CN109190171A - A method of the vehicle movement model optimization based on deep learning - Google Patents
A method of the vehicle movement model optimization based on deep learning Download PDFInfo
- Publication number
- CN109190171A CN109190171A CN201810871827.8A CN201810871827A CN109190171A CN 109190171 A CN109190171 A CN 109190171A CN 201810871827 A CN201810871827 A CN 201810871827A CN 109190171 A CN109190171 A CN 109190171A
- Authority
- CN
- China
- Prior art keywords
- deep learning
- vehicle
- angle
- steering wheel
- mass center
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/15—Vehicle, aircraft or watercraft design
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Geometry (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Aviation & Aerospace Engineering (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Computer Hardware Design (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Automation & Control Theory (AREA)
- Steering Control In Accordance With Driving Conditions (AREA)
Abstract
The method of the present invention relates to a kind of vehicle movement model optimization based on deep learning, the nonlinear function between steering wheel angle α and mass center drift angle β is fitted by way of deep learning, such a process finally can be obtained, it is known that steering wheel angle α can solve β mass center drift angle.This method mainly trains the mass center drift angle information being collected into a large amount of vehicle body steering wheel information, speed, centripetal acceleration and vehicle movement model by building deep learning neural network training model, find the nonlinear correspondence relation between steering wheel angle and mass center drift angle information, to which the steering wheel angle information in vehicle movement model be optimized, more accurate direction of motion information is obtained, the accuracy of motion model dead reckoning is further increased.
Description
Technical field
The present invention relates to intelligent driving technical fields, and in particular to a kind of vehicle movement model optimization based on deep learning
Method.
Background technique
When using simple vehicle movement model, yaw angle θ cannot accurately describe the deflection angle of vehicle real motion
Degree, simple vehicle movement model, as shown in Figure 1,
Wherein, θ is yaw angle, it is the anticlockwise angle of opposite X-axis；
V is the speed in the direction θ；
(x, y) is the position coordinates of vehicle.
And real vehicles kinematics model, as shown in Fig. 2,
Wherein, ψ is the yaw angle of vehicle；β is the mass center drift angle of vehicle.
If β mass center drift angle can be found out, so that it may solve the exact position of vehicle with real vehicles kinematics model
Estimation.Normal practice is then to solve β mass center using calculation formula every 10 ° of detection front tyre drift angles by calibration steering wheel
, there are two aspects, the first calibration when being stationary vehicle in the problem of drift angle, this mode is brought, data acquisition be it is discrete, it is intermediate poor
Value needs to be acquired with integral, and error is relatively large；Second is that β mass center drift angle and speed, centripetal acceleration, tire cornering power have
It closes, is continuous dynamic process, it is single correctly to solve β mass center drift angle from static demarcating method.
Summary of the invention
The present invention for the technical problems in the prior art, provides a kind of vehicle movement model based on deep learning
The method of optimization fits the nonlinear function between steering wheel angle α and mass center drift angle β, most by way of deep learning
Such a process can be obtained eventually, it is known that steering wheel angle α can solve β mass center drift angle.This method is mainly by building depth
Neural network training model is practised to being collected into a large amount of vehicle body steering wheel information, speed, centripetal acceleration and vehicle movement model
In mass center drift angle information trained, find the nonlinear correspondence relation between steering wheel angle and mass center drift angle information, from
And optimized the steering wheel angle information in vehicle movement model, more accurate direction of motion information is obtained, further
Improve the accuracy of motion model dead reckoning.
The technical scheme to solve the above technical problems is that
A method of the vehicle movement model optimization based on deep learning, comprising the following steps:
According to the inertia mass of vehicle, speed, mass center drift angle, yaw velocity, the yaw rotation inertia of vehicle, vehicle matter
The heart determines vehicle basic exercise equation to the distance of front and back axle, the cornering stiffness of front and back wheel and front vehicle wheel drift angle；
Influence according to the characteristic of steering system to dynamics of vehicle performance, by the steering system equation of motion and vehicle basic exercise
Equations simultaneousness obtains steering system about the relevant vehicle movement equation of steering wheel angle and mass center drift angle；
With the nonlinear function of method the fitting steering wheel angle and mass center drift angle of deep learning, the model of deep learning
Input is steering wheel angle, speed, centripetal acceleration, and output is mass center drift angle, uses mean square deviation as measurement index.
Further, it is contemplated that the retardance of steering wheel angle opposing front wheels steering angle, the deep learning model selection band
The network structure of time series characteristic.
Further, the deep learning model includes input layer, hidden layer, accepts layer and output layer；The undertaking layer is used
Carry out storage state, there is system and adapt to time-varying characteristics, to be adapted to the prediction model of settling time sequence.
Further, the training set of the deep learning model includes the height of acquisition under different operating conditions and scene condition
Precision inertial guidance data and vehicle body signal data, and turned according to what the High Accuracy Inertial data and the analysis of vehicle body signal data obtained
To disk corner, speed, centripetal acceleration.
Further, the training method of the deep learning model includes:
Each Log data content is divided into training sample set and test sample collection,
It is concentrated in training sample, extracting adjacent 50 data is a sample, wherein first 49 are set as independent variable X, latter 1
For target value y,
And so on, finally it is divided into N number of training sample, forms training matrix；
Similarly, test matrix is formed.
The beneficial effects of the present invention are:
1, many because being known as of β mass center drift angle are influenced, simple static calibration can introduce large error, and the present invention is with vehicle
Based on steering maneuver theory, β mass center drift angle is factor-related with which when studying vehicle movement, sorts out associated non-thread
Property function formula.
2, data are dynamically acquired, comprising different operating conditions, different scenes, by mass data training, fit with
The smallest nonlinear function of real vehicles kinematic error.
3, static demarcating is discrete, and the nonlinear function that the present invention solves is continuous.
Detailed description of the invention
Fig. 1 is simple vehicle movement model schematic；
Fig. 2 is real vehicles kinematics model schematic diagram；
Fig. 3 is the lateral deviation power motion model schematic diagram of car plane movement and tire；
Fig. 4 is steering system equivalent model schematic diagram；
Fig. 5 is the aligning torque schematic diagram for acting on front-wheel；
Fig. 6 is deep learning model；
Fig. 7 is that deep learning model calculates step schematic diagram.
Specific embodiment
The principle and features of the present invention will be described below with reference to the accompanying drawings, and the given examples are served only to explain the present invention, and
It is non-to be used to limit the scope of the invention.
The present invention provides a kind of method of vehicle movement model optimization based on deep learning, including the following contents:
1, vehicle basic exercise equation is established.The sport dynamics that the steering given corresponding to some inputs lower vehicle are special
Property, vehicle basic exercise equation when drawing any given front wheel slip angle δ by mechanical equation, it is not fixed by it with respect to ground
The influence of the position and its course of coordinate system.
2, steering system model and the equation of motion are established.Influence of the characteristic of steering system to dynamics of vehicle performance is derived, it will
The steering system equation of motion and first step vehicle basic exercise equations simultaneousness, it is inclined about steering wheel angle α and β mass center can to obtain steering system
The relevant vehicle movement equation in angle.Prove the non-linear relation between steering wheel angle α and β mass center drift angle by speed ν and sideway
Angular speed γ influences, because yaw velocity γ is relevant with to cadion-acceleration η, so the relationship of steering wheel angle α and β mass center drift angle
Also it is influenced from speed ν and to cadion-acceleration η.
3, with the nonlinear function of the method fitting steering wheel angle α and β mass center drift angle of deep learning.In different works
Under condition and scene condition, High Accuracy Inertial data and vehicle body signal data are acquired, the mode input of deep learning is that steering wheel turns
Angle, speed, centripetal acceleration, output are mass center drift angles, use mean square deviation as measurement index, level off to 0 or the smaller the better, this
The nonlinear function of sample fitting is closer to real motion result.
Specifically, this method specifically includes:
Firstly, establishing vehicle basic exercise equation
Fig. 3 is the lateral deviation power motion model schematic diagram of car plane movement and tire.It can be able to lower mechanics mould as shown in Figure 3
Type:
Wherein, m is the inertia mass of vehicle, and v is speed, and β is mass center drift angle, and γ is yaw velocity, CfAnd CrFor front and back
The lateral deviation power of wheel, I are the yaw rotation inertia of vehicle, lfAnd lrFor vehicle centroid to the distance of front and back axle, KfAnd KrFor front and back
The cornering stiffness of wheel, δ are front vehicle wheel drift angle.
Formula (1.3) and formula (1.4) are substituted into formula (1.1) and formula (1.2) respectively, can be obtained:
Above equation just becomes the basic exercise equation of description vehicle.
Secondly, establishing steering system model and the equation of motion
Fig. 4 is steering system equivalent model schematic diagram, and Fig. 5 is the aligning torque schematic diagram for acting on front-wheel.
As shown in Fig. 4 Fig. 5, following mechanical equation can be obtained:
Wherein, IsTo turn to inertia, KsFor coefficient of elasticity, CsFor damped coefficient, α is steering wheel angle, and δ is that front vehicle wheel is inclined
Angle, KfFor front tyre cornering stiffness, βfThe side drift angle that front-wheel generates, ξ are the sum of pneumatic trail and positive caster offset.
Formula (2.1) and formula (2.2) are substituted into formula (2.3) respectively, can be obtained:
For known α, front wheel steering angle δ can be obtained by formula (2.4), be influenced for research steering system characteristic, ignored and turned
The process rotated rapidly to disk, that is to say, that at this timeWithVery little, and IsAnd CsAlso very little, in formula (2.4)WithIt is negligible.By formula (2.4), formula (1.5) and formula (1.6) simultaneous, just obtained and meanwhile consider steering system about
The vehicle movement equation group of steering wheel angle:
Non-linear relation between provable steering wheel angle α and β mass center drift angle is by speed v and yaw velocity γ shadow
It rings, because yaw velocity γ is relevant with to cadion-acceleration η, so the relationship of steering wheel angle α and β mass center drift angle is also by speed v
It is influenced with yaw velocity γ, to cadion-acceleration η.
Third, it was demonstrated that the opposite operating lag with steering wheel angle α of front wheel steering angle δ
Formula (2.4), formula (1.5) and formula (1.6) are slightly deformed, the vehicle accordingly with steering wheel angle α can be obtained
The equation of motion is as follows:
Strictly, when steering wheel is quickly operated, in formula (2.11)WithIt cannot be ignored, according to
These equations it is contemplated that, front wheel steering angle δ can be made opposite and steering wheel angle α around the front-wheel Inertia and damping term of stub
Operating lag needs to consider the importance of time series when selection is using neural network fit non-linear function.
4th, with the nonlinear function of the method fitting steering wheel angle α and β mass center drift angle of deep learning
Under different operating conditions and scene condition, High Accuracy Inertial data and vehicle body signal data are acquired, deep learning
Mode input is steering wheel angle, speed, centripetal acceleration, and output is mass center drift angle, uses mean square deviation as measurement index, examines
Consider the retardance of steering wheel angle α opposing front wheels steering angle sigma, select the network structure classics Elman with time series characteristic,
It is illustrated in fig. 6 shown below.
Network structure classics Elman with time series characteristic is a kind of local regression network, its primary structure is by defeated
Enter layer, hidden layer, accept layer, output layer composition, there is stronger computing capability and function of associate memory, it is in inside by holding
It connects layer and carrys out storage state, there is system and adapt to time-varying characteristics, to be adapted to the prediction model of settling time sequence.This is pre-
The calculating step for surveying model is as shown in Figure 7.
The mode of model training is that each Log data content is divided into training sample set and test sample collection, in training sample
This concentration, extracting adjacent 50 data is a sample, wherein first 49 are set as independent variable X, latter 1 is target value y, successively
Analogize, be finally divided into N number of training sample, forms training matrix.Same test set forms test matrix also referring to such division.
The foregoing is merely presently preferred embodiments of the present invention, is not intended to limit the invention, it is all in spirit of the invention and
Within principle, any modification, equivalent replacement, improvement and so on be should all be included in the protection scope of the present invention.
Claims (5)
1. a kind of method of the vehicle movement model optimization based on deep learning, which comprises the following steps:
Extremely according to the inertia mass of vehicle, speed, mass center drift angle, yaw velocity, the yaw rotation inertia of vehicle, vehicle centroid
The distance of front and back axle, the cornering stiffness of front and back wheel and front vehicle wheel drift angle determine vehicle basic exercise equation；
Influence according to the characteristic of steering system to dynamics of vehicle performance, by the steering system equation of motion and vehicle basic exercise equation
Simultaneous obtains steering system about the relevant vehicle movement equation of steering wheel angle and mass center drift angle；
With the nonlinear function of method the fitting steering wheel angle and mass center drift angle of deep learning, the mode input of deep learning
It is steering wheel angle, speed, centripetal acceleration, output is mass center drift angle, uses mean square deviation as measurement index.
2. a kind of method of the vehicle movement model optimization based on deep learning according to claim 1, which is characterized in that examine
Consider the retardance of steering wheel angle opposing front wheels steering angle, the network of the deep learning model selection with time series characteristic
Structure.
3. a kind of method of the vehicle movement model optimization based on deep learning according to claim 2, which is characterized in that institute
Deep learning model is stated to include input layer, hidden layer, accept layer and output layer；The undertaking layer is used to storage state, makes system
With time-varying characteristics are adapted to, to be adapted to the prediction model of settling time sequence.
4. a kind of method of the vehicle movement model optimization based on deep learning according to claim 3, which is characterized in that institute
The training set for stating deep learning model includes the High Accuracy Inertial data and vehicle body of acquisition under different operating conditions and scene condition
Signal data, and the steering wheel angle, speed, centripetal obtained according to the High Accuracy Inertial data and the analysis of vehicle body signal data
Acceleration.
5. a kind of method of the vehicle movement model optimization based on deep learning according to claim 4, which is characterized in that institute
The training method for stating deep learning model includes:
Each Log data content is divided into training sample set and test sample collection,
It is concentrated in training sample, extracting adjacent 50 data is a sample, wherein first 49 are set as independent variable X, latter 1 is mesh
Numerical value y is marked,
And so on, finally it is divided into N number of training sample, forms training matrix；
Similarly, test matrix is formed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810871827.8A CN109190171B (en) | 2018-08-02 | 2018-08-02 | Vehicle motion model optimization method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810871827.8A CN109190171B (en) | 2018-08-02 | 2018-08-02 | Vehicle motion model optimization method based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109190171A true CN109190171A (en) | 2019-01-11 |
| CN109190171B CN109190171B (en) | 2022-06-17 |
Family
ID=64920589
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810871827.8A Active CN109190171B (en) | 2018-08-02 | 2018-08-02 | Vehicle motion model optimization method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109190171B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110378201A (en) * | 2019-06-05 | 2019-10-25 | 浙江零跑科技有限公司 | A kind of hinged angle measuring method of multiple row vehicle based on side ring view fisheye camera input |
| CN111090959A (en) * | 2019-12-31 | 2020-05-01 | 湖南行必达网联科技有限公司 | Vehicle load spectrum acquisition method and system |
| CN111144383A (en) * | 2020-01-15 | 2020-05-12 | 河南理工大学 | Method for detecting vehicle deflection angle |
| CN114394130A (en) * | 2021-12-27 | 2022-04-26 | 中国矿业大学 | Coal mine auxiliary transport vehicle positioning method and positioning system |
| WO2022193940A1 (en) * | 2021-03-18 | 2022-09-22 | 北京航迹科技有限公司 | Vehicle speed measurement method and apparatus, vehicle-mounted computer device and storage medium |
| RU2815190C1 (en) * | 2023-04-27 | 2024-03-12 | Общество с ограниченной ответственностью "ЭвоКарго" | Method for setting parameters of dynamic model of wheeled vehicle |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106218633A (en) * | 2016-08-02 | 2016-12-14 | 大连理工大学 | Four motorized wheels electric automobile stability control method based on Q study |
| CN106649983A (en) * | 2016-11-09 | 2017-05-10 | 吉林大学 | Vehicle dynamics model modeling method used in unmanned vehicle high-velocity motion planning |
| US20180113204A1 (en) * | 2017-12-21 | 2018-04-26 | GM Global Technology Operations LLC | Systems and methods for aligning an inertial measurement unit in a vehicle |
| CN108082182A (en) * | 2017-12-18 | 2018-05-29 | 长春工业大学 | A kind of active rear steer control method for widening vehicle steadily domain |
-
2018
- 2018-08-02 CN CN201810871827.8A patent/CN109190171B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106218633A (en) * | 2016-08-02 | 2016-12-14 | 大连理工大学 | Four motorized wheels electric automobile stability control method based on Q study |
| CN106649983A (en) * | 2016-11-09 | 2017-05-10 | 吉林大学 | Vehicle dynamics model modeling method used in unmanned vehicle high-velocity motion planning |
| CN108082182A (en) * | 2017-12-18 | 2018-05-29 | 长春工业大学 | A kind of active rear steer control method for widening vehicle steadily domain |
| US20180113204A1 (en) * | 2017-12-21 | 2018-04-26 | GM Global Technology Operations LLC | Systems and methods for aligning an inertial measurement unit in a vehicle |
Non-Patent Citations (1)
| Title |
|---|
| 曹初等: "车辆质心侧偏角的估计与仿真分析", 《科技经济导刊》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110378201A (en) * | 2019-06-05 | 2019-10-25 | 浙江零跑科技有限公司 | A kind of hinged angle measuring method of multiple row vehicle based on side ring view fisheye camera input |
| CN111090959A (en) * | 2019-12-31 | 2020-05-01 | 湖南行必达网联科技有限公司 | Vehicle load spectrum acquisition method and system |
| CN111090959B (en) * | 2019-12-31 | 2023-06-20 | 湖南行必达网联科技有限公司 | Vehicle load spectrum acquisition method and system |
| CN111144383A (en) * | 2020-01-15 | 2020-05-12 | 河南理工大学 | Method for detecting vehicle deflection angle |
| CN111144383B (en) * | 2020-01-15 | 2023-03-28 | 河南理工大学 | Method for detecting vehicle deflection angle |
| WO2022193940A1 (en) * | 2021-03-18 | 2022-09-22 | 北京航迹科技有限公司 | Vehicle speed measurement method and apparatus, vehicle-mounted computer device and storage medium |
| CN114394130A (en) * | 2021-12-27 | 2022-04-26 | 中国矿业大学 | Coal mine auxiliary transport vehicle positioning method and positioning system |
| CN114394130B (en) * | 2021-12-27 | 2022-11-11 | 中国矿业大学 | Coal mine auxiliary transport vehicle positioning method and positioning system |
| RU2815190C1 (en) * | 2023-04-27 | 2024-03-12 | Общество с ограниченной ответственностью "ЭвоКарго" | Method for setting parameters of dynamic model of wheeled vehicle |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109190171B (en) | 2022-06-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109190171A (en) | A method of the vehicle movement model optimization based on deep learning | |
| CN108482379B (en) | Wheel-hub motor driven vehicle coefficient of road adhesion and the synchronous real-time estimation system and method for road gradient | |
| CN108594652B (en) | Observer information iteration-based vehicle state fusion estimation method | |
| CN111780981B (en) | Intelligent vehicle formation lane change performance evaluation method | |
| CN109597308A (en) | Pilotless automobile model predictive controller design method based on kinetic model | |
| Ni et al. | Dynamics control of autonomous vehicle at driving limits and experiment on an autonomous formula racing car | |
| CN111845774A (en) | Automatic driving automobile dynamic trajectory planning and tracking method based on transverse and longitudinal coordination | |
| CN105946863B (en) | A kind of determining method in vehicle run stability region | |
| CN114379583B (en) | Automatic driving vehicle track tracking system and method based on neural network dynamics model | |
| CN110497915B (en) | Automobile driving state estimation method based on weighted fusion algorithm | |
| CN110341714B (en) | Method for simultaneously estimating vehicle mass center slip angle and disturbance | |
| CN115406446A (en) | Multi-axis special vehicle state estimation method based on neural network and unscented Kalman filtering | |
| Kegelman | Learning from professional race car drivers to make automated vehicles safer | |
| CN112270039A (en) | Distributed asynchronous fusion-based nonlinear state estimation method for drive-by-wire chassis vehicle | |
| Reina et al. | Odometry correction using visual slip angle estimation for planetary exploration rovers | |
| CN109033017B (en) | Vehicle roll angle and pitch angle estimation method under packet loss environment | |
| CN108394413A (en) | A kind of electronic vehicle attitude and parameter correcting method of four motorized wheels and steering | |
| CN112287289A (en) | Vehicle nonlinear state fusion estimation method for cloud control intelligent chassis | |
| Zhenhai | Soft sensor application in vehicle yaw rate measurement based on Kalman filter and vehicle dynamics | |
| Junqueira et al. | A Model-less Approach for Estimating Vehicles Sideslip Angle by a Neural Network Concept | |
| Coyte et al. | Decision tree assisted EKF for vehicle slip angle estimation using inertial motion sensors | |
| CN108413923B (en) | Vehicle roll angle and pitch angle estimation method based on robust hybrid filtering | |
| CN115809590A (en) | Estimation method for main dynamic parameters of vehicle | |
| Kuyt et al. | Mixed kinematics and camera based vehicle dynamic sideslip estimation for an rc scaled model | |
| CN115422837A (en) | Intelligent automobile dynamics prediction model based on deep Gaussian process, training data acquisition method and training method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |