CN107310550B - Road vehicles travel control method and device - Google Patents
Road vehicles travel control method and device Download PDFInfo
- Publication number
- CN107310550B CN107310550B CN201610270123.6A CN201610270123A CN107310550B CN 107310550 B CN107310550 B CN 107310550B CN 201610270123 A CN201610270123 A CN 201610270123A CN 107310550 B CN107310550 B CN 107310550B
- Authority
- CN
- China
- Prior art keywords
- road
- parameter
- graph model
- factor graph
- prediction
- 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
- 238000000034 method Methods 0.000 title claims abstract description 48
- 230000008447 perception Effects 0.000 claims abstract description 38
- 230000006870 function Effects 0.000 claims abstract description 24
- 238000012549 training Methods 0.000 claims description 21
- 238000001514 detection method Methods 0.000 claims description 9
- 238000012360 testing method Methods 0.000 claims description 7
- 238000004590 computer program Methods 0.000 claims description 5
- 238000000926 separation method Methods 0.000 claims description 4
- 230000004048 modification Effects 0.000 claims description 3
- 238000012986 modification Methods 0.000 claims description 3
- 238000007689 inspection Methods 0.000 claims 2
- 230000008569 process Effects 0.000 description 10
- 238000004422 calculation algorithm Methods 0.000 description 9
- 238000010586 diagram Methods 0.000 description 9
- 239000011159 matrix material Substances 0.000 description 6
- 230000006798 recombination Effects 0.000 description 6
- 238000005215 recombination Methods 0.000 description 6
- 230000001419 dependent effect Effects 0.000 description 5
- 239000002131 composite material Substances 0.000 description 4
- 235000013399 edible fruits Nutrition 0.000 description 3
- 238000010606 normalization Methods 0.000 description 3
- 230000003068 static effect Effects 0.000 description 3
- 230000004888 barrier function Effects 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 230000011218 segmentation Effects 0.000 description 2
- 230000001568 sexual effect Effects 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 238000007476 Maximum Likelihood Methods 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000005315 distribution function Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000002922 simulated annealing Methods 0.000 description 1
- 238000002948 stochastic simulation Methods 0.000 description 1
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
- B60W30/00—Purposes of road vehicle drive control systems not related to the control of a particular sub-unit, e.g. of systems using conjoint control of vehicle sub-units
-
- 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
- B60W40/02—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 ambient conditions
-
- 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
-
- 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
- B60W2050/0062—Adapting control system settings
- B60W2050/0075—Automatic parameter input, automatic initialising or calibrating means
-
- 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
- B60W2420/00—Indexing codes relating to the type of sensors based on the principle of their operation
- B60W2420/40—Photo, light or radio wave sensitive means, e.g. infrared sensors
- B60W2420/403—Image sensing, e.g. optical camera
-
- 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
- B60W2420/00—Indexing codes relating to the type of sensors based on the principle of their operation
- B60W2420/40—Photo, light or radio wave sensitive means, e.g. infrared sensors
- B60W2420/408—Radar; Laser, e.g. lidar
-
- 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
- B60W2555/00—Input parameters relating to exterior conditions, not covered by groups B60W2552/00, B60W2554/00
- B60W2555/20—Ambient conditions, e.g. wind or rain
-
- 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
- B60W2556/00—Input parameters relating to data
- B60W2556/45—External transmission of data to or from the vehicle
Landscapes
- Engineering & Computer Science (AREA)
- Automation & Control Theory (AREA)
- Transportation (AREA)
- Mechanical Engineering (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Traffic Control Systems (AREA)
Abstract
The present invention relates to a kind of road vehicles travel control method and devices, this method comprises: obtaining the real-time perception result of road traffic environment；The corresponding present road element parameter of various roads element is obtained according to the real-time perception result；Road scene parameter is predicted according to factor graph model and the present road element parameter；The factor graph model is used to express the Joint Distribution probability density function between road scene parameter and the corresponding road element parameter of various roads element；Traveling control instruction is generated according to the road scene parameter of prediction and is exported.Road vehicles travel control method and device provided by the invention can accurately control road vehicles traveling.
Description
Technical field
The present invention relates to Traffic Information technical fields, more particularly to traveling control field more particularly to a kind of road
Road vehicle travels control method and device.
Background technique
With universal and Internet technology the development of road vehicles, road vehicles traveling control becomes one
Kind new demand can be used for navigation hint or unmanned etc..The control of road vehicles traveling mainly passes through meter at present
Calculation machine visual identity goes out barrier, and based on road information planning travelling line documented by map datum, thus according to planning
Traffic route carry out traveling control.
However, current road vehicles travel control mode, dependent on road information documented by map datum, and
The update of map datum is lag, can not be timely updated when road information changes, this is resulted in can not be accurately
Carry out traveling control；But also driving safety problem can be brought, problem is even more serious when being particularly applied to unmanned.
Summary of the invention
Based on this, it is necessary to cause to travel dependent on map datum for current road vehicles traveling control mode
The problem for controlling inaccuracy provides a kind of road vehicles travel control method and device.
A kind of road vehicles travel control method, which comprises
Obtain the real-time perception result of road traffic environment；
The corresponding present road element parameter of various roads element is obtained according to the real-time perception result；
Road scene parameter is predicted according to factor graph model and the present road element parameter；The factor graph model is used
Joint Distribution probability density function between expression road scene parameter and the corresponding road element parameter of various roads element；
Traveling control instruction is generated according to the road scene parameter of prediction and is exported.
A kind of road vehicles travel controlling system, described device include:
Data acquisition module, for obtaining the real-time perception result of road traffic environment；According to the real-time perception result
Obtain the corresponding present road element parameter of various roads element；
Prediction module, for predicting road scene parameter according to factor graph model and the present road element parameter；Institute
Factor graph model is stated for expressing the Joint Distribution between road scene parameter and the corresponding road element parameter of various roads element
Probability density function；
Output module, for generating traveling control instruction according to the road scene parameter of prediction and exporting.
Above-mentioned road vehicles travel control method and device utilize the present road element parameter of real-time perception result
It predicts road scene parameter, to carry out traveling control according to the road scene parameter of prediction, is no longer dependent on map number
According to the accuracy of vehicle travels control is enhanced.Moreover, expressing road scene parameter and a variety of roads by factor graph
Joint Distribution probability density function between the corresponding road element parameter of road element, factor graph describe global object letter with graph structure
The global operations of huge construction program are divided into simple local operation by several factorization forms, improve the efficiency of traveling control.
Furthermore various roads element can be merged by factor graph model, it may be considered that the correlation between road element parameter,
Without assuming that independently of one another, so that vehicle travels control meets real roads scene, precise control.
Detailed description of the invention
Fig. 1 is the applied environment figure of road vehicles drive-control system in one embodiment；
Fig. 2 is the schematic diagram of internal structure of electronic equipment in one embodiment；
Fig. 3 is the flow diagram of road vehicles travel control method in one embodiment；
The flow diagram for the step of Fig. 4 is training factor graph model in one embodiment；
Fig. 5 is the process schematic for converting probability of recombination graph model in one embodiment factor graph model；
Fig. 6 is modeling, study and the reasoning process schematic diagram of factor graph model in one embodiment；
Fig. 7 is the structure learning process schematic diagram of factor graph model in one embodiment；
Fig. 8 is the structural block diagram of road vehicles travel controlling system in one embodiment；
Fig. 9 is the structural block diagram of road vehicles travel controlling system in another embodiment.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.
Fig. 1 is the applied environment figure of road vehicles drive-control system in one embodiment, which includes perception
Equipment 110, electronic equipment 120 and road vehicles 130.Wherein awareness apparatus 110 can be camera or radar；Electronics
Equipment 120, which can be placed in inside the vehicles 130, directly controls the vehicles 130, is also possible to through network remote to traffic
Tool 130 sends traveling control instruction.Road vehicles 130 are can be along the vehicles of road driving, such as car, goods
Vehicle, car, motorcycle or electric bicycle etc..
Fig. 2 is the schematic diagram of internal structure of electronic equipment 120 in one embodiment.As shown in Fig. 2, the electronic equipment includes
Processor, non-volatile memory medium, built-in storage and the network interface connected by system bus.Wherein, electronic equipment
Non-volatile memory medium is stored with operating system, further includes a kind of road vehicles travel controlling system, the road traffic
Tool travel controlling system is for realizing a kind of road vehicles travel control method.The processor of electronic equipment is for providing
Calculating and control ability, support the operation of electronic equipment.The built-in storage of electronic equipment is the road in non-volatile memory medium
Road vehicle travels control device provides running environment.Computer-readable instruction can be stored in the built-in storage, the calculating
When machine readable instruction is executed by processor, processor may make to execute a kind of road vehicles travel control method.Network connects
Mouth is for being connected to network.The electronic equipment can be mobile phone, vehicle-mounted computer or server etc..Those skilled in the art can be with
Understand, structure shown in Figure 2, only the block diagram of part-structure relevant to application scheme, is not constituted to the application
The restriction for the electronic equipment that scheme is applied thereon, specific electronic equipment may include more more or fewer than as shown in the figure
Component perhaps combines certain components or with different component layouts.
As shown in figure 3, in one embodiment, providing a kind of road vehicles travel control method, the present embodiment
It is applied to the electronic equipment 120 in above-mentioned Fig. 1 and Fig. 2 in this way to illustrate.This method specifically comprises the following steps:
Step 302, the real-time perception result of road traffic environment is obtained.
Road traffic environment refers to road vehicles locating environment when driving, and real-time perception result is then to pass through perception
The data for being used to describe the road element in road traffic environment of equipment real-time perception.Electronic equipment can be adopted by awareness apparatus
The image for collecting road traffic environment carries out semantic segmentation to road traffic environment, distinguishes different kinds of roads element, felt in real time
Know result.
Real-time perception result includes road information, can also include road signs information, information of vehicles, pedestrian information and
At least one of scene light stream etc..Wherein, road information may include lane line information and road vanishing point information.Lane line is
It is indicated on the road surface of road for the linear of the indicative regular shape of traveling to road vehicles, lane line information such as vehicle
Road line position or lane line type or the affiliated road of lane line.Road vanishing point is the point that road disappears at a distance, road vanishing point with
Upper is usually sky portion, road belonging to road vanishing point information such as road vanishing point position or road vanishing point.
Further, vehicle and pedestrian is the dynamic road element in road traffic environment, and target following mode can be used
Obtain information of vehicles and/or pedestrian information.Where information of vehicles such as travel speed, driving direction, vehicle location and vehicle
Road etc., pedestrian information such as speed of travel, direction of travel, pedestrian position and road where pedestrian etc..Scene light stream can also
To express the information of moving target in road traffic environment.
Step 304, the corresponding present road element parameter of various roads element is obtained according to real-time perception result.
Wherein, road element refers to that driving status impacts on road to road vehicles in road traffic environment
Factor, such as lane line, road vanishing point, traffic sign, vehicle or pedestrian etc..Road element parameter is by corresponding road member
What is obtained after the sensing results Parameter Expression of element characterizes the parameter of the road element feature.In one embodiment, road member
Plain parameter includes vehicle tracking parameter, pedestrian tracking parameter, lane detection result, road vanishing point testing result and traffic mark
At least one of will testing result.Here real-time perception result parameter is turned into corresponding road element in current road element
Parameter.The species number of various roads element is determining.
Step 306, road scene parameter is predicted according to factor graph model and present road element parameter；Factor graph model is used
Joint Distribution probability density function between expression road scene parameter and the corresponding road element parameter of various roads element.
Wherein, factor graph is the unified representation of Bayesian network and Markov Network, by introducing factor nodes come bright
The decomposition of joint probability distribution on factor graph really is described.Further, factor graph introduces a factor nodes set, quite
In the objective function being defined on usual Node subsets.Factor graph is a bigraph (bipartite graph), for describing the office on some variables set
The product of portion's function.
Factor graph model is the obtained mathematical model of principle training using factor graph, expresses road scene parameter and more
Joint Distribution probability density function between the kind corresponding road element parameter of road element, electronic equipment are general according to the Joint Distribution
Rate density function and present road element parameter estimate the probability of road scene parameter value, thus taking maximum probability
It is worth the road scene parameter as prediction.Estimate that maximum a-posteriori estimation side can be used when the probability of road scene parameter value
Method (MAP) or maximum Likelihood.
Road scene parameter is parameter based on when road vehicles are carried out with traveling control, in one embodiment
In, road scene parameter includes road topology relationship, crossing center, road width, the road that will be driven into and headstock side
To at least one of the angle between angle and present road and the road that will be driven into.Wherein road topology relationship is such as
Straight road, left-hand rotation road, right-hand rotation road, T-type road or crossroad etc..Crossing center can be crossing center and exist
Coordinate under bodywork reference frame.
In one embodiment, road scene parameter expressed by factor graph model and the corresponding road of various roads element
Joint Distribution probability density function, that is, objective function between element parameter can be used following formula (1) to indicate:
Wherein, θ indicates all parameters in factor graph model.R={ k, c, w, r, a } indicates road scene parameter, is to use
To describe the stochastic variable of road topology relationship and geometry；Wherein k indicates road topology relationship, and c=(x, z) indicates crossing
Coordinate of the center under bodywork reference frame, w indicate that road width, r indicate the road that will be driven into and headstock angular separation, a table
Show the angle between present road and the road that will be driven into.
T={ t1,...,tNtIndicate vehicle tracking parameter, share Nt parameter；P={ p1,...,pNpIndicate pedestrian with
Track parameter shares Np parameter；L={ l1,...,lNlIndicate road vanishing point testing result, share Nl parameter；V=
{v1,...,vNvLane detection is indicated as a result, sharing Nv parameter；S={ s1,...,sNsIndicate road traffic sign detection knot
Fruit shares Ns parameter.The observational variable of E={ T, P, L, V, S } Components graph model, is also called data evidence.{ E-T } is
Gather the expression that simplifies of { P, L, V, S }, similar, { E-P } indicates that set { T, L, V, S, } { E-L } indicates set { T, P, V, S },
{ E-V } indicates set { T, P, L, S }, and { E-S } indicates set { T, P, L, V }.
In one embodiment, electronic equipment specifically (2) can use maximum a-posteriori estimation side according to the following formula
Method predicts road scene parameter R:
R=argmaxRP (R | E, θ) formula (2)
Wherein, E indicates that present road element parameter, θ indicate all parameters in factor graph model here, and R indicates prediction
Road scene parameter, formula (2) indicates that the road scene parameter R of prediction maximizes Probability p (R | E, θ).
Step 308, traveling control instruction is generated according to the road scene parameter of prediction and exported.
Wherein, traveling control instruction is the instruction of controllable road vehicles traveling, such as control road vehicles
Driving direction, travel speed or brake etc..Electronic equipment can directly export traveling control instruction to road vehicles
Controller so that road vehicles realize automatic Pilot；Electronic equipment can also be by traveling control instruction with visual cues
Perhaps voice broadcast mode output such as prompt changes driving direction, prompt change travel speed or prompt brake etc. to mode,
To realize the navigation of road vehicles.
Above-mentioned road vehicles travel control method, is predicted using the present road element parameter of real-time perception result
Road scene parameter is no longer dependent on map datum, traffic to carry out traveling control according to the road scene parameter of prediction
The accuracy of tool traveling control is enhanced.Moreover, expressing road scene parameter and various roads element by factor graph
Joint Distribution probability density function between corresponding road element parameter, factor graph with graph structure describe global objective function because
The global operations of huge construction program are divided into simple local operation by formula decomposed form, improve the efficiency of traveling control.Furthermore
Various roads element can be merged by factor graph model, it may be considered that the correlation between road element parameter, without
It assumes that independently of one another, so that vehicle travels control meets real roads scene, precise control.
In one embodiment, road element parameter includes tracking target component；This method further include: according to factor artwork
Type and the road scene parameter prediction of prediction tracking target are relative to site of road；Step 308 includes: the road field according to prediction
Scape parameter and the tracking target of prediction generate traveling control instruction relative to site of road and export.
Wherein, track the moving targets such as target such as vehicle or pedestrian, the such as above-mentioned vehicle of tracking target component with
Track parameter T or pedestrian tracking parameter P.Tracking target refers to that the tracking such as vehicle or pedestrian target is opposite relative to site of road
In the positional relationship of road.According to the road target of prediction relative to site of road, electronic equipment be can determine on the road of traveling
Existing barrier can determine feasible travelling route in conjunction with the road scene parameter of prediction, to generate corresponding
Traveling control instruction simultaneously exports.
In one embodiment, electronic equipment specifically (3) can use maximum a-posteriori estimation side according to the following formula
Method predicts the position of vehicle in the road:
Wherein, l indicates the mark for being currently located road, lNtIndicate vehicle tNtThe mark of place road, sNtIndicate vehicle with
The expression of the Spline Model of vehicle location in track parameter；The vehicle tracking parameter of t expression current vehicle；R indicates the road of prediction
Road scenario parameters；VL indicates the set of the position of the vehicle for predicting to obtain in the road；VL indicates each tracking vehicle of prediction
Position in the road.The position of pedestrian in the road can also be acquired using identical method for pedestrian, specifically by formula
(3) information of tracking vehicle replaces with the information of tracking pedestrians in.
In the present embodiment, by predicting tracing target relative to site of road, not only examined when generating and travelling control instruction
The road scene parameter for considering prediction further accounts for the tracking target of prediction relative to site of road, can road be handed in this way
Logical tool traveling is more intelligent, can be safer when applied to unmanned scene.
As shown in figure 4, in one embodiment, before step 302, which is also wrapped
The step of including trained factor graph model, specifically comprises the following steps:
Step 402, various roads element is subjected to Parameter Expression, obtains corresponding road element parameter.
Different kinds of roads element is carried out unification and the Parameter Expression of coordinate system by electronic equipment, obtains corresponding road member
Plain parameter.If can be used particularly for lane line through the expression of the Spline Model done, by Spline Model express in parameter make
For the corresponding road element parameter of lane line.Road vanishing point can be used in road birds-eye view in the road element parameter of road vanishing point
Coordinate position.Position and its semantic expressiveness of traffic sign can be used in the road element parameter of traffic sign.Pedestrian and vehicle
It is dynamic data Deng tracking target, the parameter work of the position and its trace model that track target in each frame video image can be used
For the road element parameter for tracking target.
Step 404, probability graph model is constructed according to the corresponding road element parameter of each road element respectively.
Specifically, due to there is static road element, such as lane line or traffic sign in road traffic environment；?
There are dynamic road elements, such as light stream, pedestrian or vehicle, and static and dynamic road element can be adopted respectively here
It is modeled with different types of probability graph model.Bayesian network (BN) model foundation can be used particularly for static road element
Probability graph model can be used Hidden Markov (HMM) model for dynamic road element or Multiple reference (MRF) model built
Vertical probability graph model.
Step 406, each probability graph model is connected according to road traffic priori knowledge, obtains factor graph model.
Wherein, road traffic priori knowledge is the known information present in road traffic environment, such as vehicle when red light
Majority slows down or stops, and car speed is slower at road cross, and vehicle is overtaken other vehicles in left side road, pedestrian in road both sides,
Two cars can not be in same position etc..Road traffic priori knowledge can reflect out the pass between each road element parameter
Connection, therefore using traffic priori knowledge, associated node in the connection of each probability graph model is connected, and marks corresponding general
Rate, so that each probability graph model be united to form probability of recombination graph model.Electronic equipment according to the display rule of factor graph,
Probability of recombination graph model is expressed again, converts the probability graph model of oriented chart-pattern to the factor artwork of undirected chart-pattern
Type.As shown in figure 5, by probability of recombination graph model Bayesian network model and Hidden Markov Model carry out factor graph respectively
Expression, obtains factor graph model.
Probability graph model be complicated uncertain problem is modeled, one of the important tool of reasoning, common probability
Graph model mainly includes Bayesian network model, Markov Network model and factor graph model etc..Wherein factor graph model
It is a kind of bipartite graph for indicating function of many variables factorization structure, there is very strong ability to express.Factor graph model not only can be with
Indicate all independent sexual intercourse that Bayesian network model and Markov Network model can indicate, additionally it is possible to indicate them
The independent sexual intercourse that cannot be indicated.Moreover, Bayesian network model and Markov Network model only need to be by simple steps
Suddenly factor graph model can be converted into.
Step 408, according to road data sample training factor graph model.
Specifically, the training of factor graph model includes Structure learning and parameter learning.Structure learning includes certainty factor figure
The connection type of the node of model and side and node and side, and with the presence or absence of feedback loop etc..Electronic equipment can be according to mark
Remember that road data sample carries out the road scene parameter under different kinds of roads scene and the dependence between corresponding road element parameter
Clustering, there are the associated factor in certainty factor graph model, thus the structure of Studying factors graph model, by hierarchy factor figure
Model conversation is normalization factor figure.And then pass through some BP (Error Back Propagation, error back propagation) algorithm
Parameter in Studying factors graph model.
Wherein road data sample can be label road data sample, and label road data sample is to have marked prediction knot
The sample of fruit and observational variable.Dependence clustering specifically can pass through clustering algorithm according to Euclidean distance or mahalanobis distance
(the non-supervisory method of such as K-means) carries out clustering.Hierarchy factor graph model is increased on the basis of normalization factor graph model
Add in relation to composite variable and composite factor, wherein composite variable and composite factor are the obtainable intermediate knots in modeling process
Hierarchy factor graph model is converted normalization factor graph model by fruit, may make modeling process more orderliness orderly, also can simplify
The reasoning process of factor graph model.
To the parameter learning of factor graph model, the utilization of belief propagation (Belief propagation) mechanism specifically can be used
Sum-product algorithm (sum-product algorithm).Sum-product algorithm is common factor graph model reasoning algorithm, use and integrating
When method training factor graph model, marginal probability distribution letter is realized by carrying out message transmission between factor graph model neighborhood of nodes
Several calculating, marginal probability refer to the adduction of a certain group of probability.
In one embodiment, electronic equipment can be loaded into according to road data sample and be obtained according to road traffic priori knowledge
Priori factor graph model, using belief propagation mechanism and using sum-product algorithm carry out marginal probability distribution function calculating,
And the dependence between proof factor graph model interior joint, carried out subtracting branch according to dependence, gradually form a standard because
Subgraph model, the Structure learning of realization factor graph model, and then parameter learning is carried out according to road data sample.
Specifically, electronic equipment can to the node of the corresponding road element parameter of the road element two-by-two of factor graph model,
Such as traffic sign and vehicle, road and pedestrian etc., the analysis of dependence is carried out, this dependence indicates between node
Correlation.If between the factor Relationship Comparison it is small if directly cut away in factor graph model between respective nodes frontier juncture connection.To the factor
Graph model carries out that the observational variable of factor graph model and the cost function of incidence matrix can be constructed when Structure learning, by seeking most
Small cost function acquires incidence matrix, to be carried out subtracting branch according to incidence matrix, wherein incidence matrix is indicated in factor graph model
Dependence between node.It seeks minimizing cost function being a NP-hard (nondeterministic polynomial is difficult) problem, it can
It is realized using simulated annealing mode.
In the present embodiment, different kinds of roads element progress Parameter Expression is obtained into road element parameter and constructs probability artwork
Type connects probability graph model according to road traffic priori knowledge to obtain factor graph model, can obtain meeting road traffic priori
The factor graph model of knowledge, and then road data sample training factor graph model is utilized, aloow factor graph model accurate
Road traffic environment is described so that electronic equipment can accurate understanding road traffic environment, and then accurately to road vehicles
Traveling controlled.
In one embodiment, road data sample includes label road data sample and unmarked road data sample；
Step 408 specifically includes: according to label road data sample and unmarked road data sample, using semi-supervised learning mode pair
Factor graph model carries out Structure learning and parameter learning.
Unmarked road data sample refers to the sample for only marking observational variable and unmarked prediction result.Electronic equipment can
The similarity of label road data sample and unmarked road data sample is first determined, thus according to label road data sample institute
The prediction result of mark estimates the prediction result of unmarked road data sample, so according to label road data sample, not
Road data sample and the prediction result of estimation is marked to carry out Structure learning and parameter learning to factor graph model.Calculate similarity
It is contemplated that a variety of similarities, such as Road form similarity, scene structure similarity and semantic similarity etc., it can also be by vehicle
Diatom is merged with travelable regionally detecting result, road and lane line morphology sample characteristics dictionary is established, thus according to lane line
Characteristic vector in form sample characteristics dictionary measures similarity.
In the present embodiment, by semi-supervised learning mode training factor graph model, labeled cost can be reduced, factor graph is improved
The training effectiveness of model.
In one embodiment, factor graph model is dynamic；The road vehicles travel control method further include: receive
Collect real-time perception result；According to the real-time perception result updating factor graph model of collection.Specifically, electronic equipment collects real-time
Sensing results and corresponding prediction result are used as and do not mark as mark road data sample, or collection real-time perception result
Road data sample further carries out Structure learning and/or parameter learning to factor graph model.Electronic equipment is also based on greatly
Scale sparse matrix optimizes factor graph model.By constantly updating factor graph model, factor graph model can be kept to use
In the accuracy of road vehicles traveling control.
In a specific application scenarios, the whole flow process of road vehicles travel control method is as shown in Figure 6.Electricity
Sub- equipment collects the real-time perception for participating in the different kinds of roads element of road traffic environment as a result, real-time perception result can pass through vision
The awareness apparatus such as system or radar system perceive to obtain.Different kinds of roads element such as lane line, road vanishing point, traffic sign, row
People, vehicle etc., real-time perception result for example lane detection result, road vanishing point testing result, road traffic sign detection result and its
The semantic, parameter of pedestrian tracking device and the parameter of vehicle tracking device etc..It can recognize different kinds of roads element by semantic segmentation.
Further, real-time perception result is carried out Parameter Expression by electronic equipment, obtains road element parameter；According to each
The corresponding road element parameter of road element constructs probability graph model respectively；According to road traffic priori knowledge by each probability artwork
Type connection, obtains probability of recombination graph model, converts dynamic factor figure for probability of recombination graph model.By road scene parameter and more
Joint Distribution probability density function between the kind corresponding road element parameter of road element is as objective function.
Further, relevance of the electronic equipment based on time and space, according to real-time perception result and using semi-supervised
Mode of learning carries out Structure learning and parameter learning to factor graph model, and wherein structure learning process is as shown in Figure 7.Utilize update
Incremental data dynamic updating factor graph model, and utilize Large Scale Sparse matrix optimizing factor graph model.Example finally can be used
It is realized such as MCMC (Markov Chain Monte Carlo, the stochastic simulation) algorithm of Metropolis-Hastings algorithm
Parameter reasoning such as predicts road scene parameter.
As shown in figure 8, in one embodiment, providing a kind of road vehicles travel controlling system 800, comprising:
Data acquisition module 801, prediction module 802 and output module 803.
Data acquisition module 801, for obtaining the real-time perception result of road traffic environment；It is obtained according to real-time perception result
To the corresponding present road element parameter of various roads element.
Prediction module 802, for predicting road scene parameter according to factor graph model and present road element parameter；The factor
The Joint Distribution probability that graph model is used to express between road scene parameter and the corresponding road element parameter of various roads element is close
Spend function.
Output module 803, for generating traveling control instruction according to the road scene parameter of prediction and exporting.
Above-mentioned road vehicles travel controlling system 800, using the present road element parameter of real-time perception result come
Predict that road scene parameter is no longer dependent on map datum to carry out traveling control according to the road scene parameter of prediction,
The accuracy of vehicle travels control is enhanced.Moreover, expressing road scene parameter and various roads by factor graph
Joint Distribution probability density function between the corresponding road element parameter of element, factor graph describe global objective function with graph structure
Factorization form, the global operations of huge construction program are divided into simple local operation, improve traveling control efficiency.Again
Person can be merged various roads element by factor graph model, it may be considered that the correlation between road element parameter, and
It does not assume that independently of one another, so that vehicle travels control meets real roads scene, precise control.
In one embodiment, prediction module 802 is also used to pre- according to factor graph model and the road scene parameter of prediction
Tracking target is surveyed relative to site of road.Output module 803 is also used to according to the road scene parameter of prediction and the tracking of prediction
Target generates traveling control instruction relative to site of road and exports.
In the present embodiment, by predicting tracing target relative to site of road, not only examined when generating and travelling control instruction
The road scene parameter for considering prediction further accounts for the tracking target of prediction relative to site of road, can road be handed in this way
Logical tool traveling is more intelligent, can be safer when applied to unmanned scene.
As shown in figure 9, in one embodiment, road vehicles travel controlling system 800 further include: factor graph model
Training module 804 obtains corresponding road element parameter for various roads element to be carried out Parameter Expression；According to each road
Element corresponding road element parameter in road constructs probability graph model respectively；According to road traffic priori knowledge by each probability graph model
Connection, obtains factor graph model；According to road data sample training factor graph model.
In the present embodiment, different kinds of roads element progress Parameter Expression is obtained into road element parameter and constructs probability artwork
Type connects probability graph model according to road traffic priori knowledge to obtain factor graph model, can obtain meeting road traffic priori
The factor graph model of knowledge, and then road data sample training factor graph model is utilized, aloow factor graph model accurate
Road traffic environment is described so that electronic equipment can accurate understanding road traffic environment, and then accurately to road vehicles
Traveling controlled.
In one embodiment, road data sample includes label road data sample and unmarked road data sample；
Factor graph model training module 804 is also used to be supervised according to label road data sample and unmarked road data sample using half
It superintends and directs mode of learning and Structure learning and parameter learning is carried out to factor graph model.
In the present embodiment, by semi-supervised learning mode training factor graph model, labeled cost can be reduced, factor graph is improved
The training effectiveness of model.
In one embodiment, road vehicles travel controlling system 800 further include: factor graph model modification module
805, for collecting real-time perception result；According to the real-time perception result updating factor graph model of collection.
In one embodiment, road scene parameter includes road topology relationship, crossing center, road width, i.e.
By at least one of the angle between the road driven into and headstock angular separation and present road and the road that will be driven into；
Road element parameter include vehicle tracking parameter, pedestrian tracking parameter, lane detection result, road vanishing point testing result and
At least one of road traffic sign detection result.
Those of ordinary skill in the art will appreciate that realizing all or part of the process in above-described embodiment method, being can be with
Relevant hardware is instructed to complete by computer program, which can be stored in a computer-readable storage and be situated between
In matter, the program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Wherein, storage medium above-mentioned can be
The non-volatile memory mediums such as magnetic disk, CD, read-only memory (Read-Only Memory, ROM) or random storage note
Recall body (Random Access Memory, RAM) etc..
Each technical characteristic of above embodiments can be combined arbitrarily, for simplicity of description, not to above-described embodiment
In each technical characteristic it is all possible combination be all described, as long as however, the combination of these technical characteristics be not present lance
Shield all should be considered as described in this specification.
Only several embodiments of the present invention are expressed for above embodiments, and the description thereof is more specific and detailed, but can not
Therefore it is construed as limiting the scope of the patent.It should be pointed out that for those of ordinary skill in the art,
Under the premise of not departing from present inventive concept, various modifications and improvements can be made, and these are all within the scope of protection of the present invention.
Therefore, the scope of protection of the patent of the invention shall be subject to the appended claims.
Claims (14)
1. a kind of road vehicles travel control method, which comprises
Obtain the real-time perception result of road traffic environment；
The corresponding present road element parameter of various roads element is obtained according to the real-time perception result；
Road scene parameter is predicted according to factor graph model and the present road element parameter；Wherein, the factor graph model
For expressing the Joint Distribution probability density function between road scene parameter and the corresponding road element parameter of various roads element；
The factor graph model is connected each probability graph model according to road traffic priori knowledge and according to road data sample training
It obtains；The probability graph model is constructed respectively according to the corresponding road element parameter of various roads element；
Traveling control instruction is generated according to the road scene parameter of prediction and is exported.
2. the method according to claim 1, wherein the road element parameter includes tracking target component；Institute
State method further include:
According to the factor graph model and the road scene parameter prediction of prediction tracking target relative to site of road；
It is described that traveling control instruction is generated according to the road scene parameter of prediction and is exported, comprising:
Traveling control instruction and defeated is generated relative to site of road according to the road scene parameter of prediction and the tracking target of prediction
Out.
3. the method according to claim 1, wherein it is described obtain road traffic environment real-time perception result it
Before, the method also includes:
Various roads element is subjected to Parameter Expression, obtains corresponding road element parameter；
Probability graph model is constructed respectively according to the corresponding road element parameter of each road element；
Each probability graph model is connected according to road traffic priori knowledge, obtains factor graph model；
According to factor graph model described in road data sample training.
4. according to the method described in claim 3, it is characterized in that, the road data sample includes label road data sample
With unmarked road data sample；The factor graph model according to road data sample training, comprising:
According to label road data sample and unmarked road data sample, using semi-supervised learning mode to the factor artwork
Type carries out Structure learning and parameter learning.
5. the method according to claim 1, wherein the method also includes:
Collect the real-time perception result；
The factor graph model is updated according to the real-time perception result of collection.
6. the method according to claim 1, wherein the road scene parameter includes road topology relationship, road
Mouth center, road width, the road that will be driven into and headstock angular separation and present road and the road that will be driven into
Between at least one of angle；The road element parameter includes vehicle tracking parameter, pedestrian tracking parameter, lane line inspection
Survey at least one of result, road vanishing point testing result and road traffic sign detection result.
7. a kind of road vehicles travel controlling system, which is characterized in that described device includes:
Data acquisition module, for obtaining the real-time perception result of road traffic environment；It is obtained according to the real-time perception result
The corresponding present road element parameter of various roads element；
Prediction module, for predicting road scene parameter according to factor graph model and the present road element parameter；Wherein, institute
Factor graph model is stated for expressing the Joint Distribution between road scene parameter and the corresponding road element parameter of various roads element
Probability density function；The factor graph model is connected each probability graph model according to road traffic priori knowledge and according to road
Circuit-switched data sample training obtains；The probability graph model is according to the corresponding road element parameter difference structure of various roads element
It builds；Output module, for generating traveling control instruction according to the road scene parameter of prediction and exporting.
8. device according to claim 7, which is characterized in that the prediction module is also used to according to the factor graph model
Road scene parameter prediction tracking target with prediction is relative to site of road；
The output module is also used to raw relative to site of road according to the road scene parameter of prediction and the tracking target of prediction
At traveling control instruction and export.
9. device according to claim 7, which is characterized in that described device further include: factor graph model training module is used
In various roads element is carried out Parameter Expression, corresponding road element parameter is obtained；According to the corresponding road of each road element
Road element parameter constructs probability graph model respectively；Each probability graph model is connected according to road traffic priori knowledge, is obtained
Factor graph model；According to factor graph model described in road data sample training.
10. device according to claim 9, which is characterized in that the road data sample includes label road data sample
Sheet and unmarked road data sample；The factor graph model training module is also used to according to label road data sample and does not mark
Remember road data sample, Structure learning and parameter learning are carried out to the factor graph model using semi-supervised learning mode.
11. device according to claim 7, which is characterized in that described device further include: factor graph model modification module,
For collecting the real-time perception result；The factor graph model is updated according to the real-time perception result of collection.
12. device according to claim 7, which is characterized in that the road scene parameter includes road topology relationship, road
Mouth center, road width, the road that will be driven into and headstock angular separation and present road and the road that will be driven into
Between at least one of angle；The road element parameter includes vehicle tracking parameter, pedestrian tracking parameter, lane line inspection
Survey at least one of result, road vanishing point testing result and road traffic sign detection result.
13. a kind of electronic equipment, including memory and processor, the memory are stored with computer program, which is characterized in that
The computer program realizes the step of any one of claims 1 to 6 the method when being executed by the processor.
14. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the computer program
The step of method described in any one of claims 1 to 6 is realized when being executed by processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610270123.6A CN107310550B (en) | 2016-04-27 | 2016-04-27 | Road vehicles travel control method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610270123.6A CN107310550B (en) | 2016-04-27 | 2016-04-27 | Road vehicles travel control method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107310550A CN107310550A (en) | 2017-11-03 |
| CN107310550B true CN107310550B (en) | 2019-09-17 |
Family
ID=60184988
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610270123.6A Active CN107310550B (en) | 2016-04-27 | 2016-04-27 | Road vehicles travel control method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107310550B (en) |
Families Citing this family (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE102017221634B4 (en) * | 2017-12-01 | 2019-09-05 | Audi Ag | Motor vehicle with a vehicle guidance system, method for operating a vehicle guidance system and computer program |
| CN108304580A (en) * | 2018-03-05 | 2018-07-20 | 上海思贤信息技术股份有限公司 | A kind of major event method for early warning and system towards urban grid management |
| CN108944947A (en) * | 2018-07-15 | 2018-12-07 | 北京三快在线科技有限公司 | The prediction technique and device of steer decision |
| CN109117802B (en) * | 2018-08-21 | 2021-10-29 | 东北大学 | Ship detection method for large-scene high-resolution remote sensing image |
| CN109359689B (en) * | 2018-10-19 | 2021-06-04 | 科大讯飞股份有限公司 | Data identification method and device |
| CN109649386A (en) * | 2018-12-26 | 2019-04-19 | 成都路行通信息技术有限公司 | A kind of vehicle collision detection method based on machine learning |
| CN111469835A (en) * | 2019-01-22 | 2020-07-31 | 观致汽车有限公司 | Vehicle auxiliary driving system and method based on topological map |
| CN110766061B (en) * | 2019-10-15 | 2022-05-31 | 武汉中海庭数据技术有限公司 | Road scene matching method and device |
| CN111061255B (en) * | 2019-12-27 | 2021-05-18 | 山东国惠安创智能物联发展有限公司 | Control method and device of electric vehicle |
| CN113156479B (en) * | 2021-03-23 | 2023-03-28 | 同济大学 | Combined positioning method for outdoor country road scene |
| CN113642644B (en) * | 2021-08-13 | 2024-05-10 | 北京赛目科技有限公司 | Method and device for determining vehicle environment level, electronic equipment and storage medium |
| CN114018274B (en) | 2021-11-18 | 2024-03-26 | 阿波罗智能技术(北京)有限公司 | Vehicle positioning method and device and electronic equipment |
| CN114329342B (en) * | 2021-12-20 | 2024-02-06 | 中交第二公路勘察设计研究院有限公司 | Expressway traffic capacity calculation method based on extremum-Gaussian mixture distribution model |
| CN116048000B (en) * | 2023-03-31 | 2023-06-13 | 中科航迈数控软件(深圳)有限公司 | Modeling method, modeling device, modeling equipment and modeling storage medium for multi-axis motion numerical control system |
| CN116916049B (en) * | 2023-09-12 | 2023-12-08 | 北京青水环境科技有限公司 | Video data online acquisition and storage system based on cloud computing technology |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE102011082126B4 (en) * | 2011-09-05 | 2020-07-23 | Robert Bosch Gmbh | SAFETY DEVICE FOR MOTOR VEHICLES |
| US8793046B2 (en) * | 2012-06-01 | 2014-07-29 | Google Inc. | Inferring state of traffic signal and other aspects of a vehicle's environment based on surrogate data |
| DE102013212710A1 (en) * | 2013-05-16 | 2014-11-20 | Siemens Aktiengesellschaft | Sensor product, simulator and method for simulating sensor measurements, merging sensor measurements, validating a sensor model and designing a driver assistance system |
| KR102016551B1 (en) * | 2014-01-24 | 2019-09-02 | 한화디펜스 주식회사 | Apparatus and method for estimating position |
-
2016
- 2016-04-27 CN CN201610270123.6A patent/CN107310550B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN107310550A (en) | 2017-11-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107310550B (en) | Road vehicles travel control method and device | |
| US11934962B2 (en) | Object association for autonomous vehicles | |
| Yu et al. | Deep learning-based traffic safety solution for a mixture of autonomous and manual vehicles in a 5G-enabled intelligent transportation system | |
| US11860636B2 (en) | Providing actionable uncertainties in autonomous vehicles | |
| US11836623B2 (en) | Object detection and property determination for autonomous vehicles | |
| CN110796856B (en) | Vehicle lane change intention prediction method and training method of lane change intention prediction network | |
| US12097844B2 (en) | Constraining vehicle operation based on uncertainty in perception and/or prediction | |
| US10310087B2 (en) | Range-view LIDAR-based object detection | |
| CN112789481A (en) | Trajectory prediction for top-down scenarios | |
| US10730531B1 (en) | Machine-learning based vehicle motion control system | |
| Xiao et al. | UB‐LSTM: a trajectory prediction method combined with vehicle behavior recognition | |
| US20210139024A1 (en) | Collision monitoring using system data | |
| US20210139023A1 (en) | Collision monitoring using statistic models | |
| CN112185148A (en) | Method for evaluating possible trajectories | |
| CN103996298B (en) | A kind of driver behavior modeling method and device | |
| Bender et al. | Predicting driver intent from models of naturalistic driving | |
| Song et al. | Multi-vehicle tracking with microscopic traffic flow model-based particle filtering | |
| CN115662166A (en) | Automatic driving data processing method and automatic driving traffic system | |
| Wu et al. | Lane-GNN: Integrating GNN for Predicting Drivers' Lane Change Intention | |
| US20240119830A1 (en) | System and method for predicting a future position of a road user | |
| CN116448134B (en) | Vehicle path planning method and device based on risk field and uncertain analysis | |
| Ranacher et al. | A model to estimate and interpret the energy-efficiency of movement patterns in urban road traffic | |
| Chipka et al. | Estimation and navigation methods with limited information for autonomous urban driving | |
| Isong et al. | Deep Learning-Based Object Detection Techniques for Self-Driving Cars: an in-Depth Analysis | |
| Zyner | Naturalistic driver intention and path prediction using machine learning |
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 |