CN114872730A - Vehicle driving track prediction method and device, automobile and storage medium - Google Patents
Vehicle driving track prediction method and device, automobile and storage medium Download PDFInfo
- Publication number
- CN114872730A CN114872730A CN202210547029.6A CN202210547029A CN114872730A CN 114872730 A CN114872730 A CN 114872730A CN 202210547029 A CN202210547029 A CN 202210547029A CN 114872730 A CN114872730 A CN 114872730A
- Authority
- CN
- China
- Prior art keywords
- track
- predicted
- target vehicle
- characteristic
- information
- 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.)
- Pending
Links
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
- B60W60/00—Drive control systems specially adapted for autonomous road vehicles
- B60W60/001—Planning or execution of driving tasks
- B60W60/0015—Planning or execution of driving tasks specially adapted for safety
- B60W60/0016—Planning or execution of driving tasks specially adapted for safety of the vehicle or its occupants
-
- 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
- B60W40/10—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 vehicle motion
-
- 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
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 embodiment of the invention discloses a vehicle running track prediction method, a vehicle running track prediction device, a vehicle and a storage medium, wherein historical track information of a target vehicle to be predicted and peripheral vehicles is obtained, a first track characteristic and a kinematic characteristic of the target vehicle to be predicted and a second track characteristic of the peripheral vehicles are respectively determined on the basis of the historical track information, the first track characteristic, the kinematic characteristic and the second track characteristic, the time attention weight of the first track characteristic of the target vehicle to be predicted at the historical moment and the time attention weight of the first track characteristic of the target vehicle to be predicted at the current moment are determined on the basis of the first track characteristic, the kinematic characteristic and the second track characteristic, and the space attention weight of the peripheral vehicles and the target vehicle to be predicted is determined; and determining space-time characteristic information according to the time attention weight and the space attention weight, and outputting a track prediction result of the target vehicle to be predicted according to the space-time characteristic information through a decoder module. By considering the discrimination of the historical tracks and the influence among the vehicles, the accuracy of the prediction of the vehicle running track can be improved, and the driving safety can be ensured.
Description
Technical Field
The invention relates to the technical field of vehicle driving, in particular to a vehicle running track prediction method and device, an automobile and a storage medium.
Background
With the continuous development of social economy, the number of intelligent driving automobiles on a traffic road is gradually increased, the road traffic condition is more and more complex, and the prediction of the vehicle driving track is harder under the complex traffic environment.
At present, when an intelligent driving automobile is in a high-dynamic scene, the driving behaviors and the driving tracks of other surrounding vehicles cannot be accurately predicted, so that the driving safety is reduced and the automobile decision planning module cannot work normally.
Disclosure of Invention
The invention provides a vehicle running track prediction method and device, an automobile and a storage medium, which are used for improving the accuracy of vehicle running track prediction so as to ensure the driving safety.
According to an aspect of the present invention, there is provided a vehicle travel track method including: acquiring historical track information of a target vehicle to be predicted and peripheral vehicles, and respectively determining a first track characteristic and a kinematic characteristic of the target vehicle to be predicted and a second track characteristic of the peripheral vehicles based on the historical track information;
inputting the first track characteristic and the kinematic characteristic into a time attention module, and determining a time attention weight of the first track characteristic of the target vehicle to be predicted at the historical moment and the first track characteristic of the target vehicle to be predicted at the current moment through the time attention module;
inputting the first track characteristic, the kinematic characteristic and the second track characteristic into a spatial attention module, and determining a spatial attention weight of the surrounding vehicle and the target vehicle to be predicted through the spatial attention module;
and determining the space-time characteristic information of the target vehicle to be predicted according to the time attention weight and the space attention weight, and outputting a track prediction result of the target vehicle to be predicted according to the space-time characteristic information through a decoder module.
According to another aspect of the present invention, there is provided a vehicle travel track apparatus including: the historical track information acquisition module is used for acquiring historical track information of a target vehicle to be predicted and peripheral vehicles, and determining a first track characteristic and a kinematic characteristic of the target vehicle to be predicted and a second track characteristic of the peripheral vehicles respectively based on the historical track information;
the time attention weight determining module is used for inputting the first track characteristics and the kinematic characteristics into the time attention module, and determining the time attention weight of the first track characteristics of the target vehicle to be predicted at the historical moment and the first track characteristics of the target vehicle to be predicted at the current moment through the time attention module;
a spatial attention weight determination module, configured to input the first trajectory feature, the kinematic feature, and the second trajectory feature into a spatial attention module, and determine, by the spatial attention module, a spatial attention weight of the surrounding vehicle and the target vehicle to be predicted;
and the track prediction result output module is used for determining the space-time characteristic information of the target vehicle to be predicted according to the time attention weight and the space attention weight, and outputting the track prediction result of the target vehicle to be predicted according to the space-time characteristic information through the decoder module.
According to another aspect of the present invention, there is provided an automobile including:
at least one processor; and
a memory communicatively coupled to the at least one processor; wherein the content of the first and second substances,
the memory stores a computer program executable by the at least one processor, the computer program being executable by the at least one processor to enable the at least one processor to perform a vehicle driving trajectory method according to any of the embodiments of the present invention.
According to another aspect of the present invention, there is provided a computer-readable storage medium storing computer instructions for causing a processor to implement a vehicle driving trajectory method according to any one of the embodiments of the present invention when the computer instructions are executed.
According to the technical scheme, the historical track information of the target vehicle to be predicted and the historical track information of the surrounding vehicles can be obtained, the first track characteristic and the kinematic characteristic of the target vehicle to be predicted and the second track characteristic of the surrounding vehicles are respectively determined based on the historical track information, the time attention weight of the first track characteristic of the target vehicle to be predicted at the historical moment and the time attention weight of the first track characteristic of the target vehicle to be predicted at the current moment are determined based on the first track characteristic, the kinematic characteristic and the second track characteristic, and the space attention weight of the surrounding vehicles and the target vehicle to be predicted is determined; and determining space-time characteristic information according to the time attention weight and the space attention weight, and outputting a track prediction result of the target vehicle to be predicted according to the space-time characteristic information through a decoder module. In the vehicle running track prediction process, the embodiment of the invention combines the discrimination of the historical track information and the mutual influence among the vehicles, and can improve the accuracy of the vehicle running track prediction, thereby ensuring the driving safety.
It should be understood that the statements in this section do not necessarily identify key or critical features of the embodiments of the present invention, nor do they necessarily limit the scope of the invention. Other features of the present invention will become apparent from the following description.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the description below are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a flowchart of a method for tracking a vehicle according to an embodiment of the present invention;
FIG. 2 is a flowchart of a method for tracking a vehicle according to a second embodiment of the present invention;
fig. 3 is a schematic structural diagram of a vehicle travel track device according to a third embodiment of the invention;
fig. 4 is a schematic structural diagram of an automobile for implementing a vehicle driving track method according to a fourth embodiment of the present invention.
Detailed Description
In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or vehicle that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or vehicle.
Example one
Fig. 1 is a flowchart of a vehicle driving trajectory method according to an embodiment of the present invention, where the embodiment is applicable to a vehicle driving scenario, and the method may be implemented by a vehicle driving trajectory device, which may be implemented in a form of hardware and/or software, and the vehicle driving trajectory device may be configured in an automobile, and in particular, in a vehicle driving trajectory device configured in an automobile. As shown in fig. 1, the method includes:
s110, historical track information of the target vehicle to be predicted and peripheral vehicles is obtained, and a first track characteristic and a kinematic characteristic of the target vehicle to be predicted and a second track characteristic of the peripheral vehicles are determined respectively based on the historical track information.
The target vehicle to be predicted can be a vehicle which is in running and is provided with the vehicle running track prediction provided by the embodiment of the invention; the surrounding vehicles may be one or more vehicles selected in the vicinity of the surrounding vehicle to be predicted after the target vehicle to be predicted is determined; the history track information may be travel track information generated by a vehicle traveling along a lane during a past certain period of time, and further, the vehicle travel track information may be composed of a set of discrete two-dimensional coordinate points.
Specifically, the historical track information of the target vehicle to be predicted and the historical track information of the surrounding vehicles may be represented by the following formulas, respectively:
wherein, { X tar ,X nbs Denotes history track information of the target vehicle to be predicted and history track information of surrounding vehicles, T is 1,2, …, T obs Representing the time span of historical track information that can be observed,
the method comprises the steps of representing specific longitudinal coordinate points and specific lateral coordinate points of a vehicle in a two-dimensional space; denotes a cross-correlation operation, specifically to implement a convolution of the historical trace information.
The kinematic characteristics may be kinematic characteristics of the vehicle that the vehicle should meet certain kinematic rules and constraints during driving.
Exemplarily, S110 may include embedding the historical trajectory information of the target vehicle to be predicted and the surrounding vehicles respectively through a nonlinear activation function, and encoding the embedded historical trajectory information through a long-short term memory network to obtain a first trajectory feature of the target vehicle to be predicted and a second trajectory feature of the surrounding vehicles; and determining the kinematic characteristics of the target vehicle to be predicted based on the historical track information of the target vehicle to be predicted through the full-connection layer network.
The first track information and the second track information are coded information obtained by coding the embedded historical track information based on a long-short term memory network. Further, the embedding of the historical track information of the target vehicle to be predicted and the surrounding vehicles can be realized through the following formula:
e t ＝LeakyReLU(W e *X t +b e )；
wherein the content of the first and second substances,
respectively represents the information of the target vehicle to be predicted and the track information of the vehicles around the target vehicle to be predicted after being embedded by the nonlinear activation function, LeakyReLU (DEG) represents the nonlinear activation function,
showing the historical trajectories of the target vehicle and its surrounding vehicles, { W e ,W b Denotes the weight and bias, respectively, of the embedding function to be trained.
Furthermore, the embedded historical track information can be coded through a long-term and short-term memory network, and the coded information at each moment is stored in a database.
Specifically, a first trajectory feature of the target vehicle to be predicted and a second trajectory feature of the surrounding vehicle may be expressed by the following formula;
h t ＝LSTM(W l ·e t +b l )；
wherein the content of the first and second substances,
can respectively represent a first track characteristic of a target vehicle to be predicted and a second track characteristic of a surrounding vehicle, and the LSTM represents a long-short term memory network module, { W l ,W l Respectively representing the weight and the bias of the LSTM to be trained;
since the target vehicle to be predicted should satisfy a certain kinematic rule and constraint, additional kinematic feature extraction needs to be performed on the target vehicle to be predicted, and specifically, a single-layer Fully Connected Network (FCN) may be used to determine the kinematic feature of the target vehicle to be predicted.
Further, the kinematic characteristics of the target vehicle to be predicted can be extracted by using the following formula:
wherein the content of the first and second substances,
representing the kinematic mapping information of the target vehicle to be predicted, FC representing the full-connection layer network, W f Representing the weight of the full connectivity layer network.
And S120, inputting the first track characteristic and the kinematic characteristic into a time attention module, and determining the time attention weight of the first track characteristic of the target vehicle to be predicted at the historical moment and the first track characteristic of the target vehicle to be predicted at the current moment through the time attention module.
The time attention module can sample the importance of the track coordinates at different historical moments in the vehicle running track prediction, and then screen out the track information at the historical moments which have great influence on the prediction result and increase the time attention weight, so that the importance of the track information is deepened.
For example, determining, by the temporal attention module, temporal attention weights of the historical time trajectory and the current time trajectory of the target vehicle to be predicted may include: respectively calculating the time correlation of the first track characteristics of the target vehicle to be predicted at the current moment and the first track characteristics of each historical moment through a similarity function; and calculating the time attention weight of the first track characteristic at each historical moment and the first track characteristic at the current moment according to the time correlation by using an activation function and a normalized exponential function, wherein the time attention module comprises a similarity function, an activation function and a normalized exponential function.
Further, for the target vehicle to be predicted, the first track characteristic obtained by calculating the current time is
The first track characteristic of each historical moment is
In this embodiment, the importance of each history node may be determined by comparing the first track characteristic with the first track information of each history time, and specifically, the importance of the history node may be represented by the following formula:
specifically, calculating the time correlations between the first trajectory characteristics of the target vehicle to be predicted at the current time and the first trajectory characteristics of the target vehicle to be predicted at each historical time through the similarity function may be expressed by the following formula:
wherein f (-) represents a correlation metric function, W r Representing the linear transformation weights to be learned.
Further, the calculation of the temporal attention weight for each historical time instant from the temporal correlation may be expressed according to the following formula:
wherein alpha is i And representing the proportion of each historical node occupied in the current prediction, namely a time attention weight, wherein softmax (·) is a normalized exponential function, and tanh (·) is a hyperbolic tangent activation function.
And S130, inputting the first track characteristic, the kinematic characteristic and the second track characteristic into a space attention module, and determining the space attention weight of the surrounding vehicle and the target vehicle to be predicted through the space attention module.
The space attention module can be used for describing interaction information between vehicles during driving.
Specifically, assuming that the target vehicle to be predicted and n surrounding vehicles spatially generate an interactive behavior, each trajectory feature corresponding to the n surrounding vehicles at each historical time may be calculated according to S110-S120, which may be denoted as S:
for example, determining, by the spatial attention module, spatial attention weights of the surrounding vehicles and the target vehicle to be predicted may include: calculating the spatial correlation of the first track characteristic and the second track characteristic through a cosine distance measurement function; and calculating the spatial attention weight of the surrounding vehicle and the target vehicle to be predicted according to the spatial correlation through a normalization function, wherein the spatial attention module comprises a cosine distance measurement function and the normalization function.
The track characteristics of each vehicle comprise track characteristic information of each historical moment, namely:
specifically, the calculation of the spatial correlation between the first track feature and the second track feature by the cosine distance metric function can be represented by the following formula:
wherein g (-) is a cosine distance metric function;
specifically, calculating the spatial attention weight of the surrounding vehicle and the target vehicle to be predicted according to the spatial correlation by the normalization function can be represented by the following formula:
wherein, alpha' i Is the spatial attention weight.
And S140, determining the space-time characteristic information of the target vehicle to be predicted according to the time attention weight and the space attention weight, and outputting a track prediction result of the target vehicle to be predicted according to the space-time characteristic information through a decoder module.
The spatiotemporal characteristic information is tensor information without actual physical significance, and can comprise fitting information of vehicle tracks, interaction information among vehicles and importance information of the tracks; the decoder can be used for decoding the space-time characteristic information and converting the space-time characteristic information into a target vehicle track prediction result to be predicted.
Exemplarily, S140 may include: determining a prediction coordinate point based on a pre-configured prediction time domain and a sampling frequency; for each predicted coordinate point, calculating a context vector containing temporal attention information according to the temporal attention weight, and calculating a context vector containing spatial attention information according to the spatial attention weight; determining space-time characteristic information of the target vehicle to be predicted corresponding to the prediction coordinate points according to the context vector containing the time attention information and the context vector containing the space attention information of each prediction coordinate point through a full-connection layer network; and outputting a track prediction result of the target vehicle to be predicted according to the space-time characteristic information of each prediction coordinate point through a decoder module.
Illustratively, calculating the context vector containing the temporal attention information according to the temporal attention weight, and calculating the context vector containing the spatial attention information according to the spatial attention weight may include: calculating a context vector containing time attention information according to the first track characteristics and the time attention weight of each historical moment through a time attention module; calculating, by the spatial attention module, a context vector containing spatial attention information based on the second trajectory features of the surrounding vehicles and the spatial attention weight.
Specifically, the calculation of the context vector containing the temporal attention information by the temporal attention module according to the first trajectory characteristics and the temporal attention weight at each historical time may be represented by the following formula:
wherein v is t Representing a context vector containing time attention information, k representing the number of track features contained in the track of each historical moment, for example, a preset time domain T pred 5 seconds, the sampling frequency is 0.2 seconds, and k is calculated to be 25;
specifically, the calculation of the context vector containing the spatial attention information by the spatial attention module according to the second trajectory characteristics of the surrounding vehicles and the spatial attention weight can be represented by the following formula:
wherein, v' t Representing a context vector containing temporal attention information,
specifically, the spatiotemporal feature information of the target vehicle to be predicted, which is determined to correspond to the prediction coordinate point, may be represented by the following formula:
h final ＝W v v t +W v′ v′ t ；
wherein h is final To include time attentionSpatiotemporal feature information of force and spatial attention.
Specifically, the coordinate points of the future trajectory predicted by the LSTM decoder may be represented by the following formula:
Y tar ＝LSTM(h final ；W final ,b fianl )；
wherein the content of the first and second substances,
the method comprises the steps of obtaining historical track information of a target vehicle to be predicted and peripheral vehicles, respectively determining a first track characteristic and a kinematic characteristic of the target vehicle to be predicted and a second track characteristic of the peripheral vehicles based on the historical track information, determining a time attention weight of the first track characteristic of the target vehicle to be predicted at a historical moment and the first track characteristic of the target vehicle to be predicted at a current moment based on the first track characteristic, the kinematic characteristic and the second track characteristic, and determining a space attention weight of the peripheral vehicles and the target vehicle to be predicted; and determining space-time characteristic information according to the time attention weight and the space attention weight, and outputting a track prediction result of the target vehicle to be predicted according to the space-time characteristic information through a decoder module. Different spatial influences are reflected by applying different spatial weights to surrounding vehicles; a time attention mechanism is designed to analyze the correlation between the current state and the historical state of the target vehicle to be predicted, and time weights of different historical moments are calculated to reflect different time influences; the accuracy of the vehicle running track prediction can be improved, and therefore the driving safety is ensured.
Example two
Fig. 2 is a flowchart of a vehicle driving trajectory method according to a second embodiment of the present invention, and in this embodiment, based on the above embodiments, a "outputting, by a decoder module, a trajectory prediction result of a target vehicle to be predicted according to spatio-temporal feature information of each prediction coordinate point" is further optimized. Wherein explanations of the same or corresponding terms as those of the above embodiments are omitted. Referring to fig. 3, the method for vehicle driving trajectory provided by the embodiment specifically includes the following steps:
s210, obtaining historical track information of the target vehicle to be predicted and surrounding vehicles, and determining a first track characteristic and a kinematic characteristic of the target vehicle to be predicted and a second track characteristic of the surrounding vehicles respectively based on the historical track information.
S220, inputting the first track characteristic and the kinematic characteristic into a time attention module, and determining the time attention weight of the first track characteristic of the target vehicle to be predicted at the historical moment and the first track characteristic of the target vehicle to be predicted at the current moment through the time attention module.
And S230, inputting the first track characteristic, the kinematic characteristic and the second track characteristic into a space attention module, and determining the space attention weight of the surrounding vehicle and the target vehicle to be predicted through the space attention module.
And S240, determining the space-time characteristic information of the target vehicle to be predicted according to the time attention weight and the space attention weight.
S250, mapping the space-time characteristic information of the target vehicle to be predicted at each prediction coordinate point to a driving behavior space through a decoder module to obtain a driving behavior prediction result of each prediction coordinate point; determining the running track of the target vehicle to be predicted according to the driving behavior prediction result of each prediction coordinate point;
wherein the decoder module is configured based on the following driving behavior: longitudinal acceleration, uniform speed, deceleration, lateral left lane change, lane keeping and right lane change.
The driving behavior can be defined by the following formula:
M＝{m k |k＝1,2,…6}；
the driving behavior may include: the method comprises the following steps of longitudinal acceleration, uniform speed and deceleration, lateral lane change to the left, lane keeping and lane change to the right, 6 types of driving behavior prediction results can be obtained by mapping the space-time characteristic information of the target vehicle to be predicted at each prediction coordinate point to a driving behavior space.
Specifically, taking the predicted longitudinal driving behavior as an example, the predicted longitudinal driving behavior and the longitudinal driving behavior probability can be expressed by the following formulas:
B lon ＝FC(h final ；W lon )；
B lat ＝FC(h final ；W lat )；
wherein, { B lon ,B lat Denotes the predicted longitudinal driving behavior and longitudinal driving behavior probabilities, respectively, { W } lon ,W lat Respectively representing the weight of a full-connection layer network of longitudinal driving behaviors and lateral driving behaviors;
specifically, assuming that the predicted trajectory coordinates satisfy a binary gaussian distribution, the mean value thereof can be expressed by the following formula:
wherein the content of the first and second substances,
the predicted longitudinal position and the predicted lateral position are expressed separately, and the standard deviation thereof can be expressed by the following formula:
wherein the content of the first and second substances,
respectively representing the predicted longitudinal standard deviation and the predicted lateral standard, and the correlation coefficient of the binary Gaussian distribution is rho t ；
Considering the prediction time domain T pred At T e (T +1, T + T) for 5 seconds pred ) At the moment, the prediction result should satisfy the following formula:
wherein the content of the first and second substances,
indicating that the target vehicle to be predicted is at (T +1, T + T) pred ) And if the predicted position at a certain moment in the interval is within the interval, the posterior probability distribution of the future track of the target vehicle to be predicted can be represented by the following formula:
wherein the content of the first and second substances,
indicating conditional transition probabilities of the history track information of the target vehicle to be predicted and its surrounding vehicles to the predicted track of the target vehicle,
and expressing parameters of binary Gaussian distribution at each moment in a prediction time domain, and decomposing the posterior probability into the probability of calculating the driving behavior through the historical tracks of the target vehicle to be predicted and the surrounding vehicles, and further predicting the future track by combining the driving behavior.
The above steps establish a probability transfer relationship of the historical track-driving behavior-predicted track, the loss function can be constructed by Root of Mean Square Error (RMSE), and the specific construction can be represented by the following formula:
and S260, determining spatial attention weight distribution information according to the spatial attention weight of each prediction coordinate point, determining temporal attention weight distribution information according to the temporal attention weight of each prediction coordinate point, and outputting the running track, the driving behavior prediction result, the spatial attention weight distribution information and the temporal attention weight distribution information.
The embodiment can continuously send the running data into the model in an iterative mode for optimization to obtain the network parameters which enable the total loss function error to be the lowest, complete model training, and output the running track of the target vehicle to be predicted, the driving behavior prediction result, the time attention weight distribution and the space attention weight distribution when new data flow into the trained prediction model, wherein the former two responses can reflect the accuracy of the prediction result, and the latter two responses can reflect the input data importance sampling and the interactive influence of the target vehicle and the cycle.
The embodiment of the invention can also map the time-space characteristic information of the target vehicle to be predicted at each predicted coordinate point to the driving behavior space through the decoder module to obtain the driving behavior prediction result of each predicted coordinate point to predict the vehicle driving track, so that the prediction accuracy and rationality are improved, the output information not only comprises the driving behavior and the future track, but also comprises the spatial attention weight distribution and the time attention weight distribution, the analysis of the vehicle driving characteristics aiming at different traffic scenes is facilitated, the data-driven track prediction model has interpretability, and the defect of complex abstraction of the traditional deep learning method is overcome.
EXAMPLE III
Fig. 3 is a schematic structural diagram of a vehicle travel track device according to a third embodiment of the present invention. As shown in fig. 3, the apparatus includes:
a historical track information obtaining module 310, configured to obtain historical track information of a target vehicle to be predicted and surrounding vehicles, and determine a first track characteristic and a kinematic characteristic of the target vehicle to be predicted and a second track characteristic of the surrounding vehicles based on the historical track information, respectively;
a time attention weight determination module 320, configured to input the first trajectory feature and the kinematic feature to the time attention module, and determine, by the time attention module, a time attention weight of the first trajectory feature at the historical time and the first trajectory feature at the current time of the target vehicle to be predicted;
the spatial attention weight determination module 330 is configured to input the first trajectory feature, the kinematic feature, and the second trajectory feature into the spatial attention module, and determine spatial attention weights of the surrounding vehicle and the target vehicle to be predicted through the spatial attention module;
and the trajectory prediction result output module 340 is configured to determine the temporal and spatial feature information of the target vehicle to be predicted according to the temporal attention weight and the spatial attention weight, and output the trajectory prediction result of the target vehicle to be predicted according to the temporal and spatial feature information through the decoder module.
Optionally, the historical track information obtaining module 310 may include:
the historical track information coding unit is used for embedding the historical track information of the target vehicle to be predicted and the peripheral vehicles through a nonlinear activation function respectively, and coding the embedded historical track information through a long-term and short-term memory network to obtain a first track characteristic of the target vehicle to be predicted and a second track characteristic of the peripheral vehicles;
and the kinematic feature determining unit is used for determining the kinematic features of the target vehicle to be predicted based on the historical track information of the target vehicle to be predicted through the full-connection layer network.
Optionally, the temporal attention weight determination module 320 may include:
the first track characteristic and time attention weight calculation unit is used for a similarity function calculation module and is used for calculating the time correlation between the first track characteristic of the target vehicle to be predicted at the current moment and the first track characteristic of each historical moment through the similarity function;
optionally, the spatial attention weight determination module 330 may include:
the cosine distance measurement function calculation unit is used for calculating the spatial correlation of the first track characteristic and the second track characteristic through a cosine distance measurement function;
and the normalization function calculation unit is used for calculating the spatial attention weight of the surrounding vehicle and the target vehicle to be predicted according to the spatial correlation through the normalization function, wherein the spatial attention module comprises a cosine distance measurement function and the normalization function.
Optionally, the trajectory prediction result output 340 may include:
the prediction coordinate point prediction unit is used for determining a prediction coordinate point based on a pre-configured prediction time domain and a sampling frequency;
a context vector calculation unit configured to calculate, for each of the prediction coordinate points, a context vector including temporal attention information according to the temporal attention weight, and a context vector including spatial attention information according to the spatial attention weight;
the spatial-temporal characteristic information determining unit is used for determining spatial-temporal characteristic information of the target vehicle to be predicted corresponding to the prediction coordinate points according to the context vector containing the time attention information and the context vector containing the space attention information of each prediction coordinate point through the full-connection layer network;
and the track prediction result output unit is used for outputting the track prediction result of the target vehicle to be predicted according to the space-time characteristic information of each prediction coordinate point through the decoder module.
Optionally, the context vector calculating unit may include:
the context vector calculation subunit of the time attention information is used for calculating context vectors containing the time attention information according to the first track characteristics and the time attention weight of each historical moment through the time attention module;
and the context vector calculation subunit of the spatial attention information is used for calculating a context vector containing the spatial attention information according to the second track characteristics of the surrounding vehicles and the spatial attention weight through the spatial attention module.
Optionally, the trajectory prediction result output module 340 may include:
the mapping unit is used for mapping the space-time characteristic information of the target vehicle to be predicted at each predicted coordinate point to the driving behavior space through the decoder module to obtain the driving behavior prediction result of each predicted coordinate point;
the driving track prediction unit is used for determining the driving track of the target vehicle to be predicted according to the driving behavior prediction result of each prediction coordinate point;
a spatial attention weight distribution information determination unit configured to determine spatial attention weight distribution information from spatial attention weights of the respective prediction coordinate points, determine temporal attention weight distribution information from temporal attention weights of the respective prediction coordinate points, and output a travel trajectory, a driving behavior prediction result, the spatial attention weight distribution information, and the temporal attention weight distribution information; wherein the decoder module is configured based on the following driving behavior: longitudinal acceleration, uniform speed, deceleration, lateral left lane change, lane keeping and right lane change.
The vehicle running track device provided by the embodiment of the invention can execute the vehicle running track method provided by any embodiment of the invention, and has corresponding functional modules and beneficial effects of the execution method.
Example four
Fig. 4 is a schematic structural diagram of an automobile according to a fourth embodiment of the present invention. As shown in fig. 4, the automobile 10 includes at least one processor 11, and a memory communicatively connected to the at least one processor 11, such as a Read Only Memory (ROM)12, a Random Access Memory (RAM)13, and the like, wherein the memory stores a computer program executable by the at least one processor, and the processor 11 may perform various suitable actions and processes according to the computer program stored in the Read Only Memory (ROM)12 or the computer program loaded from the storage unit 18 into the Random Access Memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the automobile 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other via a bus 14. An input/output (I/O) interface 15 is also connected to bus 14.
The processor 11 may be a variety of general and/or special purpose processing components having processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), various specialized Artificial Intelligence (AI) computing chips, various processors running machine learning model algorithms, a Digital Signal Processor (DSP), and any suitable processor, controller, microcontroller, or the like. The processor 11, when executing the various methods described above, implements:
acquiring historical track information of a target vehicle to be predicted and peripheral vehicles, and determining a first track characteristic and a kinematic characteristic of the target vehicle to be predicted and a second track characteristic of the peripheral vehicles based on the historical track information respectively;
inputting the first track characteristic and the kinematic characteristic into a time attention module, and determining a time attention weight of the first track characteristic of the target vehicle to be predicted at the historical moment and the first track characteristic of the target vehicle to be predicted at the current moment through the time attention module;
inputting the first track characteristic, the kinematic characteristic and the second track characteristic into a space attention module, and determining a space attention weight of a surrounding vehicle and a target vehicle to be predicted through the space attention module;
and determining the space-time characteristic information of the target vehicle to be predicted according to the time attention weight and the space attention weight, and outputting a track prediction result of the target vehicle to be predicted through a decoder module according to the space-time characteristic information.
Computer programs for implementing the methods of the present invention can be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the computer programs, when executed by the processor, cause the functions/acts specified in the flowchart and/or block diagram block or blocks to be performed. A computer program can execute entirely on a machine, partly on a machine, as a stand-alone software package partly on a machine and partly on a remote machine or entirely on a remote machine or server.
In the context of the present invention, a computer-readable storage medium may be a tangible medium that can contain, or store a computer program for use by or in connection with an instruction execution system, apparatus, or device. A computer readable storage medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. Alternatively, the computer readable storage medium may be a machine readable signal medium. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
It should be understood that various forms of the flows shown above may be used, with steps reordered, added, or deleted. For example, the steps described in the present invention may be executed in parallel, sequentially, or in different orders, and are not limited herein as long as the desired results of the technical solution of the present invention can be achieved.
The above-described embodiments should not be construed as limiting the scope of the invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may be made in accordance with design requirements and other factors. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A method for predicting a travel locus of a vehicle, comprising:
acquiring historical track information of a target vehicle to be predicted and peripheral vehicles, and respectively determining a first track characteristic and a kinematic characteristic of the target vehicle to be predicted and a second track characteristic of the peripheral vehicles based on the historical track information;
inputting the first track characteristic and the kinematic characteristic into a time attention module, and determining a time attention weight of the first track characteristic of the target vehicle to be predicted at the historical moment and the first track characteristic of the target vehicle to be predicted at the current moment through the time attention module;
inputting the first track characteristic, the kinematic characteristic and the second track characteristic into a spatial attention module, and determining a spatial attention weight of the surrounding vehicle and the target vehicle to be predicted through the spatial attention module;
and determining the space-time characteristic information of the target vehicle to be predicted according to the time attention weight and the space attention weight, and outputting a track prediction result of the target vehicle to be predicted according to the space-time characteristic information through a decoder module.
2. The method according to claim 1, wherein the obtaining of historical trajectory information of a target vehicle to be predicted and a surrounding vehicle, the determining of a first trajectory characteristic and a kinematic characteristic of the target vehicle to be predicted and a second trajectory characteristic of the surrounding vehicle based on the historical trajectory information, respectively, comprises:
embedding historical track information of the target vehicle to be predicted and the surrounding vehicles through a nonlinear activation function, and coding the embedded historical track information through a long-term and short-term memory network to obtain a first track characteristic of the target vehicle to be predicted and a second track characteristic of the surrounding vehicles;
and determining the kinematic characteristics of the target vehicle to be predicted based on the historical track information of the target vehicle to be predicted through a full-connection layer network.
3. The method of claim 1, wherein the determining, by the temporal attention module, temporal attention weights for the historical time trajectory and the current time trajectory of the target vehicle to be predicted comprises:
respectively calculating the time correlation of the first track characteristics of the target vehicle to be predicted at the current moment and the first track characteristics of the target vehicle to be predicted at each historical moment through a similarity function;
and calculating the time attention weight of the first track characteristic at each historical moment and the first track characteristic at the current moment according to the time correlation through an activation function and a normalized exponential function, wherein the time attention module comprises a similarity function, an activation function and a normalized exponential function.
4. The method of claim 1, wherein the determining, by the spatial attention module, spatial attention weights of the surrounding vehicle and the target vehicle to be predicted comprises:
calculating the spatial correlation of the first track feature and the second track feature through a cosine distance metric function;
and calculating the spatial attention weight of the surrounding vehicle and the target vehicle to be predicted according to the spatial correlation through a normalization function, wherein the spatial attention module comprises a cosine distance measurement function and a normalization function.
5. The method according to claim 1, wherein the determining the spatiotemporal feature information of the target vehicle to be predicted according to the temporal attention weight and the spatial attention weight, and outputting the trajectory prediction result of the target vehicle to be predicted according to the spatiotemporal feature information through a decoder module comprises:
determining a prediction coordinate point based on a pre-configured prediction time domain and a sampling frequency;
for each prediction coordinate point, calculating a context vector containing temporal attention information according to the temporal attention weight, and calculating a context vector containing spatial attention information according to the spatial attention weight;
determining space-time characteristic information of the target vehicle to be predicted corresponding to the prediction coordinate points according to the context vector containing the time attention information and the context vector containing the space attention information of each prediction coordinate point through a full-connection layer network;
and outputting a track prediction result of the target vehicle to be predicted according to the space-time characteristic information of each prediction coordinate point through a decoder module.
6. The method of claim 5, wherein computing the context vector containing temporal attention information according to the temporal attention weight and computing the context vector containing spatial attention information according to the spatial attention weight comprises:
calculating a context vector containing time attention information according to the first track characteristics of each historical moment and the time attention weight through the time attention module;
calculating, by the spatial attention module, a context vector including spatial attention information according to the second trajectory feature of the surrounding vehicle and the spatial attention weight.
7. The method according to claim 1, wherein the outputting, by a decoder module, a trajectory prediction result of the target vehicle to be predicted from the spatiotemporal feature information of each prediction coordinate point includes:
mapping the space-time characteristic information of the target vehicle to be predicted at each prediction coordinate point to a driving behavior space through a decoder module to obtain a driving behavior prediction result of each prediction coordinate point;
determining the running track of the target vehicle to be predicted according to the driving behavior prediction result of each prediction coordinate point;
determining spatial attention weight distribution information according to the spatial attention weight of each predicted coordinate point, determining temporal attention weight distribution information according to the temporal attention weight of each predicted coordinate point, and outputting the driving track, the driving behavior prediction result, the spatial attention weight distribution information and the temporal attention weight distribution information;
wherein the decoder module is configured based on the following driving behavior: longitudinal acceleration, uniform speed, deceleration, lateral left lane change, lane keeping and right lane change.
8. A vehicle travel track prediction apparatus characterized by comprising:
the historical track information acquisition module is used for acquiring historical track information of a target vehicle to be predicted and peripheral vehicles, and determining a first track characteristic and a kinematic characteristic of the target vehicle to be predicted and a second track characteristic of the peripheral vehicles respectively based on the historical track information;
the time attention weight determining module is used for inputting the first track characteristics and the kinematic characteristics into the time attention module, and determining the time attention weight of the first track characteristics of the target vehicle to be predicted at the historical moment and the first track characteristics of the target vehicle to be predicted at the current moment through the time attention module;
a spatial attention weight determination module, configured to input the first trajectory feature, the kinematic feature, and the second trajectory feature into a spatial attention module, and determine, by the spatial attention module, a spatial attention weight of the surrounding vehicle and the target vehicle to be predicted;
and the track prediction result output module is used for determining the space-time characteristic information of the target vehicle to be predicted according to the time attention weight and the space attention weight, and outputting the track prediction result of the target vehicle to be predicted according to the space-time characteristic information through the decoder module.
9. An automobile, characterized in that the automobile comprises:
at least one processor; and
a memory communicatively coupled to the at least one processor; wherein the content of the first and second substances,
the memory stores computer programs executable by the at least one processor to enable the at least one processor to perform a vehicle driving trajectory prediction method as claimed in any one of claims 1 to 7.
10. A computer-readable storage medium storing computer instructions for causing a processor to implement a vehicle travel track prediction method according to any one of claims 1 to 7 when executed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210547029.6A CN114872730A (en) | 2022-05-18 | 2022-05-18 | Vehicle driving track prediction method and device, automobile and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210547029.6A CN114872730A (en) | 2022-05-18 | 2022-05-18 | Vehicle driving track prediction method and device, automobile and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114872730A true CN114872730A (en) | 2022-08-09 |
Family
ID=82676731
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210547029.6A Pending CN114872730A (en) | 2022-05-18 | 2022-05-18 | Vehicle driving track prediction method and device, automobile and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114872730A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115257814A (en) * | 2022-08-19 | 2022-11-01 | 东软睿驰汽车技术（沈阳）有限公司 | Method, device, equipment and storage medium for predicting lane change of vehicle |
| CN116070780A (en) * | 2023-02-28 | 2023-05-05 | 小米汽车科技有限公司 | Evaluation method and device of track prediction algorithm, medium and vehicle |
-
2022
- 2022-05-18 CN CN202210547029.6A patent/CN114872730A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115257814A (en) * | 2022-08-19 | 2022-11-01 | 东软睿驰汽车技术（沈阳）有限公司 | Method, device, equipment and storage medium for predicting lane change of vehicle |
| CN116070780A (en) * | 2023-02-28 | 2023-05-05 | 小米汽车科技有限公司 | Evaluation method and device of track prediction algorithm, medium and vehicle |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Xu et al. | Curvelane-nas: Unifying lane-sensitive architecture search and adaptive point blending | |
| CN114872730A (en) | Vehicle driving track prediction method and device, automobile and storage medium | |
| CN112734808B (en) | Trajectory prediction method for vulnerable road users in vehicle driving environment | |
| CN113044064B (en) | Vehicle self-adaptive automatic driving decision method and system based on meta reinforcement learning | |
| CN111401557B (en) | Agent decision making method, AI model training method, server and medium | |
| CN111238523B (en) | Method and device for predicting motion trail | |
| CN111589157B (en) | AI model using method, apparatus and storage medium | |
| CN114889638A (en) | Trajectory prediction method and system in automatic driving system | |
| Liao et al. | Online prediction of lane change with a hierarchical learning-based approach | |
| CN111695737A (en) | Group target advancing trend prediction method based on LSTM neural network | |
| CN113068131A (en) | Method, device, equipment and storage medium for predicting user movement mode and track | |
| CN114021840A (en) | Channel switching strategy generation method and device, computer storage medium and electronic equipment | |
| CN114826947B (en) | Flow matrix recovery prediction method and system based on automatic encoder | |
| CN114997307A (en) | Trajectory prediction method, apparatus, device and storage medium | |
| CN113159403A (en) | Method and device for predicting pedestrian track at intersection | |
| CN112418432A (en) | Analyzing interactions between multiple physical objects | |
| CN115112141A (en) | Vehicle path planning method and system, electronic device and storage medium | |
| CN111582589A (en) | Car rental insurance prediction method, device, equipment and storage medium | |
| CN115952930A (en) | Social behavior body position prediction method based on IMM-GMR model | |
| CN112820111B (en) | Abnormal vehicle identification method and device and electronic equipment | |
| WO2022229404A1 (en) | Motion planning | |
| CN113537258B (en) | Action track prediction method and device, computer readable medium and electronic equipment | |
| CN114428889A (en) | Trajectory path binding method, model training method, device, equipment and storage medium | |
| CN117036966B (en) | Learning method, device, equipment and storage medium for point feature in map | |
| Benrachou et al. | Improving Efficiency and Generalisability of Motion Predictions With Deep Multi-Agent Learning and Multi-Head Attention |
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 |