CN109367541A - A kind of intelligent vehicle class people's lane change decision-making technique based on driver behavior pattern - Google Patents
A kind of intelligent vehicle class people's lane change decision-making technique based on driver behavior pattern Download PDFInfo
- Publication number
- CN109367541A CN109367541A CN201811195634.1A CN201811195634A CN109367541A CN 109367541 A CN109367541 A CN 109367541A CN 201811195634 A CN201811195634 A CN 201811195634A CN 109367541 A CN109367541 A CN 109367541A
- Authority
- CN
- China
- Prior art keywords
- lane change
- driver
- lane
- intelligent vehicle
- driving style
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- 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
- B60W40/09—Driving style or behaviour
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
- B60W50/00—Details of control systems for road vehicle drive control not related to the control of a particular sub-unit, e.g. process diagnostic or vehicle driver interfaces
- B60W2050/0001—Details of the control system
- B60W2050/0019—Control system elements or transfer functions
- B60W2050/0028—Mathematical models, e.g. for simulation
- B60W2050/0029—Mathematical model of the driver
-
- 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
- B60W2540/00—Input parameters relating to occupants
- B60W2540/30—Driving style
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Molecular Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Probability & Statistics with Applications (AREA)
- Automation & Control Theory (AREA)
- Transportation (AREA)
- Mechanical Engineering (AREA)
- Traffic Control Systems (AREA)
Abstract
Intelligent vehicle class people's lane change decision-making technique based on driver behavior pattern that the invention discloses a kind of, by establishing the experiment of driver's steering characteristic, extract the parameter of characterization driver's steering characteristic, then driver characteristics identifier is established using the method for K-means cluster and BP neural network, what the driver characteristics and sensor recognized according to driver characteristics identifier identified itself carries out class people lane change decision with surrounding vehicles state and environmental information, and the lane change behavior of intelligent vehicle is finally made to have the driving characteristics of class people.
Description
Technical field
The present invention relates to a kind of intelligent vehicle lane change decision-making technique more particularly to a kind of behavioral trait is turned to based on driver
Intelligent vehicle class people's lane change decision-making technique.
Background technique
Worldwide, car ownership also brings along many traffic safety problems in sustainable growth in this way.In
For state, annual traffic accident total amount probably 4,700,000 or so, and therefore and dead number accounting probably reaches 21%.
Statistics of traffic accidents in 2008 the results show that caused by China is because of the behavior mistake of driver traffic accident have 230727
It rises, accounts for the 87% of total number of accident in 2008, cause 61065 people death, 265889 people injured, account for total toll and injury respectively
The 83.1% of total number of persons and 87.2%.Obviously, and for " people-Che-road " this typical system, people is as closed-loop system
Decision-maker all produces tremendous influence to automobile and road and the other things participated, these influence major embodiment
It is all as caused by artificial origin in the traffic accident of the overwhelming majority.
In order to improve this problem, the effect for allowing people in " people-vehicle-road " this closed-loop system is needed to desalinate, even
The closed-loop system is directly become into " vehicle-road " closed-loop system.This requires automobiles itself to have certain intelligence, becomes more
Add easy driving, it might even be possible to realize that autonomous driving is realized in the manipulation for completely disengaging people.To realize this function, intelligent vehicle need to be carried out
The research of technology.
In the past few decades, huge innovation is had occurred in intelligent automobile field.Laser radar, millimetre-wave radar, camera shooting
Head, the quick change of the hardware devices such as processor, so that lane keeps auxiliary system and traffic as adaptive learning algorithms
The advanced driving assistance system development speeds such as obstruction auxiliary system become faster, while also accelerating the development speed of intelligent vehicle.
Intelligent vehicle needs the brain of people can be replaced independently to make reasonable behavior according to dynamic running environment around
Decision, and the limbs for completing by executing agency people complete corresponding decision behavior.Therefore, the accurate decision row of intelligent vehicle
To be critical core technology, and in view of the decision behavior of driver characteristics is more capable of the driving comfort of extra-high occupant
Property, the brain of whole system can be known as.So class people's lane change decision is to ensure that intelligent vehicle is exercised safely in the road
Key technology, so being of great significance to the research of intelligent vehicle class people's lane change decision problem.
Summary of the invention
In order to solve problem above of the existing technology, the present invention provides a kind of intelligence based on driver behavior pattern
Vehicle class people's lane change decision-making technique extracts the parameter of characterization driver's steering characteristic, so by establishing the experiment of driver's steering characteristic
Driver characteristics identifier is established using the method for K-means cluster and BP neural network afterwards, according to driver characteristics identifier
What the driver characteristics and sensor recognized identified itself carries out class people lane change with surrounding vehicles state and environmental information
Decision finally makes the lane change behavior of intelligent vehicle have the driving characteristics of class people.
The present invention is achieved by the following technical solutions:
A kind of intelligent vehicle class people's lane change decision-making technique based on driver behavior pattern comprising the steps of:
Step 1: establishing the experiment of driver's steering characteristic:
Emulation traffic scene is established in Carsim;Several drivers are found as experimenter, and by experimenter
It is divided into training group and test group;
Step 2: establishing driver characteristics classifier using the method that K-means is clustered:
Using the experimental data of experimenter each in the training set as cluster sample, two cluster centres are selected, and
The method clustered by K-means, is divided into two classes for experimental data, respectively radical type and conservative;
Step 3: establishing driver characteristics identifier using the method for BP neural network:
Training set of the experimental data as BP neural network of classification marker, training BP mind will be completed in the step 2
Through network;For the BP neural network that training is completed, BP neural network is surveyed using the experimental data in the test set
Examination;The BP neural network driver characteristics identifier that test passes through is used for the driving style of online recognition driver;
Step 4: establishing lane change wish identifier, the identification of driver's driving style is carried out;
Step 5: establishing lane change Interval selecting device, and for different driving style judgements and selection lane change space.
A kind of intelligent vehicle class people's lane change decision-making technique based on driver behavior pattern, to experiment number in step 1
Include following procedure according to processing is carried out:
The steering wheel angle δ and speed v in multiple turning points in experiment road are extracted respectivelytTo obtain multiple groups experiment number
According to;Direction disk rotating speed v is obtained in every group of dataδ, steering wheel angle standard deviation sigmaδWith tri- characteristic values of average speed v when turning to;
Then the corresponding three groups of data of three characteristic values are normalized.
A kind of intelligent vehicle class people's lane change decision-making technique based on driver behavior pattern, step 4 establish lane change meaning
It is willing to identifier, carrying out the identification of driver's driving style includes following procedure:
According to driver's driving style recognition result of the step 3, to the different driving styles in experimenter, if
Set different desired speed vxdesWith weight coefficient α, ω1, calculate in conjunction with above three variable element by average speed efficiency Ulv,
Average time efficiency UltgWith link length efficiency UldThe road whole effect U of compositionl, and compare the whole effect in each lane, such as
The whole effect U in fruit side lane1Greater than this lane whole effect, then assert that current vehicle has the expectation of the side lane.
A kind of intelligent vehicle class people's lane change decision-making technique based on driver behavior pattern, road whole effect UlBy with
Lower formula indicates:
Wherein, UlIt can be imitated for road, s;
w1、w2And w3It is the weight coefficient of three lane change indexs respectively;
NlvFor the regular factor of average speed efficiency:
NltgFor the regular factor of average time efficiency: Nltg=α tgdes
NldFor the regular factor of link length efficiency:
A kind of intelligent vehicle class people's lane change decision-making technique based on driver behavior pattern, step 5 are established between lane change
Gap selector, and include following procedure for different driving style judgements and selection lane change space:
When the step 4 obtains the signal of expectation lane change, in conjunction with the driving style identified in the step 3, pass through
Calculate longitudinal vehicle spacing AdWith front and back vehicle and this vehicle distance df,drSuitable lane change space is selected, A is worked asd> ALAnd df,dr
> dsWhen,
Existing can lane change space;Wherein, ALFor allow lane change longitudinal safe distance, for different driving style AL's
Value is different;dsFor allow lane change two shop safety distances, for different driving style dsValue it is different;For existing
Can lane change space, while different lane change spaces is selected according to different driving styles.
Due to the adoption of the above technical solution, the beneficial effects of the present invention are:
1, for different driver characteristics, different drive parameter (the desired speed v of selectionxdes, weight coefficient α,
ω1, longitudinal safe distance ALWith two shop safety distance ds), so that intelligent vehicle highway carry out lane-change during,
Can more class peopleization, can be further improved riding comfort in this way;
2, the present invention provides a kind of new lane change Interval selecting devices and lane change wish identifier to together form intelligent vehicle
Decision link makes in intelligent vehicle lane change decision process more safety and more class peopleization.
Detailed description of the invention
Fig. 1 is experiment road Road Course path profile;
Fig. 2 is the partial data figure for recording some driver of personnel record;
Fig. 3 is experiment path;
Fig. 4 is driver's steering characteristic classification after K-means cluster;
Fig. 5 is neural network structure figure;
Fig. 6 is the classification results for the neural network that verifying collection is completed by training;
Fig. 7 is simulation freeway traffic scene;
Fig. 8 is lane change Interval selecting strategy;
Fig. 9 is lane change safety clearance calculation amount schematic diagram;
Figure 10 is radical type driver lane change safety clearance schematic diagram;
Figure 11 is conservative driver lane change safety clearance schematic diagram.
Specific embodiment
With reference to the accompanying drawing, technical solution proposed by the invention is further elaborated and is illustrated.
Intelligent vehicle class people's lane change decision-making technique based on driver behavior pattern that the present invention provides a kind of, including it is following several
A step:
1, the experiment of driver's steering characteristic is established:
1.1, operating condition design
Emulation traffic scene is established in Carsim, including lane shape, size design, size design mainly include road
Width and the design of lane width and ambient enviroment, such as house, trees etc..
Since the driving simulator that this experiment uses is established based on Carsim simulation software, so what this was tested
Experiment scene is also to establish in Carsim.In view of behaviors more during lane change are turning behaviors, so in reality
Test the road Road Course of different turning scenes, the path such as Fig. 1 institute comprising there are many for having chosen and carrying in Carsim
Show.In view of highway is our landing scene, so, it is two lane road of individual event by lane design when lane is arranged
Face, and 8m is set by road width, wherein bicycle road width is 4m.In order to make experiment scene more true and reliable, Wo Men
It joined lawn, trees, house and other vehicles etc. in experiment scene, and these objects of reference can provide speed for driver
Spend reference.
1.2, experimenter is found
Searching possesses driver's license, and several drivers that driving age and age and male to female ratio are evenly distributed are as this
The experimenter of secondary experiment, and the experimenter selected is divided into two groups, respectively training group and test group.
21 drivers are found in this experimental design, wherein 17 drivers are training set sample, as neural network system
The training set of system, what for this 17 drivers, we were chosen is to possess the adult of motor vehicle driving license as experiment people
Member, driving age are distributed in 0-6, and age distribution is between 19 to 38 years old, and male to female ratio is in 2:1.Training set experimenter sample is such as
Shown in table 1.
1 training set experimenter's sample of table
It is left 4 drivers, is selected as test set sample, as the accuracy for removing verifying nerve network system.This four
Name driver also possesses motor vehicle driving license, and test set experimenter's sample is as shown in table 2.
2 test set experimenter's sample of table
1.3, experimentation and data acquisition
Experimenter successively tests on driving simulator, and the experimental data that transmission comes is saved and recorded.
It first can be before formally starting experiment, it is desirable that experimenter's fill message acquisition tables, to understand the base of driver
This information and drive routine habit.Then experimenter can be told first to carry out a circle test ride, allow driver to feel in this process
By with adapt to driving simulator and traffic environment, test ride one is linked up after enclosing with experimenter, it is ensured that experimenter's state prepares out
Begin the second circle (formal experiment), and acquires driving data by record personnel.The partial data for recording personnel record is as shown in Figure 2.
1.4, data processing
The experimental data of acquisition is handled, nine turning points chosen in experiment road extract nine turning points respectively
In steering wheel angle (δ) and speed (vt), to obtain nine groups of data of each driver, data are further located
It manages, obtains 3 characteristic values in every group of data, be direction disk rotating speed (v respectivelyδ), steering wheel angle standard deviation (σδ) and turn to when
Average speed (v).Then the corresponding three groups of data of three characteristic values are normalized.
The data that experiment obtains are handled, the parametric direction disk rotating speed that can characterize driver's steering characteristic is extracted
(vδ), steering wheel angle standard deviation (σδ) and turn to when average speed (v).
By taking an experimenter as an example, three kinds of data are extracted in the data first in all records about the experimenter,
Respectively time (t), steering wheel angle (δ) and speed (v);Then, these three data are being intercepted according to identical interval
9 sections therein, as shown in figure 3,9 effective turning sections are chosen in figure in thick solid curve Duan Weicong experiment path as every
9 groups of sample points of driver;Following three data finally are asked to each group of sample point according to formula (1)-(3), are direction respectively
Disk rotating speed (vδ), steering wheel angle standard deviation (σδ) and turn to when average speed (v).It is equal to the data of each experimenter
Carry out above-mentioned processing.153 groups of training samples are obtained in training set, and 36 groups of test samples are obtained in test set.
Wherein, vδFor direction disk rotating speed, °/s;δmaxFor in steering procedure, (finger ray disk corner is from zero to maximum value
In one section of turning process to zero, as shown in the thick solid curve in Fig. 3) in draw drum corner maximum value;tmaxFor from start turn
Reach maximum value (steering wheel angle δ for the first time to (steering wheel angle 0) to steering wheel anglemax) used in time;σδFor side
To disk corner standard deviation, °;N is data amount check;I is cyclic variable, i ∈ (1, n);δiIt is each in the data group of n for capacity
A steering wheel angle, i ∈ (1, n), °;It is the average value of the steering wheel angle in the data group of n for capacity, °;V is primary turns
To average speed in the process, m/s;viEach of the data group for being n for capacity speed, i ∈ (1, n), m/s;
2, driver characteristics classifier is established using the method that K-means is clustered
Using the experimental data of each driver in the training set obtained in the step 1.4) as cluster sample, two are selected
A cluster centre, and by K-means cluster method, experimental data is divided into two classes, respectively radical type (idx=1) and
Conservative (idx=2).
For 153 groups of data of 17 experimenters after data processing, cluster input is u=[vδ,σδ, v], in order to full
Classification (radical type, conservative) of the foot to experimenter, selects then two cluster centres, respectively represents radical type driver (idx=
1) and conservative driver (idx=2), therefore to cluster output be y=idx.As shown in formula (4), chooses Euclidean distance formula and make
For the measurement for measuring characterization driver's steering characteristic similitude, as cluster state variable, as shown in formula (5), it is excellent for choosing J
Change target,
Wherein A, B respectively indicate two number column matrix;N, m points are A, the line number of B matrix;I is cyclic variable, i ∈ (1,
n);J is cyclic variable, j ∈ (1, m);J is optimization aim;Γ1、Γ2Respectively weight coefficient;X (i) is to include 153 groups of data
Matrix sequence;C1, C2Respectively two cluster centre matrix sequences.
Experimenter's type after cluster is as shown in table 3, because having nine sample points for each experimenter,
It, can be by calculating nine sample points of the driver in conservative when carrying out turning to the identification of behavioral trait type to an experimenter
With the ratio occupied in radical type, it is assumed that the type of the driver when ratio is greater than 50%, such as driver 1,
It is 55.6% that have 5 sample points, which be conservative proportion, so can assert that driver 1 is conservative.Cluster result such as Fig. 4 institute
Show, fork-shaped point represents the data sample of radical type, and black circle represents the data sample of conservative, and two solid five-pointed stars represent
The cluster centre of two classifications.Cluster centre be respectively [0.1728,0.1729,0.6311] and [0.1030,0.1389,
0.3306]。
3 experimenter of table turns to behavioral trait type
3, driver characteristics identifier is established using the method for BP neural network
Training set of the experimental data as BP neural network of classification marker, training BP mind will be completed in the step 2
Through network;For the BP neural network that training is completed, use the experimental data in test set as test set to the BP nerve
Network is tested;For the BP neural network driver characteristics identifier that test passes through, it is used for online recognition driver
Driving style.
In view of discontinuous function that may be present, therefore select design containing there are two the BP neural network of hidden layer, considerations
To three driver's steering characteristic identification parameters, input layer number, n are choseni=3, output layer node number is no=1, root
It is calculated according to formula (6) and obtains node number range, selecting the first hidden node number by debugging is 5, the node of the second hidden layer
Number is 3, and neural network structure figure is as shown in Figure 5.It (includes v that the sorted 153 groups of data of classifier in 2, which will be passed through,δ,σδ,v,
Idx) as the training set of BP neural network, i.e. train_set=[vδ,σδ, v, idx], optimization aim is used as according to formula (7)
Training is optimized to BP neural network.Output function after trainingAs shown in formula (8), each layer of function is such as public
Each layer of neural network of weight coefficient shown in formula (9), after trainingNerve as shown in formula (10-12), after training
Shown in the deviation that each layer of network such as formula (13-15).
θ(3)=[- 0.7857 5.0422 0.8290] (12)
b(2)=[- 2.4343 0.2468-0.2453 2.5962-1.0204]T (13)
b(3)=[- 2.0462-0.0586-1.8793]T (14)
b(4)=0.7464 (15)
Wherein, m is the number of nodes of hidden layer 1 and hidden layer 2;miFor input layer number;moFor output layer number of nodes;A be
Number, a ∈ [1,10];J (θ, b) is the optimization aim of neural network;N is number of samples;I is cyclic variable, i ∈ (1, n);y(i)
For output sequence, y(i)=idx;It is arbitrary value for output function;u(i)For input sample u(i)=[vδ,σδ,v];λ is
Weight factor, λ=1;L is which layer, l ∈ [1, L-1];L is the total number of plies of neural network, L=4;N is number of samples;K is circulation
Variable, k ∈ (1, Sl);SlFor dimensionless book, the S as l=1l=3, the S as l=2l=5, the S as l=3l=3;J is that circulation becomes
Amount, j ∈ (1, Sl+1);Sl+1For dimensionless book, the S as l=1l+1=5, the S as l=2l+1=3, the S as l=3l+1=1;
For the weight coefficient of neural network, such as formula (shown in 10-12), searching method is to find homography by the i in the upper right corner, so
Corresponding element is found according to the jk in the lower right corner afterwards, wherein j is abscissa, and k is ordinate;For each layer of function, it is
Arbitrary value;U is input matrix u=[vδ,σδ,v];θ(1)、θ(2)And θ(3)Respectively to the weight of hidden layer 1, hidden layer 1 arrives input layer
The weight and hidden layer 2 of hidden layer 2 arrive the weight of output layer;b(2)、b(3)And b(4)It is hidden layer 1, the deviation of hidden layer 2 and output layer respectively.
Then the 36 groups of data test_set=[v concentrated with verifyingδ,σδ, v] BP neural network that training is completed is carried out
Test, verifying collection personnel essential information are as shown in table 4.Verifying collection classification results are as shown in fig. 6, driver characteristics identifier recognizes
The driving style of four drivers out is as shown in table 5, and from test result, it appear that, the driving performance of four drivers can be just
True picking out comes.Illustrate we design driver characteristics identifier be can be available.
Table 4 verifies collection personnel Basic Information Table
Number | Gender | Driving age | Age | Type |
1 | Male | 1 | 25 | It is radical |
2 | Male | 0.5 | 29 | It is conservative |
3 | Male | 5 | 33 | It is radical |
4 | Male | 2 | 27 | It is conservative |
The steering behavioral trait type of the verifying collection personnel of table 5
4, lane change wish identifier
To the different driving styles of experimenter, different desired speed v is setxdesWith weight coefficient α, ω1, in conjunction with upper
Three variable elements are stated to calculate by three major parameters, i.e. average speed efficiency Ulv, average time efficiency UltgIt is imitated with link length
It can UldThe road whole effect U of compositionl, and compare the whole effect in each lane, if the whole effect U in certain side lane1Greater than this
Lane whole effect U0, then assert that current vehicle has the expectation of the side lane.
In this step, intelligent vehicle is mainly allowed to be decided whether according to itself with surrounding vehicles and environmental information as people
There is the wish of lane change.Assuming that sensor can perceive itself surrounding vehicles information and environmental information and be transferred to after treatment
Lane change wish identifier uses.Consider that lane change wish is based primarily upon the difference of two adjacent lanes, in two road not
It can mainly be summarized as at 3 points with putting, be speed first, the average speed in adjacent two lane is different;Followed by vehicle
Between when away from the average following distance in adjacent two lane is also different;It is finally road with the presence or absence of the end.These three are all two
The difference in a lane, while being also main three elements in need of consideration during lane change.
4.1, average speed efficiency
In view of adjacent lane speed is not all a lane change index, we with reference to average speed efficiency concept, such as
Shown in formula (16).
Wherein, UlvFor average speed efficiency, s;dmaxDistance is sailed for road maximum feasible, is definite value dmax=6000m;vxdes
For driver's desired speed, m/s;γ is the constant for making equation denominator be not zero, γ=5m/s;vlμFor the average speed of road,
m/s。
It can see from formula (16), as the average speed v of roadlμCloser to the desired speed v of driverxdesWhen,
The lane change of driver thirst for it is higher, it is on the contrary then the serious hope of lane change is smaller.
So being directed to the driver of different driving styles, the driver of each type has the desired speed of oneself,
So being directed to the driver of different driving styles, different desired speeds is had chosen, height can be selected for the driver of radical type
Speed, and the speed low relative to radical type driver can be selected for conservative driver.For the verifying to design effect,
In the present invention, for the driver of radical type, the desired speed v that we selectxdes=30m/s, and driving for conservative
The person of sailing, the desired speed that we select are vxdes=20m/s.
4.2, average time efficiency
In view of adjacent lane headway is not all a lane change index, we are general with reference to average time efficiency
It reads, as shown in formula (17).
Ultg=min (α tgdes,tglμ) (17)
Wherein, UltgFor average time efficiency, s;α is headway weight coefficient;tgdesFor desired headway;tglμ
It is averaged headway for road.
From formula (17) it can be seen that, road be averaged headway be can be increased, when it reach it is desirable that
When headway, the maximum value of average time efficiency is also just reached, has also just reached us and most it is expected the effect of lane change.
So being directed to the driver of different driving styles, the driver of each type has the headway of oneself, institute
To be directed to the driver of different driving styles, different headway weight coefficients, i.e., different α, for radical type are had chosen
Driver can select relatively small headway coefficient, and the driver of conservative can be selected to drive relative to radical type
The big headway coefficient of member.In the verifying to design effect, in the present invention, for the driver of radical type, Wo Menxuan
The expectation headway coefficient selected is 1;And for the driver of conservative, the expectation headway coefficient that we select is 2.
4.3, link length efficiency
In view of the adjacent dead end street that whether there is is a lane change index, we are general with reference to link length efficiency
It reads, as shown in formula (18).
Wherein, UldFor average time efficiency, s;dminFor in view of dead end street, vehicle allows the minimum range travelled.
It can be seen that, the distance of dead end street is smaller, then link length efficiency is smaller from formula (18), and opposite road is most
The distance of head is bigger, then link length efficiency is bigger, is more suitable vehicle driving, also increases the lane change expectation of this vehicle.
4.4, road whole effect
In view of above three lane change index, we can imitate with reference to road, as shown in formula (19).When adjacent lane
Total energy effect UlGreater than the U in this lanel, then the wish of the oriented adjacent lane lane change of vehicle.
Wherein, UlIt can be imitated for road, s;w1、w2And w3It is the weight coefficient of three lane change indexs respectively;NlvFor average speed
Shown in the regular factor such as formula (20) for spending efficiency;NltgShown in regular factor such as formula (21) for average time efficiency; Nld
Shown in regular factor such as formula (22) for link length efficiency.
Nltg=α tgdes (21)
For formula (19), for the driver of different driving styles, it is contemplated that speed is in headway and desired speed
In all have a great impact, so be directed to different driver's types, selected different weight coefficient w1, for radical type
Driver w1Numerical value can be relatively smaller, the w for conservative driver1Numerical value can be big relative to radical type driver
It is some.In the verifying to design effect, in the present invention, when average speed is less than or equal to desired speed, i.e. vlμ≤vxdes,
For the driver of radical type, we select w1It is 1;And for the driver of conservative, we select w1It is 1.5;When average vehicle
When speed is greater than desired speed, i.e. vlμ> vxdes, for the driver of radical type, we select w1It is 5;And driving for conservative
The person of sailing, we select w1It is 10.
As shown in fig. 7, road is wherein unidirectional double lane, and vehicle is even for the experiment scene of simulation highway
Speed traveling, E are this vehicle, S1, S2And S3For other vehicles of surrounding.And for time traffic scene, conservative and radical type phase are calculated
Adjacent lane is in different tglμAnd vlμWhen road whole effect, such as table 6, shown in table 7.
6 conservative lane change wish table of table
The radical type lane change wish table of table 7
In table 6 and 7, the 0.59 of the upper right corner and 0.47 respectively this lane efficiency value of conservative and radical type can be seen
Out, this lane efficiency value of radical type is lower under identical operating condition, then sees the efficiency value in adjacent lane, can be seen that in two tables
Come, the efficiency value in conservative neighbour lane is greater than from the above two o'clock for the efficiency value in radical type neighbour lane, it can be seen that come,
Under the same conditions, radical type is more inclined to lane change than conservative.
5, lane change Interval selecting device is established
It is illustrated in figure 8 lane change Interval selecting strategy.
Firstly, i.e. Change=1, accounting, which is calculated, whether there is enough lane change space motion sheets when obtaining lane-change wish
Vehicle vehicle lane change, as shown in figure 9, calculating separately the vehicle headway A in the horizontal direction x according to formula (23), (24) and (25)dWith
Point arrives the distance d of vehicle centroid before rear car is right when lane change2With the distance d of the left back point of front truck when lane change to vehicle centroid1。
Wherein, AdFor the vehicle headway in the horizontal direction x, m;tcFor lane change time, s;For S1Speed in the direction x,
m/s;S when starting for lane change1The centroid position abscissa of vehicle, m;L is vehicle commander, m;vxESpeed for E in the direction x, m/s;d1
Distance of the left back point of front truck to vehicle centroid when for lane change;For S1Speed in the direction y, m/s;S when starting for lane change1
The centroid position ordinate of vehicle, m;W is vehicle width, m;vyESpeed for E in the direction y, m/s;d2Point arrives before rear car is right when for lane change
The distance of vehicle centroid;For S2Speed in the direction y, m/s;S when starting for lane change2The centroid position ordinate of vehicle,
m;For S2Speed in the direction x, m/s;S when starting for lane change2The centroid position abscissa of vehicle, m;
Then, judge A during lane changed> 1, and d1> 1, d2When > 1, then it is assumed that it is empty to possess enough lane changes
Between, as shown in Figure 10,11, respectively under current working, the available lane change space of radical type and conservative.
Finally, when calculating available lane change space, it is different according to driver's type, different lane change spaces can be selected,
Position can be selected near the shortest lane change space of preceding running time radical type driver, and conservative driver can be selected
Select the maximum lane change space lane change in lane change space.Under the operating condition, radical type and conservative has and only a kind of lane change space,
Therefore directly select the lane change space.The case where for no lane change space, selection provide deceleration instruction waiting and calculate next time.
Claims (5)
1. a kind of intelligent vehicle class people's lane change decision-making technique based on driver behavior pattern, which is characterized in that comprise the steps of:
Step 1: establishing the experiment of driver's steering characteristic:
Emulation traffic scene is established in Carsim;Several drivers are found to be divided into as experimenter, and by experimenter
Training group and test group;
Step 2: establishing driver characteristics classifier using the method that K-means is clustered:
Using the experimental data of experimenter each in the training set as cluster sample, two cluster centres are selected, and pass through
The method of K-means cluster, is divided into two classes for experimental data, respectively radical type and conservative;
Step 3: establishing driver characteristics identifier using the method for BP neural network:
Training set of the experimental data as BP neural network of classification marker, training BP nerve net will be completed in the step 2
Network;For the BP neural network that training is completed, BP neural network is tested using the experimental data in the test set;It will
Test the driving style that the BP neural network driver characteristics identifier passed through is used for online recognition driver;
Step 4: establishing lane change wish identifier, the identification of driver's driving style is carried out;
Step 5: establishing lane change Interval selecting device, and for different driving style judgements and selection lane change space.
2. a kind of intelligent vehicle class people's lane change decision-making technique based on driver behavior pattern as described in claim 1, feature
It is, carrying out processing to experimental data in the step 1 includes following procedure:
The steering wheel angle δ and speed v in multiple turning points in experiment road are extracted respectivelytTo obtain multiple groups experimental data;Often
Direction disk rotating speed v is obtained in group dataδ, steering wheel angle standard deviation sigmaδWith tri- characteristic values of average speed v when turning to;Then
The corresponding three groups of data of three characteristic values are normalized.
3. a kind of intelligent vehicle class people's lane change decision-making technique based on driver behavior pattern as described in claim 1, feature
It is, the step 4 establishes lane change wish identifier, and carrying out the identification of driver's driving style includes following procedure:
According to driver's driving style recognition result of the step 3, to the different driving styles in experimenter, setting is not
Same desired speed vxdesWith weight coefficient α, ω1, calculate in conjunction with above three variable element by average speed efficiency Ulv, average
Time efficiency UltgWith link length efficiency UldThe road whole effect U of compositionl, and compare the whole effect in each lane, if certain
The whole effect U in side lane1Greater than this lane whole effect, then assert that current vehicle has the expectation of the side lane.
4. a kind of intelligent vehicle class people's lane change decision-making technique based on driver behavior pattern as claimed in claim 3, feature
It is, the road whole effect UlIt is indicated by following formula:
Wherein, UlIt can be imitated for road, s;
w1、w2And w3It is the weight coefficient of three lane change indexs respectively;
NlvFor the regular factor of average speed efficiency:
NltgFor the regular factor of average time efficiency: Nltg=α tgdes
NldFor the regular factor of link length efficiency:
5. a kind of intelligent vehicle class people's lane change decision-making technique based on driver behavior pattern as described in claim 1, feature
It is, the step 5 establishes lane change Interval selecting device, and includes for different driving style judgements and selection lane change space
Following procedure:
When the step 4 obtains the signal of expectation lane change, in conjunction with the driving style identified in the step 3, pass through calculating
Longitudinal vehicle spacing AdWith front and back vehicle and this vehicle distance df,drSuitable lane change space is selected, A is worked asd> ALAnd df,dr> ds
When, that is, existing can lane change space;Wherein, ALFor allow lane change longitudinal safe distance, for different driving style ALValue not
Together;dsFor allow lane change two shop safety distances, for different driving style dsValue it is different;For it is existing can lane change
Space, while different lane change spaces is selected according to different driving styles.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811195634.1A CN109367541B (en) | 2018-10-15 | 2018-10-15 | Intelligent vehicle-like person lane change decision-making method based on driver behavior characteristics |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811195634.1A CN109367541B (en) | 2018-10-15 | 2018-10-15 | Intelligent vehicle-like person lane change decision-making method based on driver behavior characteristics |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109367541A true CN109367541A (en) | 2019-02-22 |
CN109367541B CN109367541B (en) | 2020-12-25 |
Family
ID=65397738
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811195634.1A Active CN109367541B (en) | 2018-10-15 | 2018-10-15 | Intelligent vehicle-like person lane change decision-making method based on driver behavior characteristics |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109367541B (en) |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110171361A (en) * | 2019-06-17 | 2019-08-27 | 山东理工大学 | A kind of consideration driver mood and drive tendentious ride safety of automobile method for early warning |
CN110427850A (en) * | 2019-07-24 | 2019-11-08 | 中国科学院自动化研究所 | Driver's super expressway lane-changing intention prediction technique, system, device |
GB2578910A (en) * | 2018-11-13 | 2020-06-03 | Jaguar Land Rover Ltd | A controller for a vehicle |
CN111547064A (en) * | 2020-05-26 | 2020-08-18 | 吉林大学 | Driving style recognition and classification method for automobile adaptive cruise system |
WO2020224623A1 (en) * | 2019-05-08 | 2020-11-12 | Byton Limited | Deep neural network based driving assistance system |
CN112129290A (en) * | 2019-06-24 | 2020-12-25 | 罗伯特·博世有限公司 | System and method for monitoring riding equipment |
CN112693458A (en) * | 2021-01-15 | 2021-04-23 | 一汽解放汽车有限公司 | Cruise control method and device, vehicle and storage medium |
CN112785863A (en) * | 2020-12-26 | 2021-05-11 | 浙江天行健智能科技有限公司 | Merging decision classification early warning method based on K-Means and entropy weighting |
CN112896185A (en) * | 2021-01-25 | 2021-06-04 | 北京理工大学 | Intelligent driving behavior decision planning method and system for vehicle-road cooperation |
GB2590877A (en) * | 2018-11-13 | 2021-07-07 | Jaguar Land Rover Ltd | A controller for a vehicle |
CN113343892A (en) * | 2021-06-24 | 2021-09-03 | 东风汽车集团股份有限公司 | Vehicle line-following driving scene extraction method |
CN114771539A (en) * | 2022-06-16 | 2022-07-22 | 小米汽车科技有限公司 | Vehicle lane change decision method, device, storage medium and vehicle |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050131588A1 (en) * | 2003-12-16 | 2005-06-16 | Nissan Motor Co., Ltd. & Drexel University | Method and system for intention estimation and operation assistance |
US20130253797A1 (en) * | 2012-03-20 | 2013-09-26 | Toyota Motor Engineering & Manufacturing North America, Inc. (Tema) | Prediction of driver-specific cruise speed using dynamic modeling |
CN103842232A (en) * | 2011-10-04 | 2014-06-04 | 罗伯特·博世有限公司 | Method for activating a driver assistance system |
CN105912814A (en) * | 2016-05-05 | 2016-08-31 | 苏州京坤达汽车电子科技有限公司 | Lane change decision model of intelligent drive vehicle |
CN105946860A (en) * | 2016-05-20 | 2016-09-21 | 武汉理工大学 | Bend vehicle speed predicting method considering driving style |
CN107813820A (en) * | 2017-10-13 | 2018-03-20 | 江苏大学 | A kind of unmanned vehicle lane-change paths planning method for imitating outstanding driver |
-
2018
- 2018-10-15 CN CN201811195634.1A patent/CN109367541B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050131588A1 (en) * | 2003-12-16 | 2005-06-16 | Nissan Motor Co., Ltd. & Drexel University | Method and system for intention estimation and operation assistance |
CN103842232A (en) * | 2011-10-04 | 2014-06-04 | 罗伯特·博世有限公司 | Method for activating a driver assistance system |
US20130253797A1 (en) * | 2012-03-20 | 2013-09-26 | Toyota Motor Engineering & Manufacturing North America, Inc. (Tema) | Prediction of driver-specific cruise speed using dynamic modeling |
CN105912814A (en) * | 2016-05-05 | 2016-08-31 | 苏州京坤达汽车电子科技有限公司 | Lane change decision model of intelligent drive vehicle |
CN105946860A (en) * | 2016-05-20 | 2016-09-21 | 武汉理工大学 | Bend vehicle speed predicting method considering driving style |
CN107813820A (en) * | 2017-10-13 | 2018-03-20 | 江苏大学 | A kind of unmanned vehicle lane-change paths planning method for imitating outstanding driver |
Cited By (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
GB2590877B (en) * | 2018-11-13 | 2021-12-29 | Jaguar Land Rover Ltd | A controller for a vehicle |
GB2578910A (en) * | 2018-11-13 | 2020-06-03 | Jaguar Land Rover Ltd | A controller for a vehicle |
GB2578910B (en) * | 2018-11-13 | 2021-05-12 | Jaguar Land Rover Ltd | A controller for a vehicle |
GB2590877A (en) * | 2018-11-13 | 2021-07-07 | Jaguar Land Rover Ltd | A controller for a vehicle |
WO2020224623A1 (en) * | 2019-05-08 | 2020-11-12 | Byton Limited | Deep neural network based driving assistance system |
CN110171361B (en) * | 2019-06-17 | 2022-09-23 | 山东理工大学 | Automobile safety early warning method based on emotion and driving tendency of driver |
CN110171361A (en) * | 2019-06-17 | 2019-08-27 | 山东理工大学 | A kind of consideration driver mood and drive tendentious ride safety of automobile method for early warning |
CN112129290A (en) * | 2019-06-24 | 2020-12-25 | 罗伯特·博世有限公司 | System and method for monitoring riding equipment |
CN110427850A (en) * | 2019-07-24 | 2019-11-08 | 中国科学院自动化研究所 | Driver's super expressway lane-changing intention prediction technique, system, device |
CN111547064A (en) * | 2020-05-26 | 2020-08-18 | 吉林大学 | Driving style recognition and classification method for automobile adaptive cruise system |
CN111547064B (en) * | 2020-05-26 | 2022-07-12 | 吉林大学 | Driving style recognition and classification method for automobile adaptive cruise system |
CN112785863A (en) * | 2020-12-26 | 2021-05-11 | 浙江天行健智能科技有限公司 | Merging decision classification early warning method based on K-Means and entropy weighting |
CN112785863B (en) * | 2020-12-26 | 2022-05-03 | 浙江天行健智能科技有限公司 | Merging decision classification early warning method based on K-Means and entropy weighting |
CN112693458A (en) * | 2021-01-15 | 2021-04-23 | 一汽解放汽车有限公司 | Cruise control method and device, vehicle and storage medium |
CN112896185A (en) * | 2021-01-25 | 2021-06-04 | 北京理工大学 | Intelligent driving behavior decision planning method and system for vehicle-road cooperation |
CN113343892A (en) * | 2021-06-24 | 2021-09-03 | 东风汽车集团股份有限公司 | Vehicle line-following driving scene extraction method |
CN114771539B (en) * | 2022-06-16 | 2023-02-28 | 小米汽车科技有限公司 | Vehicle lane change decision method and device, storage medium and vehicle |
CN114771539A (en) * | 2022-06-16 | 2022-07-22 | 小米汽车科技有限公司 | Vehicle lane change decision method, device, storage medium and vehicle |
Also Published As
Publication number | Publication date |
---|---|
CN109367541B (en) | 2020-12-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109367541A (en) | A kind of intelligent vehicle class people's lane change decision-making technique based on driver behavior pattern | |
US11084494B2 (en) | Method for detecting safety of driving behavior, apparatus, device and storage medium | |
CN106778583B (en) | Vehicle attribute identification method and device based on convolutional neural network | |
CN108313054B (en) | The autonomous lane-change decision-making technique of automatic Pilot and device and automatic driving vehicle | |
CN107169567B (en) | Method and device for generating decision network model for automatic vehicle driving | |
CN106203330A (en) | A kind of vehicle classification method based on convolutional neural networks | |
CN103871242B (en) | A kind of driving behavior overall evaluation system and method | |
CN109145939A (en) | A kind of binary channels convolutional neural networks semantic segmentation method of Small object sensitivity | |
CN109447033A (en) | Vehicle front obstacle detection method based on YOLO | |
CN107229973A (en) | The generation method and device of a kind of tactful network model for Vehicular automatic driving | |
Wang et al. | Unsupervised learning of object semantic parts from internal states of cnns by population encoding | |
Wang et al. | Evaluation of lane departure correction systems using a regenerative stochastic driver model | |
KR102313123B1 (en) | Method and device for detecting vehicle occupancy using passenger's keypoint detected through image analysis for humans' status recognition | |
CN110378210A (en) | A kind of vehicle and car plate detection based on lightweight YOLOv3 and long short focus merge distance measuring method | |
CN111301404B (en) | Vehicle control method and device, storage medium and processor | |
CN106845387A (en) | Pedestrian detection method based on self study | |
CN106203296A (en) | The video actions recognition methods of one attribute auxiliary | |
DE102020122837A1 (en) | METHOD AND DEVICE FOR CREATING A TURNING PATH IN A MULTI-SHIFT LEARNING-BASED AUTONOMOUS VEHICLE | |
JP2020061155A (en) | Sdl model introduced image extraction method | |
CN112987711A (en) | Optimization method of automatic driving regulation algorithm and simulation testing device | |
CN112381101A (en) | Infrared road scene segmentation method based on category prototype regression | |
DE102023107233A1 (en) | SELECTION OF VEHICLE ACTIONS BASED ON SIMULATED CONDITIONS | |
Koenig et al. | Bridging the gap between open loop tests and statistical validation for highly automated driving | |
CN112052829B (en) | Pilot behavior monitoring method based on deep learning | |
CN112966334A (en) | Method, system and device for generating traffic scene attribute label |
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 |