Modeling method and device of vehicle dynamics model, equipment and medium

Abstract

The embodiment of the invention discloses a modeling method, device, equipment and medium for a vehicle dynamics model, and the method comprises the steps: obtaining the historical motion state information of a current vehicle, and the current environment information and the current control information of the current moment, the environment information comprises road gradient information, road deceleration strip information and road friction force information, and the current control information comprises the road gradient information and the road deceleration strip information; the motion state information comprises speed information, course angle information and steering wheel turning angle information; inputting the current environment information, the current control information and the historical motion state information into a trained preset neural network model to obtain the motion state information of the current vehicle at the next moment, the preset neural network model establishes an association relationship among the current state information of the vehicle under the current environment information, the current control information and the motion state information of the next moment. By adopting the technical scheme, the accuracy of the dynamic model of the vehicle under different environments and different working conditions is improved.

Description

Technical Field

[0001] The embodiments of the present invention relate to the technical field of autonomous driving, and more specifically, to a method, device, equipment, and medium for modeling a vehicle dynamics model. Background Art

[0002] In the field of autonomous driving, creating a vehicle dynamics model is an important part of realizing autonomous driving simulation scenarios, which helps to improve test efficiency, reduce costs, optimize control logic, evaluate performance and safety. When establishing a vehicle dynamics model, variables such as the acceleration, speed, and displacement of the vehicle need to be determined to describe the motion state of the vehicle in time and space.

[0003] For most ordinary road scenarios with low acceleration change rate and small sideslip angle, a simple dynamics model is applicable. However, when a more accurate and comprehensive performance of the dynamics model is required, the simple dynamics model will not be able to predict and reproduce the dynamic characteristic changes caused by the environment, and a large number of model parameters need to be adjusted to make the dynamics model more accurate. However, the process of adjusting parameters is often very difficult and not easy to implement. Summary of the Invention

[0004] The embodiments of the present invention provide a method, device, equipment, and medium for modeling a vehicle dynamics model to improve the accuracy of the vehicle dynamics model in different environments and working conditions.

[0005] The specific technical solutions are as follows:

[0006] In a first aspect, the embodiments of the present invention provide a method for modeling a vehicle dynamics model, including:

[0007] Obtain the historical motion state information of the current vehicle, as well as the current environment information and current control information at the current moment. Among them, the environment information includes road slope information, road speed bump information, and road friction information, the control information includes vehicle throttle command and steering wheel angle command, and the motion state information includes speed information, heading angle information, and steering wheel angle information;

[0008] Input the current environment information, current control information, and historical motion state information into a pre-trained preset neural network model to obtain the motion state information of the current vehicle at the next moment. Among them, the preset neural network model has established the correlation relationship between the current state information, current control information of the vehicle under the current environment information and the motion state information at the next moment.

[0009] As can be seen from the above solution, by creating a vehicle dynamics model based on a trained end-to-end neural network model, the vehicle's dynamics model can better conform to the actual rigid body kinematics. Moreover, by using the environmental information of the vehicle as the input information of the neural network model, the obtained vehicle dynamics model can more accurately and comprehensively describe the vehicle's performance in different environments.

[0010] Optionally, the preset neural network model is trained in the following manner:

[0011] Obtain sample data for training the preset neural network model. The sample data includes environmental information, vehicle control information, and a motion state data set. The motion state data set includes motion state information at different times;

[0012] In each training process, input the sample data into the preset neural network model, and use the motion state data output by the preset neural network model to update the motion state information at one moment in the motion state data set to obtain a new motion state data set. Then, use the new sample motion state data set and the new control information as sample data for the next training;

[0013] When the loss function value between the motion state information output by the preset neural network model and the motion state information in the sample data reaches convergence, the training of the preset neural network model is completed.

[0014] Optionally, the preset neural network model is composed of an input layer, a gated recurrent unit (GRU) layer, a first feature extraction layer, a second feature extraction layer, and an output layer in sequence.

[0015] Optionally, the road speed bump information is represented by the acceleration information when the vehicle passes over the speed bump. The acceleration information is obtained in the following manner:

[0016] Input the slope angle information of the speed bump, the vehicle speed information when passing over the speed bump, and the throttle information into the preset speed bump model to obtain the acceleration information when the vehicle passes over the speed bump.

[0017] Optionally, the road slope information is obtained by collecting it with an inertial measurement unit (IMU) during the actual driving process of the vehicle; or, it is obtained based on a method of fusing multiple in-vehicle sensor data. Among them, the multiple sensors include IMU, GPS, radar, and / or image sensors.

[0018] In a second aspect, an embodiment of the present invention further provides a modeling device for a vehicle dynamics model, including:

[0019] A status information acquisition module, configured to acquire historical motion state information of the current vehicle, as well as current environment information and current control information at the current moment. Among them, the environment information includes road gradient information, road speed bump information, and road friction information, the control information includes vehicle throttle commands and steering wheel angle commands, and the motion state information includes speed information, heading angle information, and steering wheel angle information;

[0020] A motion state information prediction module, configured to input the current environment information, current control information, and historical motion state information into a pre-trained preset neural network model to obtain the motion state information of the current vehicle at the next moment. Among them, the preset neural network model establishes the association relationship between the current state information, current control information of the vehicle under the current environment information, and the motion state information at the next moment.

[0021] Optionally, the preset neural network model is trained in the following manner:

[0022] Obtain sample data for training the preset neural network model. Among them, the sample data includes environment information, control information of the vehicle, and a motion state data set, and the motion state data set includes motion state information at different moments;

[0023] In each training process, input the sample data into the preset neural network model, and use the motion state data output by the preset neural network model to update the motion state information at one moment of the motion state data set to obtain a new motion state data set, and use the new sample motion state data set and the new control information as sample data for the next training;

[0024] When the loss function value between the motion state information output by the preset neural network model and the motion state information in the sample data reaches convergence, the training of the preset neural network model is completed.

[0025] Optionally, the preset neural network model consists of an input layer, a gated recurrent unit (GRU) layer, a first feature extraction layer, a second feature extraction layer, and an output layer in sequence.

[0026] Optionally, the road speed bump information is represented by the acceleration information when the vehicle passes over the speed bump. The acceleration information is obtained in the following manner:

[0027] Input the gradient angle information of the speed bump, the speed information of the vehicle when passing over the speed bump, and the throttle information into a preset speed bump model to obtain the acceleration information of the vehicle when passing over the speed bump.

[0028] Optionally, the road slope information is obtained by collecting data from an inertial measurement unit (IMU) during actual vehicle driving; alternatively, it is obtained based on the fusion of multiple vehicle-mounted sensor data, where the multiple sensors include an IMU, GPS, radar, and / or an image sensor.

[0029] In a third aspect, an embodiment of the present invention provides a computer device, which includes:

[0030] At least one processor, where the processor is coupled to a memory, and the memory stores a program or instructions that, when executed by the processor, implement the method for modeling a vehicle dynamics model provided in any embodiment of the present invention.

[0031] In a fourth aspect, an embodiment of the present invention provides a readable storage medium, on which a program or instructions are stored, and when the program or instructions are executed by a processor, they implement the method for modeling a vehicle dynamics model provided in any embodiment of the present invention.

[0032] In a fifth aspect, an embodiment of the present invention provides a vehicle, which includes the device for modeling a vehicle dynamics model provided in any embodiment of the present invention, or includes the computer device provided in any embodiment of the present invention.

[0033] In a sixth aspect, an embodiment of the present invention provides a computer program, which includes program instructions, and when the program instructions are executed by a computer, they implement the method for modeling a vehicle dynamics model provided in any embodiment of the present invention. Description of the Drawings

[0034] To more clearly illustrate the technical solutions in the embodiments of the present invention or in the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0035] Figure 1a It is a schematic structural diagram of a neural network model provided in Embodiment 1 of the present invention;

[0036] Figure 1b It is a flowchart of a method for training a neural network model provided in Embodiment 1 of the present invention;

[0037] Figure 2a It is a flowchart of a method for modeling a vehicle dynamics model provided in Embodiment 2 of the present invention;

[0038] Figure 2b It is a schematic diagram for determining vehicle motion state information based on an end-to-end neural network model provided in Embodiment 2 of the present invention;

[0039] Figure 3 It is a structural block diagram of a modeling device for a vehicle dynamics model provided in Embodiment 3 of the present invention;

[0040] Figure 4 It is a structural block diagram of a computer device provided in Embodiment 4 of the present invention;

[0041] Figure 5 It is a schematic diagram of a vehicle provided in Embodiment 5 of the present invention. Detailed implementation manners

[0042] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0043] It should be noted that the terms "include" and "have" and any variations thereof in the embodiments of the present invention and the accompanying drawings are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally further includes steps or units not listed, or optionally further includes other steps or units inherent to these processes, methods, products or devices.

[0044] To more clearly explain the content of each embodiment of the present invention, the implementation principles of the embodiments of the present invention will be briefly introduced below.

[0045] An embodiment of the present invention discloses a modeling method, device, equipment and medium for a vehicle dynamics model. Among them, the establishment of the vehicle dynamics model requires defining state variables such as the acceleration, speed and displacement of the vehicle to describe the motion state of the vehicle in time and space. In this embodiment, state variables such as the acceleration, speed, position and heading angle of the vehicle during the creation of the vehicle dynamics model are obtained based on an end-to-end neural network. Specifically, the current environment information of the vehicle, the vehicle control quantity at the current moment and the vehicle state quantity at the historical moment are input into a preset neural network model that has been trained. The neural network model will output the state quantity at the next moment, so as to realize the determination of the end-to-end vehicle state information. Among them, the state variables include the speed information, heading angle information and steering wheel angle information of the current vehicle, etc. The control quantity is a control instruction for driving the vehicle, including a throttle instruction and a steering wheel angle instruction, etc. The solution of the related technology is to first input the steering wheel angle instruction and the throttle information instruction into a simple neural network model to obtain the acceleration information of the vehicle and the change rate of the steering wheel angle, and then perform integral operations on the acceleration information and the change rate of the steering wheel angle respectively through a pure mathematical integration method, so as to predict the speed, position and steering wheel angle of the vehicle at the next moment, etc. Such processing does not conform to the actual rigid body kinematics of the vehicle, and the obtained vehicle dynamics model cannot describe the highly nonlinear changes of the model caused by sudden acceleration and deceleration, or sudden steering, and also cannot predict and reproduce the changes in the dynamic characteristics of the model caused by the environment. The technical solution provided by the embodiment of the present invention creates a vehicle dynamics model based on an end-to-end neural network model that has been trained, making the vehicle dynamics model more in line with the actual rigid body kinematics. And, by using the environment information where the vehicle is located as the input information of the neural network model, the obtained vehicle dynamics model can more accurately describe the performance of the vehicle in different environments.

[0046] Next, the technical solution provided by the embodiment of the present invention will be described in detail from two aspects: the training stage and the application stage of the preset neural network model.

[0047] Embodiment 1

[0048] Figure 1a is a schematic structural diagram of a neural network model provided by Embodiment 1 of the present invention, as Figure 1aAs shown in the figure, the neural network model provided in this embodiment is successively composed of an input layer, a gated recurrent unit (GRU) layer, a first feature extraction layer, a second feature extraction layer, and an output layer. Among them, the input layer is used to input the current state information, control information, and environmental information of the vehicle. The GRU layer can effectively process time series data and capture the long-term dependence relationship of the input data. The feature extraction layer (such as the Dense layer) can learn the internal representation and features of the input data and is often used for the prediction output of the model. In this embodiment, by adopting two feature extraction layers, the complexity of the model can be increased, enabling it to learn and adapt to more complex data distributions. The output layer is used to output the predicted vehicle state information. Figure 1b It is a flowchart of a neural network model training method provided in Embodiment 1 of the present invention. As Figure 1b shown, the training method of the neural network model provided in this embodiment includes:

[0049] S110. Obtain sample data for training a preset neural network model.

[0050] In this embodiment, the sample data is the data collected during the actual driving of the vehicle, including environmental information, control information of the vehicle, and a motion state data set, and the motion state data set includes motion state information at different times. Using the data collected during the actual driving of the vehicle to train the preset neural network model can enable the trained model to learn the dynamic characteristics of the vehicle under various working conditions, so that the vehicle dynamics model obtained based on this neural network model can accurately and comprehensively describe the dynamic characteristics of the vehicle under various working conditions.

[0051] Specifically, the motion state information at each moment may include the longitudinal speed, lateral speed, heading angle, and steering wheel angle of the vehicle, etc. The control information of the vehicle is a control instruction for driving the vehicle, including driving instructions, such as throttle instructions and steering wheel angle instructions. The environmental information includes road slope information, road speed bump information, and road friction information, etc.

[0052] Among them, the road ramp information can be represented by the slope angle, which can be obtained by the vehicle collecting through the IMU (Inertial Measurement Unit) during actual driving; alternatively, it can also be obtained based on the fusion of multiple vehicle-mounted sensor data. Among them, the multi-sensors include IMU, GPS (Global Positioning System), radar, and / or image sensors. The road speed bump information can be represented by the acceleration information when the vehicle passes through the road. The acceleration information can be obtained by inputting the slope angle information of the speed bump, the speed information and the throttle information of the vehicle when passing through the speed bump into a preset speed bump model, and the output of the preset speed bump model is the acceleration information of the vehicle when passing through the speed bump. Among them, the preset speed bump model is a deep neural network model.

[0053] S120. During each training process, input the sample data into the preset neural network model, and use the motion state data output by the preset neural network model to update the motion state information at one moment in the motion state data set, obtain a new motion state data set, and use the new motion state data set and the new control information as sample data for the next training. When the loss function value between the motion state information output by the preset neural network model and the motion state information in the sample data reaches convergence, the training of the preset neural network model is completed.

[0054] In this embodiment, after being trained, the neural network model can predict the vehicle state at the next moment. During the training process, the motion state data output by the neural network model at each moment can replace the motion state information at one moment in the motion state data set. For example, the motion state information at the farthest moment in the motion state data of each moment in the motion state data set can be deleted, and the motion state data output by the neural network model at the current moment can be added to the motion state data set, so as to obtain a new motion state data set, and the new motion state data set and the new control information can be used for the next training. The loss function adopted during the training process can be the mean square error loss function.

[0055] The technical solution provided in this embodiment, by adopting the motion state data and environmental data collected by the vehicle during actual driving in the training process of the preset neural network model, can enable the trained neural network to learn the dynamic characteristics of the vehicle under various working conditions. The obtained vehicle dynamics model can accurately describe the highly nonlinear changes of the model caused by sudden acceleration and deceleration, or sudden steering, and can also accurately predict and reproduce the changes of the model dynamic characteristics caused by the environment.

[0056] After the preset neural network model is trained, it can be applied to the modeling process of the vehicle dynamics model. For the specific application process, please refer to the description of the following embodiments.

[0057] Embodiment 2

[0058] Figure 2a The flowchart of a method for modeling a vehicle dynamics model provided in Embodiment 2 of the present invention. This method can be applied to in-vehicle terminals such as in-vehicle computers and industrial personal computers (IPCs) in vehicles, and can also be applied to servers. The embodiments of the present invention do not make specific limitations in this regard. The method provided in this embodiment can be executed by a modeling device for a vehicle dynamics model, and this device can be implemented in software and / or hardware. As Figure 2a shown, the method provided in this embodiment specifically includes:

[0059] S210. Obtain the historical motion state information of the current vehicle, as well as the current environmental information and current control information at the current moment.

[0060] Among them, the motion state information includes speed information, heading angle information, steering wheel angle information, etc. The historical motion state information is the motion state information of the vehicle at a time before the current moment, such as the motion state information at the previous moment, the motion state information at the previous two moments, or the motion state information at the previous three moments of the current moment, etc. The present embodiment does not make a specific limitation on the number of historical moments used.

[0061] In this embodiment, the control information is a control instruction for driving the vehicle to travel, including a vehicle throttle instruction and a steering wheel angle instruction.

[0062] S220. Input the current environmental information, current control information, and historical motion state information into the trained preset neural network model to obtain the motion state information of the current vehicle at the next moment.

[0063] Among them, the training process of the preset neural network model can refer to the description of the above embodiments and will not be elaborated here. In this embodiment, the trained neural network model can be applied to the modeling process of the vehicle dynamics model to obtain information such as the speed, position, and heading angle of the vehicle.

[0064] Specifically, Figure 2b is a schematic diagram for determining the vehicle motion state information based on an end-to-end neural network model provided in Embodiment 2 of the present invention. As Figure 2bAs shown, the historical motion state information of the current vehicle, the current environmental information at the current moment, and the current control information are input into the neural network model, and the output of this model is the motion state information of the vehicle at the next moment, realizing the determination of the parameters of the vehicle dynamics model using an end-to-end neural network model.

[0065] The technical solution provided in this embodiment creates a vehicle dynamics model based on a trained end-to-end neural network model, which can make the vehicle dynamics model more in line with the actual rigid body kinematics. Moreover, by using the environmental information where the vehicle is located as the input information of the neural network model, the obtained vehicle dynamics model can more accurately and comprehensively describe the performance of the vehicle in different environments. Compared with the related technology that performs integral operations on acceleration information and the change rate of the steering wheel angle to predict the speed, position, and steering wheel angle of the vehicle at the next moment, the technical solution provided in this embodiment effectively improves the accuracy of the vehicle dynamics model in different environments and different working conditions, and also avoids the problems of difficult adjustment of model parameters and hard-to-implement.

[0066] Embodiment III

[0067] Figure 3 is a structural block diagram of a modeling device for a vehicle dynamics model provided in Embodiment III of the present invention. As Figure 3 shown, the device includes: a state information acquisition module 310 and a motion state information prediction module 320, where

[0068] The state information acquisition module 310 is configured to acquire the historical motion state information of the current vehicle, as well as the current environmental information and the current control information at the current moment. Among them, the environmental information includes road slope information, road speed bump information, and road friction information, the control information includes vehicle throttle instructions and steering wheel angle instructions, and the motion state information includes speed information, heading angle information, and steering wheel angle information;

[0069] The motion state information prediction module 320 is configured to input the current environmental information, the current control information, and the historical motion state information into a trained preset neural network model to obtain the motion state information of the current vehicle at the next moment. Among them, the preset neural network model establishes the correlation relationship between the current state information, the current control information of the vehicle under the current environmental information, and the motion state information at the next moment.

[0070] Optionally, the preset neural network model is trained in the following manner:

[0071] Obtain sample data for training a preset neural network model, where the sample data includes environmental information, control information of the vehicle, and a motion state data set, and the motion state data set includes motion state information at different times;

[0072] In each training process, input the sample data into the preset neural network model, and use the motion state data output by the preset neural network model to update the motion state information at one moment in the motion state data set to obtain a new motion state data set, and use the new sample motion state data set and the new control information as sample data for the next training;

[0073] When the loss function value between the motion state information output by the preset neural network model and the motion state information in the sample data reaches convergence, the training of the preset neural network model is completed.

[0074] Optionally, the preset neural network model is successively composed of an input layer, a gated recurrent unit (GRU) layer, a first feature extraction layer, a second feature extraction layer, and an output layer.

[0075] Optionally, the road speed bump information is represented by the acceleration information when the vehicle passes over the speed bump, and the acceleration information is obtained through the following method:

[0076] Input the slope angle information of the speed bump, the speed information of the vehicle when passing over the speed bump, and the throttle information into a preset speed bump model to obtain the acceleration information of the vehicle when passing over the speed bump.

[0077] Optionally, the road slope information is obtained by collecting through an inertial measurement unit (IMU) during the actual driving process of the vehicle; or, it is obtained based on a method of fusing multiple in-vehicle sensor data, where the multiple sensors include IMU, GPS, radar, and / or image sensors.

[0078] The modeling device of the vehicle dynamics model provided by the embodiments of the present invention can execute the modeling method of the vehicle dynamics model provided by any embodiment of the present invention, and has corresponding functional modules and beneficial effects for executing the method. For technical details not described in detail in the above embodiments, reference can be made to the modeling method of the vehicle dynamics model provided by any embodiment of the present invention.

[0079] Embodiment 4

[0080] Figure 4 Shown is a structural block diagram of a computer device provided for Embodiment 4 of the present invention, as Figure 4 shown, the computer device includes:

[0081] At least one processor ( Figure 4 a processor 520 is shown in

[0082] The processor 520 is coupled to the memory 510, and the memory 510 stores a program or instructions that run on the processor 520. When the program or instructions are executed by the processor 520, the modeling method of the vehicle dynamics model provided by any embodiment of the present invention is implemented.

[0083] Based on the above embodiments, another embodiment of the present invention provides a vehicle, which includes the device described in any of the above embodiments, or includes the computer device described above.

[0084] Embodiment Five

[0085] Figure 5 It is a schematic diagram of a vehicle provided for Embodiment Five of the present invention. As Figure 5 shown, the vehicle includes a speed sensor 61, an ECU (Electronic Control Unit) 62, a GPS (Global Positioning System) positioning device 63, and a T-Box (Telematics Box) 64. Among them, the speed sensor 61 is used to measure the vehicle speed and use the vehicle speed as the empirical speed for model training; the GPS positioning device 63 is used to obtain the current geographical location of the vehicle; the T-Box 64 can communicate with the server as a gateway; the ECU 62 can execute the modeling method of the vehicle dynamics model described above.

[0086] In addition, the vehicle may further include: a V2X (Vehicle-to-Everything) module 65, a radar 66, and a camera 67. The V2X module 65 is used to communicate with other vehicles, roadside devices, etc.; the radar 66 or the camera 67 is used to sense the road environment information in the front and / or other directions to obtain the original point cloud data; the radar 66 and / or the camera 67 can be configured at the front and / or the rear of the vehicle body.

[0087] Based on the above method embodiments, another embodiment of the present invention provides a readable storage medium, on which a program or instructions are stored. When the program or instructions are executed by a processor, the processor implements the modeling method of the vehicle dynamics model described in any of the above embodiments.

[0088] Those of ordinary skill in the art can understand that the drawings are only schematic diagrams of an embodiment, and the modules or processes in the drawings are not necessarily essential for implementing the present invention.

[0089] Those of ordinary skill in the art can understand that the modules in the device in the embodiments can be distributed in the device in the embodiments as described in the embodiments, or can be correspondingly changed to be located in one or more devices different from the present embodiments. The modules in the above embodiments can be combined into one module, or can be further split into multiple sub-modules.

[0090] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for modeling a vehicle dynamics model, where the dynamics model includes the motion state information of the vehicle. Characterized in that, It includes: Obtain the historical motion state information of the current vehicle, as well as the current environment information and current control information at the current moment. Among them, the environment information includes road slope information, road speed bump information, and road friction information, and the control information includes vehicle throttle commands and steering wheel angle commands. The motion state information includes speed information, heading angle information, and steering wheel angle information; Input the current environment information, current control information, and historical motion state information into a pre-trained preset neural network model to obtain the motion state information of the current vehicle at the next moment. Among them, the preset neural network model establishes the association relationship between the current state information, current control information of the vehicle under the current environment information, and the motion state information at the next moment.

2. The method according to claim 1, Characterized in that, The preset neural network model is trained in the following manner: Obtain sample data for training the preset neural network model. Among them, the sample data includes environment information, vehicle control information, and a motion state data set. The motion state data set includes motion state information at different moments; In each training process, input the sample data into the preset neural network model, and use the motion state data output by the preset neural network model to update the motion state information of one moment in the motion state data set to obtain a new motion state data set, and use the new sample motion state data set and new control information as sample data for the next training; When the loss function value between the motion state information output by the preset neural network model and the motion state information in the sample data reaches convergence, the training of the preset neural network model is completed.

3. The method according to claim 1 or 2, Characterized in that, The preset neural network model is successively composed of an input layer, a gated recurrent unit (GRU) layer, a first feature extraction layer, a second feature extraction layer, and an output layer.

4. The method according to claim 1, Characterized in that, The road speed bump information is represented by the acceleration information when the vehicle passes through the speed bump. This acceleration information is obtained in the following manner: Input the slope angle information of the speed bump, the speed information of the vehicle when passing through the speed bump, and the throttle information into a preset speed bump model to obtain the acceleration information of the vehicle when passing through the speed bump.

5. The method according to claim 1, Characterized in that, The road slope information is obtained by collecting through an inertial measurement unit (IMU) during the actual driving process of the vehicle; or obtained based on the method of fusing multiple on-vehicle sensor data. Among them, the multiple sensors include IMU, GPS, radar, and / or image sensors.

6. A modeling device for a vehicle dynamics model, Characterized in that, It includes: A status information acquisition module, configured to acquire historical motion state information of a current vehicle, current environment information and current control information at a current moment, where the environment information includes road slope information, road speed bump information and road friction information, the control information includes a vehicle throttle command and a steering wheel angle command, and the motion state information includes speed information, heading angle information and steering wheel angle information; A motion state information prediction module, configured to input the current environment information, current control information and historical motion state information into a pre-trained preset neural network model to obtain motion state information of the current vehicle at the next moment, where the preset neural network model establishes a correlation relationship between the current state information, current control information of the vehicle under the current environment information and the motion state information at the next moment.

7. The apparatus according to claim 6, wherein, the preset neural network model is trained in the following manner: Obtain sample data for training the preset neural network model, where the sample data includes environment information, control information of the vehicle and a motion state data set, and the motion state data set includes motion state information at different moments; In each training process, input the sample data into the preset neural network model, and use the motion state data output by the preset neural network model to update the motion state information of one moment in the motion state data set to obtain a new motion state data set, and use the new sample motion state data set and the new control information as sample data for the next training; When the loss function value between the motion state information output by the preset neural network model and the motion state information in the sample data reaches convergence, the training of the preset neural network model is completed.

8. The apparatus according to claim 6, wherein, the preset neural network model is sequentially composed of an input layer, a gated recurrent unit (GRU) layer, a first feature extraction layer, a second feature extraction layer and an output layer.

9. A computer device, wherein, it includes at least one processor, the processor is coupled with a memory, the memory stores a program or instruction running on the processor, and when the program or instruction is executed by the processor, the steps of the method for modeling a vehicle dynamics model according to any one of claims 1 to 5 are implemented.

10. A readable storage medium, on which a program or instruction is stored, wherein, when the program or instruction is executed by a processor, the steps of the method for modeling a vehicle dynamics model according to any one of claims 1 to 5 are implemented.

Images