CN-121995790-A - Feedforward suspension virtual evaluation system and construction method thereof

CN121995790ACN 121995790 ACN121995790 ACN 121995790ACN-121995790-A

Abstract

The invention relates to the technical field of automobile evaluation, in particular to a feedforward suspension virtual evaluation system and a construction method thereof. The method comprises the steps of constructing a digital twin road scene in a virtual simulation environment, outputting a binocular RAW image and a front road surface elevation sequence, injecting the front road surface elevation sequence into a vehicle dynamics model, reversely sending vehicle pose/state information to the virtual simulation environment by the vehicle dynamics model, carrying out zero-frame buffer serialization output on the binocular RAW image through a video injection board card, feeding the binocular RAW image into an ADAS controller of a vehicle to be tested, generating predicted road surface excitation information, sending the predicted road surface excitation information to a suspension ECU, generating a damping/stiffness control instruction according to the road surface excitation information, driving a shock absorber test bench to generate an actual damping force, and collecting the actual damping force of the test bench to replace a virtual damping force model of the vehicle dynamics model to carry out dynamics calculation at the next moment. According to the technical scheme, the development efficiency can be improved, and the development safety is improved.

Inventors

YAN LIU
LI SHIYING
XU HAOXUAN
YANG SIYU
ZHU LIJIANG
TANG JUN
YUAN YUAN
MA YUANYUAN

Assignees

中国汽车工程研究院股份有限公司

Dates

Publication Date: 20260508
Application Date: 20260123

Claims (9)

1. The method for building the feedforward suspension virtual evaluation system is characterized by comprising the following steps of: Constructing a digital twin road scene consistent with a real test field in a virtual simulation environment, and outputting a binocular RAW image and a front road surface high program sequence; Injecting the front road surface high program list into a vehicle dynamics model, and reversely transmitting vehicle pose/state information to a virtual simulation environment by the vehicle dynamics model; Zero frame buffer serialization output is carried out on the binocular RAW image through a video injection board card, the zero frame buffer serialization output is fed into an ADAS controller of a vehicle to be tested, the ADAS controller generates predicted road surface excitation information, the predicted road surface excitation information is sent to a suspension ECU, and the road surface excitation information comprises a road surface elevation sequence, an obstacle position and estimated impact strength; And the suspension ECU generates damping/stiffness control instructions according to the road surface excitation information, drives the shock absorber test bench to generate actual damping force, and acquires the actual damping force of the test bench to replace a virtual damping force model of the vehicle dynamics model to perform dynamics calculation at the next moment.
2. The method for building a virtual evaluation system for a feed-forward suspension according to claim 1, wherein the dynamics calculation process at the next time is expressed as: Wherein, the The complete state vector of the vehicle is the current time Carsim; a front road surface high program list which is output in real time for the virtual simulation environment; the real damping force four-way vector is measured and fed back in real time by the force sensor for the shock absorber test bench.
3. The method for building a feedforward suspension virtual evaluation system according to claim 1, wherein a digital twin road scene which is highly consistent with a real test field is built in a virtual simulation environment, and a binocular RAW image and a forward road surface elevation sequence are output, comprising the following contents: the method comprises the steps of acquiring a target test field point cloud by adopting a mobile laser scanning vehicle, unmanned aerial vehicle aerial survey and a knapsack/hand-push type three-dimensional scanner, denoising and registering the target test field point cloud, deriving static grid collision and texture of a visual model, generating a logic road of a logic model, and importing the visual model and the logic model into a virtual engine for integration. A binocular virtual camera consisting of two monocular RGB sensors is configured in a virtual simulation environment, When the vehicle runs, the virtual simulation environment queries the elevation profile data of the road surface within a certain distance in front in real time according to the current position and speed of the vehicle, wherein the elevation data comprises the absolute elevation of the left and right track road surfaces of each sampling point and the road surface attachment coefficient of the corresponding sampling point.
4. The method of claim 1, wherein the vehicle dynamics model comprises a vehicle body, a suspension, a tire and a steering subsystem, and the semi-active/active suspension model is configured, and damping force or rigidity input interfaces of the semi-active/active suspension model need to be reserved for control signals of a suspension ECU.
5. The method for building a virtual evaluation system for a feedforward suspension according to claim 1, wherein the estimated road surface impact strength formula is: Wherein, the To estimate the impact strength of the road surface; is the current body speed; the acceleration is the current vehicle body acceleration; to predict the road elevation; the current road elevation; is a weight coefficient.
6. The method for building the feedforward suspension virtual evaluation system according to claim 1, wherein the zero frame buffer serialization output is performed on the binocular RAW image through a hardware-level video injection board card, the binocular RAW image is fed into an ADAS controller of a vehicle to be tested, the ADAS controller generates predicted road surface excitation information and sends the predicted road surface excitation information to a suspension ECU, and the method comprises the following steps: and the ADAS controller receives the serial link signals, performs image preprocessing, lane line recognition, obstacle detection and road surface elevation prediction based on a binocular SGBM/ELAS algorithm, and sends the road surface elevation sequence and the obstacle position within a certain distance in front to the suspension ECU through CAN/Ethernet.
7. The method for building the feedforward suspension virtual assessment system according to claim 6, wherein the suspension ECU fuses the pre-aiming road surface information and the vehicle state of the vehicle dynamics model, runs a feedforward control algorithm to output a damping adjustment PWM or current command and an air spring air pressure command, and drives the rack after translation by the real-time machine.
8. The method for building the feedforward suspension virtual evaluation system according to claim 1, wherein the upper computer is used for unified management of scene configuration, data recording, three-dimensional animation playback and HIL debugging, performance evaluation and control strategy optimization.
9. A feed-forward suspension virtual assessment system, characterized in that a feed-forward suspension virtual assessment system construction method according to any one of claims 1-8 is applied, comprising: The virtual simulation environment is used for constructing and running a virtual road scene consistent with the geometric characteristics and the physical characteristics of the real test field, and outputting vehicle pose, synchronous binocular vision images and front road elevation pre-aiming data; the vehicle dynamics model is deployed on real time and is used for running the whole vehicle dynamics model and receiving the road elevation pre-aiming data so as to calculate the vehicle state and the suspension stroke; in real time, running the automobile dynamics model and related sub-models, and communicating with the CARLA, the ADAS controller and the suspension ECU controller; The video injection board is used for converting the binocular vision image into a serial signal and embedding a hardware time stamp synchronous with the virtual simulation environment; The ADAS controller is used for receiving the serial signals, executing three-dimensional matching and road elevation estimation, and outputting front road excitation pre-aiming information; the suspension ECU is used for generating a damper damping control instruction according to the road surface excitation pre-aiming information and the vehicle state data; The shock absorber hardware is arranged on the annular rack and is used for receiving the damping control instruction and a suspension stroke speed signal from the vehicle dynamics model, driving the real shock absorber to generate an actual damping force, and feeding back the actual damping force to the vehicle dynamics model to replace the built-in damping force of the model.

Description

Feedforward suspension virtual evaluation system and construction method thereof Technical Field The invention relates to the technical field of automobile evaluation, in particular to a feedforward suspension virtual evaluation system and a construction method thereof. Background The automotive industry is primarily moving to intellectualization and electrodynamic, and active suspension and semi-active suspension systems play an important role in improving the riding comfort and the steering stability of vehicles. Among other things, road pre-sighting technology provides a powerful support for further optimization of suspension system performance. The road pre-aiming technology mainly uses forward perception sensors, such as cameras, laser radars and the like, to accurately acquire detailed information of a road ahead in advance, and covers key elements such as road surface unevenness, obstacle positions, characteristics and the like. Based on the information acquired in advance, the suspension system can rapidly and accurately adjust damping or rigidity parameters of the suspension system so as to adapt to the change of the road conditions to be faced in advance, thereby remarkably improving the running performance of the vehicle under various complex road conditions and bringing more stable and comfortable traveling experience to drivers and passengers. In the development flow of the road pre-aiming vision feed-forward suspension system, the system verification link directly relates to the reliability and stability of the system performance. Traditionally, this verification process has been highly dependent on a large number of real vehicle tests. However, real vehicle testing presents a number of challenges in practical operation that are difficult to overcome. From the cost perspective, real vehicle testing requires a tremendous amount of capital investment, covering many aspects of purchasing and refitting the test vehicle, outfitting professional test equipment, renting test sites, and paying test personnel. The high cost makes many enterprises and research institutions face huge economic pressure when relevant tests are carried out, limits the scale and frequency of the tests, and further influences the development progress and performance optimization effect of the system. In terms of the test period, the flow of the real vehicle test is extremely tedious and time-consuming for a long time. From careful preparation of road and environmental conditions, to various operations and monitoring during test execution, to subsequent data collection and analysis, each link requires a significant amount of time. Especially when faced with complex and changeable test requirements, the test period is generally further prolonged, which results in low efficiency of the whole research and development process, and difficulty in meeting the requirement of the market for quick iteration of new products. Poor reproducibility is also a significant drawback of real vehicle testing. Because the road condition of a real road is complex and various, and is dynamically influenced by various factors such as weather, traffic flow and the like, the specific road excitation and driving scene are difficult to realize when the specific road excitation and driving scene are required to be accurately reproduced. The method makes it difficult to obtain consistent test results in different test stages or under different test conditions, so that great difficulty is brought to iteration and optimization of a control strategy, and the accuracy and effectiveness of research and development are reduced. More importantly, the real vehicle test under the extreme working condition has serious safety risks. For example, under extreme conditions such as high speed, sharp turns, severe road conditions, etc., the performance and stability of the vehicle face a great challenge, and once an accident occurs in the testing process, the testing vehicle may be seriously damaged, and the life safety of the testing personnel may be endangered. The safety risk has made enterprises and research institutions extremely careful when conducting extreme condition tests, and even in some cases have to discard some necessary test items, thereby affecting the overall verification and performance improvement of the system. In addition to the problems with real vehicle testing, existing test schemes also suffer from significant shortcomings. The conventional Prescan +carsim scheme is too simplified in the design of a camera model to output original bayer format image data (RAW image). In practical application, ADAS domain control needs to be subjected to advanced processing and analysis based on original image data so as to realize more accurate environment perception and decision control. Because the RAW image cannot be provided by the scheme, the ADAS domain control cannot form complete closed loop control, and the performance exertion and test effect of the roa