CN106384507A - Travel time real-time estimation method based on sparse detector - Google Patents
Travel time real-time estimation method based on sparse detector Download PDFInfo
- Publication number
- CN106384507A CN106384507A CN201610832685.5A CN201610832685A CN106384507A CN 106384507 A CN106384507 A CN 106384507A CN 201610832685 A CN201610832685 A CN 201610832685A CN 106384507 A CN106384507 A CN 106384507A
- Authority
- CN
- China
- Prior art keywords
- detector
- data
- time
- real
- traffic flow
- 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
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
- G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
- G08G1/0108—Measuring and analyzing of parameters relative to traffic conditions based on the source of data
- G08G1/0116—Measuring and analyzing of parameters relative to traffic conditions based on the source of data from roadside infrastructure, e.g. beacons
-
- 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
Landscapes
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses a travel time real-time estimation method based on a sparse detector. The method comprises the steps: taking the traffic flow speed data of the detector as a classification attribute variable, and carrying out the clustering of detectors; selecting a detector with the highest data validity from each class as a key detector; taking the detection data of all key detectors as the input, taking the corresponding travel time data of a to-be-researched road segment, carrying out the training of a regression analysis model, and obtaining a travel time real-time estimation model; and finally carrying out the travel time real-time estimation through employing the obtained travel time real-time estimation model. The method can obtain the accurate travel time real-time estimation through the detection data of a few key detectors, can greatly reduce the maintenance cost of the detectors, and can reduce the requirements for the performance and storage spaces of calculation equipment of a traffic management and information system.
Description
Technical field
The present invention relates to a kind of journey time real-time estimation method only needing using a small amount of detector data, belong to intelligence
Technical field of transportation.
Background technology
In recent years, China's Expressway Development is rapid, reaches by building highway mileage open to traffic to the end of the year 2015
120000 kilometers.But the continuous increase with freeway network scale, Freeway Traffic Volume sharp increase, freeway traffic
Congestion problems are increasingly serious.In the case, relevant governmental administration section progressively adopts advanced traffic administration and information system
ATMIS (Advanced traffic management and information system) carries out informationization to highway
With intelligent management to tackle traffic jam issue.
Express highway section journey time real-time estimation is the corn module in ATMIS.Journey time letter in real time
Breath can provide real-time traffic circulation state for traffic administration person, traffic trip person, thus carrying out traffic behavior evaluation for them
There is provided foundation with communications policy.For example traffic administration person can change determination event according to journey time and position occurs;Traveler
Can be according to real-time travel time information plan or adjustment trip route.
At present, the travel time estimation of express highway section is set up in traffic flow basic data, and traffic flow data master
Detector (including coil checker, microwave detector etc.) to be passed through obtains.Existing express highway section journey time is real-time
Method of estimation generally requires the traffic flow data of all detectors in section.But due to by under vile weather, road surface breakage, roadbed
Heavy grade affects, and the detector that highway is laid easily damages, and causes partial data to lose efficacy.In the case, using existing side
The motorway journeys time Estimate result poor accuracy that method obtains.On the other hand, the traffic flow data of general detector every
30s collects once, and greatly, the traffic flow data using all detectors estimates the number to one side equipment for the journey time to data volume
Have higher requirements according to memory space, also bring the problems such as operation efficiency is low, and operation result is delayed.Additionally, prior art needs
Realizing travel time estimation, the operational maintenance maintenance cost of these detectors is very high for substantial amounts of detector.
Content of the invention
The technical problem to be solved is to overcome prior art not enough, provide a kind of based on sparse detector
Journey time real-time estimation method is it is only necessary to the detection data of a small amount of detector can obtain accurate travel time estimation in real time
Value, cost of implementation is relatively low.
The present invention specifically employs the following technical solutions solution above-mentioned technical problem:
Based on the journey time real-time estimation method of sparse detector, including model training stage and estimation stages;Model
Training stage includes:
Step A, traffic flow speed number within a period of time for the one group of detector laid in advance is collected on section to be ground
According to;
Step B, using described traffic flow speed data as categorical attribute variable, described detector is clustered;And from
Each apoplexy due to endogenous wind chooses a data validity highest detector as crucial detector;
Detection data within least one time cycle for each detector that step C, acquisition step A are laid in advance,
And calculate the corresponding travel time data in section to be ground according to these detection datas;
Step D, each moment being gathered using step C all key detectors detection data as input, to treat
Grind the corresponding travel time data in section as desired output, regression analysis model is trained, obtains journey time real-time
Estimate model;
Estimation stages include:
Step E, by all for a certain moment key detectors real-time detector data input travel time real-time estimation model,
The output section as to be ground of journey time real-time estimation model is in the travel time estimation value in this moment.
Preferably, using self organizing neural network, described detector is clustered.
Preferably, described regression analysis model is support vector machine.
Compared to existing technology, technical solution of the present invention has the advantages that:
1) utilize the inventive method, highway administration department only need to safeguard the normal operation of crucial detector with regard to energy by emphasis
There is provided accurate link travel time information to traffic trip person, thus substantially reducing maintenance and the maintenance cost of detector;
2) traffic administration and the performance of information system arithmetic facility can be reduced using the inventive method and memory space requires.
Brief description
Fig. 1 is the schematic flow sheet of journey time real-time estimation method of the present invention;
Fig. 2 is Loop detector layout schematic diagram;
Fig. 3 is the cluster result schematic diagram of I880N express highway section detector;
Fig. 4 is the true journey time of I880N express highway section and the contrast schematic diagram estimating journey time.
Specific embodiment
Below in conjunction with the accompanying drawings technical scheme is described in detail:
Carry out the problem that road travel time estimation is brought, the present invention for prior art using a large amount of detector data
Thinking be:First using the traffic flow speed data of detector as categorical attribute variable, detector is clustered;And from every
One apoplexy due to endogenous wind chooses a data validity highest detector as crucial detector;Then with the detection of all key detectors
Data, as input, using the corresponding travel time data in section to be ground as desired output, is trained to regression analysis model,
Obtain journey time real-time estimation model;Finally carry out journey time using obtained journey time real-time estimation model real-time
Estimate.By the program, it is effectively reduced the detector number needed for travel time estimation, that is, utilize sparse detector to realize fast
Speed accurately journey time real-time estimation.
The journey time real-time estimation method based on sparse detector for the present invention, as shown in figure 1, it includes model training rank
Section and estimation stages;The model training stage includes:
Step A, collect traffic flow speed data within a period of time for the one group of detector laid in advance in section to be ground;
First as shown in Fig. 2 a number of detector is laid on section to be ground, can adopt of the prior art herein
Loop detector layout mode;Or directly using existing detector on section to be ground.Then collect each detector at one section
In time, (present invention preferably collects the detection data on one week interior 24 hours 5 working day, and choosing working days evidence is due to work
Day traffic flow spatial-temporal distribution characteristic substantially, thus the travel pattern of readily identified detector and classified) traffic flow velocity
Degrees of data.The detector of arrangement connects computer, walks the flow on studied section, occupation rate and speed etc. for collection and counts
According to the data of collection exports every 5min to the data base of institute's connection computer.
Step B, using described traffic flow speed data as categorical attribute variable, described detector is clustered;And from
Each apoplexy due to endogenous wind chooses a data validity highest detector as crucial detector;
The detectable data of detector includes the data such as flow, occupation rate and speed, and why the present invention is according to traffic flow
Speed data is clustered, and is because speed is the tolerance traffic flow modes the most effective characteristic parameter.
The present invention can adopt existing various clustering methods, such as K-means cluster, ant colony clustering, artificial neural network
Cluster etc..Consider the factor such as whether simple the need of specified cluster classification number and algorithm, present invention preferably employs
Self organizing neural network (self-organized mapping, abbreviation SOM) is carrying out the cluster of detector.
SOM can carry out unsupervised learning cluster to data.Its thought is very simple, is substantially a kind of only input
Layer -- the neutral net of hidden layer.The class that one needs of one of hidden layer node on behalf are polymerized to.Using " competition is learned during training
The mode of habit ", the sample of each input finds the node of and its coupling in hidden layer, is referred to as its activation node,
Also it is " winning neuron ".And then stochastic gradient descent method is used to update the parameter of activation node.Meanwhile, and activation node
The point closing on is also according to them apart from the distance of activation nodes suitably undated parameter.With the most frequently used K-Means cluster side
Method compares, and SOM has advantages below:
(1) K-Means needs to fix the number of class in advance, that is, the value of K.SOM is then without some sections in hidden layer
Point can not have any input data to belong to it.So, K-Means is subject to initialized impact than larger.
(2) K-means is, after each input data finds a most like class, only to update the parameter of this class.SOM is then
The node closing on can be updated.So K-mean is affected by noise data, and ratio is larger, the accuracy of SOM may compare k-
Means low (because also have updated neighbor node).
(3) visualization of SOM is relatively good.Graceful topological relation figure.
Due to providing self organizing neural network workbox in MATLAB, therefore can be directly by each inspection in step A
The traffic flow speed data surveying device imports in MATLAB, that is, using traffic flow speed as categorical attribute variable, runs MATLAB's
Self organizing neural network workbox, you can obtain the cluster result of above-mentioned all detectors.
Choose one respectively from each apoplexy due to endogenous wind and there is the detector of higher data effectiveness as crucial detector.Vehicle is examined
Survey device data validity be evaluated as prior art, specifically can be found in document [Chengcheng Xu, Andrew P.Tarko,
Wei Wang,Pan Liu.Predicting crash likelihood and severity on freeways with
real-time loop detector data.Accident analysis and prevention.57(2013),30-
39.].The present invention is when judging the data validity of detector, it is preferred to use the discrimination standard of invalid data below:(1) put down
All occupation rate is more than 100%;(2) average speed is more than 0 and traffic flow rate is equal to 0;(3) more than 0 but occupation rate is traffic flow rate
0;(4) average speed is more than 160.9km/h;(5) more than 0 but traffic flow rate is equal to 0 to occupation rate.Meet one of conditions above, that is,
Can determine that as invalid data.
Detection data within least one time cycle for each detector that step C, acquisition step A are laid in advance,
And calculate the corresponding travel time data in section to be ground according to these detection datas;
Gather the inspection such as at least one week daily flow of 24 hours in section to be ground, occupation rate and speed using all of detector
Survey data, and calculate corresponding travel time data according to these detection datas.Detection data using detector calculates row
The journey time is known in the art general knowledge, and here is omitted.
Step D, each moment being gathered using step C all key detectors detection data as input, to treat
Grind the corresponding travel time data in section as desired output, regression analysis model is trained, obtains journey time real-time
Estimate model;
The detection data of all key detectors is contained, with every in the time cycle in the detection data that step C is obtained
The detection data of individual moment all keys detector as input sample, using corresponding travel time data as desired output,
Regression analysis model can be trained, after the completion of training, can get journey time real-time estimation model.The present invention is permissible
Using regression analysis models such as support vector machine, feedforward neural network, deep learning neutral nets it is contemplated that the complexity of algorithm
And the complexity obtaining, it is preferred to use support vector machine are as regression analysis model.For example, can directly invoke in MATLAB
Support vector machine regression workbox.Estimation stages include:
Step E, by all for a certain moment key detectors real-time detector data input travel time real-time estimation model,
The output section as to be ground of journey time real-time estimation model is in the travel time estimation value in this moment;
After obtaining journey time real-time estimation model, the detection data that the crucial detector of input any time is gathered is
Real-time estimation can be carried out to the journey time of this moment express highway section.
In order to verify the effect of the inventive method, carry out simulating, verifying experiment.This experiment adopts California, USA
The true traffic flow data of the I880N highway collection of state San Francisco Bay Area.
Select the long 36.9mile in I880N highway section first and lay detector 109, select in April, 2015
Traffic flow speed data to all detectors on April 17 I880N highway section on the 13rd, and it is stored in data base.To detect
The traffic flow speed data of device imports the self organizing neural network workbox running MATLAB in MATLAB, obtains 109 detections
The cluster result of device, as shown in Figure 3.Each hexagon of in figure represents a classification, and the numeral in hexagon represents in the category
The number of detector.Two larger for similarity classifications are merged, and selects one from each classification and there is higher number
According to accuracy detector as crucial detector, altogether screening obtains 10 crucial detectors.
Collect I880N express highway section in two weeks the traffic flow of the journey time of daily 24 hours and crucial detector account for
There are rate and data on flows, and be stored in data base.Above-mentioned data is imported in MATLAB, using the data of first week as training set,
Run the Support vector regression workbox in MATLAB, through training, obtain three key parameter C of support vector machine, ε and
The value of σ is respectively 32,0.01,0.25.Using the supporting vector machine model having trained, using the data of second week as test
Collection, the crucial detector traffic flow data of input any time can estimate the journey time in this moment.Table 1 shows using
Several groups of journey times that the supporting vector machine model training estimates.
Table 1 travel time estimation
Crucial detector traffic fluxion using I880N express highway section April 18 to each moment in April 30 in 2015
According to the supporting vector machine model that input has trained, obtained true journey time and the comparing result estimating journey time
As shown in Figure 4.
According to true journey time and estimation journey time result, using root-mean-square value (RME) and root-mean-square error
(RMSE) two indices to be being evaluated, RME and RMSE obtaining is respectively 0.032 and 0.052, and error amount very little can connect
In the range of being subject to, illustrate that the inventive method reliability and accuracy are higher.
Claims (5)
1. the journey time real-time estimation method based on sparse detector is it is characterised in that include model training stage and estimation
Stage;The model training stage includes:
Step A, traffic flow speed data within a period of time for the one group of detector laid in advance is collected on section to be ground;
Step B, using described traffic flow speed data as categorical attribute variable, described detector is clustered;And from each
Apoplexy due to endogenous wind chooses a data validity highest detector as crucial detector;
Detection data within least one time cycle for each detector that step C, acquisition step A are laid in advance, and root
Calculate the corresponding travel time data in section to be ground according to these detection datas;
Step D, each moment being gathered using step C all key detectors detection data as input, with Dai Yan road
The corresponding travel time data of section, as desired output, is trained to regression analysis model, obtains journey time real-time estimation
Model;
Estimation stages include:
Step E, by all for a certain moment key detectors real-time detector data input travel time real-time estimation model, stroke
The output section as to be ground of time real-time estimation model is in the travel time estimation value in this moment.
2. method as claimed in claim 1 is it is characterised in that clustered to described detector using self organizing neural network.
3. method as claimed in claim 1 is it is characterised in that described regression analysis model is support vector machine.
4. method as claimed in claim 1 is it is characterised in that when judging the data validity of detector, invalid data
Discrimination standard is to meet one of following condition:(1) average occupancy is more than 100%;(2) average speed is more than 0 and traffic flow
Rate is equal to 0;(3) more than 0 but occupation rate is 0 to traffic flow rate;(4) average speed is more than 160.9km/h;(5) occupation rate is more than 0
But traffic flow rate is equal to 0.
5. method as claimed in claim 1 it is characterised in that traffic flow speed data collected by step A be one week interior five
The traffic flow speed data on 24 hours working days.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610832685.5A CN106384507A (en) | 2016-09-20 | 2016-09-20 | Travel time real-time estimation method based on sparse detector |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610832685.5A CN106384507A (en) | 2016-09-20 | 2016-09-20 | Travel time real-time estimation method based on sparse detector |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106384507A true CN106384507A (en) | 2017-02-08 |
Family
ID=57936733
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610832685.5A Pending CN106384507A (en) | 2016-09-20 | 2016-09-20 | Travel time real-time estimation method based on sparse detector |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106384507A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107545785A (en) * | 2017-07-21 | 2018-01-05 | 华南理工大学 | A kind of river channel running method based on big data |
CN108417029A (en) * | 2018-02-11 | 2018-08-17 | 东南大学 | City road network travel time estimation method based on adaptive multitask deep learning |
CN109417767A (en) * | 2017-04-27 | 2019-03-01 | 北京嘀嘀无限科技发展有限公司 | For determining the system and method for estimating arrival time |
CN109712398A (en) * | 2019-01-22 | 2019-05-03 | 江苏智通交通科技有限公司 | Motorway journeys time Estimate Model Parameter Optimization method |
CN113450566A (en) * | 2021-06-22 | 2021-09-28 | 中科曙光(南京)计算技术有限公司 | Urban traffic flow prediction method |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH0850695A (en) * | 1994-08-03 | 1996-02-20 | Nippon Signal Co Ltd:The | Determining method for time to be required for travel |
US20060161334A1 (en) * | 2004-12-28 | 2006-07-20 | Denso Corporation | Average vehicle speed computation device and car navigation device |
CN101436347A (en) * | 2008-12-09 | 2009-05-20 | 北京交通大学 | Prediction method for rapid road travel time |
CN101551945A (en) * | 2009-05-13 | 2009-10-07 | 东南大学 | Traffic security early warning method of traffic control system on highway |
CN102831431A (en) * | 2012-02-05 | 2012-12-19 | 四川大学 | Detector training method based on hierarchical clustering |
CN103198672A (en) * | 2013-03-27 | 2013-07-10 | 大连海事大学 | Method for laying urban road network traffic flow detectors |
-
2016
- 2016-09-20 CN CN201610832685.5A patent/CN106384507A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH0850695A (en) * | 1994-08-03 | 1996-02-20 | Nippon Signal Co Ltd:The | Determining method for time to be required for travel |
US20060161334A1 (en) * | 2004-12-28 | 2006-07-20 | Denso Corporation | Average vehicle speed computation device and car navigation device |
CN101436347A (en) * | 2008-12-09 | 2009-05-20 | 北京交通大学 | Prediction method for rapid road travel time |
CN101551945A (en) * | 2009-05-13 | 2009-10-07 | 东南大学 | Traffic security early warning method of traffic control system on highway |
CN102831431A (en) * | 2012-02-05 | 2012-12-19 | 四川大学 | Detector training method based on hierarchical clustering |
CN103198672A (en) * | 2013-03-27 | 2013-07-10 | 大连海事大学 | Method for laying urban road network traffic flow detectors |
Non-Patent Citations (4)
Title |
---|
CHENGCHENG XU 等: "Predicting crash likelihood and severity on freeways with real-time loop detector data", 《ACCIDENT ANALYSIS & PREVENTION》 * |
GUIYAN JIANG 等: "Travel time prediction for urban arterial road", 《PROCEEDINGS OF THE 2003 IEEE INTERNATIONAL CONFERENCE ON INTELLIGENT TRANSPORTATION SYSTEMS》 * |
TANG, JJ 等: "Travel Time Estimation Using Freeway Point Detector Data Based on Evolving Fuzzy Neural Inference System", 《PLOS ONE》 * |
桑丽 等: "旅行时间数据采集及预测技术综述", 《2008第四届中国智能交通年会论文集》 * |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109417767A (en) * | 2017-04-27 | 2019-03-01 | 北京嘀嘀无限科技发展有限公司 | For determining the system and method for estimating arrival time |
CN109417767B (en) * | 2017-04-27 | 2020-09-18 | 北京嘀嘀无限科技发展有限公司 | System and method for determining estimated time of arrival |
US11398002B2 (en) | 2017-04-27 | 2022-07-26 | Beijing Didi Infinity Technology And Development Co., Ltd. | Systems and methods for determining an estimated time of arrival |
CN107545785A (en) * | 2017-07-21 | 2018-01-05 | 华南理工大学 | A kind of river channel running method based on big data |
CN108417029A (en) * | 2018-02-11 | 2018-08-17 | 东南大学 | City road network travel time estimation method based on adaptive multitask deep learning |
CN108417029B (en) * | 2018-02-11 | 2020-11-06 | 东南大学 | Urban road network travel time estimation method based on self-adaptive multitask deep learning |
CN109712398A (en) * | 2019-01-22 | 2019-05-03 | 江苏智通交通科技有限公司 | Motorway journeys time Estimate Model Parameter Optimization method |
WO2020151293A1 (en) * | 2019-01-22 | 2020-07-30 | 江苏智通交通科技有限公司 | Parameter optimization method for expressway travel time estimation model |
CN109712398B (en) * | 2019-01-22 | 2021-05-28 | 江苏智通交通科技有限公司 | Expressway travel time estimation model parameter optimization method |
CN113450566A (en) * | 2021-06-22 | 2021-09-28 | 中科曙光(南京)计算技术有限公司 | Urban traffic flow prediction method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107248283B (en) | A kind of urban area road network evaluation of running status method considering section criticality | |
CN103839409B (en) | Based on the traffic flow modes method of discrimination of multibreak facial vision sensing cluster analysis | |
Kim et al. | Urban traffic flow prediction system using a multifactor pattern recognition model | |
CN106384507A (en) | Travel time real-time estimation method based on sparse detector | |
CN105303197B (en) | A kind of vehicle follow the bus safety automation appraisal procedure based on machine learning | |
CN102663264B (en) | Semi-supervised synergistic evaluation method for static parameter of health monitoring of bridge structure | |
CN101216998B (en) | An urban traffic flow information amalgamation method of evidence theory based on fuzzy rough sets | |
Dia et al. | Development and evaluation of arterial incident detection models using fusion of simulated probe vehicle and loop detector data | |
Xia et al. | A clustering approach to online freeway traffic state identification using ITS data | |
CN102592453B (en) | Real-time traffic condition judging method based on time window | |
CN108492555A (en) | A kind of city road net traffic state evaluation method and device | |
CN104021672B (en) | A kind of method and apparatus obtaining traffic congestion index | |
CN109191849B (en) | Traffic jam duration prediction method based on multi-source data feature extraction | |
CN106023344A (en) | Driving style estimation method based on driving pattern transition probability | |
CN106652441A (en) | Urban road traffic condition prediction method based on spatial-temporal data | |
CN106856049B (en) | Key intersection demand aggregation analysis method based on bayonet number plate identification data | |
CN112613225B (en) | Intersection traffic state prediction method based on neural network cell transmission model | |
CN107180534B (en) | The express highway section average speed estimation method of support vector regression fusion | |
CN110047291A (en) | A kind of Short-time Traffic Flow Forecasting Methods considering diffusion process | |
CN107437339A (en) | Variable information advices plate control method for coordinating and system under a kind of information guidance | |
CN109785618A (en) | Short-term traffic flow prediction method based on combinational logic | |
CN106530704A (en) | Floating car aggregation detection method based on multivariate data fusion | |
CN112884014A (en) | Traffic speed short-time prediction method based on road section topological structure classification | |
CN116052435B (en) | Urban road congestion influence range definition and road influence calculation method | |
CN110400462A (en) | Track traffic for passenger flow monitoring and pre-alarming method and its system based on fuzzy theory |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170208 |