CN113954863B - Vehicle track prediction method based on dual-attention mechanism and improved Social Gan - Google Patents
Vehicle track prediction method based on dual-attention mechanism and improved Social Gan Download PDFInfo
- Publication number
- CN113954863B CN113954863B CN202111044320.3A CN202111044320A CN113954863B CN 113954863 B CN113954863 B CN 113954863B CN 202111044320 A CN202111044320 A CN 202111044320A CN 113954863 B CN113954863 B CN 113954863B
- Authority
- CN
- China
- Prior art keywords
- vehicle
- track
- gan
- target vehicle
- attention
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 230000007246 mechanism Effects 0.000 title claims abstract description 29
- 238000000034 method Methods 0.000 title claims abstract description 24
- 238000011176 pooling Methods 0.000 claims abstract description 32
- 230000002452 interceptive effect Effects 0.000 claims abstract description 4
- 230000002093 peripheral effect Effects 0.000 claims description 14
- 230000001133 acceleration Effects 0.000 claims description 12
- 230000003993 interaction Effects 0.000 claims description 8
- 230000004913 activation Effects 0.000 claims description 3
- 238000012360 testing method Methods 0.000 description 11
- 238000012549 training Methods 0.000 description 9
- 230000006870 function Effects 0.000 description 7
- NTJOBXMMWNYJFB-UHFFFAOYSA-N amisulpride Chemical compound CCN1CCCC1CNC(=O)C1=CC(S(=O)(=O)CC)=C(N)C=C1OC NTJOBXMMWNYJFB-UHFFFAOYSA-N 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000012795 verification Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000006073 displacement reaction Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 238000000714 time series forecasting Methods 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
- B60W60/00—Drive control systems specially adapted for autonomous road vehicles
- B60W60/001—Planning or execution of driving tasks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Automation & Control Theory (AREA)
- Human Computer Interaction (AREA)
- Transportation (AREA)
- Mechanical Engineering (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention provides a vehicle track prediction method based on a dual-attention mechanism and an improved Social Gan, which is used for inputting track characteristics of a target vehicle and surrounding vehicles into a vehicle track prediction model to predict a running track of the target vehicle at a future moment, wherein the vehicle track prediction model comprises an encoding layer, a pooling layer, a driving intention prediction module and a decoding layer, the pooling layer comprises two attention modules and a pooling module for improving the Social Gan, one attention module acts on the track characteristics of the encoded target vehicle, performs attention weighting on related variables influencing the track of the target vehicle, the other attention module acts on the track characteristics of the encoded surrounding vehicles, performs attention weighting on importance of the surrounding vehicles influencing the target vehicle, and the pooling module for improving the Social Gan is used for extracting interactive characteristics of the surrounding vehicles. The method and the device effectively improve the prediction accuracy and can be better applied to the field of automatic driving.
Description
Technical Field
The invention belongs to the technical field of automatic driving, and relates to a vehicle track prediction method based on a double-attention mechanism and an improved Social Gan.
Background
The development of automatic driving technology promotes the construction pace of smart cities, and has great influence on road traffic safety and urban traffic management.
In order to ensure safe running of the automatic driving automobile, the running track of the surrounding vehicles needs to be predicted, and the automatic driving automobile can plan a path in advance by utilizing the predicted surrounding vehicle track, so that the running safety of the automobile can be improved, and the riding comfort of passengers can also be improved.
In the prior art, only track information of surrounding vehicles is used for prediction, and interaction information of the vehicles and the road cannot be effectively utilized, so that the prediction accuracy is low.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a vehicle track prediction method based on a double-attention mechanism and improved Social Gan, and the accuracy of surrounding vehicle track prediction is improved.
The present invention achieves the above technical object by the following means.
Based on a double-attention mechanism and a vehicle track prediction method for improving the Social Gan, constructing a vehicle track prediction model for predicting the running track of the target vehicle at the future moment based on the double-attention mechanism and the improved Social Gan;
the vehicle track prediction model based on the dual-attention mechanism and the improved Social Gan comprises an encoding layer, a pooling layer, a driving intention prediction module and a decoding layer;
The coding layer comprises n+1 LSTM encoders, wherein the 1 LSTM encoders code the input track characteristics of the target vehicle, and the n LSTM encoders code the input track characteristics of n surrounding vehicles;
The pooling layer comprises two attention modules and a pooling module for improving the Social Gan, wherein one attention module acts on the track characteristics of the coded target vehicle, performs attention weighting on related variables affecting the track of the target vehicle, the other attention module acts on the track characteristics of the coded peripheral vehicle, performs attention weighting on the importance of the peripheral vehicle affecting the target vehicle, and the pooling module for improving the Social Gan is used for extracting the interaction characteristics of the peripheral vehicle;
The driving intention prediction module is used for predicting the transverse and longitudinal driving intention of the target vehicle in the next time period and decoding the driving intention by an LSTM decoder of the decoding layer.
Further, the pooling module of the improved Social Gan is configured to extract interaction features of surrounding vehicles, specifically: calculating the relative position and the relative speed of the current position of the target vehicle and the previous time period, and the relative position and the relative speed of each surrounding vehicle and the target vehicle; the relative positions and relative speeds of the target vehicle and the surrounding vehicles are independently processed by the multi-layer perceptron, and the interactive features of the surrounding vehicles are extracted by using maximum pooling.
Further, the driving intention prediction module inputs a complete track obtained by splicing the track of the target vehicle after the attention weighting and the track of the pooled surrounding vehicles.
Still further, the driving intention prediction module is composed of two groups of LSTM layers and Softmax layers, wherein the LSTM layers encode complete tracks and multi-classify the tracks by using Softmax activation functions to obtain lateral driving intention and longitudinal driving intention of the next time period.
Still further, the LSTM decoder inputs a track with driving intention formed by splicing the predicted driving intention with the complete track.
Still further, the driving intention is determined by setting a trajectory prediction time and an acceleration threshold value.
Further, the track characteristics of the target vehicle include position information, lane information, speed and acceleration information, and lateral and longitudinal driving intention of the target vehicle.
Further, the track characteristic of the nearby vehicle is position information of the nearby vehicle.
Compared with the prior art, the invention has the beneficial effects that:
The invention adopts a double-attention mechanism to respectively act on the related variables of the track data of the target vehicle and the position information of the surrounding vehicles, and can effectively extract the characteristics with larger influence on the track of the target vehicle by re-distributing weights; the method comprises the steps that a pooling module of the Social Gan is improved, surrounding vehicles of a target vehicle are identified at any time t, and the relative position and the relative speed of the target vehicle, the relative position and the relative speed of each surrounding vehicle and the target vehicle are calculated; the relative position and the relative speed are connected with LSTM hidden states of each vehicle, are independently processed by a multi-layer perceptron, and use maximum pooling to calculate pooling tensors of the output target vehicles, and finally generate a plurality of possible prediction tracks according to probability through a plurality of LSTM networks; the vehicle track prediction method based on the double-attention mechanism and the improved Social Gan can improve the accuracy of predicting the surrounding vehicle tracks.
Drawings
FIG. 1 is a flow chart of a vehicle trajectory prediction method based on a dual-attention mechanism and modified Social Gan according to the present invention;
FIG. 2 is a schematic view of an extracted surrounding vehicle according to the present invention;
FIG. 3 is a diagram of a vehicle trajectory prediction model according to the present invention;
FIG. 4 is a block diagram of a pooling module in a vehicle trajectory model according to the present invention.
Detailed Description
For a more detailed and complete description of the present invention, reference should be made to the detailed description of the embodiments and accompanying drawings. It will be apparent that the described embodiments are only some, but not all, of the embodiments of the present specification. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are intended to be within the scope of the present disclosure.
As shown in fig. 1, the vehicle track prediction method based on the dual-attention mechanism and the improved Social Gan of the present invention specifically comprises the following steps:
s1, preprocessing a vehicle track data set to construct a training set, a verification set and a test set
Firstly, acquiring position information, lane information, speed and acceleration information of a vehicle (including a target vehicle and surrounding vehicles) from a vehicle track data set; secondly, extracting the transverse and longitudinal driving intentions of the target vehicle according to the driving intention standard, wherein the transverse driving intentions are divided into lane keeping, lane changing leftwards and lane changing rightwards, and the longitudinal driving intentions are divided into acceleration, deceleration and uniform speed; the position information, lane information, speed and acceleration information, and the longitudinal and transverse driving intention of the target vehicle are taken as input track characteristics of the target vehicle.
Specific driving intention criteria are: setting track prediction time, wherein the lane id of the target vehicle changes in the prediction time, namely the lane change is realized, and otherwise, the lane keeping state is realized. Setting an acceleration threshold value as a standard for judging the longitudinal driving intention type of the target vehicle, and considering that the vehicle accelerates when the acceleration of the target vehicle is greater than the set threshold value; when the target vehicle acceleration is smaller than the negative threshold value, the vehicle is considered to run at a reduced speed; the rest of the conditions consider that the vehicle is running at a constant speed.
As shown in fig. 2, a surrounding vehicle searching area is divided, surrounding vehicles of the target vehicle are searched, wherein a dark vehicle is the target vehicle, an elliptic curve is the surrounding vehicle searching area, light vehicles contained by the elliptic curve are the surrounding vehicles of the target vehicle, and light vehicles outside the rest elliptic curves are other vehicles which do not participate in track prediction of the target vehicle; and taking the position information of the surrounding vehicle as the input track characteristic of the surrounding vehicle.
A track data set containing only the input track features of the target vehicle and the input track features of the surrounding vehicles is set as 7:1: the scale of 2 is divided into a training set, a validation set and a test set. The test sets are further divided into a lane keeping test set, a left lane changing test set and a right lane changing test set according to different transverse driving intention types.
S2, constructing a vehicle track prediction model based on a double-attention mechanism and improved Social Gan
The vehicle track prediction model based on the dual-attention mechanism and the improved Social Gan is shown in FIG. 3, and comprises an input layer, an encoding layer, a pooling layer, a driving intention prediction module, a decoding layer and an output layer; the input layer is used for inputting the track of the target vehicle and the track of the surrounding vehicles into a dual-attention mechanism and improving a vehicle track prediction model of the Social Gan; the coding layer comprises n+1 LSTM encoders, one LSTM encoder codes the input track characteristics of the target vehicle, and n LSTM encoders code the input track characteristics of n surrounding vehicles; the pooling layer comprises two attention modules and a pooling module for improving the Socian Gan, wherein one attention module acts on the track characteristics of the coded target vehicle, performs attention weighting on related variables affecting the track of the target vehicle, such as vehicle speed, acceleration and the like, the other attention module acts on the track characteristics of the coded peripheral vehicle, performs attention weighting on the importance of the peripheral vehicle affecting the target vehicle (for the prior art, see Shih S Y,Sun F K,Lee H.Temporal pattern attention for multivariate time series forecasting[J].Machine Learning,2019,108(8):1421-1441.), for improving the pooling module of the Socian Gan for extracting the interaction characteristics of the peripheral vehicle (for the prior art, see Gupta A,Johnson J,Fei-Fei L,et al.Social gan:Socially acceptable trajectories with Ganerative adversarial networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.2018:2255-2264.), for transmitting the attention weighted importance of the peripheral vehicle track characteristics to the pooling module of the improved Socian Gan, since the attention mechanism highlights the influence relationship between the peripheral vehicle and the target vehicle, the interaction characteristics between the vehicles can be extracted more easily during the pooling operation, and the driving intention prediction module is used for predicting the transverse and longitudinal driving intention of the target vehicle in the next time period, decoded by the LSTM decoder of the decoding layer and finally outputted by the output layer.
The specific process for constructing the vehicle track prediction model is as follows:
s2.1, encoding the input track characteristics of the target vehicle and the input track characteristics of surrounding vehicles by using an LSTM encoder;
s2.2, a double-attention mechanism (two attention modules) is applied, wherein one attention module acts on the track characteristics of the coded target vehicle, performs attention weighting on related variables affecting the track of the target vehicle, such as vehicle speed, acceleration and the like, and the other attention module acts on the track characteristics of the coded peripheral vehicle, so as to perform attention weighting on the importance of the peripheral vehicle affecting the target vehicle;
S2.3, improved Social Gan pooling Module
Calculating the relative position and the relative speed of the current position of the target vehicle and the previous time period, and the relative position and the relative speed of each surrounding vehicle and the target vehicle; the relative position, relative speed of the target vehicle and the surrounding vehicles are connected with the LSTM hidden state of each vehicle, independently processed by a multi-layer perceptron (MLP), and the maximum pooling (MAX) is used to extract the interactive features of the surrounding vehicles, as shown in FIG. 4;
S2.4, splicing the track of the target vehicle after attention weighting and the track of the pooled surrounding vehicles to form a complete track, and then transmitting the track to a driving intention prediction module, wherein the driving intention prediction module consists of two groups of LSTM layers and Softmax layers and is used for predicting the transverse driving intention and the longitudinal driving intention of the next time period respectively, the LSTM layers encode the complete track and multi-classify the complete track by using a Softmax activation function to obtain the predicted driving intention;
S2.5, splicing the predicted driving intention with the complete track to form a track with the driving intention, transmitting the track into an LSTM decoder, wherein the decoder consists of a plurality of LSTM networks, and outputting a plurality of possible tracks.
S3, training a vehicle track prediction model based on a double-attention mechanism and an improved Social Gan, and using Root Mean Square Error (RMSE) as a loss function;
The specific steps of training the vehicle track prediction model are as follows: training a vehicle track prediction model, taking a predicted track with the highest probability as an output value, taking a Root Mean Square Error (RMSE) as a loss function, and taking a negative log likelihood function (NLL) as a key index of a verification set on each time step; wherein the loss function is:
where T pred is the predicted time step, (x t,yt) is the true position coordinates of the target vehicle and (x' t,y′t) is the predicted target vehicle position coordinates.
The vehicle track prediction model based on the double-attention mechanism and the improved Social Gan is trained by utilizing a training set, the accuracy of the vehicle track prediction model is calculated by using a verification set in the training process, and the model is prevented from being fitted by combining the change characteristics of a loss function in the training process.
S4, storing the vehicle track prediction model after training is finished.
S5, testing a vehicle track prediction model based on a double-attention mechanism and an improved Social Gan;
and using the test set as the input of the trained vehicle track prediction model, predicting the possible track of the target vehicle, and testing the accuracy of the vehicle track prediction model.
S6, comparing and analyzing the vehicle track prediction model with other popular vehicle track prediction models;
Comparing the test results of the vehicle track prediction model based on the dual-attention mechanism and the improved Social Gan, including Root Mean Square Error (RMSE), average Displacement Error (ADE) and Final Displacement Error (FDE), with the test results of other vehicle track prediction models; the historical trajectory data with a time step of 3 is used to predict the trajectory for the next 5 time steps.
The present invention performs a number of experiments to evaluate the effectiveness and advancement of the present invention, and the comparison results with other models are shown in table 1:
Table 1 RMSE value comparison
As shown in table 1, the vehicle track prediction model based on the dual-attention mechanism and the modified Social Gan proposed by the present invention is superior to all other methods in all prediction time steps.
Compared with the existing track prediction method, the method adopts a double-attention mechanism to respectively act on the related variables of the track data of the target vehicle and the position information of the surrounding vehicles, and can effectively extract the characteristics with larger influence on the track of the target vehicle by re-distributing weights; the method comprises the steps that a pooling module of the Social Gan is improved, and at any time t, the relative position and the relative speed of the current position of a target vehicle and the previous time period, and the relative position and the relative speed of each peripheral vehicle and the target vehicle are calculated; the relative position and relative speed of the target vehicle and the surrounding vehicles are connected with the LSTM hidden state of each vehicle, are independently processed by a multi-layer perceptron (MLP), and use maximum pooling (MAX) to calculate and output the pooling tensor of the interaction between the surrounding vehicles; finally, a plurality of possible prediction tracks are generated according to probability through a plurality of LSTM networks.
The foregoing is merely exemplary of the present application and is not intended to limit the present application. Various modifications and variations of the present application will be apparent to those skilled in the art. Any modification, equivalent replacement, improvement, etc. which come within the spirit and principles of the application are to be included in the scope of the claims of the present application.
Claims (6)
1. The vehicle track prediction method based on the double-attention mechanism and the improved Social Gan is characterized by constructing a vehicle track prediction model based on the double-attention mechanism and the improved Social Gan to predict the running track of a target vehicle at the future moment;
the vehicle track prediction model based on the dual-attention mechanism and the improved Social Gan comprises an encoding layer, a pooling layer, a driving intention prediction module and a decoding layer;
The coding layer comprises n+1 LSTM encoders, wherein the 1 LSTM encoders code the input track characteristics of the target vehicle, and the n LSTM encoders code the input track characteristics of n surrounding vehicles;
The pooling layer comprises two attention modules and a pooling module for improving the Social Gan, wherein one attention module acts on the track characteristics of the coded target vehicle, performs attention weighting on related variables affecting the track of the target vehicle, the other attention module acts on the track characteristics of the coded peripheral vehicle, performs attention weighting on the importance of the peripheral vehicle affecting the target vehicle, and the pooling module for improving the Social Gan is used for extracting the interaction characteristics of the peripheral vehicle;
the driving intention prediction module is used for predicting the transverse and longitudinal driving intention of the target vehicle in the next time period and decoding the transverse and longitudinal driving intention through an LSTM decoder of the decoding layer;
The driving intention prediction module inputs a complete track formed by splicing the track of the target vehicle after the attention is weighted and the track of the surrounding vehicles after the pooling;
the driving intention prediction module consists of two groups of LSTM layers and a Softmax layer, wherein the LSTM layers encode complete tracks and perform multi-classification by using a Softmax activation function to obtain the transverse driving intention and the longitudinal driving intention of the next time period.
2. The vehicle track prediction method based on the dual-attention mechanism and the modified Social Gan according to claim 1, wherein the pooling module of the modified Social Gan is configured to extract interaction features of surrounding vehicles, specifically: calculating the relative position and the relative speed of the current position of the target vehicle and the previous time period, and the relative position and the relative speed of each surrounding vehicle and the target vehicle; the relative positions and relative speeds of the target vehicle and the surrounding vehicles are independently processed by the multi-layer perceptron, and the interactive features of the surrounding vehicles are extracted by using maximum pooling.
3. The dual-attention mechanism and modified-society Gan based vehicle trajectory prediction method of claim 1, wherein the LSTM decoder inputs a trajectory with driving intent formed by stitching the predicted driving intent with a complete trajectory.
4. The dual-attention mechanism and modified-society Gan based vehicle trajectory prediction method of claim 3, wherein the driving intention is determined by setting a trajectory prediction time and an acceleration threshold.
5. The dual-attention-mechanism and modified-society-Gan-based vehicle trajectory prediction method according to claim 1, wherein the trajectory characteristics of the target vehicle include position information, lane information, speed and acceleration information, and lateral-longitudinal driving intention of the target vehicle.
6. The dual-attention-mechanism and modified-society-Gan-based vehicle trajectory prediction method according to claim 1, wherein the trajectory characteristic of the nearby vehicle is position information of the nearby vehicle.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111044320.3A CN113954863B (en) | 2021-09-07 | 2021-09-07 | Vehicle track prediction method based on dual-attention mechanism and improved Social Gan |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111044320.3A CN113954863B (en) | 2021-09-07 | 2021-09-07 | Vehicle track prediction method based on dual-attention mechanism and improved Social Gan |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113954863A CN113954863A (en) | 2022-01-21 |
CN113954863B true CN113954863B (en) | 2024-06-11 |
Family
ID=79460983
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111044320.3A Active CN113954863B (en) | 2021-09-07 | 2021-09-07 | Vehicle track prediction method based on dual-attention mechanism and improved Social Gan |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113954863B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116890881B (en) * | 2023-09-08 | 2023-12-08 | 摩尔线程智能科技(北京)有限责任公司 | Vehicle lane change decision generation method and device, electronic equipment and storage medium |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP3650297A1 (en) * | 2018-11-08 | 2020-05-13 | Bayerische Motoren Werke Aktiengesellschaft | Method and apparatus for determining information related to a lane change of a target vehicle, method and apparatus for determining a vehicle comfort metric for a prediction of a driving maneuver of a target vehicle and computer program |
CN112766561A (en) * | 2021-01-15 | 2021-05-07 | 东南大学 | Generating type confrontation track prediction method based on attention mechanism |
CN112949597A (en) * | 2021-04-06 | 2021-06-11 | 吉林大学 | Vehicle track prediction and driving manipulation identification method based on time mode attention mechanism |
WO2021134172A1 (en) * | 2019-12-30 | 2021-07-08 | 华为技术有限公司 | Trajectory prediction method and related device |
CN113313320A (en) * | 2021-06-17 | 2021-08-27 | 湖北汽车工业学院 | Vehicle track prediction method based on residual attention mechanism |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US11989952B2 (en) * | 2019-12-10 | 2024-05-21 | Samsung Electronics Co., Ltd. | Systems and methods for trajectory prediction |
-
2021
- 2021-09-07 CN CN202111044320.3A patent/CN113954863B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP3650297A1 (en) * | 2018-11-08 | 2020-05-13 | Bayerische Motoren Werke Aktiengesellschaft | Method and apparatus for determining information related to a lane change of a target vehicle, method and apparatus for determining a vehicle comfort metric for a prediction of a driving maneuver of a target vehicle and computer program |
WO2021134172A1 (en) * | 2019-12-30 | 2021-07-08 | 华为技术有限公司 | Trajectory prediction method and related device |
CN112766561A (en) * | 2021-01-15 | 2021-05-07 | 东南大学 | Generating type confrontation track prediction method based on attention mechanism |
CN112949597A (en) * | 2021-04-06 | 2021-06-11 | 吉林大学 | Vehicle track prediction and driving manipulation identification method based on time mode attention mechanism |
CN113313320A (en) * | 2021-06-17 | 2021-08-27 | 湖北汽车工业学院 | Vehicle track prediction method based on residual attention mechanism |
Non-Patent Citations (1)
Title |
---|
蔡英凤等.基于注意力机制的车辆行为预测.江苏大学学报.2020,第41卷(第2期),125-130. * |
Also Published As
Publication number | Publication date |
---|---|
CN113954863A (en) | 2022-01-21 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111931902B (en) | Generating countermeasure network model and vehicle track prediction method using generating countermeasure network model | |
CN111275711B (en) | Real-time image semantic segmentation method based on lightweight convolutional neural network model | |
CN111930110A (en) | Intent track prediction method for generating confrontation network by combining society | |
CN113044064B (en) | Vehicle self-adaptive automatic driving decision method and system based on meta reinforcement learning | |
CN111339867A (en) | Pedestrian trajectory prediction method based on generation of countermeasure network | |
CN113076599A (en) | Multimode vehicle trajectory prediction method based on long-time and short-time memory network | |
CN114202120A (en) | Urban traffic travel time prediction method aiming at multi-source heterogeneous data | |
CN114898550B (en) | Pedestrian track prediction method and system | |
CN113954863B (en) | Vehicle track prediction method based on dual-attention mechanism and improved Social Gan | |
CN108319909A (en) | A kind of driving behavior analysis method and system | |
CN109086926B (en) | Short-time rail transit passenger flow prediction method based on combined neural network structure | |
CN114495499B (en) | Multi-target intelligent internet vehicle cooperative optimization control method | |
CN113159403B (en) | Intersection pedestrian track prediction method and device | |
CN115049009A (en) | Track prediction method based on semantic fusion representation | |
CN114372570A (en) | Multi-mode vehicle trajectory prediction method | |
CN101964061B (en) | Binary kernel function support vector machine-based vehicle type recognition method | |
CN114379595A (en) | Vehicle track prediction method based on vehicle behavior characteristics and environment interaction information | |
CN116415200A (en) | Abnormal vehicle track abnormality detection method and system based on deep learning | |
CN114368387B (en) | Attention mechanism-based driver intention recognition and vehicle track prediction method | |
CN115862322A (en) | Vehicle variable speed limit control optimization method, system, medium and equipment | |
CN117237475A (en) | Vehicle traffic track generation method and device based on diffusion generation model | |
CN116432454A (en) | Character selection-based automatic driving automobile decision planning method | |
CN116946183A (en) | Commercial vehicle driving behavior prediction method considering driving capability and vehicle equipment | |
CN115438786A (en) | Robust neural network training method based on sample-driven target loss function optimization | |
CN113191539B (en) | High-density composite scene track prediction method based on heterogeneous graph aggregation network |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant |