CN115743101A - Vehicle track prediction method, and track prediction model training method and device - Google Patents
Vehicle track prediction method, and track prediction model training method and device Download PDFInfo
- Publication number
- CN115743101A CN115743101A CN202211593201.8A CN202211593201A CN115743101A CN 115743101 A CN115743101 A CN 115743101A CN 202211593201 A CN202211593201 A CN 202211593201A CN 115743101 A CN115743101 A CN 115743101A
- Authority
- CN
- China
- Prior art keywords
- track
- vehicle
- point data
- trajectory
- data
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Abstract
The embodiment of the application provides a vehicle track prediction method, a track prediction model training method and a device, relates to the technical field of intelligent driving, and comprises the following steps: the method comprises the steps of obtaining a first track representation sequence of a first target vehicle, inputting the first track representation sequence into a vehicle track prediction model to obtain a second track representation sequence output by the vehicle track prediction model, optimizing each track point data in the second track representation sequence according to track point data of the first target vehicle at the time t to obtain a predicted target track of the first target vehicle from the time t +1 to the time t + n, and performing intelligent driving control on the second target vehicle according to the target track. And predicting the track of the vehicle by adopting the trained vehicle track prediction model, and optimizing the predicted track. The accuracy of the track prediction of the vehicle can be effectively improved, and the driving safety is further improved.
Description
Technical Field
The application relates to the technical field of intelligent driving, in particular to a vehicle trajectory prediction method, a trajectory prediction model training method and a device.
Background
The track prediction means that a track prediction model is established by adopting historical track data of an object, a known part of a track is used as the input of the prediction model, and a road where the next moment of the track is located is obtained through deduction operation of the model, and belongs to one of position prediction.
At present, a dynamic model, a Kalman filtering algorithm model and other modes are usually adopted for trajectory prediction of a vehicle, however, when the vehicle runs on a road, the running state is constrained by the vehicle state, road traffic rules and the like, the establishment process of the dynamic model is relatively complex and has higher difficulty, and the Kalman filtering algorithm model based on kinematics has certain advantages and higher calculation efficiency in a short time, but the prediction precision is obviously reduced when the prediction is carried out in a longer time domain.
Disclosure of Invention
The embodiment of the application provides a vehicle track prediction method, a track prediction model training method and a device, which can improve the accuracy of vehicle track prediction.
In a first aspect, an embodiment of the present application provides a vehicle trajectory prediction method, including:
the method comprises the steps of obtaining a first track characterization sequence of a first target vehicle, wherein the first track characterization sequence comprises track point data of the first target vehicle from t-m moment to t moment, and the track point data comprises: the position and speed of the trajectory point;
inputting the first track representation sequence into a vehicle track prediction model to obtain a second track representation sequence output by the vehicle track prediction model, wherein the second track representation sequence comprises predicted track point data of the first target vehicle from t +1 to t + n; both m and n are integers greater than 1;
optimizing each track point data in the second track characterization sequence according to the track point data of the first target vehicle at the time t to obtain a predicted target track from the time t +1 to the time t + n of the first target vehicle;
and carrying out intelligent driving control on a second target vehicle according to the target track.
Optionally, the optimizing each trajectory point data in the second trajectory representation sequence according to the trajectory point data of the first target vehicle at the time t to obtain the predicted target trajectory of the first target vehicle from the time t +1 to the time t + n includes:
taking each track point data in the second track representation sequence as observation data, and acquiring estimated track point data corresponding to each track point data in a plurality of motion modes according to the observation data and the track point data of the first target vehicle at the time t;
obtaining optimized data of each track point according to the estimated track point data in a plurality of motion modes corresponding to the data of each track point;
and obtaining the predicted target track from the t +1 moment to the t + n moment of the first target vehicle according to the optimized track point data.
Optionally, the step of taking each track point data in the second track characterization sequence as observation data, obtaining estimated track point data of the first target vehicle at the time t in multiple motion modes corresponding to each track point data according to the observation data and the track point data of the first target vehicle at the time t includes:
acquiring initial estimated track point data under a plurality of motion modes corresponding to each track point data according to a state equation corresponding to each motion mode and the track point data of the first target vehicle at the time t;
respectively obtaining a gain matrix of each initial estimation track point data under each motion mode;
and obtaining estimated track point data in a plurality of motion modes corresponding to each track point data according to the observation data, the gain matrix of each initial track point data in each motion mode and the initial estimated track point data in the plurality of motion modes corresponding to each track point data.
Optionally, the obtaining optimized data of each track point according to the estimated track point data in the multiple motion modes corresponding to each data of the track point includes:
and according to the weight corresponding to each motion mode, weighting the estimated track point data under the plurality of motion modes corresponding to each track point data to obtain each optimized track point data.
In a second aspect, an embodiment of the present application provides a vehicle trajectory prediction model training method, including:
acquiring historical track data of M sample vehicles;
generating a sample data set according to historical track point data of M sample vehicles from t-M to t + n, wherein the sample data set comprises: generating a first sample track characterization sequence by using the historical track data of each sample vehicle, and generating a label track characterization sequence, wherein the first sample track characterization sequence comprises the historical track point data of the sample vehicle from the time t-m to the time t, and the label track characterization sequence comprises the historical track point data of the sample vehicle from the time t +1 to the time t + n; the historical track point data comprises: the position and speed of the trajectory point; m, M and n are integers greater than 1;
and training a vehicle track prediction model by using the sample data set to obtain the trained vehicle track prediction model.
Optionally, the training the vehicle trajectory prediction model by using the sample data set to obtain a trained vehicle trajectory prediction model includes:
dividing the sample data set into a training set and a test set, the training set comprising: a first sample trajectory characterization sequence of M1 sample vehicles, and a tag trajectory characterization sequence, the test set comprising: the first sample track characterization sequences of the M2 sample vehicles and the label track characterization sequences, wherein M1+ M2 is not more than M;
during training of the vehicle trajectory prediction model:
if the first loss function value of the vehicle track prediction model obtained based on the training set is larger than a preset value, updating parameters of the vehicle track prediction model;
if the first loss function value is smaller than or equal to a preset value and a second loss function value of the vehicle track prediction model obtained based on the test set is larger than the preset value, increasing a feature extraction layer in the vehicle track prediction model;
and if the first loss function value is smaller than or equal to a preset value and the second loss function value is larger than or equal to a preset value, determining that the training of the vehicle track prediction model is finished.
Optionally, the generating a sample data set according to historical track point data of the M sample vehicles from the time t-M to the time t + n includes:
preprocessing historical track point data of M sample vehicles from the time t-M to the time t + n;
and generating the sample data set according to the preprocessed historical track point data of the sample vehicle from the time t-m to the time t + n.
In a third aspect, an embodiment of the present application provides a vehicle trajectory prediction apparatus, including:
the acquisition module is used for acquiring a first track representation sequence of a first target vehicle, wherein the first track representation sequence comprises track point data of the first target vehicle from t-m moment to t moment, and the track point data comprises: the position and speed of the trace points;
the processing module is used for inputting the first track representation sequence into a vehicle track prediction model to obtain a second track representation sequence output by the vehicle track prediction model, and the second track representation sequence comprises predicted track point data of the first target vehicle from t +1 moment to t + n moment; both m and n are integers greater than 1;
the optimization module is used for optimizing each track point data in the second track characterization sequence according to the track point data of the first target vehicle at the time t to obtain a predicted target track of the first target vehicle from the time t +1 to the time t + n;
and the control module is used for carrying out intelligent driving control on a second target vehicle according to the target track.
Alternatively, the trajectory prediction device of the vehicle may execute the trajectory prediction method of the vehicle described in any one of the first aspect.
In a fourth aspect, an embodiment of the present application provides a vehicle trajectory prediction model training device, including:
the acquisition module is used for acquiring historical track data of the M sample vehicles;
the processing module is used for generating a sample data set according to historical track point data of the M sample vehicles from the time t-M to the time t + n, and the sample data set comprises: generating a first sample track characterization sequence by using the historical track data of each sample vehicle, and generating a label track characterization sequence, wherein the first sample track characterization sequence comprises the historical track point data of the sample vehicle from the time t-m to the time t, and the label track characterization sequence comprises the historical track point data of the sample vehicle from the time t +1 to the time t + n; the historical track point data comprises: the position and speed of the trace points; each of M, M and n is an integer greater than 1;
and the training module is used for training the vehicle track prediction model by using the sample data set to obtain the trained vehicle track prediction model.
Alternatively, the trajectory prediction model training device of the vehicle may execute the trajectory prediction model training method of the vehicle according to any one of the second aspects.
In a fifth aspect, an embodiment of the present application provides an electronic device, including: a memory and a processor;
the memory is used for storing computer instructions; the processor is configured to execute the computer instructions stored by the memory to implement the method of any one of the first and/or second aspects.
In a sixth aspect, embodiments of the present application provide a computer-readable storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the method of any one of the first and/or second aspects.
In a seventh aspect, an embodiment of the present application provides a computer program product, which includes a computer program that, when executed by a processor, implements the method of any one of the first and/or second aspects.
According to the vehicle track prediction method and the track prediction model training method and device, a first track representation sequence of a first target vehicle is obtained, the first track representation sequence is input into a vehicle track prediction model, a second track representation sequence output by the vehicle track prediction model is obtained, each track point data in the second track representation sequence is optimized according to track point data of the first target vehicle at the time t, the predicted target track of the first target vehicle from the time t +1 to the time t + n is obtained, and intelligent driving control is performed on the second target vehicle according to the target track. And predicting the track of the vehicle by adopting the trained vehicle track prediction model, and optimizing the predicted track. The accuracy of the track prediction of the vehicle can be effectively improved, and the driving safety is further improved.
Drawings
FIG. 1 provides a scene schematic for an embodiment of the present application;
fig. 2 is a first flowchart illustrating a vehicle trajectory prediction method according to an embodiment of the present disclosure;
FIG. 3 is a schematic diagram of a vehicle trajectory prediction model provided by an embodiment of the present application;
fig. 4 is a second flowchart illustrating a vehicle trajectory prediction method according to an embodiment of the present application;
FIG. 5 is a schematic diagram of optimization of prediction results provided by an embodiment of the present application;
FIG. 6 is a first flowchart illustrating a vehicle trajectory prediction model training method according to an embodiment of the present disclosure;
FIG. 7 is a second flowchart illustrating a vehicle trajectory prediction model training method according to an embodiment of the present disclosure;
FIG. 8 is a schematic diagram of vehicle trajectory prediction provided by an embodiment of the present application;
fig. 9 is a schematic structural diagram of a vehicle trajectory prediction device according to an embodiment of the present application;
FIG. 10 is a schematic structural diagram of a vehicle trajectory prediction model training apparatus according to an embodiment of the present application;
fig. 11 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some embodiments of the present application, but not all embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments in the present application without making any creative effort belong to the protection scope of the present application.
In the embodiments of the present application, the terms "first", "second", and the like are used to distinguish the same or similar items with basically the same functions and actions, and the order of the items is not limited. Those skilled in the art will appreciate that the terms "first," "second," etc. do not denote any order or quantity, nor do the terms "first," "second," etc. denote any order or importance.
It should be noted that in the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described herein as "exemplary" or "e.g.," is not necessarily to be construed as preferred or advantageous over other embodiments or designs. Rather, use of the word "exemplary" or "such as" is intended to present relevant concepts in a concrete fashion.
The vehicle with the advanced assistant driving system can utilize the vehicle-mounted sensor to sense the surrounding environment in the driving process of the vehicle, identify static and dynamic objects in the surrounding environment and feed back the static and dynamic objects to potential dangers in the environment of a driver, so that the comfort and the safety in the driving process of the vehicle are effectively improved. The method is an important part of a high-grade auxiliary driving system and is also a necessary operation for realizing automatic driving of L3 and above.
The traditional trajectory prediction comprises a dynamics model, a Kalman filtering algorithm model, a machine learning model and the like. A dynamic model is built for vehicles around the self vehicle for track prediction, the model is difficult to build accurately, and the prediction precision is generally low.
When the vehicle runs on a road, the running state is constrained by the vehicle state, road traffic rules and the like, a Kalman filtering equation is established based on a kinematic state equation to predict the track, the prediction precision has certain advantages in a short time and high calculation efficiency, but the precision is obviously reduced when the prediction is carried out in a longer time domain.
In view of the above, the application provides a vehicle trajectory prediction method, a vehicle trajectory prediction model training method and a vehicle trajectory prediction model training device, which are capable of effectively improving the accuracy of vehicle trajectory prediction by predicting the vehicle trajectory by using a deep learning trajectory prediction model, optimizing the predicted trajectory by combining with a kinematic state equation, and utilizing the advantage of deep learning in a long-term domain.
The following describes the technical solutions of the present invention and how to solve the above technical problems with specific embodiments. The following several specific embodiments may be combined with each other, and details of the same or similar concepts or processes may not be repeated in some embodiments. Embodiments of the present invention will be described below with reference to the accompanying drawings.
Fig. 1 is a schematic view of a scenario provided in an embodiment of the present application, as shown in fig. 1, including a vehicle 101, a vehicle 102, and a sensor 103 mounted in the vehicle 101.
The sensors 103 in the vehicle 101 may collect historical trajectory data of the vehicle 102, e.g., the vehicle 102 has a historical 4 second position and speed. The sensors 103 may input the collected historical trajectory data of the vehicle 102 into a processing unit of the vehicle 101, which may derive predicted trajectory data of the vehicle 102, e.g., a position and a speed 5 seconds into the future, based on the received data and a pre-trained vehicle trajectory prediction model.
The vehicle 101 may perform intelligent driving control of the vehicle 101 based on the predicted trajectory data obtained by the processing unit. For example, if the vehicle 101 can determine from the predicted trajectory data that the vehicle 102 will turn within 3 seconds, the vehicle 101 can take braking action to increase the separation between the vehicle 101 and the vehicle 102.
Optionally, in this embodiment of the present application, the vehicle 102 may be any vehicle around the vehicle 101, the types of the vehicle 101 and the vehicle 102 may be vehicles, and may also be other types of vehicles, and the embodiment of the present application does not limit the types of the vehicles.
Alternatively, in the embodiment of the present application, the vehicle trajectory prediction model may be stored in a storage unit of the vehicle 101 in advance.
Alternatively, the sensor 103 may comprise an industrial camera, a lighting device camera mount, or the like, and the lighting device comprises a light sensor, an auxiliary light source. The embodiment of the present application does not limit the type of the sensor 103.
Fig. 1 briefly illustrates an application scenario of the embodiment of the present application, and a vehicle 101 in fig. 1 is taken as an example to describe a trajectory prediction method of the vehicle in the embodiment of the present application in detail.
Fig. 2 is a schematic flowchart of a method for predicting a trajectory of a vehicle according to an embodiment of the present disclosure, as shown in fig. 2, including the following steps:
s201, obtaining a first track representation sequence of the first target vehicle, wherein the first track representation sequence comprises track point data of the first target vehicle from t-m to t.
In the embodiment of the present application, as shown in fig. 1, the first target vehicle may be any vehicle on the road except for the vehicle 101. The first trajectory characterization sequence may be trajectory point data representing m instants of time of the first target vehicle, m being an integer greater than 1. The track points may be the track of the first target vehicle at any one time, and the track point data includes the position and speed of the track points.
Wherein the position of the track point comprises the sum of the track point in the x directiony-direction position, track point velocity includes the velocity of the track point in the x-direction and the y-direction. For example, for any time t, the trajectory point data at time t may be represented as C t ＝{x t ,y t ,v xt ,v yt }。
In the embodiment of the application, a first track characterization sequence of a first target vehicle can be obtained through a sensor mounted on the vehicle.
S202, inputting the first track representation sequence into a vehicle track prediction model to obtain a second track representation sequence output by the vehicle track prediction model, wherein the second track representation sequence comprises track point data of the predicted first target vehicle from t +1 moment to t + n moment.
In the embodiment of the application, the vehicle trajectory prediction model may be a network model based on deep learning.
Illustratively, as shown in fig. 3, the vehicle trajectory prediction model according to the embodiment of the present application is composed of two main components, namely, a Gate Recovery Unit (GRU) and an Attention Mechanism (AM), and the GRU-AM prediction model is composed of an input layer, a GRU layer, an AM layer, and an output layer. Wherein the GRU layer comprises at least one GRU and the AM layer comprises at least one attention cell h i . The GRU layer may also be referred to as a feature extraction layer, and is configured to perform feature extraction and data conversion on data input by the input layer, and the AM layer may further analyze data input by the GRU layer.
In the embodiment of the application, the vehicle track prediction model can be a model which is obtained through training and meets the use requirement.
In the embodiment of the application, the first track characterization sequence can be input into an input layer of the vehicle track prediction model, the first track characterization sequence is input into the GRU layer through the input layer, and the GRU layer performs preliminary prediction on the first track characterization sequence to obtain a preliminary prediction result. And then, transmitting the preliminary prediction result into an AM layer for further prediction to obtain a more accurate prediction result, and finally outputting the more accurate prediction result through an output layer to obtain a second track representation sequence output by the vehicle track prediction model.
Optionally, in order to make the practical significance of the prediction result output by the vehicle trajectory prediction model more intuitive and readable, the prediction result of the vehicle trajectory prediction model may be subjected to inverse normalization processing, that is, inverse normalization processing is performed on the second trajectory representation sequence representation data.
And S203, optimizing each track point data in the second track representation sequence according to the track point data of the first target vehicle at the time t to obtain the predicted target track from the time t +1 to the time t + n of the first target vehicle.
In the embodiment of the application, because the vehicle track prediction model is obtained by training according to the data actually sampled by the vehicle, the influence caused by noise is difficult to avoid in the data sampling process. Therefore, the second trajectory representation sequence output by the vehicle trajectory prediction model may have an error with the actual trajectory point of the first target vehicle, and the second trajectory representation sequence output by the vehicle trajectory prediction model needs to be further optimized to obtain a more accurate second trajectory representation sequence.
In the embodiment of the application, the second trajectory representation sequence output by the vehicle trajectory prediction model can be optimized based on the kinematic state equation.
Illustratively, a Kalman filtering equation can be established based on a kinematic state equation, and a second trajectory representation sequence output by the vehicle trajectory prediction model is optimized to obtain a target trajectory from the t +1 moment to the t + n moment of the first target vehicle.
For example, according to the track point data of the first target vehicle at the time t, a Kalman filtering equation is combined to obtain a predicted value of the first target vehicle from the time t +1 to the time t + n, track point data of each corresponding time in the second track characterization sequence is used as an observed quantity, track point data of each optimized time is obtained, and then the predicted target track of the first target vehicle from the time t +1 to the time t + n is obtained.
And S204, carrying out intelligent driving control on the second target vehicle according to the target track.
In the embodiment of the application, when the target track of the first target vehicle is obtained, the driving strategy of the second target vehicle can be correspondingly adjusted according to the target track.
For example, continuing to refer to fig. 1, if the vehicle 101 obtains the target trajectory of the vehicle 102, it may be determined that the vehicle 102 will turn after 3 seconds according to the target trajectory, and the vehicle 101 may adopt braking or steering to avoid the influence of turning of the vehicle 102, and optionally, the second target vehicle may be a heavy-duty car.
According to the vehicle track prediction method, the first track representation sequence of the first target vehicle is obtained and is input into the vehicle track prediction model, the second track representation sequence output by the vehicle track prediction model is obtained, each track point data in the second track representation sequence is optimized according to the track point data of the first target vehicle at the time t, the predicted target track of the first target vehicle from the time t +1 to the time t + n is obtained, and intelligent driving control is conducted on the second target vehicle according to the target track. And predicting the track of the vehicle by adopting the trained vehicle track prediction model, and optimizing the predicted track. The accuracy of the track prediction of the vehicle can be effectively improved, and the driving safety is further improved.
Fig. 4 is a second flowchart of the vehicle trajectory prediction method provided in the embodiment of the present application, and further illustrates a method for optimizing each trajectory point data in the second trajectory characterization sequence on the basis of the embodiment shown in fig. 2, as shown in fig. 4, the method includes the following steps:
s401, taking each track point data in the second track representation sequence as observation data, and obtaining estimated track point data corresponding to each track point data in a plurality of motion modes according to the observation data and the track point data of the first target vehicle at the time t.
In the embodiment of the application, the second track representation sequence output by the vehicle track prediction model can be optimized according to the Kalman estimation idea, and the estimated track point data under a plurality of motion modes corresponding to each track point data is obtained. The specific method can be as follows:
a1: and acquiring initial estimated track point data under a plurality of motion modes corresponding to each track point data according to the state equation corresponding to each motion mode and the track point data of the first target vehicle at the time t.
In the embodiment of the application, a certain motion state is assumed to be calculated in a kalman filtering algorithm, and if the assumed state is incorrect, a new error is brought, so that when the second trajectory representation sequence is optimized, the motion state of the first target vehicle is divided into a static mode, a constant speed mode and a uniform acceleration mode.
According to a state equation in a Kalman filtering algorithm and track point data of a vehicle at the time t, initial estimation track point data of different motion modes at the time t +1 can be obtained.
Illustratively, the state equation is as follows:
wherein, the first and the second end of the pipe are connected with each other,
initial estimation of trajectory point data for time t +1, F t+1|t Is the state transition matrix from time t to time t +1, X t And (5) obtaining trace point data at the time t.
And substituting the state transition matrixes of different motion modes into a state equation to obtain initial estimated track point data of different motion modes at the moment of t + 1.
The state transition matrix for different motion modes is as follows:
a static mode:
a uniform speed mode:
a uniform acceleration mode:
wherein, Δ t is the difference between the t +1 time and the t time.
In the embodiment of the application, the value of delta t is changed, and initial estimation track point data of different motion modes at the time from t +2 to t + n can be obtained according to the state equation, namely, the initial estimation track point data under a plurality of motion modes corresponding to each track point data.
a2: and respectively obtaining a gain matrix of each initial estimation track point data under each motion mode.
In the embodiment of the present application, the gain matrix is a kalman gain matrix, and may be determined in the following manner:
s1: and respectively obtaining the variance matrixes of the initial estimation track point data at the t +1 moment in different motion modes.
Determined according to the following formula: variance matrix of initial estimated trajectory point data at time t +1
Wherein, P t+1|t Is the variance matrix from time t to time t +1,
as a transpose of the state transition matrix, P t Is a variance matrix at time t, which can be an identity matrix, Q t Is the noise matrix at time t.
S2: and obtaining a gain matrix of the initial estimation track point data at the t +1 moment according to the variance matrix of the initial estimation track point data at the t +1 moment in different motion modes.
Determining a gain matrix for the initial estimated trajectory point data at time t +1 according to the following equation:
wherein H t+1 For the transition matrix at time t +1,
transposed matrix being the transformation matrix at time t +1, R t+1 the noise matrix at time t + 1.
S3: a gain matrix is obtained for each initial estimated trajectory point data.
The variance matrix of the initial estimated trajectory point data at time t +1 is updated according to the following formula.
P t+1 ＝P t+1|t (I-K t+1 H t+1 )
The process from S1 to S3 is repeated to obtain the gain matrix of the initial estimated trajectory point data corresponding to each time, that is, the gain matrix of each initial estimated trajectory point data in each motion mode can be obtained.
a3: and obtaining estimated track point data under a plurality of motion modes corresponding to each track point data according to the observation data, the gain matrix of each initial estimated track point data under each motion mode and the initial estimated track point data under a plurality of motion modes corresponding to each track point data.
In the embodiment of the application, the estimated track point data under the multiple motion modes corresponding to each track point data can be obtained according to each track point data in the observation data.
For example, taking the time t +1 as an example, the estimated trajectory point data in multiple motion modes corresponding to the time t +1 may be determined according to the following formula.
Wherein, Z t+1 And the locus point data at the t +1 moment in the observation data.
Changing the time in the above formula, the estimated track point data in multiple motion modes corresponding to each time can be obtained, that is, the estimated track point data in multiple motion modes corresponding to each track point data can be obtained.
S402, obtaining optimized each track point data according to the estimated track point data in the plurality of motion modes corresponding to each track point data.
In the embodiment of the application, the estimated track point data in a plurality of motion modes can be comprehensively processed to obtain optimized track point data.
Specifically, according to the weight corresponding to each motion mode, weighting processing is performed on the estimated track point data in the multiple motion modes corresponding to each track point data, so as to obtain optimized each track point data.
For example, each trajectory point data may be determined by the following formula:
wherein the content of the first and second substances,
for the optimized ith trace point data,
ith estimated trajectory point data, α i The weight corresponding to the ith motion pattern.
And S403, obtaining a predicted target track from the t +1 moment to the t + n moment of the first target vehicle according to the optimized point data of each track.
In the embodiment of the application, each optimized track point data is obtained, and each optimized track point data can be connected according to a time sequence to obtain the predicted target track from the t +1 moment to the t + n moment of the first target vehicle.
In summary, referring to fig. 5, in the embodiment of the present application, each trajectory point data output by the vehicle trajectory prediction model is optimized by using a kalman filter method, so as to obtain optimized data corresponding to each trajectory point data in 3 motion modes, and according to weights corresponding to the 3 motion modes, the optimized data corresponding to each trajectory point data in the 3 motion modes is subjected to weighted calculation, so as to finally obtain a predicted target trajectory from time t +1 to time t + n of the first target vehicle.
According to the vehicle trajectory prediction model, optimization processing is carried out on each trace point data output by the vehicle trajectory prediction model, errors of model output results caused by noise influence of sample data in a model training process can be reduced, accuracy of trajectory prediction of a vehicle is improved, and driving safety of the vehicle is improved.
The method for predicting the trajectory of the vehicle by using the vehicle trajectory prediction model provided in the embodiment of the present application is described above, and a method for training the vehicle trajectory prediction model will be described below.
Fig. 6 is a flowchart illustrating a method for training a vehicle trajectory prediction model according to an embodiment of the present disclosure, where an execution subject according to an embodiment of the present disclosure may be an electronic device or a model training platform. The platform may be a platform located in the cloud, or a platform deployed in a distributed manner (e.g., partially deployed in a cloud environment, partially deployed in an edge environment, etc.). As shown in fig. 6, the method comprises the following steps:
s601, obtaining historical track data of M sample vehicles.
In the embodiment of the application, the historical track data refers to data obtained according to the historical running track of the sample vehicle, and comprises position coordinates (x, y) and corresponding speed (v) of the sample vehicle at any moment x 、v y )。
In the embodiment of the application, the historical track data of the N sample vehicles can be obtained according to the running tracks of a plurality of roads and a plurality of vehicles.
In the embodiment of the application, tracks with abnormal data may exist in the acquired historical track data of the N sample vehicles. For example, a sample vehicle trajectory data point is particularly rare; there is a track drift, i.e. the moving distance is very long in a short time; there is a point where the signal is highly delayed, i.e., the moving distance is very short for a long time; the obvious outliers existing in the data, namely a large distance exists between a certain point and other points of adjacent time, and the abnormity obviously exists.
In the embodiment of the application, the track with abnormal data in the historical track data of the N sample vehicles can be filtered and deleted, the track with data missing is corrected by an interpolation method, and the track is completed to obtain the historical track data of the M sample vehicles.
And S602, generating a sample data set according to historical track point data of the M sample vehicles from the time t-M to the time t + n.
In the embodiment of the application, when obtaining the historical track data of M sample vehicles, the historical track in any end duration in the historical track data can be used as data for training the vehicle track prediction model. Namely, the historical track point data of the M sample vehicles from the time t-M to the time t + n is used as training data.
In the embodiment of the present application, sample data and corresponding tag data are generally required for training a model. Therefore, historical track point data of the M sample vehicles from the time t-M to the time t + n can be preprocessed, and a sample data set is generated according to the preprocessed historical track point data of the sample vehicles from the time t-M to the time t + n. The sample data set includes sample data and corresponding tag data.
Specifically, a first sample trajectory representation sequence and a label trajectory representation sequence can be generated by using the historical trajectory data of each sample vehicle, wherein the first sample trajectory representation sequence comprises the historical trajectory point data of the sample vehicle from the time t-m to the time t, and the label trajectory representation sequence comprises the historical trajectory point data of the sample vehicle from the time t +1 to the time t + n; the historical track point data comprises: the position and speed of the trace points.
Optionally, in order to reduce errors in the model training process, normalization processing may be performed on historical track point data of the M sample vehicles from the time t-M to the time t + n, and all data may be converted into data in a range of [0,1 ].
The normalization process may be performed according to the following formula:
in the formula, X is a certain track characteristic value in the raw data, such as X or Y in world coordinates, max and min represent the maximum value and the minimum value of the characteristic value, and X represents a normalized characteristic value.
And S603, training the vehicle track prediction model by using the sample data set to obtain the trained vehicle track prediction model.
In the embodiment of the application, the first sample trajectory representation of each sample vehicle in the sample data can be respectively input into the vehicle trajectory prediction model to obtain the prediction result output by the vehicle trajectory prediction model.
And obtaining the prediction result and the value of the loss function of the corresponding label track representation sequence, and finishing model training if the value of the loss function meets a preset threshold value. If the value of the loss function does not meet the preset threshold value, the parameters of the vehicle track prediction model can be updated according to the value of the loss function, and the training process is repeated by using the vehicle track prediction model with the updated parameters until the trained vehicle track prediction model is obtained.
Alternatively, the loss function merit calculation may use the root mean square error RMSE, i.e., the square root of the sum of the squares of the difference between the predicted and tag values and the ratio of the numbers.
Wherein real t Is a label value, pre t Is a predicted value.
Optionally, the optimizer of the vehicle trajectory prediction model may be an Adam optimizer, the learning rate may be set to 0.001, the batch size may be set to 256, and the activation function may be set to a leak Relu function.
According to the training method of the vehicle track prediction model, the accuracy of the vehicle track prediction model for vehicle track prediction can be improved by training the vehicle track prediction model.
Fig. 7 is a second flowchart of a method for training a vehicle trajectory prediction model according to an embodiment of the present application, where on the basis of the embodiment shown in fig. 6, a process of training the model is further described, and as shown in fig. 7, the method includes the following steps:
s701, dividing the sample data set into a training set and a testing set.
In the embodiment of the application, in order to improve the accuracy of training the vehicle track prediction model, the sample data set can be divided into a training set and a testing set, the model is trained according to the training set, and the trained model is tested according to the testing set.
Illustratively, the sample data set may be divided into a training set and a test set according to a ratio of 7: a first sample trajectory characterization sequence of M1 sample vehicles, and a tag trajectory characterization sequence, comprising in the test set: the first sample track characterization sequences of the M2 sample vehicles and the label track characterization sequences, wherein M1+ M2 is not more than M.
S702, training the vehicle track prediction model based on the training set.
In the embodiment of the application, the first sample trajectory characterization sequence of each sample vehicle in the training set can be respectively input into the vehicle trajectory prediction model to obtain the prediction result of the vehicle trajectory prediction model.
And obtaining a prediction result and a value of a loss function of the corresponding label track representation sequence, and finishing model training if the value of the loss function meets a preset threshold value. If the value of the loss function does not meet the preset threshold, the parameters of the vehicle trajectory prediction model can be updated according to the value of the loss function, and the training process is repeatedly performed by using the vehicle trajectory prediction model with the updated parameters until the obtained loss function meets the preset threshold.
And bringing the vehicle track prediction model meeting the requirements of the training set into the test set for testing.
And S703, testing the vehicle track prediction model trained by the training set based on the test set to obtain the trained vehicle track prediction model.
In the embodiment of the application, the first sample trajectory characterization sequence of each sample vehicle in the test set can be respectively input into the vehicle trajectory prediction model to obtain the prediction result of the vehicle trajectory prediction model.
And obtaining a prediction result and a value of a loss function of the corresponding label track representation sequence, and obtaining a trained vehicle track prediction model if the value of the loss function meets a preset threshold value.
If the loss function does not meet the preset threshold, the number of layers of GRUs in the vehicle track prediction model can be increased, and the steps from S701 to S703 are executed again until the prediction result output by the vehicle track prediction model in the test set and the value of the loss function of the corresponding label track representation sequence meet the preset threshold, so that the trained vehicle track prediction model is obtained.
To sum up, in the trajectory prediction method and the trajectory prediction model training method for a vehicle provided in the embodiments of the present application, as shown in fig. 8, the vehicle trajectory prediction model is trained through the historical trajectory data of the target vehicle, a predicted trajectory output by the vehicle trajectory prediction model is obtained, the predicted trajectory is optimized by using kalman filtering, and the optimized predicted trajectory is output.
According to the predicted track output by the vehicle track prediction model, the accuracy of track prediction in a long time domain can be improved. The prediction track is optimized through Kalman filtering, so that the influence caused by noise in the training process can be reduced, the accuracy of vehicle track prediction is further improved, and the driving safety is improved.
On the basis of the above embodiments, the present application also provides a vehicle trajectory prediction device.
Fig. 9 is a schematic structural diagram of a vehicle trajectory prediction apparatus 90 according to an embodiment of the present application, including:
an obtaining module 901, configured to obtain a first trajectory representation sequence of a first target vehicle, where the first trajectory representation sequence includes trajectory point data of the first target vehicle from a time t-m to a time t, and the trajectory point data includes: the position and speed of the trace point.
The processing module 902 is configured to input the first trajectory representation sequence into a vehicle trajectory prediction model to obtain a second trajectory representation sequence output by the vehicle trajectory prediction model, where the second trajectory representation sequence includes predicted trajectory point data of the first target vehicle from time t +1 to time t + n; and both m and n are integers greater than 1.
And the optimizing module 903 is configured to optimize each trajectory point data in the second trajectory representation sequence according to the trajectory point data of the first target vehicle at the time t, so as to obtain a predicted target trajectory from the time t +1 to the time t + n of the first target vehicle.
And the control module 904 is configured to perform intelligent driving control on the second target vehicle according to the target track.
Optionally, the optimization module 903 is further configured to use each trajectory point data in the second trajectory representation sequence as observation data, and obtain estimated trajectory point data in multiple motion modes corresponding to each trajectory point data according to the observation data and the trajectory point data of the first target vehicle at time t. And obtaining each optimized track point data according to the estimated track point data in the plurality of motion modes corresponding to each track point data. And obtaining the predicted target track from the t +1 moment to the t + n moment of the first target vehicle according to the optimized track point data.
Optionally, the optimization module 903 is further configured to obtain initial estimated trajectory point data in multiple motion modes corresponding to each trajectory point data according to the state equation corresponding to each motion mode and the trajectory point data of the first target vehicle at the time t. And respectively obtaining a gain matrix of each initial estimation track point data under each motion mode. And obtaining estimated track point data in a plurality of motion modes corresponding to each track point data according to the observation data, the gain matrix of each initial track point data in each motion mode and the initial estimated track point data in the plurality of motion modes corresponding to each track point data.
Optionally, the optimizing module 903 is further configured to perform weighting processing on the estimated track point data in the multiple motion modes corresponding to each track point data according to the weight corresponding to each motion mode, so as to obtain each optimized track point data.
The trajectory prediction device of the vehicle provided in the embodiment of the present application may implement the technical solution of the trajectory prediction method of the vehicle provided in any one of the embodiments, and the implementation principle and the technical effect are similar, which are not described herein again.
On the basis of the above embodiments, the embodiment of the present application further provides a vehicle trajectory prediction model training device.
Fig. 10 is a schematic structural diagram of a trajectory prediction apparatus 100 of a vehicle according to an embodiment of the present application, including:
an obtaining module 1001 is configured to obtain historical trajectory data of M sample vehicles.
The processing module 1002 is configured to generate a sample data set according to historical track point data of the M sample vehicles from the time t-M to the time t + n.
The sample data set comprises: a first sample trajectory characterization sequence generated using historical trajectory data for each of the sample vehicles, and a tag trajectory characterization sequence. The first sample track characterization sequence comprises historical track point data of the sample vehicle from t-m to t, and the label track characterization sequence comprises historical track point data of the sample vehicle from t +1 to t + n; the historical track point data comprises: the position and speed of the trace points; and M, M and n are integers which are all more than 1.
And the training module 1003 is configured to train the vehicle trajectory prediction model by using the sample data set, so as to obtain a trained vehicle trajectory prediction model.
Optionally, the training module 1003 is further configured to divide the sample data set into a training set and a test set. The training set includes: a first sample trajectory characterization sequence of M1 sample vehicles, and a tag trajectory characterization sequence. The test set includes: the first sample track characterization sequences of the M2 sample vehicles and the label track characterization sequences, wherein M1+ M2 is not more than M.
Optionally, the training module 1003 is further configured to, in the process of training the vehicle trajectory prediction model: and if the first loss function value of the vehicle track prediction model obtained based on the training set is larger than a preset value, updating the parameters of the vehicle track prediction model. And if the first loss function value is smaller than or equal to a preset value and a second loss function value of the vehicle track prediction model obtained based on the test set is larger than the preset value, increasing a feature extraction layer in the vehicle track prediction model. And if the first loss function value is smaller than or equal to a preset value and the second loss function value is larger than or equal to a preset value, determining that the training of the vehicle track prediction model is finished.
Optionally, the processing module 1002 is further configured to preprocess historical track point data of the M sample vehicles from the time t-M to the time t + n. And generating the sample data set according to the preprocessed historical track point data of the sample vehicle from the time t-m to the time t + n.
The vehicle trajectory prediction model training device provided by the embodiment of the application can execute the technical scheme of the vehicle trajectory prediction model training method provided by any one of the embodiments, the implementation principle and the technical effect are similar, and details are not repeated here.
Fig. 11 is a schematic structural diagram of an electronic device 110 according to an embodiment of the present disclosure. As shown in fig. 11, the electronic device 110 may include: at least one processor 1101, a memory 1102.
A memory 1102 for storing programs. In particular, the program may include program code comprising computer operating instructions. Memory 1102 may include high-speed RAM memory, and may also include non-volatile memory (non-volatile memory), such as at least one disk memory.
The processor 1101 is configured to execute the computer-executable instructions stored in the memory 1102 to implement the training method of the vehicle trajectory prediction model or the technical solution of the vehicle trajectory prediction method embodiment described in the foregoing method embodiment.
The processor 1101 may be a Central Processing Unit (CPU), an Application Specific Integrated Circuit (ASIC), or one or more Integrated circuits configured to implement the embodiments of the present Application.
Optionally, when the electronic device 110 executes the technical solution of the embodiment of the trajectory prediction method of the vehicle, the electronic device 110 may further include a sensor 1103.
The sensor 1103 is configured to collect vehicle trajectory data in front of the road, and transmit the collected vehicle trajectory data to the memory 1102, so as to implement the technical solution of the embodiment of the vehicle trajectory prediction method described in the foregoing embodiment of the method.
Optionally, the electronic device 110 may further include a communication interface 1104 so that the external device may be communicatively interacted with through the communication interface 1104, and the external device may be, for example, a user terminal (e.g., a mobile phone, a tablet). In a specific implementation, if the communication interface 1104, the memory 1102 and the processor 11011 are implemented independently, the communication interface 1104, the memory 1102 and the processor 1101 may be connected to each other through a bus and perform communication with each other. The bus may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, an Extended ISA (EISA) bus, or the like. Buses may be divided into address buses, data buses, control buses, etc., but do not represent only one bus or type of bus.
Optionally, in a specific implementation, if the communication interface 1104, the memory 1102 and the processor 1101 are integrated into a chip, the communication interface 1104, the memory 1102 and the processor 1101 may complete communication through an internal interface.
In an embodiment of the present application, a computer-readable storage medium is further provided, where a computer program is stored, and when the computer program is executed by a processor, the computer program implements the above-mentioned training method for a vehicle trajectory prediction model or the technical solution of the embodiment of the vehicle trajectory prediction method, and the implementation principle and the technical effect are similar, which are not described herein again.
The embodiment of the present application further provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the computer program implements the above-mentioned technical solution of the embodiment of the vehicle trajectory prediction model training method or the vehicle trajectory prediction method, and the implementation principle and the technical effect of the computer program are similar, and are not described herein again.
In one possible implementation, the computer-readable medium may include Random Access Memory (RAM), read-Only Memory (ROM), compact disc Read-Only Memory (CD-ROM) or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer. Also, any connection is properly termed a computer-readable medium. For example, if the software is transmitted from a website, server, or other remote source using a coaxial cable, fiber optic cable, twisted pair, digital Subscriber Line (DSL), or wireless technologies such as infrared, radio, and microwave, then the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are included in the definition of medium. Disk and Disc, as used herein, includes Disc, laser Disc, optical Disc, digital Versatile Disc (DVD), floppy disk and blu-ray Disc where disks usually reproduce data magnetically, while discs reproduce data optically with lasers. Combinations of the above should also be included within the scope of computer-readable media.
The embodiment of the present application further provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, the computer program implements the above-mentioned technical solution of the embodiment of the vehicle trajectory prediction model training method or the vehicle trajectory prediction method, and the implementation principle and the technical effect of the computer program are similar, and are not described herein again.
In the Specific implementation of the terminal device or the server, it should be understood that the Processor may be a Central Processing Unit (CPU), other general purpose processors, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), and the like. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of a method disclosed in connection with the embodiments of the present application may be directly implemented by a hardware processor, or may be implemented by a combination of hardware and software modules in a processor.
Those skilled in the art will appreciate that all or a portion of the steps of any of the above-described method embodiments may be performed by hardware associated with program instructions. The foregoing program may be stored in a computer-readable storage medium, and when executed, performs all or a portion of the steps of the above-described method embodiments.
The technical scheme of the application can be stored in a computer readable storage medium if the technical scheme is realized in a software form and is sold or used as a product. Based on this understanding, all or part of the technical solutions of the present application may be embodied in the form of a software product, which is stored in a storage medium and includes a computer program or several instructions. The computer software product enables a computer device (which may be a personal computer, a server, a network device, or a similar electronic device) to perform all or part of the steps of the method described in the embodiments of the present application.
Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present application.
Claims (10)
1. A vehicle trajectory prediction method, characterized by comprising:
the method comprises the steps of obtaining a first track characterization sequence of a first target vehicle, wherein the first track characterization sequence comprises track point data of the first target vehicle from t-m moment to t moment, and the track point data comprises: the position and speed of the trace points;
inputting the first track representation sequence into a vehicle track prediction model to obtain a second track representation sequence output by the vehicle track prediction model, wherein the second track representation sequence comprises predicted track point data of the first target vehicle from t +1 to t + n; both m and n are integers greater than 1;
optimizing each track point data in the second track characterization sequence according to the track point data of the first target vehicle at the time t to obtain a predicted target track from the time t +1 to the time t + n of the first target vehicle;
and carrying out intelligent driving control on a second target vehicle according to the target track.
2. The method according to claim 1, wherein the optimizing each trajectory point data in the second trajectory characterization sequence according to the trajectory point data of the first target vehicle at the time t to obtain the predicted target trajectory of the first target vehicle from the time t +1 to the time t + n comprises:
taking each track point data in the second track representation sequence as observation data, and acquiring estimated track point data corresponding to each track point data in a plurality of motion modes according to the observation data and the track point data of the first target vehicle at the time t;
obtaining optimized data of each track point according to the estimated track point data under a plurality of motion modes corresponding to the data of each track point;
and obtaining a predicted target track from the t +1 moment to the t + n moment of the first target vehicle according to each optimized track point data.
3. The method according to claim 2, wherein the step of obtaining estimated trajectory point data in a plurality of motion modes corresponding to each trajectory point data by using each trajectory point data in the second trajectory characterization sequence as observation data according to the observation data and the trajectory point data of the first target vehicle at time t comprises:
acquiring initial estimated track point data under a plurality of motion modes corresponding to each track point data according to a state equation corresponding to each motion mode and the track point data of the first target vehicle at the time t;
respectively obtaining a gain matrix of each initial estimation track point data under each motion mode;
and obtaining estimated track point data in a plurality of motion modes corresponding to each track point data according to the observation data, the gain matrix of each initial track point data in each motion mode and the initial estimated track point data in the plurality of motion modes corresponding to each track point data.
4. The method according to claim 2, wherein the obtaining each optimized trajectory point data according to the estimated trajectory point data in the plurality of motion modes corresponding to each trajectory point data comprises:
and according to the weight corresponding to each motion mode, weighting the estimated track point data under the multiple motion modes corresponding to each track point data to obtain optimized each track point data.
5. A vehicle trajectory prediction model training method is characterized by comprising the following steps:
acquiring historical track data of M sample vehicles;
generating a sample data set according to historical track point data of M sample vehicles from t-M to t + n, wherein the sample data set comprises: generating a first sample track characterization sequence by using the historical track data of each sample vehicle, and generating a label track characterization sequence, wherein the first sample track characterization sequence comprises the historical track point data of the sample vehicle from the time t-m to the time t, and the label track characterization sequence comprises the historical track point data of the sample vehicle from the time t +1 to the time t + n; the historical track point data comprises: the position and speed of the trace points; each of M, M and n is an integer greater than 1;
and training a vehicle track prediction model by using the sample data set to obtain the trained vehicle track prediction model.
6. The method of claim 5, wherein training a vehicle trajectory prediction model using the sample data set to obtain a trained vehicle trajectory prediction model comprises:
dividing the sample data set into a training set and a test set, wherein the training set comprises: a first sample trajectory characterization sequence of M1 sample vehicles, and a tag trajectory characterization sequence, the test set comprising: the first sample track characterization sequences of the M2 sample vehicles and the label track characterization sequences, wherein M1+ M2 is not more than M;
in training the vehicle trajectory prediction model:
if the first loss function value of the vehicle track prediction model obtained based on the training set is larger than a preset value, updating parameters of the vehicle track prediction model;
if the first loss function value is smaller than or equal to a preset value and a second loss function value of the vehicle track prediction model obtained based on the test set is larger than the preset value, increasing a feature extraction layer in the vehicle track prediction model;
and if the first loss function value is smaller than or equal to a preset value and the second loss function value is larger than or equal to a preset value, determining that the training of the vehicle track prediction model is finished.
7. A vehicle trajectory prediction device characterized by comprising:
the acquisition module is used for acquiring a first track representation sequence of a first target vehicle, wherein the first track representation sequence comprises track point data of the first target vehicle from t-m moment to t moment, and the track point data comprises: the position and speed of the trace points;
the processing module is used for inputting the first track representation sequence into a vehicle track prediction model to obtain a second track representation sequence output by the vehicle track prediction model, wherein the second track representation sequence comprises predicted track point data of the first target vehicle from t +1 to t + n; both m and n are integers greater than 1;
the optimization module is used for optimizing each track point data in the second track characterization sequence according to the track point data of the first target vehicle at the moment t to obtain a predicted target track from the moment t +1 to the moment t + n of the first target vehicle;
and the control module is used for carrying out intelligent driving control on the second target vehicle according to the target track.
8. A vehicle trajectory prediction model training device, comprising:
the acquisition module is used for acquiring historical track data of the M sample vehicles;
the processing module is used for generating a sample data set according to historical track point data of the M sample vehicles from the time t-M to the time t + n, wherein the sample data set comprises: generating a first sample track characterization sequence by using the historical track data of each sample vehicle, and generating a label track characterization sequence, wherein the first sample track characterization sequence comprises the historical track point data of the sample vehicle from the time t-m to the time t, and the label track characterization sequence comprises the historical track point data of the sample vehicle from the time t +1 to the time t + n; the historical track point data comprises: the position and speed of the trace points; each of M, M and n is an integer greater than 1;
and the training module is used for training the vehicle track prediction model by using the sample data set to obtain the trained vehicle track prediction model.
9. An electronic device, comprising:
a memory for storing a computer program;
a processor for executing the computer program to implement the method of any of claims 1-4 or 5-6.
10. A computer-readable storage medium, having stored thereon a computer program for execution by a processor to implement the method of any one of claims 1-4 or claims 5-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211593201.8A CN115743101A (en) | 2022-12-13 | 2022-12-13 | Vehicle track prediction method, and track prediction model training method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211593201.8A CN115743101A (en) | 2022-12-13 | 2022-12-13 | Vehicle track prediction method, and track prediction model training method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115743101A true CN115743101A (en) | 2023-03-07 |
Family
ID=85345491
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211593201.8A Pending CN115743101A (en) | 2022-12-13 | 2022-12-13 | Vehicle track prediction method, and track prediction model training method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115743101A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116125819A (en) * | 2023-04-14 | 2023-05-16 | 智道网联科技(北京)有限公司 | Track correction method, track correction device, electronic device and computer-readable storage medium |
| CN116558513A (en) * | 2023-07-06 | 2023-08-08 | 中国电信股份有限公司 | Indoor terminal positioning method, device, equipment and medium |
| CN116597397A (en) * | 2023-07-17 | 2023-08-15 | 腾讯科技（深圳）有限公司 | Model training method and device for predicting vehicle track and storage medium |
-
2022
- 2022-12-13 CN CN202211593201.8A patent/CN115743101A/en active Pending
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116125819A (en) * | 2023-04-14 | 2023-05-16 | 智道网联科技(北京)有限公司 | Track correction method, track correction device, electronic device and computer-readable storage medium |
| CN116558513A (en) * | 2023-07-06 | 2023-08-08 | 中国电信股份有限公司 | Indoor terminal positioning method, device, equipment and medium |
| CN116558513B (en) * | 2023-07-06 | 2023-10-03 | 中国电信股份有限公司 | Indoor terminal positioning method, device, equipment and medium |
| CN116597397A (en) * | 2023-07-17 | 2023-08-15 | 腾讯科技（深圳）有限公司 | Model training method and device for predicting vehicle track and storage medium |
| CN116597397B (en) * | 2023-07-17 | 2023-10-24 | 腾讯科技（深圳）有限公司 | Model training method and device for predicting vehicle track and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115743101A (en) | Vehicle track prediction method, and track prediction model training method and device | |
| CN107529650B (en) | Closed loop detection method and device and computer equipment | |
| CN114202743A (en) | Improved fast-RCNN-based small target detection method in automatic driving scene | |
| CN112949519B (en) | Target detection method, device, equipment and storage medium | |
| CN112990065A (en) | Optimized YOLOv5 model-based vehicle classification detection method | |
| CN114913493A (en) | Lane line detection method based on deep learning | |
| CN110826457B (en) | Vehicle detection method and device under complex scene | |
| CN116861262B (en) | Perception model training method and device, electronic equipment and storage medium | |
| CN113379719A (en) | Road defect detection method, road defect detection device, electronic equipment and storage medium | |
| CN112101113A (en) | Lightweight unmanned aerial vehicle image small target detection method | |
| CN116009397A (en) | Brake pressure model training method, device, equipment and storage medium | |
| CN113514796B (en) | Passive positioning method, system and medium | |
| CN115439692A (en) | Image processing method and device, electronic equipment and medium | |
| CN115457365A (en) | Model interpretation method and device, electronic equipment and storage medium | |
| Jiangzhou et al. | Research on real-time object detection algorithm in traffic monitoring scene | |
| CN115146478A (en) | Running condition construction method and device based on optimization algorithm and related equipment | |
| CN114677655A (en) | Multi-sensor target detection method and device, electronic equipment and storage medium | |
| CN112749710A (en) | Target detection and intelligent driving method, device, equipment and storage medium | |
| CN112927117B (en) | Block chain-based vehicle management communication method, management system, device and medium | |
| CN117191068B (en) | Model training method and device, and track prediction method and device | |
| CN116740662B (en) | Axle recognition method and system based on laser radar | |
| CN115271001B (en) | Vehicle driving condition identification method and device, vehicle and storage medium | |
| CN115049895B (en) | Image attribute identification method, attribute identification model training method and device | |
| Kang et al. | Research on the Real-time Vehicle Detection Technology Based on Bic-YOLOv5s | |
| CN115861379A (en) | Video tracking method for twin network to update target template based on local credible template |
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 |