CN113997954B - Method, device and equipment for predicting vehicle driving intention and readable storage medium - Google Patents
Method, device and equipment for predicting vehicle driving intention and readable storage medium Download PDFInfo
- Publication number
- CN113997954B CN113997954B CN202111437983.1A CN202111437983A CN113997954B CN 113997954 B CN113997954 B CN 113997954B CN 202111437983 A CN202111437983 A CN 202111437983A CN 113997954 B CN113997954 B CN 113997954B
- Authority
- CN
- China
- Prior art keywords
- driving
- vehicle
- probability
- information
- candidate
- 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 53
- 230000004927 fusion Effects 0.000 claims abstract description 19
- 238000012549 training Methods 0.000 claims description 63
- 238000012545 processing Methods 0.000 claims description 9
- 238000004590 computer program Methods 0.000 claims description 3
- 238000004891 communication Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 5
- 238000013459 approach Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 206010039203 Road traffic accident Diseases 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000006698 induction Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012216 screening 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
- B60W60/00—Drive control systems specially adapted for autonomous road vehicles
- B60W60/001—Planning or execution of driving tasks
- B60W60/0027—Planning or execution of driving tasks using trajectory prediction for other traffic participants
Abstract
The application discloses a method, a device, equipment and a readable storage medium for predicting a vehicle driving intention, which are characterized in that a plurality of candidate lanes and the position information of each candidate lane are determined according to the current position of a vehicle, then the historical driving information, the historical orientation information and the historical position information of the vehicle in a preset time period before the current moment are acquired, the historical driving information and the position information of the plurality of candidate lanes are input into a first intention prediction model to obtain the first probability of the vehicle driving into each candidate lane, the historical orientation information and the position information of the plurality of candidate lanes are input into a second intention prediction model to obtain the second probability of the vehicle driving into each candidate lane, the historical position information and the position information of the plurality of candidate lanes are input into a third intention prediction model to obtain the third probability of the vehicle driving into each candidate lane, and then the first probability, the second probability and the third probability are input into a fusion intention prediction model to determine the driving intention of the vehicle.
Description
Technical Field
The present application relates to the field of automatic driving technologies, and in particular, to a method, an apparatus, a device, and a readable storage medium for predicting a vehicle driving intention.
Background
With the continuous development of technology, the automatic driving technology becomes an important research direction of the automobile industry. When the automatic driving vehicle runs, in order to safely pass through complex and changeable traffic scenes, not only the positions and the motion states of surrounding vehicles are estimated, but also the running intention of the surrounding vehicles is predicted in time, so that the running route can be adjusted by predicting the running intention of the surrounding vehicles, and traffic accidents are avoided. In the process of predicting the vehicle driving intention, the related vehicle features are more, the direct construction of a model of the mapping from the vehicle features to the intention is very difficult according to the traditional modeling method, the construction of the model for predicting the vehicle driving intention is very complex, and tuning is inconvenient, so that the accuracy of the vehicle driving intention prediction is lower, and therefore, how to predict the vehicle driving intention so as to improve the accuracy of the vehicle driving intention prediction is a problem which is always focused by people.
Disclosure of Invention
In view of the above, the present application provides a method, apparatus, device, and readable storage medium for predicting a vehicle driving intention, so as to improve accuracy of vehicle driving intention prediction.
In order to achieve the above object, the following solutions have been proposed:
a vehicle travel intention prediction method, comprising:
determining a plurality of candidate lanes and position information of each candidate lane according to the current position of the vehicle;
acquiring historical driving information, historical orientation information and historical position information of a vehicle in a preset time period before the current moment;
inputting the historical driving information and the position information of a plurality of candidate lanes into a first driving intention prediction model to obtain a first probability of the vehicle driving into each candidate lane;
inputting the history orientation information and the position information of a plurality of candidate lanes into a second driving intention prediction model to obtain a second probability of the vehicle driving into each candidate lane;
inputting the historical position information and the position information of a plurality of candidate lanes into a third driving intention prediction model to obtain a third probability of the vehicle driving into each candidate lane;
and inputting the first probability, the second probability and the third probability into a fusion driving intention prediction model to determine the driving intention of the vehicle.
Optionally, the first, second, and third travel intention prediction models each include: an input layer, a historical driving state determining layer and a probability predicting layer which are sequentially cascaded;
A training process of the first, second, and third travel intent prediction models, comprising:
acquiring training samples of a vehicle and training position information of a plurality of candidate lanes through an input layer;
determining, by a historical driving state determination layer, a historical driving state of the vehicle based on the training sample;
predicting, by a probability prediction layer, a probability of a vehicle driving into each candidate lane based on the historical driving state and training position information of a plurality of candidate lanes;
aiming at each candidate lane, taking the predicted probability of the vehicle driving into the candidate lane as a target, and updating parameters of the model;
wherein the training samples of the first driving intention prediction model comprise driving information training samples;
the training samples of the second driving intention prediction model comprise orientation information training samples;
the training samples of the third driving intent prediction model include position information training samples.
Optionally, the historical driving information and the position information of the plurality of candidate lanes are input into a first driving intention prediction model to obtain a first probability of the vehicle driving into each candidate lane, including:
Inputting the historical driving information and the position information of a plurality of candidate lanes into a first driving intention prediction model;
processing the historical driving information by utilizing a historical driving state determining layer of the first driving intention prediction model to obtain a historical driving state;
and processing the historical driving state and the position information of the plurality of candidate lanes by utilizing a probability prediction layer of the first driving intention prediction model so as to output a first probability of driving the vehicle into each candidate lane.
Optionally, after acquiring the historical driving information, the historical heading information and the historical position information of the vehicle in a preset time period before the current moment, the method further includes:
selecting the driving data from the history driving information according to a preset time interval to serve as target history driving data;
the step of inputting the history driving information and the position information of the plurality of candidate lanes into a first driving intention prediction model to obtain a first probability of the vehicle driving into each candidate lane, comprising:
and inputting the target historical driving data and the position information of the plurality of candidate lanes into a first driving intention prediction model to obtain a first probability of the vehicle driving into each candidate lane.
Optionally, the location information of the candidate lane includes: coordinate information of the center line of the candidate lane.
Optionally, the historical location information includes: positional information of the vehicle offset from the center line.
Optionally, the method further comprises:
and sending the driving intention to a vehicle decision module so that the vehicle decision module controls the vehicle to run according to the driving intention.
A vehicle travel intention prediction apparatus comprising:
the candidate lane determining unit is used for determining a plurality of candidate lanes and the position information of each lane according to the current position of the vehicle;
an information acquisition unit that acquires history travel information, history orientation information, and history position information of the vehicle in a preset period of time before a current time;
the first probability prediction unit is used for inputting the historical driving information and the position information of the plurality of candidate lanes into a first driving intention prediction model to obtain a first probability of the vehicle driving into each candidate lane;
the second probability prediction unit is used for inputting the history orientation information and the position information of the plurality of candidate lanes into a second driving intention prediction model to obtain a second probability of the vehicle driving into each candidate lane;
The third probability prediction unit is used for inputting the historical position information and the position information of the plurality of candidate lanes into a third driving intention prediction model to obtain a third probability of the vehicle driving into each candidate lane;
and the driving intention prediction unit is used for inputting the first probability, the second probability and the third probability into a fusion driving intention prediction model to determine the driving intention of the vehicle.
A vehicle travel intention prediction apparatus comprising: a memory and a processor;
the memory is used for storing programs;
the processor is configured to execute the program to implement the steps of the vehicle traveling intention prediction method as described above.
A readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the vehicle travel intention prediction method as described above.
From the above technical solution, it can be seen that, according to the method, apparatus, device and readable storage medium for predicting a vehicle driving intention provided by the embodiments of the present application, according to a current position of a vehicle, position information of a plurality of candidate lanes and each candidate lane is determined, then, history driving information, history orientation information and history position information of the vehicle in a preset period of time before the current moment are acquired, the history driving information and the position information of the plurality of candidate lanes are input into a first intention prediction model to obtain a first probability that the vehicle drives into each candidate lane, the history orientation information and the position information of the plurality of candidate lanes are input into a second intention prediction model to obtain a second probability that the vehicle drives into each candidate lane, the history position information and the position information of the plurality of candidate lanes are input into a third intention prediction model to obtain a third probability that the vehicle drives into each candidate lane, and then, the first probability, the second probability and the third probability are input into a fusion intention prediction model to determine the driving intention of the vehicle. According to the method, the characteristics of the vehicle are divided into the running information, the orientation information and the position information, different running intention prediction models are established according to different characteristics, the prediction results of the different running intention prediction models are input into the fusion running intention prediction model, the running intention of the vehicle is finally obtained, and the complexity of the model is simplified by establishing the different running intention prediction models according to different characteristics, so that the accuracy of the running intention prediction of the vehicle is improved to a certain extent.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings that are required to be used in the embodiments or the description of the prior art will be briefly described below, and it is obvious that the drawings in the following description are only embodiments of the present application, and that other drawings can be obtained according to the provided drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flowchart of a method for predicting a driving intention of a vehicle according to an embodiment of the present application;
FIG. 2 is a schematic diagram of an alternative architecture of a prediction model of vehicle driving intention according to an embodiment of the present application;
FIG. 3 is a flowchart of a training method of a prediction model of vehicle driving intention according to an embodiment of the present application;
FIG. 4 is a flowchart of a prediction method of a prediction model of a vehicle driving intention according to an embodiment of the present application;
fig. 5 is a schematic structural diagram of a device for predicting a driving intention of a vehicle according to an embodiment of the present application;
fig. 6 is a block diagram showing a hardware configuration of a vehicle travel intention prediction apparatus according to an embodiment of the present application.
Detailed Description
The following description of the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
Fig. 1 is a flowchart of a method for predicting a vehicle driving intention according to an embodiment of the present application, where the method may include:
step S100, determining a plurality of candidate lanes and position information of each candidate lane according to the current position of the vehicle.
Specifically, the current position of the vehicle may be obtained through a semantic map, and a plurality of candidate lanes and position information of each candidate lane are determined, where the position information of the candidate lane may include: coordinate information of the center line of the candidate lane.
For example, the current position of the vehicle can be acquired through a semantic map, when the vehicle runs on three lanes of a lane which only allows straight running and the vehicle runs on a middle lane, the three candidate lanes can be determined to be the current lane, the left lane and the right lane respectively; when the vehicle is traveling at an intersection, the candidate lane may include: left turn driving-in lane, right turn driving-in lane, straight driving-in lane, turning-around driving-in lane, etc.
Step S101, acquiring historical driving information, historical orientation information and historical position information of the vehicle in a preset time period before the current moment.
Specifically, the relevant information of the vehicle can be obtained through real-time induction of the vehicle. Because the running information, the orientation information and the position information of the vehicle in the related information of the vehicle can provide a better basis for judging the running intention of the vehicle to a certain extent, the historical running information, the historical orientation information and the historical position information of the vehicle in a preset time period before the current moment can be acquired, such as the historical running information, the historical orientation information and the historical position information of the vehicle in 4s before the current moment are acquired. The historical driving information may include historical kinematic features such as historical speed, historical acceleration and historical angular speed of the vehicle, and the historical position information may include position information of the vehicle deviated from a central line.
Step S102, the historical driving information and the position information of a plurality of candidate lanes are input into a first driving intention prediction model, and the first probability of the vehicle driving into each candidate lane is obtained.
Specifically, the history driving information and the position information of a plurality of candidate lanes obtained in the steps are used as input, a first driving intention prediction model is input, and the first probability of the vehicle driving into each candidate lane can be obtained. The analysis of the parameters in the historical driving information may determine a change situation of the driving information of the vehicle in a preset time period before the current time, for example, the analysis of the driving speed parameters of the vehicle in the historical driving information may determine a change situation of the driving speed of the vehicle in the preset time period before the current time, where the change situation may include: the speed is unchanged, the speed is increased, the speed is reduced, and the like, and different changing conditions can provide reference for judging that the vehicle is driven into a certain candidate lane to a certain extent.
Step S103, the history orientation information and the position information of a plurality of candidate lanes are input into a second driving intention prediction model, and the second probability of the vehicle driving into each candidate lane is obtained.
Specifically, the history orientation information and the position information of a plurality of candidate lanes are input into the second driving intention prediction model, so that the probability of driving into each candidate lane can be obtained. Wherein, through to the historical orientation of vehicle, can confirm the change condition of the orientation of vehicle in the preset time quantum, the change condition can include: the vehicle can be gradually deflected leftwards, gradually deflected rightwards, the direction is unchanged, and the like, and different changing conditions can provide reference for judging that the vehicle is driven into a certain candidate lane to a certain extent.
Step S104, the historical position information and the position information of a plurality of candidate lanes are input into a third driving intention prediction model, and a third probability that the vehicle is driven into each candidate lane is obtained.
Specifically, by inputting the history position information and the position information of the plurality of candidate lanes into the third driving intention prediction model, the third probability that the vehicle is driving into each candidate lane can be obtained. By means of the historical position information of the vehicle, the position of the vehicle in a lane within a preset time period can be determined, so that a rough historical track of the vehicle is obtained, and different historical tracks can be used for providing references for judging that the vehicle is driven into a certain candidate lane to a certain extent. The above-mentioned history position information may include position information of the vehicle deviated from the center line, and the position information of the vehicle may be determined by selecting a point on the center line of the lane as an origin, and establishing a Frenet coordinate system with the center line as the x-axis, thereby determining the position information of the vehicle in the Frenet coordinate system.
The execution sequence of step S102, step S103 and step S104 may be executed simultaneously, or may be executed in any order, and no matter what order is adopted, the implementation of the embodiment of the present application is not affected.
Step S105, inputting the first probability, the second probability and the third probability into a fusion driving intention prediction model, and determining the driving intention of the vehicle.
Specifically, the first probability, the second probability and the third probability of the vehicle driving into each candidate lane can be obtained through the steps, and the driving intention of the vehicle can be determined by inputting and fusing all the first probability, the second probability and the third probability into the driving intention prediction model.
The fusion driving intention prediction model can adopt a network structure of a multi-layer perceptron so as to determine the driving intention of the vehicle. The driving intention of the vehicle may be which candidate lane the vehicle will drive into. After three groups of probabilities are input into the fusion driving intention prediction model, the fusion driving intention prediction model can utilize the three groups of probabilities to obtain the final probability of the vehicle driving into each candidate lane, and the candidate lane corresponding to the maximum probability is selected from the final probabilities by the final probability, so that the fusion driving intention prediction model can be used as the lane into which the vehicle is to be driven, and the driving intention of the vehicle is obtained.
In the training process of the fusion driving intention prediction model, the first probability training sample, the second probability training sample and the third probability training sample can be utilized to input the fusion driving intention prediction model, the final probability of the vehicle driving into each candidate lane is predicted, and the fusion driving intention prediction model is trained by taking the predicted final probability of the vehicle driving into the candidate lane as a target, wherein the predicted final probability of the vehicle driving into the candidate lane approaches to the final probability label of the vehicle driving into the candidate lane.
In the above embodiment, a method for predicting a vehicle driving intention is provided, in which a plurality of candidate lanes and position information of each candidate lane are determined according to a current position of a vehicle, then, history driving information, history orientation information and history position information of the vehicle in a preset time period before the current time are acquired, the history driving information and the position information of the plurality of candidate lanes are input into a first intention prediction model to obtain a first probability that the vehicle is driven into each candidate vehicle, the history orientation information and the position information of the plurality of candidate lanes are input into a second intention prediction model to obtain a second probability that the vehicle is driven into each candidate lane, the history position information and the position information of the plurality of candidate lanes are input into a third intention prediction model to obtain a third probability that the vehicle is driven into each candidate lane, and then, the first probability, the second probability and the third probability are input into a fusion intention prediction model to determine the driving intention of the vehicle. According to the method, the characteristics of the vehicle are divided into the running information, the orientation information and the position information, different running intention prediction models are established according to different characteristics, the prediction results of the different running intention prediction models are input into the fusion running intention prediction model, the running intention of the vehicle is finally obtained, and the complexity of the model is simplified by establishing the different running intention prediction models according to different characteristics, so that the accuracy of the running intention prediction of the vehicle is improved to a certain extent.
Referring to fig. 2, fig. 2 is an optional architecture of a driving intent prediction model according to an embodiment of the present application, where the architecture of the driving intent prediction model may include: the system comprises an input layer, a historical driving state determining layer and a probability predicting layer which are sequentially cascaded.
The input layer can acquire information of a vehicle and position information of a plurality of candidate lanes, the history driving state determining layer can determine the history driving state of the vehicle by using the information of the vehicle, and the probability predicting layer can predict and obtain the probability of the vehicle driving into each candidate lane by using the history driving state and the position information of the plurality of candidate lanes.
Specifically, in some embodiments of the present application, the first driving intent prediction model, the second driving intent prediction model, and the third driving intent prediction model may use the architecture of the driving intent prediction models to predict and obtain the probability of the vehicle driving into each candidate lane.
When the first, second, and third travel intention prediction models each employ the optional architecture of the travel intention prediction model mentioned in the above embodiments, a training process of an optional travel intention prediction model provided by an embodiment of the present application will be described with reference to fig. 2 and 3, and the training process of the first, second, and third travel intention prediction models may include:
Step 200, acquiring training samples of the vehicle and training position information of a plurality of candidate lanes through an input layer.
The training samples of the first driving intention prediction model are driving information training samples, the training samples of the second driving intention prediction model are orientation information training samples, and the training samples of the third driving intention prediction model are position information training samples.
Specifically, after the input layer obtains the training samples of the vehicle and the training position information of the plurality of candidate lanes, the training samples of the vehicle and the training position information of the plurality of candidate lanes may be vectorized, and the vectorized training samples of the vehicle and the training position information of the plurality of candidate lanes may be output to the historical driving state determination layer. The probability of a vehicle driving into each candidate lane is marked during training, wherein the probability of the candidate lane into which the vehicle finally drives is 1, and the probability of the candidate lanes into which the rest vehicles do not drive is 0.
Step S201, determining, by the historical driving state determining layer, a historical driving state of the vehicle based on the training sample.
Specifically, the historical driving state determining layer may determine the historical driving state of the vehicle based on the training sample obtained by the input layer in the above step. The historical driving state determining layer can select a network structure of the long-short-period memory model to process training samples, so that the historical driving state of the vehicle is determined. The historical driving state of the vehicle may include a kinematic characteristic change trend of the vehicle, an orientation state change trend of the vehicle, and a position change trend of the vehicle, among others.
Step S202, predicting the probability of the vehicle driving into each candidate lane based on the historical driving state and training position information of a plurality of candidate lanes by a probability prediction layer.
Specifically, the probability prediction layer may predict and obtain the probability of the vehicle driving into each candidate lane by using the training position information of the plurality of candidate lanes obtained by the input layer and the historical driving state of the vehicle determined by the historical driving state determination layer. The probability prediction layer can select a network architecture of the multi-layer perceptron to process data, and finally predicts the probability of the vehicle driving into each candidate lane.
Step S203, aiming at each candidate lane, the predicted probability of the vehicle driving into the candidate lane approaches to the probability label of the vehicle driving into the candidate lane, and updating the parameters of the model.
Specifically, the probability of the vehicle driving into each candidate lane can be predicted through the steps, and the predicted probability of the vehicle driving into the candidate lane can be used as a target for each candidate lane, and parameters of the model can be updated by taking the probability label of the predicted probability of the vehicle driving into the candidate lane approaching to the probability label of the vehicle driving into the candidate lane.
For example, when the vehicle runs on a lane which only allows straight running, the lane is three lanes and the vehicle runs on a middle lane, the candidate lane can comprise a current lane, a right lane and a left lane, and the lane which is finally driven by the vehicle is the right lane, so that the probability of marking the vehicle to drive into the right lane is 1, the probability of driving into the current lane and the probability of driving into the left lane are both 0, the probability of predicting the vehicle to drive into the current lane is 0.3, the probability of driving into the right lane is 0.6, the probability of driving into the left lane is 0.1, and then the model parameters can be updated aiming at the current lane by taking the probability of driving into the current lane, which is predicted to be 0.3, approaching to the probability of driving into the current lane, which is 0 label 0 as a target; similarly, the model parameters are updated for the right lane with the predicted probability of the vehicle driving into the right lane 0.6 approaching the probability of the vehicle driving into the right lane label 1 as a target, and the model parameters are updated for the left lane with the predicted probability of the vehicle driving into the left lane 0.1 approaching the probability of the vehicle driving into the left lane label 0 as a target.
In the above embodiment, the historical driving state of the vehicle may be determined by the historical driving state determining layer based on the training sample, and then the probability predicting layer predicts the probability of the vehicle driving into each candidate lane based on the historical driving state and the training position information of a plurality of candidate lanes, and for each candidate lane, the predicted probability of the vehicle driving into the candidate lane approaches to the probability label of the vehicle driving into the candidate lane as a target, so as to update the parameters of the model and realize the optimization of the model.
FIG. 2 illustrates an alternative network structure of travel intent prediction models, in some embodiments of the application, the network structure employed by the first, second, and third travel intent prediction models may each include: the system comprises an input layer, a historical driving state determining layer and a probability predicting layer which are sequentially cascaded. Based on this, referring to fig. 4, the process of predicting the probability of the vehicle driving into each candidate lane by the first, second, and third driving intention prediction models may include:
and step S300, inputting the history information and the position information of a plurality of candidate lanes into a driving intention prediction model.
Wherein when the travel intention prediction model is a first travel intention prediction model, the history information may include history travel information; when the travel intention prediction model is the second travel intention prediction model, the history information may include history orientation information; when the travel intention prediction model is the third travel intention prediction model, the above-described history information may include history position information.
Specifically, the driving intention prediction model acquires the history driving information of the vehicle and the position information of a plurality of candidate lanes by using an input layer, can vectorize the history information of the vehicle and the position information of the plurality of candidate lanes, and then outputs the vectorized history information of the vehicle and the position information of the plurality of candidate lanes to a history driving state determination layer.
Step 301, the history information is processed by using the history driving state determining layer to obtain the history driving state.
Specifically, when the driving intention prediction model is the first driving intention prediction model, the historical driving information acquired in the above steps can be processed by using the historical driving state determining layer to obtain the historical driving state of the vehicle; when the driving intention prediction model is the second driving intention prediction model, the historical driving state determination layer is utilized to process the historical orientation information acquired in the steps so as to obtain the historical driving state of the vehicle; when the travel intention prediction model is the third travel intention prediction model, the historical position information acquired in the above steps can be processed by using the historical travel state determination layer to obtain the historical travel state of the vehicle.
Step S302, the historical driving state and the position information of a plurality of candidate lanes are processed by utilizing a probability prediction layer so as to output the probability of the vehicle driving into each candidate lane.
Specifically, by using the historical driving state of the vehicle obtained in the above step, the probability prediction layer may be used to process the historical driving state and the position information of a plurality of candidate lanes, and output the probability of the vehicle driving into each candidate lane.
In the above embodiment, the first, second and third driving intention prediction models may be used to analyze the driving information, the direction information and the position information of the vehicle respectively, and finally determine the probability of the vehicle driving into each candidate lane respectively.
In the above embodiment, the historical driving information, the historical orientation information and the historical position information of the vehicle in the preset time period before the current time are obtained, and the obtained information is all relevant data of the vehicle in the preset time period, but in the process of predicting the driving intention of the vehicle, the determination of the historical driving state of the vehicle can be realized by obtaining the data of the time node in the preset time period, so that the calculation amount in the model prediction is reduced. Based on this, in some embodiments of the present application, the time interval may be preset, and the data may be screened according to the preset time interval.
Specifically, after the historical traveling information, the historical orientation information, and the historical position information of the vehicle in a preset period of time before the current time are acquired, traveling data may be selected from the historical traveling information as target historical traveling data, orientation data may be selected from the historical orientation information as target historical orientation data, and position data may be selected from the historical position information as target historical position data at preset time intervals.
After the target historical driving data is obtained, the target historical driving data and the position information of a plurality of candidate lanes can be input into a first driving intention prediction model to obtain a first probability of the vehicle driving into each candidate lane; after the obtained target history orientation data, the target history orientation data and the position information of a plurality of candidate lanes can be input into a second driving intention prediction model to obtain a second probability of the vehicle driving into each candidate lane; after the target historical position data is obtained, the target historical position data and the position information of a plurality of candidate lanes can be input into a third driving intention prediction model, so that a third probability of the vehicle driving into each candidate lane is obtained.
For example, the history travel information, the history orientation information, and the history position information of the vehicle in 4s before the current time are acquired, at this time, the preset time interval is 0.1s, the history travel data of 40 vehicles are selected as the target history travel data of the vehicle from the history travel information of the vehicle in 4s before the current time at the time interval of 0.1s, the history orientation data of 40 vehicles are selected as the target history orientation data of the vehicle from the history orientation information of the vehicle in 4s before the current time at the time interval of 0.1s, and the history position data of 40 vehicles are selected as the target history position data of the vehicle from the history position information of the vehicle in 4s before the current time at the time interval of 0.1 s.
In the above embodiment, the specific amount of data is selected from the historical driving information, the historical orientation information and the historical position information of the vehicle in the preset time period before the current moment in a preset time interval mode, and the selected mode is selected according to the preset time interval, so that the data change trend of the vehicle can be reflected to a certain extent, the screening of the data is realized, and the calculation amount of the model is reduced.
By the traveling intention of the vehicle predicted in the above embodiment, a reference can be provided for the decision of the vehicle. Based on this, in some embodiments of the present application, the predicted travel intent of the vehicle may also be sent to the vehicle decision module.
Specifically, the driving intention of the vehicle may be sent to a vehicle decision module, and the vehicle decision module controls the vehicle to travel according to the driving intention. Since the candidate lanes into which other vehicles are to be driven can be determined in the driving intention of the vehicle, a future driving path of the vehicle can be planned in advance to realize the control of the vehicle.
A description will be given below of a vehicle travel intention prediction apparatus provided in an embodiment of the present application, and a vehicle travel intention prediction apparatus described below and a vehicle travel intention prediction method described above may be referred to correspondingly to each other.
Fig. 5 is a schematic structural diagram of a vehicle driving intention prediction apparatus according to an embodiment of the present application, where the vehicle driving intention prediction apparatus may include:
a candidate lane determining unit 10 for determining a plurality of candidate lanes and position information of each lane according to a current position of the vehicle;
an information acquisition unit 20 that acquires history travel information, history orientation information, and history position information of the vehicle in a preset period of time before the current time;
A first probability prediction unit 30, configured to input the historical driving information and the position information of the plurality of candidate lanes into a first driving intention prediction model, so as to obtain a first probability of the vehicle driving into each candidate lane;
a second probability prediction unit 40, configured to input the history orientation information and the position information of the plurality of candidate lanes into a second driving intention prediction model, so as to obtain a second probability of the vehicle driving into each candidate lane;
a third probability prediction unit 50, configured to input the historical position information and position information of a plurality of candidate lanes into a third driving intention prediction model, so as to obtain a third probability of the vehicle driving into each candidate lane;
and a travel intention prediction unit 60 configured to input the first probability, the second probability, and the third probability into a fusion travel intention prediction model, and determine a travel intention of the vehicle.
In the above embodiment, there is provided a method of predicting the driving intention of a vehicle, in which the candidate lane determining unit 10 determines a plurality of candidate lanes and the position information of each candidate lane according to the current position of the vehicle, the information acquiring unit 20 acquires the history driving information, the history orientation information and the history position information of the vehicle in a preset period of time before the current time, the first probability predicting unit 30 inputs the history driving information and the position information of the plurality of candidate lanes into the first intention predicting model to obtain the first probability of the vehicle driving into each candidate lane, the second probability predicting unit 40 inputs the history orientation information and the position information of the plurality of candidate lanes into the second intention predicting model to obtain the second probability of the vehicle driving into each candidate lane, the third probability predicting unit 50 inputs the history position information and the position information of the plurality of candidate lanes into the third intention predicting model to obtain the third probability of the vehicle driving into each candidate lane, and the first probability, the second probability and the third probability are input into the intention predicting model by the driving intention predicting unit 60 to determine the driving intention of the vehicle. According to the method, the characteristics of the vehicle are divided into the running information, the orientation information and the position information, different running intention prediction models are established according to different characteristics, the prediction results of the different running intention prediction models are input into the fusion running intention prediction model, the running intention of the vehicle is finally obtained, and the complexity of the model is simplified by establishing the different running intention prediction models according to different characteristics, so that the accuracy of the running intention prediction of the vehicle is improved to a certain extent.
Optionally, the vehicle travel intention prediction apparatus may further include:
the model training module is configured to train and obtain the first, second and third driving intention prediction models, where each of the first, second and third driving intention prediction models may include: an input layer, a historical driving state determining layer and a probability predicting layer which are sequentially cascaded;
a training process of the first, second, and third travel intent prediction models, comprising:
acquiring training samples of a vehicle and training position information of a plurality of candidate lanes through an input layer;
determining, by a historical driving state determination layer, a historical driving state of the vehicle based on the training sample;
predicting, by a probability prediction layer, a probability of a vehicle driving into each candidate lane based on the historical driving state and training position information of a plurality of candidate lanes;
aiming at each candidate lane, taking the predicted probability of the vehicle driving into the candidate lane as a target, and updating parameters of the model;
Wherein the training samples of the first driving intention prediction model comprise driving information training samples;
the training samples of the second driving intention prediction model comprise orientation information training samples;
the training samples of the third driving intent prediction model include position information training samples.
Optionally, the step of the first probability prediction unit 30 performing the step of inputting the historical driving information and the position information of the plurality of candidate lanes into the first driving intention prediction model to obtain the first probability of the vehicle driving into each candidate lane may include:
inputting the historical driving information and the position information of a plurality of candidate lanes into a first driving intention prediction model;
processing the historical driving information by utilizing a historical driving state determining layer of the first driving intention prediction model to obtain a historical driving state;
and processing the historical driving state and the position information of the plurality of candidate lanes by using a first probability prediction layer of the first driving intention prediction model so as to output the probability of driving the vehicle into each candidate lane.
Optionally, the vehicle travel intention prediction apparatus may further include:
a target data determining unit, configured to select, according to a preset time interval, driving data from the historical driving information, as target historical driving data;
The step of the first probability prediction unit 30 performing the step of inputting the history driving information and the position information of the plurality of candidate lanes into a first driving intention prediction model to obtain a first probability of the vehicle driving into each candidate lane may include:
and inputting the target historical driving data and the position information of the plurality of candidate lanes into a first driving intention prediction model to obtain a first probability of the vehicle driving into each candidate lane.
Optionally, the vehicle travel intention prediction apparatus may further include:
and the driving intention sending unit is used for sending the driving intention to the vehicle decision module so that the vehicle decision module can control the vehicle to run according to the driving intention.
The embodiment of the application also provides a device for predicting the running intention of a vehicle, fig. 6 shows a hardware structure block diagram of the device for predicting the running intention of the vehicle, and referring to fig. 6, the hardware structure of the device for predicting the running intention of the vehicle may include: at least one processor 1, at least one communication interface 2, at least one memory 3 and at least one communication bus 4;
in the embodiment of the application, the number of the processor 1, the communication interface 2, the memory 3 and the communication bus 4 is at least one, and the processor 1, the communication interface 2 and the memory 3 complete the communication with each other through the communication bus 4;
Processor 1 may be a central processing unit CPU, or a specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement embodiments of the present application, etc.;
the memory 3 may comprise a high-speed RAM memory, and may further comprise a non-volatile memory (non-volatile memory) or the like, such as at least one magnetic disk memory;
wherein the memory stores a program, the processor is operable to invoke the program stored in the memory, the program operable to: each processing flow in the vehicle running intention prediction method is realized.
The embodiment of the present application also provides a storage medium storing a program adapted to be executed by a processor, the program being configured to: each processing flow in the vehicle running intention prediction method is realized.
Finally, it is further noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
In the present specification, each embodiment is described in a progressive manner, and each embodiment focuses on the difference from other embodiments, and each embodiment may be combined with each other, and the same similar parts may be referred to each other.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. A vehicle travel intention prediction method, characterized by comprising:
determining a plurality of candidate lanes and position information of each candidate lane according to the current position of the vehicle, wherein the position information of the candidate lanes comprises coordinate information of the center line of the candidate lane;
acquiring historical driving information, historical orientation information and historical position information of a vehicle in a preset time period before the current moment;
Inputting the historical driving information and the position information of a plurality of candidate lanes into a first driving intention prediction model to obtain a first probability of the vehicle driving into each candidate lane;
inputting the history orientation information and the position information of a plurality of candidate lanes into a second driving intention prediction model to obtain a second probability of the vehicle driving into each candidate lane;
inputting the historical position information and the position information of a plurality of candidate lanes into a third driving intention prediction model to obtain a third probability of the vehicle driving into each candidate lane;
and inputting the first probability, the second probability and the third probability into a fusion driving intention prediction model to determine the driving intention of the vehicle.
2. The method of claim 1, wherein the first, second, and third travel intent prediction models each comprise: an input layer, a historical driving state determining layer and a probability predicting layer which are sequentially cascaded;
a training process of the first, second, and third travel intent prediction models, comprising:
Acquiring training samples of a vehicle and training position information of a plurality of candidate lanes through an input layer;
determining, by a historical driving state determination layer, a historical driving state of the vehicle based on the training sample;
predicting, by a probability prediction layer, a probability of a vehicle driving into each candidate lane based on the historical driving state and training position information of a plurality of candidate lanes;
aiming at each candidate lane, taking the predicted probability of the vehicle driving into the candidate lane as a target, and updating parameters of the model;
wherein the training samples of the first driving intention prediction model comprise driving information training samples;
the training samples of the second driving intention prediction model comprise orientation information training samples;
the training samples of the third driving intent prediction model include position information training samples.
3. The method of claim 2, wherein inputting the historical driving information and the location information of the plurality of candidate lanes into a first driving intent prediction model to obtain a first probability of the vehicle driving into each candidate lane comprises:
inputting the historical driving information and the position information of a plurality of candidate lanes into a first driving intention prediction model;
Processing the historical driving information by utilizing a historical driving state determining layer of the first driving intention prediction model to obtain a historical driving state;
and processing the historical driving state and the position information of the plurality of candidate lanes by utilizing a probability prediction layer of the first driving intention prediction model so as to output a first probability of driving the vehicle into each candidate lane.
4. The method according to claim 1, further comprising, after acquiring the historical travel information, the historical heading information, and the historical position information of the vehicle for a preset period of time before the current time, the steps of:
selecting the driving data from the history driving information according to a preset time interval to serve as target history driving data;
the step of inputting the history driving information and the position information of the plurality of candidate lanes into a first driving intention prediction model to obtain a first probability of the vehicle driving into each candidate lane, comprising:
and inputting the target historical driving data and the position information of the plurality of candidate lanes into a first driving intention prediction model to obtain a first probability of the vehicle driving into each candidate lane.
5. The method of claim 1, wherein the location information of the candidate lane includes: coordinate information of the center line of the candidate lane.
6. The method of claim 1, wherein the historical location information comprises: positional information of the vehicle offset from the center line.
7. The method of any one of claims 1-6, further comprising:
and sending the driving intention to a vehicle decision module so that the vehicle decision module controls the vehicle to run according to the driving intention.
8. A vehicle travel intention prediction apparatus, characterized by comprising:
the system comprises a candidate lane determining unit, a vehicle driving unit and a vehicle driving unit, wherein the candidate lane determining unit is used for determining a plurality of candidate lanes and position information of each lane according to the current position of a vehicle, and the position information of the candidate lanes comprises coordinate information of a candidate lane central line;
an information acquisition unit that acquires history travel information, history orientation information, and history position information of the vehicle in a preset period of time before a current time;
the first probability prediction unit is used for inputting the historical driving information and the position information of the plurality of candidate lanes into a first driving intention prediction model to obtain a first probability of the vehicle driving into each candidate lane;
the second probability prediction unit is used for inputting the history orientation information and the position information of the plurality of candidate lanes into a second driving intention prediction model to obtain a second probability of the vehicle driving into each candidate lane;
The third probability prediction unit is used for inputting the historical position information and the position information of the plurality of candidate lanes into a third driving intention prediction model to obtain a third probability of the vehicle driving into each candidate lane;
and the driving intention prediction unit is used for inputting the first probability, the second probability and the third probability into a fusion driving intention prediction model to determine the driving intention of the vehicle.
9. A vehicle travel intention prediction apparatus characterized by comprising: a memory and a processor;
the memory is used for storing programs;
the processor for executing the program to realize the respective steps of the vehicle running intention prediction method according to any one of claims 1 to 7.
10. A readable storage medium having stored thereon a computer program, characterized in that the computer program, when executed by a processor, implements the steps of the vehicle travel intention prediction method according to any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111437983.1A CN113997954B (en) | 2021-11-29 | 2021-11-29 | Method, device and equipment for predicting vehicle driving intention and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111437983.1A CN113997954B (en) | 2021-11-29 | 2021-11-29 | Method, device and equipment for predicting vehicle driving intention and readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113997954A CN113997954A (en) | 2022-02-01 |
| CN113997954B true CN113997954B (en) | 2023-11-21 |
Family
ID=79930795
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111437983.1A Active CN113997954B (en) | 2021-11-29 | 2021-11-29 | Method, device and equipment for predicting vehicle driving intention and readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113997954B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115331439B (en) * | 2022-08-09 | 2023-08-18 | 山东旗帜信息有限公司 | Expressway interchange traffic flow prediction method based on vehicle history image |
| CN115230743B (en) * | 2022-09-23 | 2022-12-02 | 毫末智行科技有限公司 | Method, system, electronic device and storage medium for predicting obstacle trajectory |
| CN117542205B (en) * | 2024-01-10 | 2024-03-12 | 腾讯科技（深圳）有限公司 | Lane guiding method, device, equipment and storage medium |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20130089351A (en) * | 2012-02-02 | 2013-08-12 | 자동차부품연구원 | A system to detect and to decide the lateral lane change intention of drivers |
| CN107031631A (en) * | 2016-02-04 | 2017-08-11 | 株式会社斯巴鲁 | The travel controlling system of vehicle |
| CN107944624A (en) * | 2017-11-17 | 2018-04-20 | 南京大学 | A kind of unmanned vehicle crossing Driving Decision-making method based on Hidden Markov Model |
| DE102017216202A1 (en) * | 2017-09-13 | 2019-03-14 | Bayerische Motoren Werke Aktiengesellschaft | Method of predicting an optimal lane on a multi-lane road |
| CN110400490A (en) * | 2019-08-08 | 2019-11-01 | 腾讯科技（深圳）有限公司 | Trajectory predictions method and apparatus |
| CN112163521A (en) * | 2020-09-29 | 2021-01-01 | 东软集团股份有限公司 | Vehicle driving behavior identification method, device and equipment |
| CN112415552A (en) * | 2020-11-17 | 2021-02-26 | 北京百度网讯科技有限公司 | Vehicle position determining method and device and electronic equipment |
| CN112562328A (en) * | 2020-11-27 | 2021-03-26 | 腾讯科技（深圳）有限公司 | Vehicle behavior prediction method and device |
| CN112614373A (en) * | 2020-12-29 | 2021-04-06 | 厦门大学 | BiLSTM-based weekly vehicle lane change intention prediction method |
| CN112749740A (en) * | 2020-12-30 | 2021-05-04 | 北京优挂信息科技有限公司 | Method, device, electronic equipment and medium for determining vehicle destination |
| CN113147766A (en) * | 2021-05-21 | 2021-07-23 | 蔚来汽车科技(安徽)有限公司 | Lane change prediction method and device for target vehicle |
| CN113516846A (en) * | 2021-06-24 | 2021-10-19 | 长安大学 | Vehicle lane change behavior prediction model construction, prediction and early warning method and system |
-
2021
- 2021-11-29 CN CN202111437983.1A patent/CN113997954B/en active Active
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20130089351A (en) * | 2012-02-02 | 2013-08-12 | 자동차부품연구원 | A system to detect and to decide the lateral lane change intention of drivers |
| CN107031631A (en) * | 2016-02-04 | 2017-08-11 | 株式会社斯巴鲁 | The travel controlling system of vehicle |
| DE102017216202A1 (en) * | 2017-09-13 | 2019-03-14 | Bayerische Motoren Werke Aktiengesellschaft | Method of predicting an optimal lane on a multi-lane road |
| CN107944624A (en) * | 2017-11-17 | 2018-04-20 | 南京大学 | A kind of unmanned vehicle crossing Driving Decision-making method based on Hidden Markov Model |
| CN110400490A (en) * | 2019-08-08 | 2019-11-01 | 腾讯科技（深圳）有限公司 | Trajectory predictions method and apparatus |
| CN112163521A (en) * | 2020-09-29 | 2021-01-01 | 东软集团股份有限公司 | Vehicle driving behavior identification method, device and equipment |
| CN112415552A (en) * | 2020-11-17 | 2021-02-26 | 北京百度网讯科技有限公司 | Vehicle position determining method and device and electronic equipment |
| CN112562328A (en) * | 2020-11-27 | 2021-03-26 | 腾讯科技（深圳）有限公司 | Vehicle behavior prediction method and device |
| CN112614373A (en) * | 2020-12-29 | 2021-04-06 | 厦门大学 | BiLSTM-based weekly vehicle lane change intention prediction method |
| CN112749740A (en) * | 2020-12-30 | 2021-05-04 | 北京优挂信息科技有限公司 | Method, device, electronic equipment and medium for determining vehicle destination |
| CN113147766A (en) * | 2021-05-21 | 2021-07-23 | 蔚来汽车科技(安徽)有限公司 | Lane change prediction method and device for target vehicle |
| CN113516846A (en) * | 2021-06-24 | 2021-10-19 | 长安大学 | Vehicle lane change behavior prediction model construction, prediction and early warning method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113997954A (en) | 2022-02-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113997954B (en) | Method, device and equipment for predicting vehicle driving intention and readable storage medium | |
| US11403526B2 (en) | Decision making for autonomous vehicle motion control | |
| US11243532B1 (en) | Evaluating varying-sized action spaces using reinforcement learning | |
| US11726477B2 (en) | Methods and systems for trajectory forecasting with recurrent neural networks using inertial behavioral rollout | |
| US11480972B2 (en) | Hybrid reinforcement learning for autonomous driving | |
| CN110834644B (en) | Vehicle control method and device, vehicle to be controlled and storage medium | |
| Nishi et al. | Merging in congested freeway traffic using multipolicy decision making and passive actor-critic learning | |
| JP2022516383A (en) | Autonomous vehicle planning | |
| JP2021526478A (en) | Vehicle control system, how to control the vehicle, and non-temporary computer-readable memory | |
| US11351996B2 (en) | Trajectory prediction of surrounding vehicles using predefined routes | |
| Schulz et al. | Learning interaction-aware probabilistic driver behavior models from urban scenarios | |
| JP5378002B2 (en) | Vehicle motion estimation device, vehicle motion estimation method, and vehicle motion estimation program | |
| JP5070574B2 (en) | Local traffic prediction program generation device, local traffic prediction device, local traffic prediction program generation method, local traffic prediction method and program | |
| Meghjani et al. | Context and intention aware planning for urban driving | |
| CN110843768B (en) | Automatic parking control method, device and equipment for automobile and storage medium | |
| Feher et al. | Q-learning based reinforcement learning approach for lane keeping | |
| CN115560771A (en) | Sampling-based path planning method and device and automatic driving equipment | |
| CN116045998A (en) | Context-aware path planning for autonomously driven vehicles using dynamic step search | |
| CN115923833A (en) | Personifying decision control method and device for vehicle, vehicle and storage medium | |
| US11436504B1 (en) | Unified scene graphs | |
| CN115062202A (en) | Method, device, equipment and storage medium for predicting driving behavior intention and track | |
| CN115454082A (en) | Vehicle obstacle avoidance method and system, computer readable storage medium and electronic device | |
| CN114670867A (en) | Multi-vehicle track prediction system based on hierarchical learning and potential risk model | |
| JP2023531927A (en) | Driving decision-making method, driving decision-making device, and chip | |
| Pop et al. | Evaluation of the Use of an Intelligent System in the Calibration of a Refined Car-Following Model |
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 |