CN112907956A - Expressway lane-level running speed prediction method based on space-time information - Google Patents
Expressway lane-level running speed prediction method based on space-time information Download PDFInfo
- Publication number
- CN112907956A CN112907956A CN202110124221.XA CN202110124221A CN112907956A CN 112907956 A CN112907956 A CN 112907956A CN 202110124221 A CN202110124221 A CN 202110124221A CN 112907956 A CN112907956 A CN 112907956A
- Authority
- CN
- China
- Prior art keywords
- traffic state
- data
- time
- information
- speed
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/052—Detecting movement of traffic to be counted or controlled with provision for determining speed or overspeed
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
- G08G1/0125—Traffic data processing
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
- G08G1/0125—Traffic data processing
- G08G1/0129—Traffic data processing for creating historical data or processing based on historical data
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
- G08G1/0137—Measuring and analyzing of parameters relative to traffic conditions for specific applications
Abstract
The invention relates to a method for predicting the lane-level running speed of a highway based on space-time information, which comprises the following steps: s1: selecting a highway section to be predicted, and performing sub-section division and lane division; s2: acquiring a traffic state data set of each sub-road section, and performing deep data fusion to obtain a composite traffic state data set of each sub-road section; s3: extracting the composite traffic state data set of each sub-road section according to a time sequence order, a space sequence order and a period sequence order to obtain a traffic state fusion data sequence set representing space-time information, and inputting the traffic state fusion data sequence set into a prediction model for iterative training to obtain a speed prediction model; s4: and predicting the lane-level running speed of the expressway to be predicted by using the trained speed prediction model. Compared with the prior art, the method has the advantages of high accuracy, good stability and the like.
Description
Technical Field
The invention relates to the technical field of information prediction, in particular to a method for predicting the lane-level running speed of a highway based on space-time information.
Background
In recent years, highway construction has begun to shift from the high-speed growth phase to the high-quality development phase. The construction of wisdom high-speed can effectively improve management level and operating efficiency, is the must way of the high-quality development of highway trade. The intelligent management of the expressway depends on the early perception and study and judgment of the traffic state of the expressway, the traffic state of the expressway is accurately predicted, and the intelligent management system is favorable for the traffic management, the abnormal detection and the congestion management of the expressway.
Currently, in terms of prediction of highway traffic flow running speed, Time Series (Time Series) based prediction is the most widely used method. The method utilizes the single traffic state data of a certain road section to learn the traffic state data in the past time interval and predict the traffic state of the road section in the future designated time.
The traffic state prediction based on the time sequence is wide in application and strong in practicability, but the prediction accuracy is to be improved as the prediction method only represents the traffic state information of the time dimension and does not fully represent the space dimension information presenting similar characteristics in the space; on the other hand, the prediction method only depends on the prediction of the running speed by a single detector, and the fluctuation of the prediction result is large and unstable. The two aspects are main bottlenecks for restricting the prediction method, so that a new method for predicting the running speed of the traffic flow on the highway is urgently needed to be provided, and the accuracy and the stability of prediction are improved.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a method for predicting the running speed of a highway lane grade based on space-time information, which improves the accuracy and stability of prediction.
The purpose of the invention can be realized by the following technical scheme:
a method for predicting the lane-level running speed of a highway based on space-time information comprises the following steps:
s1: selecting a highway section to be predicted, and performing sub-section division and lane division;
s2: acquiring a traffic state data set of each sub-road section, and performing deep data fusion to obtain a composite traffic state data set of each sub-road section;
s3: extracting the composite traffic state data set of each sub-road section according to a time sequence order, a space sequence order and a period sequence order to obtain a traffic state fusion data sequence set representing space-time information, and inputting the traffic state fusion data sequence set into a prediction model for iterative training to obtain a speed prediction model;
s4: and predicting the lane-level running speed of the expressway to be predicted by using the trained speed prediction model.
Further, the spatio-temporal information includes temporal information, spatial information, and period information.
Further, step S3 specifically includes:
s301: constructing a speed prediction model, wherein the speed prediction model comprises a convolution long-time and short-time prediction network and an attention extraction network;
s302: acquiring three traffic state fusion data sequences respectively used for representing space-time information;
s303: inputting the three traffic state fusion data sequences into a convolution duration prediction network respectively, and predicting to obtain running speed matrixes of three corresponding sub-road sections;
s304: inputting the three running speed prediction matrixes into an attention extraction network together, and predicting to obtain a final running speed matrix of the corresponding sub-road section;
s305: and repeating the steps S302-S304, and performing iterative training on the speed prediction model to prevent obtaining the speed prediction model with the optimal parameters.
Furthermore, the traffic state fusion data sequence set RallThe expression of (a) is:
Rall＝{Rspace,Rtime,Rperiod}
wherein R istimeA sequence of traffic state fusion data for corresponding sub-sections operating over a period of time in the past, for characterizing time information, RspaceFusion data sequence of traffic states of spatially adjacent lanes for corresponding sub-segments for characterizing spatial information, RperiodFusing data sequences for the traffic states of the corresponding sub-road sections at the same moment in the historical period, wherein the data sequences are used for representing period information;
the convolution duration prediction network output running speed matrix AallThe expression of (a) is:
Aall＝{Aspace,Atime,Aperiod}
wherein A istimeFor predicting the operating speed of the corresponding sub-section from the time information, AspaceFor predicting the operating speed of the corresponding sub-section from the spatial information, AperiodThe operation speed of the corresponding sub-section is predicted according to the cycle information.
Furthermore, the traffic state data sets comprise lane information, time information and traffic state values, the traffic state values are obtained by measurement of a collecting tool, each traffic state data set comprises at least two traffic state values, the lane information is a lane number corresponding to the traffic state value, and the time information is a time number of the traffic state value measured by the collecting tool.
Furthermore, the traffic state value comprises an average speed value, headway distance, space occupancy, time occupancy and/or a 85-minute speed value.
Furthermore, the acquisition tools are millimeter wave radars, each sub-section corresponds to one acquisition tool, and the sampling frequency of the millimeter wave radars is set to be 1 min.
Further, in step S2, the depth data fusion is implemented by a fusion model of spatio-temporal information representation learning, the fusion model includes an input layer, an output layer and an intermediate layer, the input layer inputs the traffic state data of the sub-road section, the intermediate layer sequentially includes a convolutional layer, an active layer and a full-link layer, and the output layer outputs the fused composite traffic state data.
Furthermore, the convolution long-time and short-time prediction network comprises at least two prediction networks formed by connecting ConvLstm networks.
Furthermore, the attention extracting network comprises a pooling layer, a full-link layer and a computing layer, wherein the pooling layer is used for down-sampling input, the full-link layer is used for splicing and compressing pooled data, the computing layer is used for multiplying the data of the full-link layer with the input data, and a final prediction result is obtained through pooling again.
Compared with the prior art, the invention has the following advantages:
1) the method is improved from the perspective of multi-dimensional representation of the time-space information and multi-source fusion of data, can increase priori knowledge by adding various types of information, improves the prediction effect and stability, and based on a method for representing and learning the time-space information, the time traffic data, the space traffic data and the periodic traffic data are fused to predict the lane-level traffic running speed of the highway, so that the stability and the accuracy of a prediction algorithm are enhanced, and the method can be quickly applied to a roadside acquisition system of the highway in China to realize efficient and stable lane speed prediction;
2) the invention is suitable for the existing traffic state acquisition equipment at the side of the highway, does not need additional hardware cost investment, and can save expensive investment on the highway;
3) the method and the system can obtain accurate and stable lane-level highway traffic speed prediction results, can support highway management departments to carry out highway traffic management, anomaly detection and congestion control, and can realize the advanced perception and study and judgment of highway lane traffic states, thereby effectively improving the fine management level and the operation efficiency of the highway.
Drawings
FIG. 1 is a schematic flow diagram of the present invention;
FIG. 2 is a schematic diagram of the invention for establishing a sub-segment traffic status value data set;
FIG. 3 is a schematic diagram of a fused sub-link traffic status value data sequence in accordance with the present invention;
FIG. 4 is a schematic diagram of the present invention for constructing a lane-level speed prediction model based on spatiotemporal information characterization learning;
fig. 5 is a schematic diagram illustrating a comparison between a predicted lane-level traffic running speed and a real speed, where fig. 5a is a real speed value and fig. 5b is a predicted speed value.
Detailed Description
The invention is described in detail below with reference to the figures and specific embodiments. It is to be understood that the embodiments described are only a few embodiments of the present invention, and not all embodiments. All other embodiments, which can be obtained by a person skilled in the art without any inventive step based on the embodiments of the present invention, shall fall within the scope of protection of the present invention.
Examples
The invention provides a method for predicting the lane-level running speed of a highway based on space-time information, which predicts from the aspects of multi-dimensional representation of the space-time information and multi-source fusion of data, wherein the space-time information comprises time information, space information and periodic information. The time information is traffic state information of a predicted highway lane running in a past period of time, the space information is traffic state information of a lane spatially adjacent to the predicted highway lane, and the cycle information is traffic state information of the predicted highway lane at the same time in a history cycle.
The method comprises the following steps:
s1: and selecting the highway section to be predicted, and performing sub-section division and lane division.
S2: the method comprises the steps of obtaining a traffic state data set of each sub-road section, carrying out deep data fusion, and obtaining a composite traffic state data set of each sub-road section, wherein the traffic state data set comprises lane information, time information and a traffic state value, the traffic state value is obtained by measurement of an acquisition tool, each traffic state data set comprises at least two traffic state values, the lane information is a lane number corresponding to the traffic state value, and the time information is a time number of the acquisition tool for measuring the traffic state value.
The deep data fusion is realized through a fusion model of space-time information representation learning, the fusion model comprises an input layer, an output layer and an intermediate layer, the input layer inputs the traffic state data of the sub-road section, the intermediate layer sequentially comprises a convolution layer, an activation layer and a full connection layer, and the output layer outputs the fused composite traffic state data.
S3: extracting the composite traffic state data set of each sub-road section according to a time sequence order, a space sequence order and a period sequence order to obtain a traffic state fusion data sequence set representing space-time information, inputting the traffic state fusion data sequence set into a prediction model for iterative training to obtain a speed prediction model, wherein the steps specifically comprise:
s301: constructing a speed prediction model, wherein the speed prediction model comprises a convolution long-time and short-time prediction network and an attention extraction network;
s302: acquiring three traffic state fusion data sequences respectively used for representing space-time information;
s303: inputting the three traffic state fusion data sequences into a convolution duration prediction network respectively, and predicting to obtain running speed matrixes of three corresponding sub-road sections;
s304: inputting the three running speed prediction matrixes into an attention extraction network together, and predicting to obtain a final running speed matrix of the corresponding sub-road section;
s305: and repeating the steps S302-S304, and performing iterative training on the speed prediction model to prevent obtaining the speed prediction model with the optimal parameters.
Traffic state fusion data sequence set RallThe expression of (a) is:
Rall＝{Rspace,Rtime,Rperiod}
wherein R istimeA sequence of traffic state fusion data for corresponding sub-sections operating over a period of time in the past, for characterizing time information, RspaceFusion data sequence of traffic states of spatially adjacent lanes for corresponding sub-segments for characterizing spatial information, RperiodFusing data sequences for the traffic states of the corresponding sub-road sections at the same moment in the historical period, wherein the data sequences are used for representing period information;
operation speed matrix A of convolution long-time and short-time prediction network outputallThe expression of (a) is:
Aall＝{Aspace,Atime,Aperiod}
wherein A istimeFor predicting the operating speed of the corresponding sub-section from the time information, AspaceFor predicting the operating speed of the corresponding sub-section from the spatial information, AperiodThe operation speed of the corresponding sub-section is predicted according to the cycle information.
The convolution long-time and short-time prediction network comprises a prediction network formed by connecting at least two ConvLstm networks.
The attention extraction network comprises a pooling layer, a full-connection layer and a calculation layer, wherein the pooling layer carries out down-sampling on input, the full-connection layer carries out splicing and compression on pooled data, the calculation layer carries out multiplication calculation on data of the full-connection layer and the input data, and a final prediction result is obtained through pooling again.
S4: and predicting the lane-level running speed of the expressway to be predicted by using the trained speed prediction model.
As shown in fig. 1, in the present embodiment, the execution includes the following steps:
step 1: selecting a predicted highway section, dividing the selected highway section into a plurality of sub-sections by proper length, setting proper acquisition frequency and acquisition range for an acquisition tool, measuring not less than 2 traffic state values on the divided sub-sections by using the acquisition tool, and respectively obtaining space-time information data on each sub-section.
The acquisition range of the existing various highway traffic state acquisition equipment is about 50-300 meters, and the acquisition frequency can reach millisecond level. The traffic state collecting device has a limited collecting range, and in order to comprehensively and systematically collect a large-range highway network, the collecting device is usually reasonably arranged at multiple positions of a highway. The layout and site selection of the acquisition equipment and the determination of the required sampling frequency are very important in traffic flow running speed prediction, and according to actual needs, the millimeter wave radar is selected as the acquisition equipment, and one millimeter wave radar is arranged at an interval of about 200 meters on a highway section as the acquisition equipment. Meanwhile, the higher the acquisition frequency of the equipment is, the higher the data accuracy is, the higher the value of the traffic state information is, and the higher the accuracy which can be achieved by prediction is. For this purpose, the sampling frequency is selected to be 1min according to actual needs.
General acquisition equipment acquires various types of traffic state variables under a specified sampling frequency. The average speed, which is the traffic state variable most often collected by the collection device, generally represents the average of the speeds of passing vehicles over the collection time interval. The traffic volume generally represents the number of vehicles passing through the collection time interval, and similar variables include small-sized vehicle traffic volume, medium-sized vehicle traffic volume, and the like. The space occupancy generally represents the percentage of the total length of the road segment that all vehicles occupy on the known road segment measured during that time period. In addition, the traffic state of a highway section is also related to the climate conditions.
The specific process of the step is as follows:
step 101: dividing the expressway into S lanes in the transverse direction according to lanes; in the longitudinal direction, according to a fixed acquisition range set by acquisition equipment, the expressway with the length of L is divided into M intervals at equal intervals, and the traffic state acquisition equipment is respectively placed in each interval. The method specifically comprises the following steps: in the range of an expressway, a representative road section is selected to be about 3km, 20 upright posts with the height of 7 meters are arranged at intervals of 100-150 meters in the road section, and each upright post is provided with a 24GHz millimeter wave radar device. The direction of the millimeter wave radar stand column is fixed, and the measuring range is guaranteed to cover all lanes.
Step 102: according to the collection frequency of the traffic state collection equipment, data collection is carried out on subintervals under the sub-lanes according to the consistent sampling frequency, and various traffic state data of lane levels under the time T are recorded respectively, and the method specifically comprises the following steps: recording lane-level traffic state values collected by the millimeter wave radar within 7 continuous days by taking 1 minute as a unit, wherein the traffic state values comprise 6 indexes of average speed value, headway time, headway distance, space occupancy, time occupancy and 85-quantile speed value of a lane per minute, finally forming various traffic state data sets of each sub-road section at different time periods, and generating the following traffic state data sets:
Q＝Si×Tj×Nz
wherein S isiNumber of lanes, TjIs the number of times, NzIs the number of the measured traffic status value. Fig. 2 shows a traffic state value data of a plurality of types of acquired expressway sub-sections including spatio-temporal information.
Step 2: as shown in fig. 3, the multiple traffic state data of the sub-road sections in all time ranges are deeply fused to obtain the fused composite traffic state data of the sub-road sections of the expressway, which specifically includes: and establishing a fusion model of spatio-temporal information representation learning, which comprises an input layer, an output layer and an intermediate layer. The input layer of the fusion model is a sub-road section traffic state data set Q, the middle layer of the fusion model at least comprises a convolution layer, an activation layer and a full-connection layer, and the output layer of the fusion model obtains a fused sub-road section traffic state composite data set R, wherein the middle layer and the activation layer of the fusion model are located behind the convolution layer, and the full-connection layer is located at the last of the fusion model.
The average speed of a section of a highway is affected by a number of variables. The severe weather such as dense fog, rainfall, snowfall and the like can obviously influence the average speed of the road section; the average speed of the road segment will also vary accordingly with a significant increase in traffic flow on the detection date, particularly during holidays. Therefore, semantic information contained in a plurality of variables is subjected to deep fusion representation, so that the content of input information of a prediction model is increased, and the traffic speed can be predicted more accurately.
The deep fusion model of multiple traffic states is to perform multiple weighting calculations on a speed state value matrix of a certain road segment and other related traffic state variable matrices within a fixed time interval, as shown in fig. 3. And according to the python language and the keras deep learning framework, performing the convolution fusion model calculation on each section of the highway in each time interval to obtain the composite traffic state tensor. And finally, outputting the composite traffic state value data containing the multi-class traffic semantic information according to the depth fusion model.
And step 3: as shown in fig. 4, according to the sub-link traffic state composite data set R, the fused composite data is sequentially input into the prediction model according to the time sequence relationship, the space sequence relationship and the periodic sequence relationship data for iterative training, so as to obtain the prediction model of the highway lane-level running speed, and the specific process includes: 1) the prediction model takes time, space and period information as multiple inputs and carries out preliminary prediction in a prediction network of three convolution duration; 2) outputting three predicted operation speed matrixes based on the established convolution long-time and short-time prediction network; 3) and (4) performing attention extraction on the three predicted running speed matrixes, and performing convolution fusion to output a finally predicted running speed matrix.
The specific execution process of the step comprises the following steps:
step 301: establishing a speed prediction model based on characterization learning, and extracting the fused traffic state value data as an input value of a data set according to the following method: sequentially extracting multi-traffic state value data of the road section at each past moment according to a time sequence order to serve as input of time sequence information; sequentially extracting multi-traffic state value data of the upstream road section of the road section according to the spatial sequence order to be used as input of spatial sequence information; and sequentially extracting the multi-traffic state value data of the road section at the same time in each week in the history according to the cycle sequence order as the input of the cycle sequence information.
The input of the speed prediction model is a traffic state composite data set R of the predicted expressway sub-sectionallThe concrete expression formula of the data set is as follows:
Rall＝{Rspace,Rtime,Rperiod}
the average speed of a section of a highway exhibits similar laws in time, space and periodic dimensions. In the time dimension, the traffic speed of the road section presents similar historical change rules by taking days as a unit; similarly, in the spatial dimension, the traffic speed value of the expressway shows similar historical rules among road sections; in the dimension of the cycle, the traffic speed values of the motorways exhibit similar variation laws in the surrounding units. Therefore, fusion representation of historical time sequence, spatial sequence and periodic sequence information is helpful for accurately predicting the traffic speed.
Step 302: establishing a lane-level traffic flow running speed prediction model under the representation of space-time information: firstly, inputting time information, space information and period information into a convolution duration prediction network together, and outputting three prediction speed matrixes, wherein the specific expression formula of the prediction speed matrix is as follows:
Aall＝{Aspace,Atime,Aperiod}
step 303: the three predicted speed matrices obtained in step 302 are input to the attention extraction network together to obtain the speed matrix of the final predicted sub-road section, and the speed value of the road section at the future moment is related to the information of the time dimension and the information of the upstream space dimension, and meanwhile, valuable data can be extracted from the periodic information. Therefore, the output time dimension predicted value, space velocity predicted value and period dimension predicted value are weighted by a certain weighting method, and a final velocity prediction tensor is obtained, namely the velocity predicted value of the road section in a certain time interval in the future is represented.
Step 304: and (3) dividing a data set, training a model, and sequentially selecting time, space and period information for prediction after determining a speed value at a time interval to be predicted. The selected data set is divided into a training set and a test set, wherein the ratio of the amount of samples in the training set to the amount of samples in the test set is 0.8: 0.2. And each sample comprises data and a label, the label is a speed value matrix of the expressway sub-road section in the future time to be predicted, the speed prediction model is subjected to iterative training according to the method in the step, and after the model with smaller error is obtained in the test set, the parameters of the prediction model are stored, so that the speed prediction model with the optimal parameters is obtained.
And 4, step 4: according to the speed prediction model with the optimal parameters obtained in the step S04, inputting the spatiotemporal information data of the sub-road section to be predicted into the model to realize accurate prediction of the highway traffic speed, and the prediction result is shown in fig. 5.
While the invention has been described with reference to specific embodiments, the invention is not limited thereto, and those skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope of the invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (10)
1. A method for predicting the lane-level running speed of a highway based on space-time information is characterized by comprising the following steps:
s1: selecting a highway section to be predicted, and performing sub-section division and lane division;
s2: acquiring a traffic state data set of each sub-road section, and performing deep data fusion to obtain a composite traffic state data set of each sub-road section;
s3: extracting the composite traffic state data set of each sub-road section according to a time sequence order, a space sequence order and a period sequence order to obtain a traffic state fusion data sequence set representing space-time information, and inputting the traffic state fusion data sequence set into a prediction model for iterative training to obtain a speed prediction model;
s4: and predicting the lane-level running speed of the expressway to be predicted by using the trained speed prediction model.
2. The method as claimed in claim 1, wherein the spatiotemporal information includes temporal information, spatial information and periodic information.
3. The method for predicting the lane-level running speed of the highway based on the spatio-temporal information as claimed in claim 2, wherein the step S3 specifically comprises:
s301: constructing a speed prediction model, wherein the speed prediction model comprises a convolution long-time and short-time prediction network and an attention extraction network;
s302: acquiring three traffic state fusion data sequences respectively used for representing space-time information;
s303: inputting the three traffic state fusion data sequences into a convolution duration prediction network respectively, and predicting to obtain running speed matrixes of three corresponding sub-road sections;
s304: inputting the three running speed prediction matrixes into an attention extraction network together, and predicting to obtain a final running speed matrix of the corresponding sub-road section;
s305: and repeating the steps S302-S304, and performing iterative training on the speed prediction model to prevent obtaining the speed prediction model with the optimal parameters.
4. The method as claimed in claim 3, wherein the traffic state fusion data sequence set R is a set of traffic state fusion data sequencesallThe expression of (a) is:
Rall＝{Rspace,Rtime,Rperiod}
wherein R istimeA sequence of traffic state fusion data for corresponding sub-sections operating over a period of time in the past, for characterizing time information, RspaceFusion data sequence of traffic states of spatially adjacent lanes for corresponding sub-segments for characterizing spatial information, RperiodFusing data sequences for the traffic states of the corresponding sub-road sections at the same moment in the historical period, wherein the data sequences are used for representing period information;
the convolution duration prediction network output running speed matrix AallThe expression of (a) is:
Aall＝{Aspace,Atime,Aperiod}
wherein A istimeFor predicting the operating speed of the corresponding sub-section from the time information, AspaceFor predicting the operating speed of the corresponding sub-section from the spatial information, AperiodThe operation speed of the corresponding sub-section is predicted according to the cycle information.
5. The method as claimed in claim 1, wherein the traffic state data sets include lane information, time information and traffic state values, the traffic state values are measured by a collecting means, each of the traffic state data sets includes at least two traffic state values, the lane information is a lane number corresponding to the traffic state value, and the time information is a time number of the traffic state value measured by the collecting means.
6. The method as claimed in claim 5, wherein the traffic state values include an average speed value, a headway distance, a space occupancy, a time occupancy and/or a 85-minute speed value.
7. The method for predicting the lane-level running speed of the highway based on the spatio-temporal information as claimed in claim 5, wherein the collection tools are millimeter wave radars, each sub-segment corresponds to one collection tool, and the sampling frequency of the millimeter wave radars is set to be 1 min.
8. The method as claimed in claim 1, wherein in step S2, the depth data fusion is implemented by a fusion model of spatio-temporal information characterization learning, the fusion model includes an input layer, an output layer and an intermediate layer, the input layer inputs traffic status data of sub-segments, the intermediate layer sequentially includes a convolutional layer, an active layer and a full link layer, and the output layer outputs the fused composite traffic status data.
9. The method as claimed in claim 3, wherein the convolutional long-and-short term prediction network comprises at least two ConvLstm networks connected to form a prediction network.
10. The method as claimed in claim 3, wherein the attention extraction network comprises a pooling layer, a full-link layer and a computing layer, the pooling layer down-samples the input data, the full-link layer splices and compresses the pooled data, the computing layer multiplies the full-link layer data by the input data, and the final prediction result is obtained by pooling again.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110124221.XA CN112907956B (en) | 2021-01-29 | 2021-01-29 | Expressway lane-level running speed prediction method based on space-time information |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110124221.XA CN112907956B (en) | 2021-01-29 | 2021-01-29 | Expressway lane-level running speed prediction method based on space-time information |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112907956A true CN112907956A (en) | 2021-06-04 |
| CN112907956B CN112907956B (en) | 2022-09-20 |
Family
ID=76120783
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110124221.XA Active CN112907956B (en) | 2021-01-29 | 2021-01-29 | Expressway lane-level running speed prediction method based on space-time information |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112907956B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114822019A (en) * | 2022-03-31 | 2022-07-29 | 阿里巴巴（中国）有限公司 | Traffic information processing method and device |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20160045390A (en) * | 2014-10-17 | 2016-04-27 | 주식회사 케이티 | Road velocity prediction method, server and system |
| CN109544911A (en) * | 2018-10-30 | 2019-03-29 | 中山大学 | A kind of city road net traffic state prediction technique based on LSTM-CNN |
| CN109754605A (en) * | 2019-02-27 | 2019-05-14 | 中南大学 | A kind of traffic forecast method based on attention temporal diagram convolutional network |
| CN110827546A (en) * | 2019-11-21 | 2020-02-21 | 银江股份有限公司 | Signalized intersection road section short-term speed prediction method |
| CN111540199A (en) * | 2020-04-21 | 2020-08-14 | 浙江省交通规划设计研究院有限公司 | High-speed traffic flow prediction method based on multi-mode fusion and graph attention machine mechanism |
| CN112201036A (en) * | 2020-09-25 | 2021-01-08 | 同济大学 | Urban expressway travel speed short-time prediction method based on increment-CNN |
-
2021
- 2021-01-29 CN CN202110124221.XA patent/CN112907956B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20160045390A (en) * | 2014-10-17 | 2016-04-27 | 주식회사 케이티 | Road velocity prediction method, server and system |
| CN109544911A (en) * | 2018-10-30 | 2019-03-29 | 中山大学 | A kind of city road net traffic state prediction technique based on LSTM-CNN |
| CN109754605A (en) * | 2019-02-27 | 2019-05-14 | 中南大学 | A kind of traffic forecast method based on attention temporal diagram convolutional network |
| CN110827546A (en) * | 2019-11-21 | 2020-02-21 | 银江股份有限公司 | Signalized intersection road section short-term speed prediction method |
| CN111540199A (en) * | 2020-04-21 | 2020-08-14 | 浙江省交通规划设计研究院有限公司 | High-speed traffic flow prediction method based on multi-mode fusion and graph attention machine mechanism |
| CN112201036A (en) * | 2020-09-25 | 2021-01-08 | 同济大学 | Urban expressway travel speed short-time prediction method based on increment-CNN |
Non-Patent Citations (1)
| Title |
|---|
| 陈凯勋: "基于注意力机制的短时交通流速度预测模型研究", 《中国优秀硕士学位论文全文数据库》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114822019A (en) * | 2022-03-31 | 2022-07-29 | 阿里巴巴（中国）有限公司 | Traffic information processing method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112907956B (en) | 2022-09-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110570651B (en) | Road network traffic situation prediction method and system based on deep learning | |
| CN109285346B (en) | Urban road network traffic state prediction method based on key road sections | |
| CN109544932B (en) | Urban road network flow estimation method based on fusion of taxi GPS data and gate data | |
| EP1921589B1 (en) | Traffic information interpolation system | |
| CN113313303A (en) | Urban area road network traffic flow prediction method and system based on hybrid deep learning model | |
| CN108335485B (en) | License plate recognition data-based method for predicting dynamic simulation congestion of traffic at major incident | |
| CN111160753B (en) | Road network node importance assessment method based on knowledge graph | |
| CN110310474A (en) | A kind of vehicle flowrate prediction technique and device based on space-time residual error network | |
| CN113538898A (en) | Multisource data-based highway congestion management and control system | |
| CN110782120B (en) | Method, system, equipment and medium for evaluating traffic flow model | |
| CN109410577A (en) | Adaptive traffic control sub-area division method based on Spatial Data Mining | |
| CN108648445B (en) | Dynamic traffic situation prediction method based on traffic big data | |
| CN114049765B (en) | Urban road network traffic flow OD estimation method based on automatic vehicle number plate identification data | |
| CN114783183A (en) | Monitoring method and system based on traffic situation algorithm | |
| CN112907956B (en) | Expressway lane-level running speed prediction method based on space-time information | |
| CN114463972A (en) | Road section interval traffic analysis and prediction method based on ETC portal communication data | |
| CN116504076A (en) | Expressway traffic flow prediction method based on ETC portal data | |
| CN110738855A (en) | Road traffic flow condition prediction method for data sparse time periods | |
| CN114596709B (en) | Data processing method, device, equipment and storage medium | |
| CN116913088A (en) | Intelligent flow prediction method for expressway | |
| CN109712402B (en) | Mobile object running time prediction method and device based on meta-path congestion mode mining | |
| KR102359902B1 (en) | Crossroads LOS Prediction Method Based on Big Data and AI, and Storage Medium Having the Same | |
| Anderson et al. | Travel time estimation in urban road networks | |
| CN111524358B (en) | Regional radiation sexual communication flow prediction method | |
| CN116580563B (en) | Markov chain-based regional congestion traffic source prediction method, device and equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |