus function.

[0050] Step S4: Using the LQR method and the weighted coefficient matrix output by the artificial neural network, the feedback gain matrix is ​​obtained to determine the magnitude of the active control force. The attitude of the suspension after deformation is predicted based on the precise discretization method, which is used for online supervised training of the artificial neural network. When solving the Riccati equation to obtain the matrix, the weighted coefficient matrix output by the artificial neural network is used as part of the equation parameters for learning-based LQR parameter adjustment. When predicting the attitude of the suspension after deformation based on the precise discretization method, the matrix, state variables, and optimal control force in the current state are considered, and Taylor series expansion is used for approximate calculation. When conducting online supervised training of the artificial neural network, the smooth L1 loss function is used to minimize the change in vehicle attitude in the vertical direction between two consecutive time points.

[0051] Step S5: Based on the LQR solution, dynamically adjust the air spring stiffness of the suspension, and repeat steps S1 to S5 for continuous pre-control; when dynamically adjusting the air spring stiffness of the suspension, consider the changes in the air spring cavity volume and effective force-bearing area after the suspension attitude changes, and calculate the new equivalent stiffness.

[0052] The present invention also provides an automotive suspension pre-control system based on LQR and artificial neural networks. The automotive suspension pre-control system based on LQR and artificial neural networks can be implemented by executing the process steps of the automotive suspension pre-control method based on LQR and artificial neural networks. That is, those skilled in the art can understand the automotive suspension pre-control method based on LQR and artificial neural networks as a preferred embodiment of the automotive suspension pre-control system based on LQR and artificial neural networks.

[0053] Example 2:

[0054] The present invention also provides a vehicle suspension pre-control system based on LQR and artificial neural networks, the system comprising the following modules:

[0055] Module M1: Based on binocular depth estimation technology, it pre-acquires the undulation information of the road surface ahead, and combines it with all attitude and motion parameters of the vehicle and suspension system acquired in real time by on-board sensors; the road undulation information is calculated based on the depth images acquired by the binocular camera, while taking into account the current vehicle speed to determine the time it takes for the tires to reach each road undulation position.

[0056] Module M2: Based on the acquired parameters, construct the dynamic model of the suspension system and describe it through state-space expression; when constructing the dynamic model of the suspension system, consider the sprung mass, unsprung mass, equivalent stiffness of the air spring, damper, active control external force, tire equivalent stiffness, and sprung mass displacement and unsprung mass displacement parameters.

[0057] Module M3: Design an artificial neural network that takes vehicle parameters and road surface height as input, and learns to calculate and output a weighted coefficient matrix for LQR control. The designed artificial neural network has a multi-layer fully connected structure, including an input layer, at least three hidden layers, and an output layer. The hidden layers are activated using the ReLU function, and the output layer is designed with specific dimensions to generate the weighted coefficient matrix. The output layer of the artificial neural network includes two sub-output layers: one for generating a symmetric positive semi-definite weighted sparse matrix, and the other for outputting a symmetric positive definite weighted coefficient matrix after activation by the Softplus function.

[0058] Module M4: Utilizing the LQR system and combining the weighted coefficient matrix output by the artificial neural network, the feedback gain matrix is ​​obtained, thereby determining the magnitude of the active control force. Based on the precise discretization system, the attitude of the suspension after deformation is predicted, serving as online supervised training for the artificial neural network. When solving the Riccati equation to obtain the matrix, the weighted coefficient matrix output by the artificial neural network is used as part of the equation parameters for learning-based LQR parameter adjustment. When predicting the attitude of the suspension after deformation based on the precise discretization system, the matrix, state variables, and optimal control force in the current state are considered, and Taylor series expansion is used for approximate calculation. During online supervised training of the artificial neural network, the smooth L1 loss function is used to minimize the vehicle's attitude change in the vertical direction between two consecutive time points.

[0059] Module M5: Based on the LQR solution, dynamically adjust the air spring stiffness of the suspension and repeat modules M1 to S5 for continuous pre-control; when dynamically adjusting the air spring stiffness of the suspension, consider the changes in the air spring cavity volume and effective force-bearing area after the suspension attitude changes, and calculate the new equivalent stiffness.

[0060] Example 3:

[0061] This invention relates to the field of automotive technology and provides a pre-control method for automotive suspension based on LQR and artificial neural networks. It includes: acquiring road surface undulation information in advance using a binocular camera, and then establishing a dynamic model of the vehicle suspension system using attitude and motion parameters obtained from onboard sensors, expressed using linear space equations. Subsequently, an artificial neural network is designed to calculate a weighted coefficient matrix based on vehicle parameters and road surface height using a learning method, and the optimal suspension parameters are obtained based on a linear quadratic regulator (LQR) to achieve pre-control of the suspension. Finally, the attitude of the suspension after deformation is calculated using a precise discretization method, thereby achieving online supervised training of the artificial neural network.

[0062] This invention provides a pre-control method for vehicle suspension based on LQR and artificial neural networks. The method aims to obtain road surface undulation information in advance, combine it with all parameters of the vehicle and suspension system at the current moment, use artificial neural networks to calculate a weighted coefficient matrix based on learning, and then use a linear quadratic regulator to perform optimal control of the suspension system to ensure driving safety and comfort.

[0063] This invention provides a method for pre-controlling automotive suspension based on LQR and artificial neural networks, such as... Figure 1 As shown, the method includes the following steps:

[0064] S1: Based on the binocular depth estimation method, road surface undulation information is obtained, and based on the on-board sensors, all attitude and motion parameters of the vehicle and suspension system are obtained.

[0065] In a preferred embodiment of the present invention, the process of obtaining road surface undulation information in step S1 is denoted as z. r At the current vehicle speed, the tire travels over a road surface with an undulation of z. r The time required to reach the desired position is denoted as t; all attitude and motion parameters of the vehicle and suspension system, including sprung mass m, are... s Unsprung mass m u equivalent stiffness k of air spring s Damper b s Actively control external force F a Tire equivalent stiffness k t Displacement of the spring mass zs Unsprung mass displacement z u .

[0066] S2: Construct a dynamic model of the suspension system and establish a state-space expression by selecting appropriate state variables.

[0067] As a preferred embodiment of the present invention, such as Figure 2 As shown, step S2 involves constructing a dynamic model of the suspension system and establishing a state-space expression by selecting appropriate state variables. The specific implementation method is as follows:

[0068] S21: Based on all the parameters obtained in step S1, the dynamic expression of the suspension system can be obtained according to Newton's kinematics principle:

[0069]

[0070] S22: Since the system has a total of 4 energy storage elements, the number of state variables is also 4. Let x1 = z s -z u , x3=z u -z r , Therefore, the dynamic expression is transformed into a state-space expression. Where x represents a vector consisting of four state variables, and the external force F a This represents the control input of the system. A, B, and M represent 4×4, 4×1, and 4×1 matrices, respectively. The specific expressions for each part are as follows:

[0071]

[0072] S3: Design an artificial neural network to calculate the weighted coefficient matrix in a learning manner based on vehicle parameters and road surface height.

[0073] In a preferred embodiment of the present invention, the design of the artificial neural network in step S3, which calculates the weighting coefficient matrix by learning based on vehicle parameters and road surface height, is specifically implemented as follows:

[0074] S31: The network input is a 10-dimensional vector, containing x1, x2, x3, x4, m respectively. s ,m u ,k s ,k t ,b s ,z r Subsequently, three fully connected layers with hidden dimensions of 64, 256, and 512 were constructed sequentially. The ReLU function was used for activation between every two fully connected layers, resulting in a 512-dimensional intermediate output vector.

[0075] S32: Design fully connected layers with output dimensions of 16 and 1 respectively. The former is used to generate a 4×4 matrix Q. tmp After all, it can be done in one step. The weighted sparse matrix Q is calculated, and after being activated by the Softplus function, the weighted coefficient matrix R is output.

[0076] More specifically, the Q matrix output in step S32 is a symmetric positive semi-definite matrix, and the R matrix is ​​a symmetric positive definite matrix.

[0077] S4: Based on the LQR method, the feedback gain matrix is ​​obtained, thereby obtaining the magnitude of the active control external force. Based on the precise discretization method, the attitude of the suspension after deformation is calculated, which is used to supervise the training of the artificial neural network.

[0078] In a preferred embodiment of the present invention, step S4, based on the LQR method, solves for the feedback gain matrix, thereby obtaining the magnitude of the active control external force. Based on a precise discretization method, the attitude of the suspension after deformation is calculated for supervised training of the artificial neural network. The specific implementation is as follows:

[0079] S41: Based on the weighted coefficient matrices Q and R obtained in step S3, solve for the P matrix in the Riccati equation. The specific equation expression is as follows:

[0080] PA+A T P+Q-PBR -1 B T P=0

[0081] S42: Based on the obtained P matrix, the feedback gain matrix K = R is further obtained. -1 B T P, and the optimal external force F in the current state. a =-Kx.

[0082] S43: Based on the A and B matrices, state variable x, and optimal control force F under the current state. a Based on the idea of ​​precise discretization, the new state variable x of the suspension after time t is further derived. next The specific calculation method is as follows:

[0083] G = e At H=∫0te Aτ dτ·B

[0084] x next =Gx+HF a

[0085] More specifically, step S43 calculates eAt Taylor series expansion can be used, and the first few terms can be selected for approximate calculation based on the actual accuracy requirements.

[0086] S44: To minimize the change in vehicle attitude in the vertical direction between the two time points, the designed artificial neural network is trained under supervision using smooth L1 loss. The specific loss function expression is as follows:

[0087]

[0088] S5: Based on the solution results, update the air spring stiffness of the suspension, and reacquire all attitude and motion parameters of the vehicle and suspension system to perform the next round of pre-control.

[0089] In a preferred embodiment of the present invention, the specific method for updating the air spring stiffness of the suspension based on the solution result in step S5 is as follows: when the suspension attitude changes, the stiffness is adjusted according to the current air chamber volume V of the air spring. e and the current effective force-bearing area A e Calculate the new equivalent stiffness k at the current moment. t The specific calculation formula is as follows:

[0090]

[0091] This invention implements a pre-control method for automotive suspension based on LQR and artificial neural networks. Tests show that using this pre-control method significantly improves the overall suspension control compared to not using it. Specific experimental road conditions and test results are as follows: Figure 3 and Figure 4 As shown. Given the significant role of active suspension control in ensuring driving safety and passenger comfort, this invention has important practical value, especially applicable to improving the safety and efficiency of automotive driving systems and enhancing vehicle ride comfort.

[0092] Those skilled in the art can understand this embodiment as a more specific description of Embodiment 1 and Embodiment 2.

[0093] Those skilled in the art will understand that, besides implementing the system and its various devices, modules, and units provided by this invention in the form of purely computer-readable program code, the same functions can be achieved entirely through logical programming of the method steps, making the system and its various devices, modules, and units of this invention function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, the system and its various devices, modules, and units provided by this invention can be considered as a hardware component, and the devices, modules, and units included therein for implementing various functions can also be considered as structures within the hardware component; alternatively, the devices, modules, and units for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0094] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A pre-control method for automobile suspension based on LQR and artificial neural networks, characterized in that, The method includes the following steps: Step S1: Based on binocular depth estimation technology, the undulation information of the road surface ahead is obtained in advance, and combined with all attitude parameters and motion parameters of the vehicle and suspension system obtained in real time by the on-board sensors; Step S2: Based on the obtained parameters, construct a dynamic model of the suspension system and describe it using state-space expressions; Step S3: Design an artificial neural network that takes vehicle parameters and road surface height as input, and calculates and outputs a weighted coefficient matrix for LQR control through learning. Step S4: Using the LQR method and the weighted coefficient matrix output by the artificial neural network, the feedback gain matrix is ​​obtained, thereby determining the magnitude of the active control external force. Based on the precise discretization method, the attitude of the suspension after deformation is predicted, which is used for online supervised training of the artificial neural network. Step S5: Based on the LQR solution, dynamically adjust the air spring stiffness of the suspension, and repeat steps S1 to S5 for continuous pre-control. In step S4, when solving the Riccati equation to obtain the matrix, the weighted coefficient matrix output by the artificial neural network is used as part of the equation parameters to perform learning-based LQR parameter adjustment. When predicting the attitude of the suspension after deformation based on the precise discretization method, the matrix, state variables and optimal control force under the current state are considered, and Taylor series expansion is used for approximate calculation. When training an artificial neural network in an online supervised manner, the smooth L1 loss function is used to minimize the change in the vehicle's attitude in the vertical direction between two consecutive time points.

2. The vehicle suspension pre-control method based on LQR and artificial neural networks according to claim 1, characterized in that, The road surface undulation information mentioned in step S1 is calculated based on the depth image acquired by the binocular camera, while taking into account the current vehicle speed to determine the time it takes for the tire to reach each road surface undulation.

3. The vehicle suspension pre-control method based on LQR and artificial neural networks according to claim 1, characterized in that, When constructing the dynamic model of the suspension system in step S2, the sprung mass, unsprung mass, equivalent stiffness of the air spring, damper, active control force, equivalent stiffness of the tire, and sprung mass displacement and unsprung mass displacement parameters are considered.

4. The vehicle suspension pre-control method based on LQR and artificial neural networks according to claim 1, characterized in that, The artificial neural network designed in step S3 has a multi-layer fully connected structure, including an input layer, at least three hidden layers and an output layer. The hidden layers are activated by the ReLU function, and the output layer is designed with specific dimensions to generate the weighted coefficient matrix according to the requirements. The output layer of the artificial neural network includes two sub-output layers: one for generating a symmetric positive semi-definite weighted sparse matrix, and the other for outputting a symmetric positive definite weighted coefficient matrix after activation by the Softplus function.

5. The vehicle suspension pre-control method based on LQR and artificial neural networks according to claim 1, characterized in that, In step S5, when dynamically adjusting the air spring stiffness of the suspension, the changes in the air spring cavity volume and effective force-bearing area after the suspension attitude changes are considered, and a new equivalent stiffness is calculated.

6. A vehicle suspension pre-control system based on LQR and artificial neural networks, characterized in that, The system includes the following modules: Module M1: Based on binocular depth estimation technology, it pre-acquires information on the undulation of the road surface ahead, and combines it with all attitude and motion parameters of the vehicle and suspension system acquired in real time by onboard sensors; Module M2: Based on the acquired parameters, construct a dynamic model of the suspension system and describe it using state-space expressions; Module M3: Design an artificial neural network that takes vehicle parameters and road surface height as input, and calculates and outputs a weighted coefficient matrix for LQR control through learning. Module M4: Using the LQR system and the weighted coefficient matrix output by the artificial neural network, the feedback gain matrix is ​​obtained, thereby determining the magnitude of the active control force and predicting the attitude of the suspension after deformation based on the precise discretization system, which is used for online supervised training of the artificial neural network. Module M5: Based on the LQR solution results, dynamically adjust the air spring stiffness of the suspension, and repeat modules M1 to S5 for continuous pre-control. When solving the Riccati equation in module M4 to obtain the matrix, the weighted coefficient matrix output by the artificial neural network is used as part of the equation parameters for learning-based LQR parameter adjustment. When predicting the attitude of the suspension after deformation based on the precise discretization system, the matrix, state variables and optimal control force under the current state are considered, and Taylor series expansion is used for approximate calculation. When training an artificial neural network in an online supervised manner, the smooth L1 loss function is used to minimize the change in the vehicle's attitude in the vertical direction between two consecutive time points. When dynamically adjusting the air spring stiffness of the suspension in module M5, the changes in the air spring cavity volume and effective force-bearing area after the suspension attitude changes are considered, and a new equivalent stiffness is calculated.

7. The automotive suspension pre-control system based on LQR and artificial neural networks according to claim 6, characterized in that, The road surface undulation information in module M1 is calculated based on the depth image acquired by the binocular camera, while taking into account the current vehicle speed to determine the time it takes for the tire to reach each road surface undulation. When constructing the dynamic model of the suspension system in module M2, the sprung mass, unsprung mass, equivalent stiffness of the air spring, damper, active control force, equivalent stiffness of the tire, and sprung mass displacement and unsprung mass displacement parameters are considered.

8. The automotive suspension pre-control system based on LQR and artificial neural networks according to claim 6, characterized in that, The artificial neural network designed in module M3 has a multi-layer fully connected structure, including an input layer, at least three hidden layers and an output layer. The hidden layers are activated by the ReLU function, and the output layer is designed with specific dimensions to generate the weighted coefficient matrix according to the requirements. The output layer of the artificial neural network includes two sub-output layers: one for generating a symmetric positive semi-definite weighted sparse matrix, and the other for outputting a symmetric positive definite weighted coefficient matrix after activation by the Softplus function.

Images