CN113753059B - Method for predicting takeover capacity of driver under automatic driving system - Google Patents
Method for predicting takeover capacity of driver under automatic driving system Download PDFInfo
- Publication number
- CN113753059B CN113753059B CN202111116691.8A CN202111116691A CN113753059B CN 113753059 B CN113753059 B CN 113753059B CN 202111116691 A CN202111116691 A CN 202111116691A CN 113753059 B CN113753059 B CN 113753059B
- Authority
- CN
- China
- Prior art keywords
- driver
- driving
- takeover
- taking over
- grade
- 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
Links
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
- B60W40/00—Estimation or calculation of non-directly measurable driving parameters for road vehicle drive control systems not related to the control of a particular sub unit, e.g. by using mathematical models
- B60W40/08—Estimation or calculation of non-directly measurable driving parameters for road vehicle drive control systems not related to the control of a particular sub unit, e.g. by using mathematical models related to drivers or passengers
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
- B60W60/00—Drive control systems specially adapted for autonomous road vehicles
- B60W60/005—Handover processes
- B60W60/0053—Handover processes from vehicle to occupant
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Mechanical Engineering (AREA)
- Transportation (AREA)
- Automation & Control Theory (AREA)
- Theoretical Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Human Computer Interaction (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses a method for predicting the taking over capacity of a driver under an automatic driving system, belongs to the field of intelligent traffic, and aims to solve the technical problems that the popularization rate of the existing automatic driving is continuously improved, accidents of automatic driving automobiles occur frequently, and attention of existing research on the driver is insufficient. The prediction method comprises the following steps: 1. establishing a takeover capability prediction system which comprises driving capability, an initial state and a real-time state; 2. taking over performance includes alertness, sensitivity and operation stability, and taking over performance scoring is carried out on the driver; 3. constructing a prediction model, firstly determining a characteristic index discrimination threshold, establishing a takeover capability grade determination model, and calculating a comprehensive takeover capability determination value of a driver at a k grade through an actual measurement value of a driving characteristic index and the grade determination threshold; 4. and optimizing and correcting the pipe capacity grade judgment model. The invention establishes a driver state recognition system, predicts the takeover capacity, displays the takeover capacity grade in real time and promotes the system intellectualization.
Description
Technical Field
The invention belongs to the field of intelligent transportation, and particularly relates to a method for predicting the taking over capacity of a driver under a sudden situation in real time.
Background
With the trends of non-occupation of drivers, high speed of vehicle driving, and traffic environment traffic flow densification, an automatic driving technology aiming at comprehensively improving the driving safety performance of the vehicle and greatly reducing the operation load of the driver becomes a leading-edge and hot problem of research in the related fields of international automobile engineering, information science and the like. The current research is mainly developed from influencing factors of automatic driving takeover and driver takeover performance, and no related research provides a takeover capability evaluation method based on driver state recognition. The recent L3-level automatic driving automobile industry standard passed by the European economic Commission of United nations indicates that a driver state recognition system should be equipped with the L3-level automatic driving automobile, and the system should judge whether the driver has the ability to take over the driving task in real time. When the taking over capacity is weak, the system immediately sends out special warning to ensure the driving safety.
Currently, an auto-driving automobile of L3 or above level has a V2X (vehicle to electric) function, and can collect environmental information such as weather, instruments, road conditions, traffic condition tables, and the like, and determine a context awareness capability by comparing the cognition and actual conditions of a driver. The automatic driving system can give visual, auditory and tactile take-over warnings in emergency situations, collects steering wheel rotation data, accelerator and brake pedal data and the like in manual driving, and is used for testing reaction time and evaluating take-over performance. The brain wave acquisition device of brain Link pro can record fatigue degree and emotional condition. The above is the basis for the driver takeover capability prediction under the autopilot system.
Disclosure of Invention
The invention aims to solve the technical problems that the existing automatic driving popularizing rate is continuously improved, the accidents of automatic driving are frequent, and the existing research is lack of attention to drivers, and establish a driver state identification system, predict the taking over capacity, display the grade of the taking over capacity in real time, promote the system intellectualization and ensure the road traffic safety.
The method for predicting the taking over capacity of the driver under the automatic driving system is realized according to the following steps:
step one, establishing a three-stage driver takeover capacity prediction system;
the driver taking over capacity prediction system comprises driving capacity, an initial state and a real-time state, wherein the initial state is obtained through a situational awareness test and a reaction speed before driving, the real-time state comprises a fatigue index and an emotional state, and the driving capacity, the initial state and the real-time state are scored;
step two, taking over a performance evaluation method;
the taking over performance comprises alertness, sensitivity and operation stability; wherein alertness is by first gaze time (t) e ) Representing, wherein the first watching time refers to the time from the taking over request to the first watching of the road by the driver; take over time (t) for sensitivity h ) The representation is that the taking-over time refers to the time taken for the driver to turn the steering wheel by 2 degrees or press 10% of the brake pedal after the taking-over request is sent out; vehicle speed standard deviation(s) after take-over for stability of operation v ) Acceleration standard deviation(s) a ) Steering wheel angle standard deviation(s) θ ) Lateral deviation standard deviation(s) x ) Reverse rotation rates (SSRs) and steering entropy(s) e ) Common representation, namely evaluating alertness, sensitivity and operation stability to obtain a performance score taken over by a driver;
step three, constructing a driver takeover capability prediction model;
(1) Feature index discrimination threshold determination
The method comprises the steps that a driver takeover capacity prediction model predicts the takeover capacity of a driver by taking driving capacity, reaction speed before driving, situational awareness, fatigue index and emotional state as driving characteristic indexes i, firstly, the driving capacity, the reaction speed, the situational awareness, the fatigue index and the emotional state under various (different) takeover capacities are selected from a database, the classification prediction model is trained, ROC curve graphs of all driving characteristic indexes i of different levels k of the takeover capacity of the driver are obtained, and a discrimination threshold value of each driving characteristic index i under the level k is obtained through calculation
(2) Takeover capability grade judgment model
(1) According to the judgment threshold value of the driving characteristic index i under the k levelThereby obtaining TP k 、TN k 、FP k And FN k Wherein TP k Is the number of true instances in the k class, TN k Number of true counterexamples at k level, FP k For the number of false positive cases at k level, FN k The number of false counterexamples in the k level;
calculating the accuracy rate of the driving characteristic index i under the k level through a formula (1)
(2) The area under the ROC curve (AUC) is an important evaluation index for measuring the discrimination ability of the characteristic index to different states, so that the comprehensive discrimination accuracy of the driving characteristic index i for taking over the capability level k by the driver is represented as:
(3) the relative weight of the driving characteristic index i according to the comprehensive judgment accuracy of the driving characteristic index i of the driver takeover capability grade kA linear assignment is performed, the formula is as follows:
whereinAs the importance of the driving characteristic index i at the k level,the driving characteristic index i is weighted under the k level,as the importance of the driving characteristic index j in the k-rank,the weight of the driving characteristic index j under the k level is taken;
(4) actual measurement value X based on driving characteristic index i i (i.e., measured values of driving ability, reaction speed, situational awareness, fatigue index, and emotional state) and a level determination threshold value T i (lower limit node of the level of the group of data driving characteristic indexes i) calculating a comprehensive judgment value E for taking over capacity of the driver under the k level k When E is k >1, judging that the level of the taking over capacity of the driver is k level, and predicting to obtain the taking over capacity of the driver:
step four, taking over capacity grade correction model
Optimizing and correcting the management capacity grade judgment model, comparing the management capacity grade judgment with the management performance, adopting a similarity difference value interval R as a screening standard of the individual characteristic information of the driver, wherein a similarity S calculation formula is shown as a formula (5), and when the similarity is in the R, the group of data is classified into a training set of a classification prediction model to correct the model;
in the formula: d Vr For the hierarchical nodes taking over the capacity r level and r +1 level, D Vr+1 For the hierarchical nodes taking over the r +1 and r +2 levels of capacity, T i The lower limit node of the grade of the group of data driving characteristic indexes i,T i ' Upper limit node, T, of the grade of the set of data driving characteristic index i Vr The performance score is actually taken over for the driver.
The method comprises the steps of firstly establishing a three-stage takeover capacity prediction system, testing driving capacity before driving a vehicle for the first time, testing situational awareness and reaction time before running an automatic driving system each time, and recording fatigue degree and emotional condition in the running process of the automatic driving system; then, establishing a takeover performance evaluation scheme, evaluating the takeover performance according to the reaction condition of the driver and the running condition of the vehicle after takeover after the takeover prompt is sent, and taking the five indexes of each driver as a group of data to be contained in a database; secondly, establishing an ROC (ROC) takeover capability rating model, evaluating the perception capability of a driver on the basis of the data characteristics in the database and the five indexes of the situational awareness system, and displaying the takeover capability level of the driver in real time; and finally, establishing a model self-learning mechanism, taking the taking-over performance as an evaluation basis of the taking-over capacity, and after each taking-over task is completed, bringing the five indexes before sending out a prompt and the taking-over performance after sending out the prompt into an individual database for personalized correction of the model, so that the subsequently predicted taking-over capacity level is closer to the real condition of a driver.
Drawings
FIG. 1 is a schematic diagram illustrating a method for taking over performance evaluation by a driver in step two according to the embodiment;
FIG. 2 is a flow chart of the construction of the prediction model of the driver's takeover capability in step three of the embodiment;
FIG. 3 is a flowchart illustrating the capability level determination in step three according to an embodiment;
FIG. 4 is a flowchart of taking over the capability level correction model in step four of the embodiment.
Detailed Description
The first specific implementation way is as follows: the method for predicting the takeover capacity of the driver under the automatic driving system is implemented according to the following steps:
step one, establishing a three-stage driver takeover capability prediction system;
the driver taking over capacity prediction system comprises driving capacity, an initial state and a real-time state, wherein the initial state is obtained through a situational awareness test and a reaction speed before driving, the real-time state comprises a fatigue index and an emotional state, and the driving capacity, the initial state and the real-time state are scored;
step two, taking over a performance evaluation method;
the taking over performance comprises alertness, sensitivity and operation stability; wherein alertness is by first gaze time (t) e ) Representing, wherein the first watching time refers to the time from the taking over request to the first watching of the road by the driver; take over time (t) for sensitivity h ) The representation is that the taking-over time refers to the time taken for the driver to turn the steering wheel by 2 degrees or press 10% of the brake pedal after the taking-over request is sent out; vehicle speed standard deviation(s) after take-over for stability of operation v ) Acceleration standard deviation(s) a ) Steering wheel angle standard deviation(s) θ ) Standard deviation of lateral deviation(s) x ) Reverse rotation rates (SSRs) and steering entropy(s) e ) Performing common characterization, namely evaluating alertness, sensitivity and operation stability to obtain a driver takeover performance score;
step three, constructing a driver takeover capability prediction model;
(1) Feature index discrimination threshold determination
The method comprises the steps that a driver takeover capacity prediction model predicts the takeover capacity of a driver by taking driving capacity, reaction speed before driving, situational awareness, fatigue index and emotional state as driving characteristic indexes i, firstly, the driving capacity, the reaction speed, the situational awareness, the fatigue index and the emotional state under various (different) takeover capacities are selected from a database, the classification prediction model is trained, ROC curve graphs of all driving characteristic indexes i of different levels k of the takeover capacity of the driver are obtained, and a discrimination threshold value of each driving characteristic index i under the level k is obtained through calculation
(2) Takeover capability grade judgment model
(1) According to the judgment threshold value of the driving characteristic index i under the k levelThereby obtaining TP k 、TN k 、FP k And FN k Wherein TP k Number of true instances at k level, TN k Number of true counterexamples at k level, FP k For the number of false positive cases at k level, FN k The number of false counterexamples at the k level;
calculating the accuracy rate of the driving characteristic index i under the k level through a formula (1)
(2) The area under the ROC curve (AUC) is an important evaluation index for measuring the discrimination ability of the characteristic index to different states, and therefore, the comprehensive discrimination accuracy of the driving characteristic index i for the driver to take over the capability level k is represented as:
(3) the relative weight of the driving characteristic index i according to the comprehensive judgment accuracy of the driving characteristic index i of the driver takeover capability grade kA linear assignment is performed, the formula is as follows:
whereinAs the importance of the driving characteristic index i at the k level,the driving characteristic index i is weighted under the k level,as the importance of the driving characteristic index j in the k rank,the weight of the driving characteristic index j under the k level is taken;
(4) actual measurement value X based on driving characteristic index i i (i.e., measured values of driving ability, reaction speed, situational awareness, fatigue index, and emotional state) and a level determination threshold value T i (the upper line node of the level of the group of data driving characteristic indexes i) calculates a comprehensive judgment value E of taking over capacity of the driver under the k level k When E is k >1, judging that the driver takeover capability grade is k grade:
step four, taking over capacity grade correction model
Optimizing and correcting the management capacity grade judgment model, comparing the management capacity grade judgment with the management performance, adopting a similarity difference value interval R as a screening standard of the individual characteristic information of the driver, wherein a similarity S calculation formula is shown as a formula (5), and when the similarity is in the R, the group of data is classified into a training set of a classification prediction model to correct the model;
in the formula: d Vr To take over the hierarchical nodes of the r and r +1 levels of capacity,D Vr+1 for the hierarchical nodes taking over the r +1 and r +2 levels of capacity, T i A lower limit node, T, of the grade of the group of data of driving characteristic indexes i i ' is the upper limit node, T, of the grade of the group of data driving characteristic indexes i Vr The performance score is actually taken over for the driver.
The embodiment provides a method for predicting the takeover capacity of a driver under an automatic driving system, which can ensure that the physical and mental conditions and the sensing operation capacity of the driver are in a visible, adjustable and controllable environment, improve the takeover fluency and the driving stability of the driver and ensure the driving safety.
The second embodiment is as follows: the difference between the present embodiment and the first embodiment is that the driving ability in the first step includes the basic situation and the driving experience of the driver, wherein the basic situation of the driver includes age, sex, driving mileage and driving time per month (h/month); the driving experience includes driving behavior, driving skill, and danger perception.
The present embodiment may acquire the driving experience of the driver in the form of a questionnaire.
The third concrete implementation mode: the difference between the first embodiment and the second embodiment is that the reaction speed test before driving in the first step adopts comprehensive evaluation of visual stimulation reaction time, auditory stimulation reaction time and tactile stimulation reaction time, wherein the visual stimulation is the time for the driver to operate an accelerator pedal, a brake pedal and a steering wheel according to visual prompt contents; the auditory stimulation is the time for the driver to adjust the steering wheel and step on the brake pedal after hearing the prompt sound; the tactile stimulation is the time taken for a driver to adjust the steering wheel and step on the brake pedal after receiving the vibration of the steering wheel.
The fourth concrete implementation mode: the third difference between this embodiment and the third embodiment is that the visual stimulus responses account for 51% of the response speed score weight, the auditory stimulus responses account for 30% of the response speed score weight, and the tactile stimulus responses account for 19% of the response speed score weight in the response speed score.
The fifth concrete implementation mode: the present embodiment is different from the first to fourth embodiments in that the real-time state described in the first step is acquired by the brain wave acquiring device.
The sixth specific implementation mode: the difference between this embodiment and one of the first to fifth embodiments is that the fixation time in step two is collected by an eye tracker.
The seventh embodiment: the difference between the embodiment and one of the first to sixth specific embodiments is that in the second step, the alertness accounts for 20% of the weight of the driver for taking over the performance score, the sensitivity accounts for 30% of the weight of the driver for taking over the performance score, the vehicle speed standard deviation accounts for 10% of the weight of the driver for taking over the performance score, the acceleration standard deviation accounts for 10% of the weight of the driver for taking over the performance score, the steering wheel turning angle standard deviation accounts for 10% of the weight of the driver for taking over the performance score, the lateral deviation standard deviation accounts for 10% of the weight of the driver for taking over the performance score, the inversion rate accounts for 5% of the weight of the driver for taking over the performance score, and the steering entropy accounts for 5% of the weight of the driver for taking over the performance score.
The specific implementation mode is eight: this embodiment differs from one of the first to seventh embodiments in that the reaction speed described in step three is tested by the FDS driving simulator.
The specific implementation method nine: the difference between this embodiment and the first to eighth embodiments is that the threshold is calculated in the process of determining the discrimination threshold of the feature index in the third stepThe method comprises the steps of searching a maximum value point of a johnson index (TPR-FPR) of each driving characteristic index i under a level k to obtain a discrimination threshold value of the driving characteristic index i under the level k
The detailed implementation mode is ten: the difference between this embodiment and one of the first to ninth embodiments is that the value of R in step four is 0.8 to 1.5.
Example (b): the method for predicting the takeover capability of the driver under the automatic driving system is implemented according to the following steps:
step one, establishing a three-stage driver takeover capability prediction system;
the driver taking over capacity prediction system comprises driving capacity, an initial state and a real-time state, wherein the initial state is obtained through a situational awareness test and a reaction speed test before driving, and the real-time state comprises a fatigue index and an emotional state;
(1) Evaluation of Driving ability
Only the driving age and the driving mileage are selected from the basic information, and the specific scoring standards are shown in the following table:
TABLE 2-1 basic information Scoring Table
The driving ability scoring method based on the driving experience questionnaire comprises the following steps: in a questionnaire, 6 driving behaviors and 6 danger perceptions are carried out, the answer of each question is 1 point obtained frequently, 2 points obtained occasionally and 3 points obtained never; the driving skills 7 are respectively 0,1,2,3,4 and 5 points from completely inoperable to completely operable.
The driving experience questionnaire example questions:
1. driving behavior, whether there is too close a preceding vehicle to the past during driving so as to cause difficulty in parking in an emergency?
A. Often b, occasionally c, never.
2. Driving skill, i can reasonably change lanes when driving on busy road conditions?
3. Danger perception, whether there is a following operation to drive the car without turning a turn signal?
A. Often b, occasionally c, never.
The weighting process is performed for the driving skill in the driving experience.
T2-driving skill score;
ti — score of driving skill ith topic.
The final driving experience score T is:
t1-driving behavior scoring;
t2-driving skill score;
t3 — risk perception score.
And calculating according to the calculating method, wherein the score of the driving experience questionnaire is divided into 60 points, and the final score result is multiplied by 1/6 and standardized to a tenth system.
And finally, grading by combining a basic information grading table and a driving experience questionnaire grading method, and taking the average value of the two grading results as a final grading value.
The rating criteria for drivability were as follows:
TABLE 2-2 rating results sheet
(2) Reaction speed test
SCANeR is selected for reaction speed test TM The studio software is combined with an FDS driving simulator, the visual stimulation in the process of taking over is simulated through the display of a computer and the graphics or characters, the sound prompt in an automobile is simulated through a sound, and the touch stimulation is simulated through the vibration of a steering wheel. And recording the stimulation sending time and the action time of the tested person through a script in the software, and calculating the time difference between the stimulation sending time and the action time as corresponding response time.
The visual stimulation is that after the experimenter sees the graphic prompt on the screen, corresponding operations such as an accelerator pedal, a brake pedal and a steering wheel are carried out according to the prompt content, different visual stimulations respectively appear once at random, and three visual stimulation reaction experiments are carried out in total; the auditory stimulation is that after the tested person hears a beep sound, the steering wheel is finely adjusted, and the brake pedal is stepped on; the tactile stimulation is that the tested person slightly adjusts the steering wheel and presses the brake pedal after receiving the vibration of the steering wheel. Each subject was subjected to 5 reaction rate test experiments. The reaction rate test protocol is shown in the following table:
TABLE 2-3 reaction Rate testing protocol
The reaction rate scoring method is shown in the following table:
TABLE 2-4 reaction Rate evaluation Table
The reaction rate rating scale was as follows:
TABLE 2-5 reaction Rate rating Table
(3) Situational awareness measurement
The driver judges three questions which appear randomly according to the current scene, the answering time exceeds 8min, and the answering interface is automatically closed. The experimental program automatically records the answer to be tested and the answering time, judges the wrong answer and gives a score.
The situational awareness score is the sum of the perception, understanding and prediction scores of three questions and the additional score of answering time.
The example questions are as follows:
1. how the weather conditions of the current scene are
A. Good B, good C and poor
2. The maximum speed allowed for the current road section is
A. 20 km/h B, 30 km/h C, 40 km/h D, 50 km/h
3. How to use the light when driving in such weather conditions
A. Using dipped headlight B, no light C, high beam D, fog light
Specific scoring methods are shown in the following table:
TABLE 2-6 Scenario consciousness scoring sheet
TABLE 2-7 situational awareness rating sheet
(4) Physiological index detection
A brain wave acquisition device of brain wave of brain link pro is selected for detecting the fatigue degree and the emotional condition, and the brain wave acquisition device is used for recording and detecting 8 kinds of brain waves of delta, theta, alpha, beta, gamma and the like which are the most important brain waves in real time. The fatigue degree and emotional condition output by the instrument can be standardized to a complete system, and the corresponding rating method is shown in the following table.
TABLE 2-8 fatigue index rating sheet
Tables 2-9 emotional condition rating tables
Step two, taking over a performance evaluation method;
the taking over performance comprises alertness, sensitivity and operation stability; wherein alertness is by first gaze time (t) e ) The first fixation time refers to the time from the sending of the takeover request to the first fixation time (t) of the driver e ) Collecting by an eye tracker; take over time (t) for sensitivity h ) Characterized by the take-over time beingThe time it takes for the finger joint pipe request to be sent out until the driver turns the steering wheel by 2 degrees or depresses the brake pedal by 10%; vehicle speed standard deviation(s) after take-over for stability of operation v ) Acceleration standard deviation(s) a ) Steering wheel angle standard deviation(s) θ ) Standard deviation of lateral deviation(s) x ) Reverse rotation rates (SSRs) and steering entropy(s) e ) Common characterization, take-over time (t) h ) Standard deviation of velocity(s) v ) Acceleration standard deviation(s) a ) Steering wheel angle standard deviation(s) θ ) Lateral deviation standard deviation(s) x ) The reversing rate (SRRs) and the Steering Entropy (SE) are collected by an automatic driving system, and the taking-over performance score of a driver is obtained by evaluating the alertness, the sensitivity and the operation stability;
the method for taking over the performance scoring by the driver is shown as the following table:
TABLE 2-10 Take-over ability score sheet
The performance evaluation criteria taken over by the driver are shown in the following table:
TABLE 2-11 take-over performance evaluation chart
Step three, constructing a driver takeover capability prediction model;
(1) Feature index discrimination threshold determination
The method comprises the steps that a driver takeover capacity prediction model predicts the takeover capacity of a driver by taking driving capacity, reaction speed before driving, situational awareness, fatigue index and emotional state as driving characteristic indexes i, firstly, the driving capacity, the reaction speed, the situational awareness, the fatigue index and the emotional state under various (different) takeover capacities are selected from a database, the classification prediction model is trained, ROC curve graphs of all driving characteristic indexes i of different levels k of the takeover capacity of the driver are obtained, and judgment of all the driving characteristic indexes i under the level k is obtained through calculationThreshold value
(2) Takeover capability grade judgment model
(1) According to the judgment threshold value of the driving characteristic index i under the k levelThereby obtaining TP k 、TN k 、FP k And FN k Wherein TP k Is the number of true instances in the k class, TN k Number of true counter-instances at k level, FP k For the number of false positive cases at k level, FN k The number of false counterexamples in the k level;
in the embodiment, the taking over performance is divided into four specific grades, and each index is divided into four pending grades respectively. Take fatigue level at three levels as an example: if the fatigue degree of a group of data is three levels and the performance of the take-over is three levels, the data is classified as a real example TP 3 If the takeover performance is not three levels, the data is classified as false positive FN 3 (ii) a If the fatigue degree is not three levels and the performance of the take-over is not three levels, the data is classified as true counter-example TN 3 If the performance of taking over is three levels, the data is classified as false negative FP 3
Calculating False Positive Rate (FPR) and True Positive Rate (TPR) as horizontal and vertical coordinates to obtain ROC curve;
TABLE 3-1 Classification confusion matrix
Calculating the accuracy rate of the driving characteristic index i under the k level through a formula (1)
(2) The area under the ROC curve (AUC) is an important evaluation index for measuring the discrimination ability of the characteristic index to different states, and therefore, the comprehensive discrimination accuracy of the driving characteristic index i for the driver to take over the capability level k is represented as:
(3) the relative weight of the driving characteristic index i according to the comprehensive judgment accuracy of the driving characteristic index i of the driver taking over capacity level kA linear assignment is performed, the formula is as follows:
whereinAs the importance of the driving characteristic index i at the k level,the driving characteristic index i is weighted under the k level,as the importance of the driving characteristic index j in the k rank,the weight of the driving characteristic index j under the k level is taken;
(4) actual measurement value X based on driving characteristic index i i (i.e., measured values of driving ability, reaction speed, situational awareness, fatigue index, and emotional state) and a level determination threshold value T i (the group of data of driving characteristic indexes i are on the upper line node of the grade) calculating the comprehensive judgment value E of taking over capacity of the driver under the k grade k When E is k >1, judging that the level of the taking over capacity of the driver is k level, and predicting to obtain the taking over capacity of the driver:
step four, taking over capacity grade correction model
Optimizing and correcting the management capacity grade judgment model, comparing the management capacity grade judgment with the management performance, adopting a similarity difference value interval R as a screening standard of the individual characteristic information of the driver, wherein a similarity S calculation formula is shown as a formula (5), and when the similarity is in the R, the group of data is classified into a training set of a classification prediction model to correct the model;
in the formula: d Vr For the hierarchical nodes taking over the capacity r level and r +1 level, D Vr+1 For the hierarchical nodes taking over the r +1 and r +2 levels of capacity, T i ' Upper limit node, T, of the grade of the set of data driving characteristic index i Vr The performance score is actually taken over for the driver.
In the embodiment, 46 volunteers holding the motor vehicle driving license for more than one year are invited to participate in the simulation driving experiment, and five indexes related to the experiment are recorded and are included in the database. The taking capacity was rated as 12 good persons, 21 good persons, 8 medium persons and 5 bad persons.
Determining the driver characteristic index discrimination threshold value under each takeover capacity discrimination grade by applying an ROC curve method, and comprehensively determining the weight of each characteristic indexThe weight distribution of each characteristic index under different judging levels is as follows:
TABLE 2-12 Driving characteristics index weight distribution Table
Transverse and longitudinal comparison analysis is carried out on weight distribution, so that the fatigue degree and the reaction speed are main influence factors for judging the pipe taking capacity, and the fatigue degree of a driver has obvious influence on the performance of the pipe taking. Along with the reduction of the taking over capacity judgment level, the weight distribution of the index is gradually increased, and the influence on the taking over capacity judgment of the driver is more obvious, so that the real-time state monitoring of the driver is taken into consideration, and the fatigue driving of the driver is prevented by the caution. The driving ability and situational awareness comprehensive index weight distribution is small, and the influence of the driving ability and situational awareness on the judgment of the management ability is small. The weight distribution of the two is gradually reduced along with the reduction of the level of the judgment of the takeover capacity, which shows that the distinction degree between the driving capacity and the situational awareness is not obvious between the low level of the judgment of the takeover capacity.
In order to check the effectiveness of the correction model, the data of three driving of 1 volunteer is selected as an example, 5 characteristic index data are adopted through a simulation driving experiment, the management capacity grade is predicted and compared with the actual management performance grade, and the experimental data are as follows:
tables 2-13 correction of model test data
The experimental result shows that as the times of applying the takeover capability prediction system are increased, the result predicted by the model is closer to the actual takeover performance, and the takeover capability grade correction model is effective.
Claims (10)
1. The method for predicting the takeover capability of the driver under the automatic driving system is characterized by being realized according to the following steps:
step one, establishing a three-stage driver takeover capability prediction system;
the driver takeover capability prediction system comprises driving capability, an initial state and a real-time state, wherein the initial state is obtained through a situational awareness test and a reaction speed before driving, the real-time state comprises a fatigue index and an emotional state, and the driving capability, the initial state and the real-time state are scored;
step two, taking over a performance evaluation method;
the taking over performance comprises alertness, sensitivity and operation stability; wherein the alertness is by the first gaze time t e Representing, wherein the first watching time refers to the time from the taking over request to the first watching of the road by the driver; take-over time t for sensitivity h The representation is that the taking-over time refers to the time taken for the driver to turn the steering wheel by 2 degrees or press 10% of the brake pedal after the taking-over request is sent out; standard deviation s of vehicle speed after taking over for operation stability v Standard deviation of acceleration s a Steering wheel angle standard deviation s θ Standard deviation s of lateral deviation x Reverse rotation rates SSRs and steering entropy s e Performing common characterization, namely evaluating alertness, sensitivity and operation stability to obtain a driver takeover performance score;
step three, constructing a driver takeover capability prediction model;
(1) Feature index discrimination threshold determination
The method comprises the steps of firstly selecting driving capacity, reaction speed, situational awareness, fatigue index and emotional state under different taking over capacities from a database, training a classification prediction model, obtaining an ROC curve graph of each driving characteristic index of different driving characteristic indexes of different driving capacity grades k, and calculating to obtain a discrimination threshold value of each driving characteristic index i under the grade k
(2) Takeover capability grade judgment model
(1) According to the judgment threshold value of the driving characteristic index i under the k gradeThereby obtaining TP k 、TN k 、FP k And FN k Wherein TP k Is the number of true instances in the k class, TN k Number of true counter-instances at k level, FP k For the number of false positive cases at k level, FN k The number of false counterexamples in the k level;
calculating the accuracy rate of the driving characteristic index i under the k level through a formula (1)
(2) The area under the ROC curve is an important evaluation index for measuring the distinguishing capability of the characteristic index to different states, so the comprehensive distinguishing accuracy of the driving characteristic index i of the driver taking over the capability level k is expressed as follows:
(3) the relative weight of the driving characteristic index i according to the comprehensive judgment accuracy of the driving characteristic index i of the driver takeover capability grade kA linear assignment is performed, the formula is as follows:
whereinAs the importance of the driving characteristic index i at the k level,the driving characteristic index i is weighted under the k level,as the importance of the driving characteristic index j in the k rank,the weight of the driving characteristic index j under the k level is taken;
(4) actual measurement value X based on driving characteristic index i i And the lower limit node T of the grade of the group of data driving characteristic indexes i i Calculating a comprehensive judgment value E of taking over capacity of the driver under the k level k When E is k >1, judging that the driver takeover capability grade is k grade:
step four, taking over capacity grade correction model
Optimizing and correcting the receiving capacity grade judging model, comparing the receiving capacity grade judging model with the receiving performance, and adopting a similarity difference value interval R as a screening standard of the individual characteristic information of the driver, wherein a similarity S calculation formula is shown as a formula (5), and when the similarity is in the R, the group of data is classified into a training set of a classification prediction model to correct the model;
in the formula: d Vr Hierarchical nodes of order r and r +1 taking over capabilities, D Vr+1 For the hierarchical nodes taking over the r +1 and r +2 levels of capacity, T i ' Upper limit node, T, of the grade of the set of data driving characteristic index i Vr The performance score is actually taken over for the driver.
2. The method of claim 1, wherein the driving ability in step one comprises the basic condition of the driver and the driving experience, wherein the basic condition of the driver comprises age, sex, driving mileage and driving time per month; the driving experience includes driving behavior, driving skill, and danger perception.
3. The method for predicting the takeover capacity of the driver under the automatic driving system according to claim 1, wherein the reaction speed test before driving in the step one adopts comprehensive evaluation of visual stimulus reaction time, auditory stimulus reaction time and tactile stimulus reaction time, wherein the visual stimulus is time taken by the driver to operate an accelerator pedal, a brake pedal and a steering wheel according to visual prompt contents; the auditory stimulation is the time for the driver to adjust the steering wheel and step on the brake pedal after hearing the prompt sound; the tactile stimulation is the time taken for a driver to adjust the steering wheel and step on the brake pedal after receiving the vibration of the steering wheel.
4. The method according to claim 3, wherein the response rate score includes 51% visual stimulus responses, 30% auditory stimulus responses, and 19% tactile stimulus responses.
5. The method as claimed in claim 1, wherein the real-time status is obtained by a brain wave acquisition device in the step one.
6. The method of predicting takeover ability of a driver under an autopilot system according to claim 1 wherein in step two the first fixation time is collected by an eye tracker.
7. The method for predicting the takeover capability of the driver under the automatic driving system as claimed in claim 1, wherein in the second step, the alertness accounts for 20% of the weight of the driver for taking over the performance score, the sensitivity accounts for 30% of the weight of the driver for taking over the performance score, the standard deviation of the vehicle speed accounts for 10% of the weight of the driver for taking over the performance score, the standard deviation of the acceleration accounts for 10% of the weight of the driver for taking over the performance score, the standard deviation of the steering wheel angle accounts for 10% of the weight of the driver for taking over the performance score, the standard deviation of the lateral deviation accounts for 10% of the weight of the driver for taking over the performance score, the reversion rate accounts for 5% of the weight of the driver for taking over the performance score, and the steering entropy accounts for 5% of the weight of the driver for taking over the performance score.
8. The method for predicting the takeover capability of the driver under the automatic driving system as claimed in claim 1, wherein the reaction speed in step three is tested by a FDS driving simulator.
9. The method for predicting the takeover ability of the driver under the automatic driving system according to claim 1, wherein the threshold is calculated in the process of determining the discrimination threshold of the three characteristic indexes in the stepThe method is characterized in that a decision threshold value of each driving characteristic index i under the level k is obtained by searching a jordan index maximum value point of each driving characteristic index i under the level k
10. The method for predicting the takeover capability of the driver under the automatic driving system according to claim 1, wherein R in the fourth step takes a value of 0.8 to 1.5.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111116691.8A CN113753059B (en) | 2021-09-23 | 2021-09-23 | Method for predicting takeover capacity of driver under automatic driving system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111116691.8A CN113753059B (en) | 2021-09-23 | 2021-09-23 | Method for predicting takeover capacity of driver under automatic driving system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113753059A CN113753059A (en) | 2021-12-07 |
CN113753059B true CN113753059B (en) | 2022-11-29 |
Family
ID=78797085
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111116691.8A Active CN113753059B (en) | 2021-09-23 | 2021-09-23 | Method for predicting takeover capacity of driver under automatic driving system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113753059B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US11463851B2 (en) * | 2021-01-15 | 2022-10-04 | Harman International Industries, Incorporated | V2X communication system with autonomous driving information |
CN114882477A (en) * | 2022-03-04 | 2022-08-09 | 吉林大学 | Method for predicting automatic driving takeover time by using eye movement information |
CN115407750B (en) * | 2022-08-12 | 2023-11-21 | 北京津发科技股份有限公司 | Evaluation method and system for decision-making capability of man-machine collaborative intelligent system |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111703434A (en) * | 2020-06-01 | 2020-09-25 | 哈尔滨工业大学 | Design method for passive control right switching auditory cue under automatic driving |
CN111915159A (en) * | 2020-07-15 | 2020-11-10 | 北方工业大学 | Personalized takeover early warning method and system based on dynamic time budget |
CN112406882A (en) * | 2020-11-11 | 2021-02-26 | 中汽研汽车检验中心(天津)有限公司 | Device for monitoring state of driver in man-machine co-driving process and method for evaluating pipe connection capability |
CN112758098A (en) * | 2019-11-01 | 2021-05-07 | 广州汽车集团股份有限公司 | Vehicle driving authority take-over control method and device based on driver state grade |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP3795441A1 (en) * | 2019-09-17 | 2021-03-24 | Aptiv Technologies Limited | Method and device for determining an estimate of the capability of a vehicle driver to take over control of a vehicle |
WO2021061212A1 (en) * | 2019-09-26 | 2021-04-01 | Hrl Laboratories, Llc | System and method for safety and efficacy override of an autonomous system |
-
2021
- 2021-09-23 CN CN202111116691.8A patent/CN113753059B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112758098A (en) * | 2019-11-01 | 2021-05-07 | 广州汽车集团股份有限公司 | Vehicle driving authority take-over control method and device based on driver state grade |
CN111703434A (en) * | 2020-06-01 | 2020-09-25 | 哈尔滨工业大学 | Design method for passive control right switching auditory cue under automatic driving |
CN111915159A (en) * | 2020-07-15 | 2020-11-10 | 北方工业大学 | Personalized takeover early warning method and system based on dynamic time budget |
CN112406882A (en) * | 2020-11-11 | 2021-02-26 | 中汽研汽车检验中心(天津)有限公司 | Device for monitoring state of driver in man-machine co-driving process and method for evaluating pipe connection capability |
Non-Patent Citations (1)
Title |
---|
基于ROC曲线和驾驶行为特征的驾驶愤怒强度判别阈值;万平等;《吉林大学学报(工学版)》;20200131;第50卷(第1期);第121-131页 * |
Also Published As
Publication number | Publication date |
---|---|
CN113753059A (en) | 2021-12-07 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN113753059B (en) | Method for predicting takeover capacity of driver under automatic driving system | |
Feng et al. | Can vehicle longitudinal jerk be used to identify aggressive drivers? An examination using naturalistic driving data | |
CN106251583B (en) | Fatigue driving discrimination method based on driving behavior and eye movement characteristics | |
Chen et al. | A graphical modeling method for individual driving behavior and its application in driving safety analysis using GPS data | |
Zhenhai et al. | Driver drowsiness detection based on time series analysis of steering wheel angular velocity | |
CN107832748A (en) | A kind of shared automobile driver changes system and method | |
CN113743471B (en) | Driving evaluation method and system | |
CN113257023B (en) | L3-level automatic driving risk assessment and takeover early warning method and system | |
CN112693469A (en) | Method and device for testing vehicle taking over by driver, electronic equipment and medium | |
CN111657923A (en) | Method and system for testing danger perception capability of driver | |
Mori et al. | Integrated modeling of driver gaze and vehicle operation behavior to estimate risk level during lane changes | |
CN112070927A (en) | Highway vehicle microscopic driving behavior analysis system and analysis method | |
CN106355329A (en) | Method and system for testing sign aligning ability of driver | |
CN114155742B (en) | Method for evaluating and early warning longitudinal driving risk of internet vehicle | |
CN112754483B (en) | Method and system for testing and training attention allocation capability of driver | |
CN115759880A (en) | Real-time on-duty adaptability evaluation system and method for bus driver | |
CN115994844A (en) | Driving safety evaluation method under road tunnel disaster scene | |
CN109858178A (en) | A kind of commercial vehicle drivers giving fatigue pre-warning method based on Intelligent bracelet | |
CN115192026A (en) | Tunnel driving load monitoring method and terminal | |
Wang et al. | The Evaluation of the Driving Capability for Drivers Based on Vehicle States and Fuzzy-ANP Model | |
CN117717340B (en) | Driver sleepiness detection method, device, equipment and medium | |
Qi | Multiple linear regression model of defensive driving technology and driver emotional intelligence improvement | |
Ma et al. | Establishing Random Forest Model Based on Visual Variables to Detect Driving Fatigue | |
Wang | Driving Fatigue Behavior Analysis Based on BP Neural Network | |
CN106344050A (en) | Testing method and system for stress tolerance of driver |
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 |