CN107123265A - A kind of traffic status of express way method of estimation based on parallel computation - Google Patents
A kind of traffic status of express way method of estimation based on parallel computation Download PDFInfo
- Publication number
- CN107123265A CN107123265A CN201710439953.1A CN201710439953A CN107123265A CN 107123265 A CN107123265 A CN 107123265A CN 201710439953 A CN201710439953 A CN 201710439953A CN 107123265 A CN107123265 A CN 107123265A
- Authority
- CN
- China
- Prior art keywords
- mrow
- msubsup
- data
- traffic
- module
- 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
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
- G08G1/0125—Traffic data processing
- G08G1/0129—Traffic data processing for creating historical data or processing based on historical data
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16Z—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS, NOT OTHERWISE PROVIDED FOR
- G16Z99/00—Subject matter not provided for in other main groups of this subclass
Landscapes
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses a kind of express highway traffic state estimation method;This method includes a kind of parallel computation frame and four submodules:Data center module, data unification processing module, online traffic flow model parameter correction module and improved Kalman filtering optimization module, data center module are used for the history and real time data for gathering and storing traffic detector;Data unification processing module is unified for same form by different types of traffic data;Online traffic flow model parameter correction module is used for the parameter correction of traffic flow model;Kalman filtering optimization module further optimizes to the model result after correction.Parallel computation frame proposed by the present invention is used to automatically process the execution sequence between above-mentioned module, compared to the existing traffic state estimation method based on Kalman filtering, the present invention has higher execution efficiency and accuracy, different types of highway fixed detector data can be supported, with good application and promotion prospect.
Description
Technical field
The present invention relates to traffic status of express way estimation and identification field, and in particular to one kind is based on parallel computation and changes
Enter the traffic status of express way method of estimation of Kalman filtering algorithm.This method is directed to traffic status of express way fixed test
The problem of device lays not enough, according to real-time and historical traffic detector data, estimates real everywhere on the specified section of highway
When traffic behavior.
Background technology
In recent years, with China's economy sustainable growth and the quick increase of vehicle guaranteeding organic quantity and utilization rate, it is existing
Operation and Management of Expressway method can not fully meet highway real-time management demand, therefore, to Expressway Implementing master
Dynamic management and control are the development trends of freeway management from now on.Highway active management and control system be based in real time and
Historical traffic status data, by control centre and mainframe computer work station, come in and gone out to vehicle highway, bottleneck road are limited
Speed, vehicle lane-changing etc. implement automatic management and control.Therefore need to obtain the accurate real-time traffic states data of freeway
Premised on (i.e. Road average-speed and section density).The fixed detector of existing highway is set including (high definition) camera
Standby, microwave equipment, coil apparatus, its distribution density are generally 5~10 kilometers one, and the most short sampling time interval of data is 1~5
Minute, it is impossible to meet the demand of Real time Adaptive Traffic Control.Therefore need to carry out expansion sample to traffic data on room and time,
Accurate traffic status of express way is obtained by traffic state estimation method, the base of highway active management and control is used as
Plinth.
Existing traffic state estimation method can be divided into method of estimation based on traffic data and based on traffic flow mould
The method of estimation of type.Method of estimation based on traffic data is by the statistical analysis technique to known traffic data to traffic behavior
Unknown section is estimated, including Time series analysis method, homing method, Smoothing fit method.Based on traffic flow model
Method the traffic behavior evolution of highway is described by traffic flow model, and by other algorithms to traffic flow model
Parameter and result of calculation optimize, to improve the accuracy of traffic flow model.In general, the side based on traffic flow model
The accuracy that method describes traffic flow change is higher, and can accurately describe the formation of traffic congestion, expands and dissipated
Journey, can preferably meet the demand of active traffic control.The existing method based on traffic flow model is mainly using second order macroscopic view
The method that traffic flow model and EKF are combined.Such method can obtain preferable estimation effect and precision,
But there is also following deficiency:
1) parameter for the traffic flow model that such method is used is more, and Partial key parameter passes through Kalman filtering algorithm
Estimated in real time, the setting of remaining parameter still relies on experience, therefore easily produces larger error.
2) because traffic flow model is nonlinear model, such method is needed by carrying out Taylor series to traffic flow model
The mode of expansion (linear approximation) calculates kalman gain.The amount of calculation of which is larger and can produce approximate evaluation error,
The precision and efficiency of method of estimation are influenceed to a certain extent.
The content of the invention
To solve the above problems, the invention discloses a kind of high speed optimized based on parallel computation and improved Kalman filter
Highway traffic state method of estimation.This method belongs to the traffic state estimation method based on traffic flow model, including a kind of parallel
Computational frame, a kind of data unification Processing Algorithm, a kind of online traffic flow model parameter correction algorithm and one kind are improved
Kalman filtering optimized algorithm.The present invention can extract necessary traffic state data from different types of fixed detector, lead to
Parallel computation frame lifting traffic behavior estimated efficiency is crossed, is handed over by on-time model parameter correction and Kalman filtering optimization lifting
The accuracy of logical state estimation, so as to meet active management and control demand.The concrete technical scheme that the present invention is used is:
1. a kind of new traffic estimations parallel computation frame, including:
Data center module, data unification processing module, online traffic flow model parameter correction module, improved karr
Graceful filtering optimization module and corresponding parallel computation flow.Wherein, data center module includes traffic database and pattern number
According to storehouse.Traffic database is used for the history and real time traffic data for collecting and storing the collection of highway fixed detector, and bears
Blame the maintenance to traffic data.Model library is used to store different types of traffic flow model, it is possible to according to different traffic conditions
From the higher traffic flow model of fitness;Different types of fixed detector data mart modeling is by data unification processing module
Section density and speed data required for other modules;Online traffic flow model parameter correction module passes through genetic algorithm and short
When traffic historical data, traffic flow model parameter is corrected, make the model calculation meet real-time traffic states change become
Gesture.Improved Kalman filtering optimization module is using real time traffic data to the further optimization of the model calculation.
Parallel computation flow between each module comprises the following steps:
1) data center obtains and stored in real time the traffic data of each detector of highway;
2) data unification processing module is used for the request of data for responding other modules.The module obtains correspondence from database
The data of period, using data unification Processing Algorithm processing data and are sent to the module of request.
3) online traffic flow model parameter correction module corrects traffic flow using online traffic flow model parameter optimization algorithm
Model, makes the result of calculation of traffic flow model meet short time traffic conditions variation tendency.Traffic flow model parameter correction is divided into just
Stage beginning and conventional stage.In initial phase, online traffic flow model parameter correction module handles mould to data unification
Block asks the traffic data of the previous day, and optimal model parameter is calculated according to previous day data.Then into conventional stage, at this
Terminate in stage until traffic estimations method is performed, be that time interval corrects (renewal) model parameter with one hour.Specific steps
For:First, to the traffic data of upper one hour of request of data unification processing module;Secondly, joined using online traffic flow model
Number optimized algorithm, model parameter is updated according to the traffic data of upper one hour;Then, based on newest model parameter, traffic flow
Model calculated the traffic behavior estimate in each sub- section of highway with 30 seconds time intervals;Finally, result of calculation is sent to
Improved Kalman filtering optimization module does further optimization.
4) improved Kalman filtering optimization module asks real time traffic data to data processing module.Utilize improved card
Kalman Filtering optimized algorithm, optimal estimated result is calculated according to real time traffic data and the model calculation.And estimate gained
Count final output result of the result as the moment.
5) optimal estimation result is sent to online traffic flow model parameter correction module by improved Kalman's optimization module,
The model computational threads of the latter as input data, and based on newest model parameter value, solve optimal estimation result next
The model result at moment.
2. the data unification Processing Algorithm described in above-mentioned steps (2), for by existing highway fixed detector
Data (referring to video frequency pick-up head data, microwave data, loop data) are unified only to include section density and average speed, between the time
It is divided into the traffic data of 30 seconds.Existing fixed detector data include the speed data, data on flows, occupation rate number of test point
According to data were updated at intervals of 1 or 5 minute, without density data.Therefore above-mentioned purpose is reached by following process step.
1) according to the demand for control of traffic active management system of automotive, by specified express highway section it is discrete be 500 meters of spacing
Sub- section;It is public at a high speed according to CFL condition (requiring that vehicle can only appear on adjacent sub- section at the next moment) and China
Bus speed highest 120km/h limitation, will be defined as 30 seconds between data renewal time;
2) interpolation method is used by the unified traffic data to meet other module demands of the traffic data of different time granularity.
2.1) for traffic flow data, using the data interpolating method based on Poisson distribution.Step is:For 1 minute
Data on flows q1min, using k=2, λ=0.5q1minPoisson distribution function calculate the magnitude of traffic flow of every 30 seconds.For 5 minutes
Data on flows q5min, using k=10, λ=0.1q5minPoisson distribution function calculate the magnitude of traffic flow of every 30 seconds.
2.2) for speed data, the place average speed of detector is converted into road-section average using following formula first
Speed:
In formula,For express highway section i fixed detector in moment k place velocity measurement, σi(k)
It is the sample variance of spot spe J,It is the place average speed of section i detector.
Secondly, carried out using the Road average-speed data at 1 minute or 5 minutes interval of cubic Hamiltonian symmetrical systems function pair
Interpolation, obtains the link average speed estimation value of 30 seconds time intervals
3) density of traffic behavior is solved.Due to there was only occupation rate data in traffic data, it is therefore desirable to according to occupation rate
Data, density calculated value solves the estimate of density.Specific solution procedure is as follows:
3.1) density in section is calculated using following speed-density formula, density calculated value is obtained:
In formula,By the section i that solves in 2) moment k flow value,By the road-section average speed solved in 3)
Degree.T is time interval (30 seconds).
3.2) occupation rate is solved according to below equationWith density calculated valueBetween optimum linearity relation a:
In formula, a*To make occupation rateAnd densityThe scalefactor value of least square relation is met,For 3-1) in
The density calculated value solved,It is occupation rate data of the section i in moment k, N is total time space-number.Final density is estimated
Evaluation
3.3) willWithExported as final result.
3. a kind of online traffic flow model parameter optimization algorithm described in above-mentioned steps (3), the purpose of the algorithm is minimum
Change the Euclidean distance between the model calculation and short-term traffic historical data, so that both are closest.Specifically perform step
Suddenly it is:
1) set section i moment k model calculation value asTraffic data value is
P is model parameter vector, according to below equation calculating parameter be p when both distance:
2) optimization aim of genetic algorithm is determined, is divided into three sub-steps:
2.1) fitness function for determining genetic algorithm is:
In formula, T is the time interval number in 1 hour, and N is the sub- section number of highway after dividing,Be each section the model calculation and traffic data between Euclidean distance.
2.2) determined to solve optimized parameter p according to below equation*:
In formula, fkAnd f (p)k-1(p) be respectively fitness function f (p) kth time and -1 iteration of kth.ε calculates for heredity
The iteration stopping condition of method.
2.3) stop condition judges:When genetic algorithm meets stop condition ε or reaches greatest iteration algebraically, algorithm knot
Beam.Corresponding model parameter value p is returned to, p now is the optimal solution p of model parameter*。
4. a kind of improved Kalman filtering optimized algorithm described in above-mentioned steps (4) carries out excellent to the result of calculation of model
Change, concrete implementation step includes:
1) initialize:Obtain the initial model estimated result of each section i on highwayTraffic dataAnd mould
The covariance sigma of the Joint Distribution of type estimated result and traffic dataFused, 0;
2) mean μ of each section in moment k model estimated result on highway is calculated1, k, and real time traffic data
Mean μ2, k:
3) according to step 2) in mean μ1, k、μ2, k, calculate highway take up an official post a section moment k model estimate knot
The variances sigma of fruit1, k, and traffic data variances sigma2, k:
4) according to step 3) in σ1, kAnd σ2, kSolve the kalman gain K at moment kk.In this step, improved karr
Graceful filtering optimization algorithm no longer solves kalman gain by way of calculating Jacobian matrix, but passes through the model calculation
Solved with the variance ratio of traffic data observation:
5) according to kalman gain more new model estimated result, formula is as follows:
In formula,For moment k optimal estimation result,For moment k the model calculation,For moment k
Traffic observation data, k be the k-1 moment optimal estimation result, H is observing matrix, is in the present invention unit matrix.
6) the more covariance of new model estimated result and real time traffic data, formula is as follows:
7) by the model estimated result of renewalExported as moment k final estimated result.
The beneficial effects of the invention are as follows:
The present invention proposes a kind of traffic status of express way estimation side based on parallel computation and improved Kalman filter
Method.With now more compared with the ripe traffic behavior algorithm for estimating (abbreviation EKF algorithms) based on EKF, this hair
It is bright advantageous in estimated accuracy and time efficiency, and the advantage in time efficiency is more obvious.Reason is as follows:(1)
In terms of estimated accuracy, the estimation of model parameter is all based on the optimal value of the parameter of traffic data in the present invention.And EKF algorithms are removed
Important models parameter is the estimation based on traffic data, and remaining model parameter is empirical value, in actual applications error compared with
Greatly;In addition, EKF algorithms need to solve kalman gain, thus meeting by the method for linearizing nonlinear traffic flow model
Produce linear estimation error.Kalman is calculated in the present invention according to the variance of the model calculation and traffic data observation to increase
Benefit, will not produce linear estimation error.(2) in terms of time efficiency, the time complexity of Kalman filtering algorithm is O
(n2.376), n is the dimension of state vector in formula.The state vector dimension of EKF algorithms includes traffic behavior variable number and model
Number of parameters.And the dimension of state vector is only traffic behavior variable number in the present invention, Model Parameter Optimization passes through online mould
Shape parameter correction module is completed, and because being parallel computation, will not be produced influence to Kalman filtering optimization, can effectively be lifted
The computational efficiency of algorithm.In addition, the present invention have independent of bottom traffic data form, favorable expandability, moderate cost spy
Point, with preferable engineering practice value.
Brief description of the drawings
Fig. 1 is frame diagram of the invention.
Fig. 2 is timing diagram of the invention.
Fig. 3 is online traffic flow model parameter optimization algorithm flow chart of the invention.
Embodiment
With reference to the accompanying drawings and detailed description, the present invention is furture elucidated, it should be understood that following embodiments are only
For illustrating the present invention rather than limitation the scope of the present invention.
It is the graph of a relation of each module in traffic state estimation method proposed by the present invention as shown in Figure 1.The traffic of the present invention
Method for estimating state includes data center module, data unification processing module, online traffic flow model parameter correction module, changed
The Kalman filtering optimization module entered.The function of each module includes:Data center module, including traffic database and model data
Storehouse.Traffic database is used for the history and real-time traffic states data for collecting and storing the collection of highway fixed detector, and
It is responsible for the maintenance to traffic data.Model library is used to store different types of traffic flow model, it is possible to according to different traffic feelings
Condition is from the higher traffic flow model of fitness;Data unification processing module is by different types of fixed detector data mart modeling
For the section density and speed data required for other modules;Online traffic flow model parameter correction module includes traffic flow model
Thread and Genetic Algorithm for Correction thread.Model thread was calculated based on newest model parameter with 30 seconds time interval more new models
As a result.Genetic Algorithm for Correction thread, for time interval calibration model parameter, made traffic flow model result of calculation meet reality with 1 hour
When traffic behavior variation tendency.Improved Kalman filtering optimization module enters one using real time traffic data to the model calculation
Step optimization.
Fig. 2 is the parallel computation call relation timing diagram between each module.Illustrate calling for each module with reference to Fig. 1 and Fig. 2
Step is as follows:
1) data center obtains and stored in real time the traffic data of each detector of highway.
2) data unification processing module is used for the request of data for responding and handling other modules.Process step is as follows:
The traffic data of request corresponding road section and correspondence period 2-1) is extracted from database.Specifically, for from ginseng
The request of the online traffic flow model parameter correction module of number initial phase, sends the traffic data of the previous day whole day.For
The request of the online traffic flow model parameter correction module in other stages, sends the traffic data of previous hour, for from changing
The request for the Kalman filtering optimization module entered, sends real time traffic data (traffic data at current time);
2-2) traffic data is converted to using data interpolating method meet module requirement density and speed data it is concurrent
Give corresponding module;
3) online traffic flow model parameter correction module is in each traffic flow model parameter of clock renewal.Specific steps
For:First, Genetic Algorithm for Correction thread uses the data and genetic algorithm that data unification processing module is returned, to traffic model
Parameter is corrected, and makes the result of the model calculation after correction and traffic data in short-term closest.Secondly, traffic flow model
Thread iterated to calculate traffic behavior result with 30 seconds time intervals, and sent result to improved Kalman filtering optimization mould
Block;
4) improved Kalman filtering optimization module asks real time traffic data to data processing module.Use improved card
Kalman Filtering algorithm and real time traffic data, are optimized to the model calculation.And it regard gained estimated result as the moment
Final output result.Meanwhile, optimal estimation result is sent to online traffic flow model parameter by Kalman filtering optimization module
Correction module;
5) the traffic flow model thread of online traffic flow model parameter correction module regard optimal estimation result as the lower moment
Input data, and based on newest model parameter value, calculate the model result at lower moment.
The operating procedure of data unification processing module of the present invention, implements according to following steps:
1) according to the demand for control of traffic active management system of automotive, by the express highway section of control object it is discrete be 500 meters
The sub- section at interval;According to CFL condition and China expressway speed highest 120km/h limitation, by data renewal time
Between be defined as 30 seconds;
2) interpolation method is used by the unified traffic data to meet other module demands of the traffic data of different time granularity.
For traffic flow data, using the data interpolating method based on Poisson distribution.Step is:For the data on flows of 1 minute
q1min, using k=2, λ=0.5q1minPoisson distribution function calculate the magnitude of traffic flow of every 30 seconds.For the data on flows of 5 minutes
q5min, using k=10, λ=0.1q5minPoisson distribution function calculate the magnitude of traffic flow of every 30 seconds.
For speed data, average link speed is converted to using following formula spot spe Js first:
In formula,For express highway section i fixed detector in moment k place velocity measurement, σi(k)
It is the sample variance of spot spe J,It is section i average speed.Secondly, using cubic Hamiltonian symmetrical systems function pair
The Road average-speed data at 1 minute or 5 minutes interval enter row interpolation, obtain the link average speed estimation of 30 seconds time intervals
Value
3) density data is solved.First, the density in section is calculated according to following speed-density formula:
In formula,By it is above-mentioned 2) in the section i that solves moment k flow value,Put down by the section solved in 2)
Equal speed.T is time interval length.Secondly, occupation rate is solved according to below equationAnd densityBetween linear relationship a:
In formula, a*To make occupation rateAnd densityThe scalefactor value of least square relation is met,To have solved
Density calculated value,It is occupation rate data of the section i in moment k, N is total time space-number.Final densities are finally solved to estimate
Evaluation.Density estimation value is
4) willWithExported as final result.
Fig. 3 illustrates that traffic flow model computational threads and genetic algorithm in online traffic flow model parameter correction module are excellent
Change the execution sequence and interactive relation of thread.The online traffic flow model parameter correction of the present invention is further illustrated with reference to Fig. 3
The implementation steps of module, it is specific as follows:
1) set section i moment k model calculation value asTraffic data value is
P is model parameter vector, according to below equation calculating parameter be p when both distance:
2) optimization aim of genetic algorithm is determined, is divided into three sub-steps:
The fitness function for 2-1) determining genetic algorithm is:
In formula, T is the time interval number in 1 hour, and N is the sub- section number of highway after dividing,
Be each section the model calculation and traffic data between Euclidean distance.
2-2) determined to solve optimized parameter p according to below equation*:
In formula, fkAnd f (p)k-1(p) be respectively fitness function f (p) kth time and -1 iteration of kth.ε calculates for heredity
The iteration stopping condition of method.The parameter " population quantity " of genetic algorithm, " greatest iteration algebraically ", " stop condition ε " is set respectively
For 100,300 and 10-3。
2-3) stop condition judges:When genetic algorithm meets stop condition ε or reaches greatest iteration algebraically, algorithm knot
Beam.Corresponding model parameter vector p is returned to, p now is the optimal solution p of model parameter*.Traffic flow model computational threads root
According to newest parameter vector p*Calculating results.
The improved Kalman filtering of the present invention optimizes the operating procedure of doorframe, implements according to following steps:
1) initialize:Obtain the initial model estimated result of each section i on highwayAnd traffic dataAnd
The Joint Distribution (according to the assumed condition of Kalman filtering algorithm, being defaulted as Gaussian Profile) of model estimate value and traffic data
Initial covariance sigmaFused, 0;
2) mean μ of each section in moment k model estimated result on highway is calculated1, k, the average of traffic data
μ2, k, μ1, kAnd μ2, kIterative calculation formula difference it is as follows:
In formula, μ1, k-1And μ2, k-1The respectively model estimated result average and the average of traffic data at k-1 moment, n be to
Total iterative steps untill moment k.
3) according to σFused, k-1The more mean μ of new model estimated result1, k, formula is as follows:
In formula, σFused, k-1For k-1 moment model estimate value and the covariance of the Joint Distribution of traffic data, σ2, kDuring for k
The variance of the traffic data at quarter, μ1, kAnd μ2, kThe respectively average of model estimated result and the average of traffic data.
4) according to μ1, kAnd μ2, kHighway is calculated to take up an official post variances sigma of the section in moment k model estimated result1, k, with
And the variances sigma of traffic data2, k, calculation formula is distinguished as follows:
In formula, σ1, k-1And σ2, k-1For the variance at k-1 moment, μ1, kAnd μ2, kRespectively the average of model estimated result and
The average of traffic data.
5) according to step 4) in σ1, kAnd σ2, kSolve kalman gain Kk, formula is as follows:
6) according to kalman gain more new model estimated result, formula is as follows:
In formula,For moment k optimal estimation result,For moment k the model calculation,For moment k
Traffic observation data, k be the k-1 moment optimal estimation result, H is observing matrix, is in the present invention unit matrix.
7) the more covariance of new model estimated result and real time traffic data, formula is as follows:
8) by the model estimated result of renewalExported as moment k final estimated result.
Technological means disclosed in the present invention program is not limited only to the technological means disclosed in above-mentioned embodiment, in addition to
Constituted technical scheme is combined by above technical characteristic.
Claims (4)
1. a kind of traffic status of express way method of estimation based on parallel computation, it is characterised in that including parallel computation frame,
The parallel computation frame includes data center module, data unification processing module, online traffic flow model parameter correction mould
Block, improved Kalman filtering optimization module and corresponding parallel computation flow,
Data center module includes traffic database and model database.Traffic database is consolidated for collecting and storing highway
Determine the history and real-time traffic states data of detector collection, and be responsible for the maintenance to traffic data, model library is used to store not
Congener traffic flow model, it is possible to according to different traffic conditions from the higher traffic flow model of fitness;
Different type fixed detector data mart modeling, according to the request of other modules, is institute by data unification Processing Algorithm module
The section density and speed data that need and the module for being sent to request;
Online traffic flow model parameter correction module traffic historical data using genetic algorithm and in short-term, to traffic flow model parameter
It is corrected, traffic flow model result of calculation is met real-time traffic states variation tendency;
Improved Kalman filtering optimization module further optimizes according to real time traffic data to the result of calculation of traffic flow model;
Parallel computation flow between above-mentioned each module, including:
1) data center obtains and stored in real time the traffic data of each detector of highway;
2) data unification processing module responds the request of data of other modules, and the data of correspondence period are obtained simultaneously from database
The module of request is sent to after handling data;
3) online traffic flow model parameter correction module asks necessary traffic data to data unification processing module, and with 1
Hour is space correction model parameter.Traffic flow model, for interval, calculated the traffic behavior in each sub- section of highway with 30 seconds
Estimate;
4) improved Kalman filtering optimization module asks real time traffic data to data processing module, according to real time traffic data
Further optimize the result of calculation of traffic flow model.And using gained optimum results as the moment final output result;
5) optimal estimation result is sent to online traffic flow model parameter correction module, the latter by improved Kalman's optimization module
Model computational threads using optimal estimation result as input data, and based on newest model parameter value, solve subsequent time
Model result.
2. a kind of traffic status of express way method of estimation based on parallel computation as claimed in claim 1, it is characterised in that
The data unification processing module is used for existing highway fixed detector data are unified for only comprising section density
And average speed data, time interval is the traffic data of 30 seconds, realizes that step includes:
1) according to the demand for control and CFL condition of traffic active management system of automotive, it is by the express highway section of control object is discrete
The sub- section at 500 meters of intervals, data renewal time is 30 seconds;
2) for traffic flow data, row interpolation is entered using the interpolation method based on Poisson distribution;
3) for speed data, average link speed is converted to using following formula spot spe Js first:
<mrow>
<msubsup>
<mi>v</mi>
<mi>i</mi>
<mrow>
<mi>a</mi>
<mi>v</mi>
<mi>g</mi>
</mrow>
</msubsup>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<msubsup>
<mi>v</mi>
<mi>i</mi>
<mrow>
<mi>p</mi>
<mi>o</mi>
<mi>s</mi>
</mrow>
</msubsup>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mfrac>
<mrow>
<msub>
<mi>&sigma;</mi>
<mi>i</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<msubsup>
<mi>v</mi>
<mi>i</mi>
<mrow>
<mi>p</mi>
<mi>o</mi>
<mi>s</mi>
</mrow>
</msubsup>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
</mrow>
In formula,For express highway section i fixed detector in moment k place velocity measurement, σi(k) it is ground
The sample variance of point speed,It is section i average speed;
Secondly, row interpolation is entered using the Road average-speed data at 1 minute or 5 minutes interval of cubic Hamiltonian symmetrical systems function pair,
Obtain the link average speed estimation value of 30 seconds time intervals
4) according to the relation between occupation rate data and density data, the estimate of density is solved;Calculation procedure includes:
4.1) the density observation in section is calculated using following speed-density formula:
<mrow>
<msubsup>
<mi>&rho;</mi>
<mi>i</mi>
<mi>k</mi>
</msubsup>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mi>T</mi>
</mfrac>
<mfrac>
<msubsup>
<mi>q</mi>
<mi>i</mi>
<mi>k</mi>
</msubsup>
<msubsup>
<mi>v</mi>
<mi>i</mi>
<mi>k</mi>
</msubsup>
</mfrac>
</mrow>
In formula,By the section i that solves in 2) moment k flow value,By the road-section average density solved in 3).T
For time interval length;
4.2) occupation rate is solved according to below equationWith observation densityBetween linear relationship a:
<mrow>
<msup>
<mi>a</mi>
<mo>*</mo>
</msup>
<mo>=</mo>
<msubsup>
<mi>min&Sigma;</mi>
<mrow>
<mi>k</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>N</mi>
</msubsup>
<msup>
<mrow>
<mo>(</mo>
<msubsup>
<mi>&rho;</mi>
<mi>i</mi>
<mi>k</mi>
</msubsup>
<mo>-</mo>
<mi>a</mi>
<mo>&CenterDot;</mo>
<msubsup>
<mi>o</mi>
<mi>i</mi>
<mi>k</mi>
</msubsup>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
In formula, a*To make occupation rateAnd densityThe scalefactor value of least square relation is met,By 4-1) in solve
Section i moment k density observation,It is occupation rate data of the section i in moment k, N is total time space-number.Density
Estimate
4.3) willWithExported as final result.
3. a kind of traffic status of express way method of estimation based on parallel computation as claimed in claim 1, it is characterised in that
The optimized algorithm of the online traffic flow model parameter correction module, realizes that step includes:
1) set section i moment k model calculation value asTraffic data value isp
Model parameter vector, according to below equation calculating parameter be p when both distance:
<mrow>
<mi>d</mi>
<mi>i</mi>
<mi>s</mi>
<mi>t</mi>
<mrow>
<mo>(</mo>
<msubsup>
<mi>y</mi>
<mi>i</mi>
<mi>k</mi>
</msubsup>
<mo>,</mo>
<msubsup>
<mover>
<mi>y</mi>
<mo>^</mo>
</mover>
<mi>i</mi>
<mi>k</mi>
</msubsup>
<mo>(</mo>
<mi>p</mi>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mo>=</mo>
<msqrt>
<mrow>
<msup>
<mrow>
<mo>(</mo>
<msubsup>
<mi>&rho;</mi>
<mi>i</mi>
<mi>k</mi>
</msubsup>
<mo>-</mo>
<msubsup>
<mover>
<mi>&rho;</mi>
<mo>^</mo>
</mover>
<mi>i</mi>
<mi>k</mi>
</msubsup>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>+</mo>
<msup>
<mrow>
<mo>(</mo>
<msubsup>
<mi>v</mi>
<mi>i</mi>
<mi>k</mi>
</msubsup>
<mo>-</mo>
<msubsup>
<mover>
<mi>v</mi>
<mo>^</mo>
</mover>
<mi>i</mi>
<mi>k</mi>
</msubsup>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
</msqrt>
</mrow>
2) optimization aim of genetic algorithm is determined, three sub-steps are decomposed into:
2.1) fitness function for determining genetic algorithm is:
<mrow>
<mi>f</mi>
<mrow>
<mo>(</mo>
<mi>p</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<msubsup>
<mi>&Sigma;</mi>
<mrow>
<mi>k</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>T</mi>
</msubsup>
<msubsup>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>N</mi>
</msubsup>
<mi>d</mi>
<mi>i</mi>
<mi>s</mi>
<mi>t</mi>
<mrow>
<mo>(</mo>
<msubsup>
<mi>y</mi>
<mi>i</mi>
<mi>k</mi>
</msubsup>
<mo>,</mo>
<msubsup>
<mover>
<mi>y</mi>
<mo>^</mo>
</mover>
<mi>i</mi>
<mi>k</mi>
</msubsup>
<mo>(</mo>
<mi>p</mi>
<mo>)</mo>
<mo>)</mo>
</mrow>
</mrow>
In formula, T is the time interval number in 1 hour, and N is the sub- section number of highway after dividing,It is
Euclidean distance between the model calculation and traffic data in each section;
2.2) determined to solve optimized parameter p according to below equation*:
<mrow>
<mo>|</mo>
<mfrac>
<mrow>
<msub>
<mi>f</mi>
<mi>k</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>p</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<msub>
<mi>f</mi>
<mrow>
<mi>k</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>p</mi>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<msub>
<mi>f</mi>
<mrow>
<mi>k</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>p</mi>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
<mo>|</mo>
<mo><</mo>
<mi>&epsiv;</mi>
</mrow>
In formula, fkAnd f (p)k-1(p) it is fitness function f (p) kth time and -1 iteration of kth.ε stops for the iteration of genetic algorithm
Only condition;
2.3) stop condition judges:According to following public when genetic algorithm meets stop condition ε or iterations reaches that maximum changes
When counting from generation to generation, algorithm terminates;Corresponding model parameter value p is returned to, p now is the optimal solution p of model parameter*。
4. a kind of traffic status of express way method of estimation based on parallel computation as claimed in claim 1, it is characterised in that
The optimized algorithm of the improved Kalman filtering optimization module, realizes that step includes:
1) initialize:Obtain the initial model estimated result of each section i on highwayAnd traffic data
2) mean μ of each section in moment k model estimated result on highway is calculated1, k, and real time traffic data is equal
Value μ2, k;
3) according to step 2) in mean μ1, k、μ2, k, calculate highway and take up an official post model estimated result of the section in moment k
Variances sigma1, k, and traffic data variances sigma2, k;
4) according to step 3) in σ1, kAnd σ2, kSolve kalman gain Kk;Formula is as follows:
<mrow>
<msub>
<mi>K</mi>
<mi>k</mi>
</msub>
<mo>=</mo>
<mfrac>
<msubsup>
<mi>&sigma;</mi>
<mrow>
<mn>1</mn>
<mo>,</mo>
<mi>k</mi>
</mrow>
<mn>2</mn>
</msubsup>
<mrow>
<msubsup>
<mi>&sigma;</mi>
<mrow>
<mn>1</mn>
<mo>,</mo>
<mi>k</mi>
</mrow>
<mn>2</mn>
</msubsup>
<mo>+</mo>
<msubsup>
<mi>&sigma;</mi>
<mrow>
<mn>2</mn>
<mo>,</mo>
<mi>k</mi>
</mrow>
<mn>2</mn>
</msubsup>
</mrow>
</mfrac>
</mrow>
5) according to kalman gain more new model estimated result, formula is as follows:
<mrow>
<msubsup>
<mover>
<mi>y</mi>
<mo>^</mo>
</mover>
<mi>i</mi>
<mi>k</mi>
</msubsup>
<mo>=</mo>
<msubsup>
<mover>
<mi>y</mi>
<mo>^</mo>
</mover>
<mi>i</mi>
<mrow>
<mi>k</mi>
<mo>|</mo>
<mi>k</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</msubsup>
<mo>+</mo>
<msub>
<mi>K</mi>
<mi>k</mi>
</msub>
<mrow>
<mo>(</mo>
<msubsup>
<mi>y</mi>
<mi>i</mi>
<mi>k</mi>
</msubsup>
<mo>-</mo>
<mi>H</mi>
<msubsup>
<mover>
<mi>y</mi>
<mo>^</mo>
</mover>
<mi>i</mi>
<mrow>
<mi>k</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</msubsup>
<mo>)</mo>
</mrow>
</mrow>
In formula,For moment k optimal estimation result,For moment k the model calculation,For moment k friendship
Survey data are taken an overall view of, k is the optimal estimation result at k-1 moment, and H is observing matrix,
6) the more covariance of new model estimated result and real time traffic data, formula is as follows:
<mrow>
<msubsup>
<mi>&sigma;</mi>
<mrow>
<mi>f</mi>
<mi>u</mi>
<mi>s</mi>
<mi>e</mi>
<mi>d</mi>
<mo>,</mo>
<mi>k</mi>
</mrow>
<mn>2</mn>
</msubsup>
<mo>=</mo>
<mrow>
<mo>(</mo>
<mi>I</mi>
<mo>-</mo>
<msub>
<mi>K</mi>
<mi>k</mi>
</msub>
<mo>)</mo>
</mrow>
<msubsup>
<mi>&sigma;</mi>
<mrow>
<mn>1</mn>
<mo>,</mo>
<mi>k</mi>
</mrow>
<mn>2</mn>
</msubsup>
</mrow>
7) by the model estimated result of renewalExported as moment k final estimated result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710439953.1A CN107123265B (en) | 2017-06-12 | 2017-06-12 | Highway traffic state estimation method based on parallel computation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710439953.1A CN107123265B (en) | 2017-06-12 | 2017-06-12 | Highway traffic state estimation method based on parallel computation |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107123265A true CN107123265A (en) | 2017-09-01 |
CN107123265B CN107123265B (en) | 2020-01-10 |
Family
ID=59729773
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710439953.1A Active CN107123265B (en) | 2017-06-12 | 2017-06-12 | Highway traffic state estimation method based on parallel computation |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107123265B (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108257385A (en) * | 2018-03-19 | 2018-07-06 | 北京工业大学 | A kind of discriminating method of the anomalous event based on public transport |
CN110119406A (en) * | 2019-04-09 | 2019-08-13 | 口碑(上海)信息技术有限公司 | The checking method and device of real-time task record |
CN110956808A (en) * | 2019-12-23 | 2020-04-03 | 北京交通大学 | Heavy truck traffic flow prediction method based on non-full-sample positioning data |
CN110969857A (en) * | 2019-12-27 | 2020-04-07 | 华为技术有限公司 | Traffic information processing method and device |
CN111613052A (en) * | 2019-02-25 | 2020-09-01 | 北京嘀嘀无限科技发展有限公司 | Traffic condition determining method and device, electronic equipment and storage medium |
CN113362600A (en) * | 2021-06-11 | 2021-09-07 | 重庆大学 | Traffic state estimation method and system |
CN113450572A (en) * | 2021-09-02 | 2021-09-28 | 西南交通大学 | Traffic state estimation method, device and equipment and readable storage medium |
CN115311844A (en) * | 2022-06-22 | 2022-11-08 | 东南大学 | Highway traffic state estimation method based on self-supervision learning support vector machine |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102629418A (en) * | 2012-04-09 | 2012-08-08 | 浙江工业大学 | Fuzzy kalman filtering-based traffic flow parameter prediction method |
CN106251630A (en) * | 2016-10-13 | 2016-12-21 | 东南大学 | A kind of laddering EKF traffic status of express way method of estimation based on multi-source data |
CN106652493A (en) * | 2016-12-29 | 2017-05-10 | 东南大学 | Intersection signal optical control method in environment of Internet of vehicles |
-
2017
- 2017-06-12 CN CN201710439953.1A patent/CN107123265B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102629418A (en) * | 2012-04-09 | 2012-08-08 | 浙江工业大学 | Fuzzy kalman filtering-based traffic flow parameter prediction method |
CN106251630A (en) * | 2016-10-13 | 2016-12-21 | 东南大学 | A kind of laddering EKF traffic status of express way method of estimation based on multi-source data |
CN106652493A (en) * | 2016-12-29 | 2017-05-10 | 东南大学 | Intersection signal optical control method in environment of Internet of vehicles |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108257385A (en) * | 2018-03-19 | 2018-07-06 | 北京工业大学 | A kind of discriminating method of the anomalous event based on public transport |
CN111613052A (en) * | 2019-02-25 | 2020-09-01 | 北京嘀嘀无限科技发展有限公司 | Traffic condition determining method and device, electronic equipment and storage medium |
CN111613052B (en) * | 2019-02-25 | 2022-03-04 | 北京嘀嘀无限科技发展有限公司 | Traffic condition determining method and device, electronic equipment and storage medium |
CN110119406A (en) * | 2019-04-09 | 2019-08-13 | 口碑(上海)信息技术有限公司 | The checking method and device of real-time task record |
CN110956808A (en) * | 2019-12-23 | 2020-04-03 | 北京交通大学 | Heavy truck traffic flow prediction method based on non-full-sample positioning data |
CN110956808B (en) * | 2019-12-23 | 2020-11-24 | 北京交通大学 | Heavy truck traffic flow prediction method based on non-full-sample positioning data |
CN110969857A (en) * | 2019-12-27 | 2020-04-07 | 华为技术有限公司 | Traffic information processing method and device |
CN113362600A (en) * | 2021-06-11 | 2021-09-07 | 重庆大学 | Traffic state estimation method and system |
CN113450572A (en) * | 2021-09-02 | 2021-09-28 | 西南交通大学 | Traffic state estimation method, device and equipment and readable storage medium |
CN115311844A (en) * | 2022-06-22 | 2022-11-08 | 东南大学 | Highway traffic state estimation method based on self-supervision learning support vector machine |
CN115311844B (en) * | 2022-06-22 | 2023-05-16 | 东南大学 | Expressway traffic state estimation method based on self-supervision learning support vector machine |
Also Published As
Publication number | Publication date |
---|---|
CN107123265B (en) | 2020-01-10 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107123265A (en) | A kind of traffic status of express way method of estimation based on parallel computation | |
CN102629418B (en) | Fuzzy kalman filtering-based traffic flow parameter prediction method | |
Kim et al. | Urban traffic flow prediction system using a multifactor pattern recognition model | |
CN108399748B (en) | Road travel time prediction method based on random forest and clustering algorithm | |
CN104809333B (en) | Capacity prediction methods and system based on Kalman filter | |
CN102568205B (en) | Traffic parameter short-time prediction method based on empirical mode decomposition and classification combination prediction in abnormal state | |
CN105741549B (en) | A kind of traffic flow forecasting method theoretical based on space copula | |
EA013657B1 (en) | Method, system and apparatus for real-time reservoir model upgrading using ensemble kalman filter | |
CN107909227A (en) | Ultra-short term predicts the method, apparatus and wind power generating set of wind power | |
CN101480143A (en) | Method for predicating single yield of crops in irrigated area | |
CN113094860B (en) | Industrial control network flow modeling method based on attention mechanism | |
CN103177289B (en) | Modeling method for noise-uncertainty complicated nonlinear dynamic system | |
CN105303835B (en) | A kind of Forecasting Approach for Short-term of road traffic stream mode | |
CN105893669A (en) | Global simulation performance predication method based on data digging | |
CN107292410A (en) | tunnel deformation prediction method and device | |
Natali et al. | Learning time-varying graphs from online data | |
CN110991776A (en) | Method and system for realizing water level prediction based on GRU network | |
CN110059392A (en) | A kind of landslide deformation prediction method | |
CN109377761A (en) | Traffic factor network establishing method based on Markov-chain model | |
CN103279030B (en) | Dynamic soft measuring modeling method and device based on Bayesian frame | |
Tang et al. | On missing traffic data imputation based on fuzzy C-means method by considering spatial–temporal correlation | |
CN115240418B (en) | Short-time traffic flow prediction method based on causal gate control-low-pass graph convolution network | |
CN115438453B (en) | Method for constructing road network facility topological structure by using observation data | |
Rivero et al. | Short-term rainfall time series prediction with incomplete data | |
CN115616333A (en) | Power distribution network line loss prediction method and system |
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 |