CN111723470B - Automatic driving control method based on calibration optimization RSS model - Google Patents

Automatic driving control method based on calibration optimization RSS model Download PDF

Info

Publication number
CN111723470B
CN111723470B CN202010456689.4A CN202010456689A CN111723470B CN 111723470 B CN111723470 B CN 111723470B CN 202010456689 A CN202010456689 A CN 202010456689A CN 111723470 B CN111723470 B CN 111723470B
Authority
CN
China
Prior art keywords
following
rss
model
dangerous
automatic driving
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.)
Active
Application number
CN202010456689.4A
Other languages
Chinese (zh)
Other versions
CN111723470A (en
Inventor
徐晓妍
王雪松
高岩
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tongji University
Original Assignee
Tongji University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Tongji University filed Critical Tongji University
Priority to CN202010456689.4A priority Critical patent/CN111723470B/en
Publication of CN111723470A publication Critical patent/CN111723470A/en
Application granted granted Critical
Publication of CN111723470B publication Critical patent/CN111723470B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/01—Detecting movement of traffic to be counted or controlled
    • G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
    • G08G1/0125—Traffic data processing
    • G08G1/0129—Traffic data processing for creating historical data or processing based on historical data
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/16—Anti-collision systems
    • G08G1/166—Anti-collision systems for active traffic, e.g. moving vehicles, pedestrians, bikes
    • 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)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Computer Hardware Design (AREA)
  • Geometry (AREA)
  • General Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Medical Informatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Artificial Intelligence (AREA)
  • Chemical & Material Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Traffic Control Systems (AREA)

Abstract

The invention relates to an automatic driving control method based on a calibration optimization RSS model, which is used for calibrating parameters of an existing responsibility sensitive safety model, and the method is a program embedded in a computer and comprises the following steps: step 1: screening dangerous events under the following scenes from the natural driving database to obtain dangerous event data under the following scenes; step 2: using the dangerous event data obtained in the step 1 to reappear dangerous events in a simulation platform to obtain rear vehicle driving track data; step 3: establishing an objective function and constructing a multi-objective optimization problem; step 4: solving an objective function to obtain an objective function solution set; step 5: and obtaining an optimal solution from the objective function solution set to serve as a final calibration result of the RSS model. Compared with the prior art, the method has the advantages of low complexity, comprehensive consideration of the safety and conservation degree of the model, high safety, practical applicability and the like.

Description

Automatic driving control method based on calibration optimization RSS model
Technical Field
The invention relates to the technical field of automatic driving decision methods, in particular to an automatic driving control method based on a calibration optimization RSS model.
Background
To ensure the Safety and reliability of an autonomous vehicle, a strict mathematical model was established, and in 2017, a Responsibility-Sensitive Safety (RSS) model has been proposed by mobile researchers, which has become the core of the automated vehicle model decision standard IEEE 2846 formulated by the Institute of Electrical and Electronics Engineers (IEEE). The RSS model calculates the safety distance between the AV and other traffic participants in real time, and helps the AV to reasonably deal with emergency scenes. The safe distance is calculated by assuming a "worst case scenario", i.e. as long as the safe distance is obeyed, the AV does not cause an accident without taking the maximum deceleration when the surrounding vehicle takes extreme actions, such as sudden deceleration.
The RSS model has the following advantages: (1) RSS is a strict mathematical model covering three main tasks of AV: sensing, planning and decision making; (2) The "worst scenario" assumption eliminates the necessity of predicting the intent of other traffic participants; (3) According to whether the traffic participation main body makes proper risk avoidance measures when the safety distance is not met, the RSS model can help judge the responsibility attribution of the accident; (4) The RSS model can fully cover various driving scenes such as following, changing lanes, overtaking and the like. The related research of the existing RSS model has the following defects: (1) Default parameters are used, and the conservation degree of the model is large. (2) The scenes used for the simulation test are all set manually, and support of actual driving behavior data is lacked.
At present, research on reducing the conservation degree of an RSS model is also continuously carried out, for example, china patent CN110928319A discloses a control method, a device, a vehicle and a storage medium for an automatic driving vehicle, wherein the control method in the patent controls the vehicle by adding a calculation process of a safety risk parameter in the RSS model and a method with the minimum safety coefficient, and the method lightens the conservation degree of the RSS model, increases the complexity of an algorithm and prolongs the processing time of the algorithm.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide the automatic driving control method based on the calibration optimization RSS model, which has low complexity, high safety and practical applicability by comprehensively considering the safety and conservation degree of the model.
The aim of the invention can be achieved by the following technical scheme:
an automatic driving control method based on a calibration optimization RSS model, which is a program embedded in a computer and comprises the following steps:
step 1: screening dangerous events under the following scenes from the natural driving database to obtain dangerous event data under the following scenes;
step 2: using the dangerous event data obtained in the step 1 to reappear dangerous events in a simulation platform to obtain rear vehicle driving track data;
step 3: establishing an objective function and constructing a multi-objective optimization problem;
step 4: solving an objective function to obtain an objective function solution set;
step 5: obtaining an optimal solution from the objective function solution set as a final calibration result of the RSS model;
step 6: and performing automatic driving control on the automatic driving vehicle by using the RSS model subjected to calibration optimization.
Preferably, the step 1 utilizes a threshold method to screen dangerous events under the following scenes from a natural driving database.
Preferably, the step 2 specifically includes:
and (3) reproducing the dangerous event in the following scene screened in the step (1) in a simulation platform, wherein the initial speed and initial position of the rear vehicle, the track of the front vehicle, the central line of the road and the width of the road are all real data of the dangerous event obtained in the step (1), setting the rear vehicle as an automatic driving vehicle AV controlled by the adaptive cruise ACC and the responsibility sensitive safety model RSS in the simulation platform, and obtaining driving track data of the rear vehicle through simulation.
Preferably, the step 3 specifically includes:
the safety performance and conservation degree of the model are synthesized, and a multi-objective function is established:
Ω＝(max f Safety ,min f Conservativeness )
wherein max f Safety For the first sub-target, min f Conservativeness Is the second sub-target.
More preferably, the first sub-target max f Safety The method comprises the following steps:
Figure BDA0002509618460000031
Figure BDA0002509618460000032
Figure BDA0002509618460000033
wherein N is the dangerous following event number extracted from natural driving data; TIT (tungsten inert gas) i A TIT value for the ith dangerous following event; TTC (TTC) i (t) is the ith dangerous heelTTC value of the vehicle event at time t; τ SC Is a simulation time step; t C * Is the threshold for TTC.
More preferably, the second sub-target min f Conservativeness The method comprises the following steps:
Figure BDA0002509618460000034
Figure BDA0002509618460000035
wherein ,Δdi For measuring the degree of conservation of event i;
Figure BDA0002509618460000036
represents the RSS safe distance d min Is greater than the actual following distance d relative Is equal to d relative In comparison, d min The larger the RSS model, the earlier the trigger time will be, forcing the AV to minimum comfort deceleration a min,brake The speed is reduced, and the conservation degree of the model is higher.
Preferably, the step 4 specifically includes:
the non-dominant ordered genetic algorithm NSGA-II with elite strategy is used for solving the multi-objective optimization problem established in the step 3.
More preferably, the NSGA-II algorithm comprises the following steps:
step 4-1: randomly generating an initial population, wherein the individuals represent a random parameter set of the RSS model;
step 4-2: mutation and crossover are carried out on randomly selected individuals to generate subgroups;
step 4-3: combining the parent and the offspring;
step 4-4: layering the individuals according to the dominant relationship by using a rapid non-dominant ranking algorithm;
step 4-5: calculating the crowding degree i of an individual layer by layer d Sorting the individuals according to the degree of congestion,selecting proper individuals to form a new parent;
step 4-6: repeating the steps 2-5 until the average relative weight change of the fitness function among the generations is smaller than a preset tolerance or the iteration number reaches the maximum value, and stopping iteration.
More preferably, the congestion degree i in the step 4-5 d The calculation method of (1) is as follows:
Figure BDA0002509618460000037
wherein ,
Figure BDA0002509618460000038
a j-th objective function value representing point i+1; />
Figure BDA0002509618460000039
The j-th objective function value at the i-1 th point is represented.
Preferably, the step 5 specifically includes:
selecting a first sub-target max f from the target function solution set obtained in the step 4 Safety The set of data with the largest value is used as the final calibration result of the RSS model.
Compared with the prior art, the invention has the following advantages:
1. the complexity is low, and the safety and conservation degree of the model are comprehensively considered: according to the automatic driving control method, the RSS model is calibrated, the algorithm complexity is low, the safety and conservation degree of the model are comprehensively considered, a multi-objective function is built, the best safety performance of the model is ensured, the conservation degree is considered, and road resources can be effectively utilized.
2. The safety is high: according to the automatic driving control method, on the basis of comprehensively considering the safety and conservation degree of the model, the solution with the highest safety performance is selected from the solution set to serve as the parameter calibration value of the RSS model, and the safety performance is not affected.
3. The method has practical applicability: according to the automatic driving control method, when the RSS model is calibrated, a simulation scene is constructed based on actual driving behavior data, and the simulation is performed by using real events, so that the calibrated model has more practical applicability.
Drawings
FIG. 1 is a flow chart of an algorithm of the present invention;
FIG. 2 is a flow chart of the NSGA-II algorithm of the present invention;
FIG. 3 is a schematic diagram of an objective function solution set obtained in an embodiment of the present invention;
FIG. 4 is a TIT distribution diagram obtained by simulation of a rear vehicle using a human driver model in an embodiment of the present invention;
FIG. 5 is a TIT distribution diagram obtained by simulating a rear vehicle using an autonomous vehicle without an RSS model ACC in an embodiment of the present invention;
FIG. 6 is a TIT distribution diagram obtained by simulating a rear vehicle using an autonomous vehicle embedded with a default RSS model ACC in an embodiment of the present invention;
FIG. 7 is a TIT distribution diagram obtained by simulating a rear vehicle using an autonomous vehicle embedded with a calibration RSS model ACC in an embodiment of the present invention;
fig. 8 is a schematic diagram of obtaining a relative distance between a rear vehicle and a front vehicle when four rear vehicle models used in the embodiment of the present invention are simulated.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are some, but not all embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, shall fall within the scope of the present invention.
An automatic driving control method based on a calibration optimization RSS model, the flow of which is shown in figure 1, comprises the following steps:
step 1: the dangerous event under the following scene is screened from the Shanghai natural driving experimental data by using a threshold method, and the threshold is set for the vehicle kinematic parameters, wherein the threshold is specifically set in the embodiment as follows: longitudinal acceleration is less than-0.45 g and collision time is less than 4s;
in order to avoid false alarm of a threshold method, manual verification is further arranged in the embodiment to determine the effectiveness of the event;
step 2: using the dangerous event data obtained in the step 1, reproducing a dangerous event in a simulation platform, and setting a rear vehicle as an automatic driving vehicle commonly controlled by ACC and RSS;
the method specifically comprises the following steps: the dangerous event under the following scene screened in the step 1 is reproduced in a MATLAB Simulink simulation platform, the initial speed and the initial position of a rear vehicle, the track of a front vehicle, the central line of a road and the width of the road are all real data of the dangerous event obtained in the step 1, and the rear vehicle is set as an automatic driving vehicle AV controlled by the adaptive cruise ACC and the responsibility sensitive safety model RSS in the simulation platform; and obtaining the motion trail data of the rear vehicle through the simulation.
The running principle of the RSS model is to calculate in real time the safe distance that the AV needs to keep with the preceding vehicle, and if the actual following distance does not meet the safe distance, the AV decelerates with a predefined minimum deceleration, i.e. a comfortable deceleration. The core idea of calculating the safe distance is: suppose that the preceding vehicle suddenly assumes the maximum deceleration a at the present moment max,brake AV is at maximum acceleration a within the reaction time ρ max,accel Acceleration is carried out with minimum comfortable deceleration a after reaction time min,brake The minimum safe distance for deceleration, which does not collide in the above case, is RSS safe distance d min The calculation formula is as follows:
Figure BDA0002509618460000051
wherein ,[x]+ ＝max{x,0}；v a and vl The speeds of the AV and the current moment of the front vehicle are respectively; ρ, a max,accel 、a min,brake and amax,brake Is four parameters to be calibrated.
Step 3: and (3) integrating the safety performance and conservation degree of the model, establishing an objective function, and constructing a multi-objective optimization problem. Where security performance is measured by the TTC's collective indicator TIT, the goal is that AV produces the minimum total TIT in all dangerous following events. The conservation degree of the model is taken into consideration to avoid that the algorithm generates an excessive following distance for obtaining a safer result, so that the utilization rate of road resources is too low.
The multi-objective function is specifically:
Ω＝(max f Safety ,min f Conservativeness )
wherein max f Safety For the first sub-target, min f Conservativeness Is the second sub-target.
First sub-target max f Safety The method comprises the following steps:
Figure BDA0002509618460000061
Figure BDA0002509618460000062
Figure BDA0002509618460000063
wherein N is the dangerous following event number extracted from natural driving data; TIT (tungsten inert gas) i The TIT value of the ith dangerous following event is a measure index of safety performance, and the accumulated TTC less than a certain threshold value in a period of time is measured, wherein the unit is s 2 ；TTC i (t) is the TTC value of the ith dangerous following event at the moment t, specifically the ratio of the relative distance between the AV at the moment t and the front vehicle to the relative speed; τ SC Is a simulation time step; t C * A threshold value for TTC;
second sub-target min f Conservativeness The method comprises the following steps:
Figure BDA0002509618460000064
Figure BDA0002509618460000065
wherein ,Δdi For measuring the degree of conservation of event i;
Figure BDA0002509618460000066
represents the RSS safe distance d min Is greater than the actual following distance d relative Is equal to d relative In comparison, d min The larger the RSS model, the earlier the trigger time will be, forcing the AV to minimum comfort deceleration a min,brake Speed reduction is carried out, so that the conservation degree of the model is higher; Δd i Metric d min Exceeding d relative To the extent that d is not present min Greater than d relative And Δd i The value is 0.
Step 4: solving an objective function by using a non-dominant ranking genetic algorithm NSGA-II with elite strategy, wherein the flow is shown in figure 2, and the specific steps comprise:
step 4-1: randomly generating an initial population, wherein the individuals represent a random parameter set of the RSS model;
step 4-2: mutation and crossover are carried out on randomly selected individuals to generate subgroups;
step 4-3: combining the parent and the offspring;
step 4-4: layering the individuals according to the dominant relationship by using a rapid non-dominant ranking algorithm;
the dominance relationship refers to: n target components f for multi-target minimization problem i (i=1, 2,) n
Figure BDA0002509618460000067
Arbitrary given two decision variables->
Figure BDA0002509618460000068
If and only if, for->
Figure BDA0002509618460000069
All have->
Figure BDA00025096184600000610
Then->
Figure BDA00025096184600000611
Innervating->
Figure BDA00025096184600000612
The specific steps of the rapid dominant ranking algorithm are: assuming the population is P, the algorithm needs to calculate two parameters n for each individual P in P p and Sp, wherein np For the number of individuals in the population that dominate p, S p Is the set of individuals in the population that are dominated by individuals p.
(1) Find all n in the population p Individual=0 and is stored in the current set F 1 In (a) and (b);
(2) For the current set F 1 Each individual i of the group of individuals is S i Traversing S i Each individual l of (1), execute n l ＝n l -1, if n l Individual l is saved in set H, =0;
(3) Record F 1 The obtained individuals are individuals of the first non-dominant layer, H is used as a current set, and the operation is repeated until the whole population is classified;
step 4-5: calculating the crowding degree i of an individual layer by layer d Sorting individuals according to the degree of congestion, and selecting proper individuals to form a new parent;
the method for calculating the crowding degree comprises the following steps:
Figure BDA0002509618460000071
wherein ,
Figure BDA0002509618460000072
a j-th objective function value representing point i+1; />
Figure BDA0002509618460000073
A j-th objective function value representing the i-1 th point;
step 4-6: repeating the steps 2-5 until the termination condition is met, namely that the average relative weight change of the fitness function among generations is smaller than a preset tolerance or the iteration number reaches the maximum value; the average relative weight variation preset margin set in this embodiment is 10 -6 The iteration number reaches 300;
step 5: selecting a first sub-target max f from a solution set of target functions Safety The set of data with the maximum value is used as a final calibration result of the RSS model, and the calibrated RSS model is obtained;
step 6: and performing automatic driving control on the automatic driving vehicle by using the RSS model subjected to calibration optimization.
A specific example is provided below:
default parameters for uncalibrated RSS model are shown in table 1.
Table 1 default parameter values for RSS model
Parameters (Unit) Value taking
ρ(s) 1
a max,accel (m/s 2 ) 2
a min,brake (m/s 2 ) -3
a max,brake (m/s 2 ) -8
Based on Shanghai natural driving experimental data, extracting 223 dangerous following events, solving a multi-objective optimization problem by using an NSGA-II algorithm to obtain a solution set containing 105 groups of parameter combinations, selecting a solution with the largest first sub-target value from the solution set generated by the NSGA-II as a final calibration result of an RSS model, wherein the corresponding parameter values and objective function values are shown in a table 2.
Table 2 RSS calibration parameter values
Figure BDA0002509618460000081
The calibrated RSS safe distance expression is as follows:
Figure BDA0002509618460000082
in this embodiment, three rear vehicle models are selected for comparison test, and the selected rear vehicle models are respectively:
(1) A human driver rear vehicle model;
(2) An autonomous vehicle AV embedded with an auto-cruise system ACC;
(3) An autonomous vehicle AV embedded with an auto-cruise system ACC in which an RSS model using default parameters is embedded;
(4) An autonomous vehicle AV with an autonomous cruise system ACC embedded with a calibrated RSS model.
The three rear vehicle models are used for simulation, and a TIT distribution map is obtained respectively, as shown in figures 4 to 7. Analysis finds that the TIT value of the automatic driving vehicle AV embedded with the calibrated RSS model is concentrated in [0,5 ]]s 2 Compared with human drivers and ACCs not embedded with the RSS model, the TIT value of the automatic driving vehicle AV embedded with the RSS model is obviously reduced.The TIT average value of the three models is 4.37s respectively 2 、4.93s 2 、2.23s 2 and 2.65s2 。
The relative distances between the rear vehicle and the front vehicle obtained when the four rear vehicle models are simulated are shown in FIG. 8, although the average TIT value of the embedded default RSS model AV is 2.23s 2 The average TIT value of the embedded calibrated RSS model AV is 2.65s 2 The two data are relatively close, but as can be seen from analysis of fig. 8, the average relative distance between the AV embedded with the default RSS model and the front vehicle is 20.86m, the average relative distance between the AV embedded with the calibrated RSS model and the front vehicle is 12.71m, the average following distance between the AV embedded with the calibrated RSS model and the front vehicle is 7.43m, so that the RSS model using the default parameters is too conservative, the following distance between the RSS model and the front vehicle is too large, and the waste of road resources is caused, and the RSS model is not suitable for the actual driving environment.
While the invention has been described with reference to certain preferred embodiments, it will be understood by those skilled in the art that various changes and substitutions of equivalents may be made and equivalents will be apparent to those skilled in the art without departing from the scope of the invention. Therefore, the protection scope of the invention is subject to the protection scope of the claims.

Claims (6)

1. An automatic driving control method based on a calibration optimization RSS model, which is a program embedded in a computer, is characterized by comprising the following steps:
step 1: screening dangerous events under the following scenes from the natural driving database to obtain dangerous event data under the following scenes;
step 2: using the dangerous event data obtained in the step 1 to reappear dangerous events in a simulation platform to obtain rear vehicle driving track data;
the step 2 specifically comprises the following steps:
the dangerous event under the following scene screened in the step 1 is reproduced in a simulation platform, the initial speed and initial position of the rear vehicle, the track of the front vehicle, the central line of the road and the width of the road are all real data of the dangerous event obtained in the step 1, the rear vehicle is set as an automatic driving vehicle AV controlled by the adaptive cruise ACC and the responsibility sensitive safety model RSS in the simulation platform, and the driving track data of the rear vehicle is obtained through simulation;
step 3: establishing an objective function and constructing a multi-objective optimization problem;
step 4: solving an objective function to obtain an objective function solution set;
step 5: obtaining an optimal solution from the objective function solution set as a final calibration result of the RSS model;
step 6: performing automatic driving control on the automatic driving vehicle by using the RSS model subjected to calibration optimization;
the step 3 specifically comprises the following steps:
the safety performance and conservation degree of the model are synthesized, and a multi-objective function is established:
Ω＝(max f Safety ,min f Conservativeness )
wherein max f Safety For the first sub-target, min f Conservativeness Is the second sub-target;
the first sub-target max f Safety The method comprises the following steps:
Figure FDA0003975019810000011
Figure FDA0003975019810000012
Figure FDA0003975019810000013
wherein N is the dangerous following event number extracted from natural driving data; TIT (tungsten inert gas) i A TIT value for the ith dangerous following event; TTC (TTC) i (t) is the TTC value of the ith dangerous following event at time t; τ SC Is a simulation time step; t C * A threshold value for TTC;
the second sub-target min f Conservativeness The method comprises the following steps:
Figure FDA0003975019810000021
Figure FDA0003975019810000022
wherein ,Δdi For measuring the degree of conservation of event i;
Figure FDA0003975019810000023
represents the RSS safe distance d min Is greater than the actual following distance d relative Is equal to d relative In comparison, d min The larger the RSS model, the earlier the trigger time will be, forcing the AV to minimum comfort deceleration a min,brake The speed is reduced, and the conservation degree of the model is higher.
2. The automatic driving control method based on the calibration optimization RSS model according to claim 1, wherein the step 1 uses a thresholding method to screen dangerous events in a following scene from a natural driving database.
3. The automatic driving control method based on the calibration optimization RSS model according to claim 1, wherein the step 4 specifically includes:
the non-dominant ordered genetic algorithm NSGA-II with elite strategy is used for solving the multi-objective optimization problem established in the step 3.
4. An automatic driving control method based on a calibration optimized RSS model according to claim 3, wherein the NSGA-II algorithm includes the following steps:
step 4-1: randomly generating an initial population, wherein the individuals represent a random parameter set of the RSS model;
step 4-2: mutation and crossover are carried out on randomly selected individuals to generate subgroups;
step 4-3: combining the parent and the offspring;
step 4-4: layering the individuals according to the dominant relationship by using a rapid non-dominant ranking algorithm;
step 4-5: calculating the crowding degree i of an individual layer by layer d Sorting individuals according to the degree of congestion, and selecting proper individuals to form a new parent;
step 4-6: repeating the steps 2-5 until the average relative weight change of the fitness function among the generations is smaller than a preset tolerance or the iteration number reaches the maximum value, and stopping iteration.
5. The automatic driving control method based on the calibration optimization RSS model according to claim 4, wherein the congestion degree i in the steps 4-5 is as follows d The calculation method of (1) is as follows:
Figure FDA0003975019810000024
wherein ,
Figure FDA0003975019810000025
a j-th objective function value representing point i+1; />
Figure FDA0003975019810000026
The j-th objective function value at the i-1 th point is represented.
6. The automatic driving control method based on the calibration optimization RSS model according to claim 1, wherein the step 5 specifically includes:
selecting a first sub-target max f from the target function solution set obtained in the step 4 Safety The set of data with the largest value is used as the final calibration result of the RSS model.
CN202010456689.4A 2020-05-26 2020-05-26 Automatic driving control method based on calibration optimization RSS model Active CN111723470B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010456689.4A CN111723470B (en) 2020-05-26 2020-05-26 Automatic driving control method based on calibration optimization RSS model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010456689.4A CN111723470B (en) 2020-05-26 2020-05-26 Automatic driving control method based on calibration optimization RSS model

Publications (2)

Publication Number Publication Date
CN111723470A CN111723470A (en) 2020-09-29
CN111723470B true CN111723470B (en) 2023-05-02

Family

ID=72565190

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010456689.4A Active CN111723470B (en) 2020-05-26 2020-05-26 Automatic driving control method based on calibration optimization RSS model

Country Status (1)

Country Link
CN (1) CN111723470B (en)

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2022130019A1 (en) * 2020-12-17 2022-06-23 Mobileye Vision Technologies Ltd. Vehicle operation safety model test system
CN113111003B (en) * 2021-01-11 2022-05-03 北京赛目科技有限公司 Scene-independent unmanned driving simulation test evaluation method and device
CN113911131A (en) * 2021-09-24 2022-01-11 同济大学 Responsibility sensitive safety model calibration method for human-vehicle conflict in automatic driving environment
CN113954874B (en) * 2021-11-03 2023-04-28 同济大学 Automatic driving control method based on improved intelligent driver model
CN114116454A (en) * 2021-11-03 2022-03-01 同济大学 Calibration method for vehicle automatic emergency brake model in mechanical and non-conflict scene
CN115017742B (en) * 2022-08-08 2022-12-13 西安深信科创信息技术有限公司 Automatic driving test scene generation method, device, equipment and storage medium
CN115080450B (en) * 2022-08-22 2022-11-11 深圳慧拓无限科技有限公司 Automatic driving test data generation method and system, electronic device and storage medium

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2016169290A1 (en) * 2015-04-21 2016-10-27 华南理工大学 Decision-making supporting system and method oriented towards emergency disposal of road traffic accidents
WO2019213980A1 (en) * 2018-05-08 2019-11-14 清华大学 Intelligent vehicle safety decision-making method employing driving safety field
CN110641470A (en) * 2019-09-30 2020-01-03 福州大学 Pure electric vehicle driving auxiliary system optimization method integrating driver preference

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9946818B2 (en) * 2013-07-30 2018-04-17 University Of Florida Research Foundation, Inc. System and method for automated model calibration, sensitivity analysis, and optimization

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2016169290A1 (en) * 2015-04-21 2016-10-27 华南理工大学 Decision-making supporting system and method oriented towards emergency disposal of road traffic accidents
WO2019213980A1 (en) * 2018-05-08 2019-11-14 清华大学 Intelligent vehicle safety decision-making method employing driving safety field
CN110641470A (en) * 2019-09-30 2020-01-03 福州大学 Pure electric vehicle driving auxiliary system optimization method integrating driver preference

Also Published As

Publication number Publication date
CN111723470A (en) 2020-09-29

Similar Documents

Publication Publication Date Title
CN111723470B (en) Automatic driving control method based on calibration optimization RSS model
CN110610260B (en) Driving energy consumption prediction system, method, storage medium and equipment
Talebpour et al. Speed harmonization: Evaluation of effectiveness under congested conditions
CN108960426A (en) Road grade Synthesize estimation system based on BP neural network
CN104851280B (en) A kind of vehicle travel control method, device, system and relevant device
CN111081067B (en) Vehicle collision early warning system and method based on IGA-BP neural network under vehicle networking environment
CN110941902A (en) Lightning stroke fault early warning method and system for power transmission line
CN114132341A (en) Ramp track planning model for automatically-driven vehicle in Internet of vehicles environment
CN113312733B (en) Method, device, equipment and storage medium for calibrating simulation model parameters of signal-controlled roundabout
Yang et al. An adaptive cruise control method based on improved variable time headway strategy and particle swarm optimization algorithm
CN113360412A (en) Test scene selection method and device, electronic equipment and storage medium
Liu et al. Trade-offs between bus and private vehicle delays at signalized intersections: Case study of a multiobjective model
CN112731806A (en) Intelligent networking automobile random model prediction control real-time optimization method
CN114896754B (en) Automatic driving system performance evaluation method oriented to logic scene full parameter space
CN113095584B (en) Intersection safety level short-time prediction method, system, terminal and readable storage medium based on traffic flow characteristics
KR102006755B1 (en) In-vehicle external data intrusion detection apparatus by comparing information entropy and operating method thereof
CN115973179A (en) Model training method, vehicle control method, device, electronic equipment and vehicle
CN113887131A (en) Responsibility sensitive safety model calibration method facing vehicle cut-in scene
EP4339051A1 (en) Driving strategy determination method and apparatus, device, and vehicle
CN112572606B (en) Working condition fusion road sense simulation method based on SVM (support vector machine) and BP (back propagation) neural network
CN115447605A (en) Autonomous vehicle, and control method and control device for same
CN114148349A (en) Vehicle personalized following control method based on generation countermeasure simulation learning
CN113673146A (en) NSGA-II-based vehicle safety multi-objective optimization method
Marzbanrad et al. Prediction of driver’s accelerating behavior in the stop and go maneuvers using genetic algorithm-artificial neural network hybrid intelligence
CN113837211A (en) Driving decision method and device

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