CN101739814A - SCATS coil data-based traffic state online quantitative evaluation and prediction method - Google Patents

Abstract

The invention discloses an SCATS coil data-based traffic state online quantitative evaluation and prediction method, which relates to the traffic state online quantitative evaluation and prediction method in the technical field of traffic. The SCATS coil data-based traffic state online quantitative evaluation and prediction method comprises the following steps: acquiring traffic parameters and uploading the traffic parameters to a traffic information centre by an SCATS coil; performing quality evaluation and control on actually-measured data by a computer; performing traffic parameter virtual time sequence construction on the actually-measured data subjected to the quality evaluation and control; performing the quality evaluation and control on virtual time sequence data; performing traffic congestion automatic identification on the virtual time sequence data of the parameters; performing short multi-stage prediction on the virtual time sequence data of the traffic parameters; and on the basis of the short multi-stage prediction on the virtual time sequence time of the traffic parameters, performing traffic congestion spreading range prediction and duration prediction and issuing prediction results to a traffic administrative department to provide more forcible basis for decision making for the traffic administrative department.

Description

Online quantitative evaluation of traffic behavior and Forecasting Methodology based on SCATS coil data

Technical field

The present invention relates to online quantitative evaluation of traffic behavior and Forecasting Methodology in the intelligent transport technology field, be specifically related to a kind of online quantitative evaluation of traffic behavior and Forecasting Methodology based on SCATS coil data.

Background technology

SCATS (the self-adaptation traffic system is coordinated in Sydney) is a kind of intersection traffic signal control system of computerizeing control.From the angle of control theory and technology, the SCATS technology maturation, multiple functional, control mode is flexible, is a kind of more outstanding whistle control system, worldwide obtained widely using.

The dynamic traffic data that SCATS obtains only are used for whistle control system but at present.Angle from data sharing, if the dynamic traffic data that it obtained can be used for the online quantitative evaluation of traffic behavior of this institute of system covering path, then can when reducing the dynamic information cost, provide stronger decision-making foundation for formulating effective traffic management measure.Yet coil installation position among the SCATS and the traffic data that is provided thereof have very strong uniqueness, and online quantitative evaluation of existing traffic behavior and Forecasting Methodology are difficult to effectively play a role under this condition.

Compare with the dynamic traffic data that inductive coil among through street and the SCOOT obtains, the inductive coil among the SCATS only can provide the dynamic data of flow and average headway at present, and therefore, the available information amount when carrying out the online quantitative evaluation of traffic behavior is less than normal.And, in SCATS, inductive coil is laid in the position of downstream, highway section near the intersection parking line, because this dynamic traffic data can't embody the influence of different queuing vehicle numbers on the highway section, therefore when the highway section degree of crowding was carried out online quantitative evaluation, its effect will decrease.

SCATS is a kind of real-time adaptive control system, and the sampling interval of its dynamic traffic data is by the signal period decision of affiliated teleseme, and the signal period is the parameter that constantly changes.The time sampling interval condition that becomes under, the dynamic traffic data do not have comparability, not only can't carry out the online quantitative evaluation of traffic behavior according to these dynamic datas, nor can be with it as the estimation of the basis realization of setting up forecast model to the future transportation state.

Summary of the invention

The technical problem to be solved in the present invention is, uniqueness at above-mentioned SCATS inductive coil dynamic traffic data, overcome existing online quantitative evaluation of traffic behavior and Forecasting Methodology and be applied to deficiency among the SCATS, a kind of effective online quantitative evaluation of traffic behavior and Forecasting Methodology based on SCATS coil data are provided, when reducing the dynamic information cost, provide stronger decision-making foundation for formulating effective traffic management measure.

The technical solution adopted for the present invention to solve the technical problems is that the concrete grammar flow process is as follows:

(1) the SCATS coil is gathered traffic parameter automatically and is uploaded to traffic information center by the network connection;

(2) traffic information center carries out quality assessment and control by computing machine to SCATS coil measured data, and comprise traffic events is detected automatically, and to vehicle supervision department's issue testing result, for it provides stronger decision-making foundation;

(3) traffic information center by computing machine to quality assessment with control after SCATS coil measured data carry out traffic parameter virtual time sequence construct;

(4) traffic information center carries out quality assessment and control by computing machine to the virtual time sequence data of SCATS traffic parameter;

(5) traffic information center carries out traffic congestion identification automatically by computing machine to the virtual time sequence data of SCATS traffic parameter, and differentiates the result to vehicle supervision department's issue, for it provides stronger decision-making foundation;

(6) traffic information center carries out multi-step prediction in short-term by computing machine to the virtual time sequence data of SCATS traffic parameter;

(7) traffic information center carries out congested in traffic range of scatter and duration prediction on the basis of multi-step prediction by computing machine in short-term to SCATS traffic parameter virtual time sequence data, and predict the outcome to vehicle supervision department issue, for it provides stronger decision-making foundation.

Wherein,

(1), analyzes based on the discernible traffic behavior type of SCATS coil data according to traffic behavior definition and classification;

(2) design detects (AutomatedIncident Detection, AID) method automatically based on the traffic events of SCATS coil data;

(3) SCATS coil measured data being carried out quality assessment and control, mainly is identification and control to obliterated data and events affecting data, for subsequent module provides more reliable data basis;

(4) research SCATS coil dynamic traffic data characteristic, virtual sampling interval scheme is proposed to break through the bottleneck that the SCATS traffic parameter can't be predicted, and structure SCATS traffic parameter virtual time sequence, design delay factor computing method, analyze the regularity of virtual time sequence under the different time yardstick, to determine rational virtual sampling interval;

(5) the virtual time sequence data to the SCATS traffic parameter carries out quality assessment and control, for follow-up traffic forecast, crowded identification etc. provides more reliable basic data;

(6) design is based on congested in traffic identification automatically (Automated Congestion Identification, the ACI) method of SCATS traffic parameter time series data;

(7) multistep forecasting method in short-term of research SCATS traffic parameter virtual time sequence is for the traffic behavior prediction provides the data basis;

(8) design congested in traffic range of scatter and duration Forecasting Methodology formulated plan of travel foundation is provided for traffic administration person formulates Managed Solution and traveler.

Described analysis is meant based on the discernible traffic behavior type of SCATS coil data, definition and classification according to traffic behavior in the continuous traffic flow, in conjunction with SCATS inductive coil installation position and dynamic traffic data characteristics, draw and to discern smooth and easy state and retardance state based on SCATS coil data and can't discern the conclusion of congestion status.

Described design comprises two steps of design with the automatic detection algorithm of traffic events chosen of basic traffic parameter based on the automatic detection algorithm of traffic events of SCATS coil data, the basic traffic parameter of wherein choosing is this traffic parameter of flow, and the automatic detection algorithm of traffic events adopts the standard deviation method.

Described SCATS coil measured data quality assessment and control are meant that the near-sighted recognition methods of horizontal time series segmentation moving average is adopted in the identification of obliterated data, adopt horizontal time series segmentation moving average to repair to the control of obliterated data; The designed automatic detection algorithm of traffic events based on SCATS coil data is adopted in the identification of event data; although the measured data under the traffic events condition can depart from normal effective range usually; but this is a kind of real data; need not repair control to it; to the identification of obliterated data and events affecting data and the purpose of control be, for subsequent module provides more reliable data basis.

The construction method of described SCATS traffic parameter virtual time sequence comprises:

(1) SCATS traffic parameter actual measurement sampling interval is to the mapping relations analysis of virtual sampling interval;

(2) definite method of the virtual sampling interval of single crossing;

(3) definite method of the unified virtual sampling interval of road network.

For the mapping from the actual samples interval to virtual sampling interval of more real realization reduced discharge, introduce this notion of the average magnitude of traffic flow.Define the ratio that the average magnitude of traffic flow is reduced discharge and signal period, unit is :/second, reduced discharge from actual samples at interval to the mapping relations of virtual sampling interval is: virtual sampling interval by its cross over actual samples at interval institute at times with the corresponding average magnitude of traffic flow sum of products; Average headway from actual samples at interval to the mapping relations of virtual sampling interval is: virtual sampling interval by its cross over actual samples at interval institute at times with the corresponding average magnitude of traffic flow and the average headway sum of products, divided by virtual sampling interval by its cross over actual samples at interval institute at times with the corresponding average magnitude of traffic flow sum of products;

Definite method of the virtual sampling interval of single crossing: the angle that real-time requires is considered from the online quantitative evaluation of traffic behavior, choose delay as virtual sampling interval definiteness mark really, it is the time of each end cycle that definition of data extracts the some time, the decision point time is the time that each virtual sampling interval finishes, definition of data extracts point and postpones to be the mistiming of decision point time prior to the data extract point time that decision point postpones to be the mistiming of data extract point time prior to the decision point time.The data extract point that the delay of each decision point equals this decision point postpones to postpone sum with decision point, the delay sum of one day 24 hours all decision points of definition is a delay factor, each virtual sampling interval correspondence a delay factor, and the virtual sampling interval of delay factor minimum is exactly the value that the method is determined;

Definite method of the unified virtual sampling interval of road network is: by the delay factor algorithm, can obtain the virtual sampling interval of each crossing delay factor minimum, and the virtual sampling interval that different crossings obtains is different.And the virtual sampling interval of a unification of whole SCATS system's needs is carried out the information issue.The present invention proposes to adopt the system delay factor two factors of stability minimum and the virtual time sequence data to determine the system virtualization sampling interval.But people relatively accept traditionally as information issuing time of 5,10,15 minutes etc. 5 minutes integral multiples at interval, therefore, determine also to should be taken into account when road network is unified sampling interval the mental habit of traveler, the integral multiple of preferably choosing 5 minutes is as the sampling interval duration.

Described SCATS traffic parameter virtual time sequence data carries out quality assessment and control is meant, the traffic parameter time series data has secular trend, same place is little in the traffic parameter data variation of different operating day or nonworkdays synchronization, the traffic parameter data value fluctuates up and down at a certain numerical value, can adopt mathematical statistic method, formulate the one day 24 hours continuous intervals that change of different data values constantly, to estimate and the control misdata, for follow-up traffic forecast, crowded identification etc. provides more reliable basic data.

Described design is meant based on the congested in traffic automatic identification algorithm of SCATS traffic parameter time series data, according to above-mentioned draw based on SCATS coil data can discern smooth and easy state and the retardance state conclusion, by analyzing flow and the variation tendency of average headway under smooth and easy and retardance state, adopt flow, average headway combination as the congested in traffic foundation of discerning, propose according to the track on the different signal phase combination highway sections, and provide corresponding combination track traffic behavior.

The multistep forecasting method in short-term of described research SCATS traffic parameter virtual time sequence is meant, specificity analysis based on traffic parameter virtual time sequence, carry out traffic parameter secular trend multi-step prediction, traffic parameter intermediate trend multi-step prediction, the prediction of the traffic parameter short-term trend of the times, for the traffic behavior prediction provides the data basis.

Congested in traffic range of scatter of described design and duration method of estimation comprise the crowded space of single road section traffic volume range of scatter Study on estimation method, the crowded duration Study on estimation method of single road section traffic volume and the crowded range of scatter Study on estimation method of regional road grid traffic, formulate plan of travel foundation is provided for traffic administration person formulates Managed Solution and traveler.

Remarkable result of the present invention is: the online quantitative evaluation of traffic behavior based on SCATS coil data provided by the invention and Forecasting Methodology have overcome the difficult problem that the online quantitative evaluation of existing traffic behavior and Forecasting Methodology can't combine with SCATS, when reducing the dynamic information cost, can significantly improve the coordination degree of traffic signals control, transport information guiding and point duty, provide stronger decision support for improving the congested in traffic effect of dredging to greatest extent.

Description of drawings

Fig. 1 is based on the online quantitative evaluation of traffic behavior and prediction process flow diagram of SCATS coil data;

Fig. 2 is flow-velocity diagram;

Fig. 3 is a SCATS covering path synoptic diagram;

Fig. 4 is the crowded differentiation logical flow chart of standard deviation method;

Fig. 5 is a SCATS covering path traffic parameter virtual time sequence construct process;

Fig. 6 is τ＜C _MinThe time delay factor concern analysis diagram;

Fig. 7 is τ＞C _MaxThe time delay factor concern analysis diagram;

Fig. 8 is a traffic parameter data secular trend prediction process flow diagram;

Fig. 9 is a traffic parameter data intermediate trend prediction process flow diagram;

Figure 10 is a traffic parameter data short-term trend of the times prediction process flow diagram;

Figure 11 is the single highway section crowded range of scatter estimation of the property sent out an often process flow diagram;

Figure 12 is that often the crowded duration of the property sent out is estimated process flow diagram in single highway section.

Embodiment

Below each module among Fig. 1 is elaborated:

1) based on the discernible traffic behavior type analysis of SCATS coil data

According to the characteristic of continuous traffic flow, the relation of flow and speed is a para-curve as can be known, and speed reduces with the increase of flow, till the flow that reaches the maximum traffic capacity, and the part of on curve, crowding, flow and speed all reduce.So the maximum point of flow, just crowded and non-crowded tr pt C point.On crowded curve, individual point is arranged, be the representative point under the congested conditions, be referred to as the D point, as shown in Figure 2.

In continuous stream, C is done the critical point of point for smooth traffic and retardance, and the critical point of D point as retardance and obstruction.Wherein, AC represents normal traffic state (smooth and easy); CD represents the slight congestion state (retardance) of slow traffic flow; DE represents serious congestion state (blocking up).

And the SCATS system is a real-time adaptive control system, SCATS covering path detecting device is laid in the intersection entrance place, the sampling interval of its inductive coil is by the signal period decision of teleseme under it, and the traffic flow of SCATS covering path has the characteristic of being interrupted traffic flow.

Fig. 3 is a SCATS covering path synoptic diagram, P ₁And P ₂The expression crossing, when wagon flow at P ₁-P ₂When travelling from left to right on the highway section, crossing P ₂Detecting device before the stop line can detect the vehicle that passes through.When not stopping queuing before the stop line, vehicle can pass through stop line smoothly in the green time, and traffic this moment is in smooth and easy state; When the queuing of parking was arranged before the stop line, vehicle began to start by stop line in the green time, and traffic this moment is in comparatively congestion state, but can not determine that retardance still is a blocked state.Though because detecting device can detect the vehicle number that passes through, it can not detect the queue length of vehicle on the highway section.Even the queuing vehicle is a lot of on the highway section, the vehicle that comes the front in the green time still may be that the upper limit is passed through stop line with the traffic capacity of crossing.

By above analysis as can be known, can not identify the traffic behavior below the C point among Fig. 2 based on SCATS covering path detector data.The present invention's definition is retardance and critical point smoothly with the corresponding B point of D point.Wherein AB represents smooth and easy traffic behavior; BC represents the retardance state of slow traffic flow.

2) based on the automatic detection algorithm of traffic events of SCATS coil data

The structure of traffic parameter virtual time sequence, make that the pairing traffic behavior of traffic parameter of adjacent time period is average, short or some slight traffic events for some time of origins, use traffic parameter virtual time sequence data to discern to come out, and make up traffic parameter virtual time sequence and can produce the regular hour delay, the real-time of event detection is not strong.Therefore, the present invention is that the basis is carried out traffic events and detected automatically with traffic parameter elapsed time sequence data.

(1) basic traffic parameter chooses

In the SCATS system, average headway is the ratio of phase place duration and flow.When traffic events took place highway section, detecting device upstream, the flow by detecting device can reduce.But SCATS is the real-time adaptive induction control system, and after flow reduced, the respective phase duration also can be less, and the increase and decrease situation of time headway is difficult to judge.And after the morning peak, along with the variation of flow, the average headway amplitude of variation is little, and therefore, the present invention only adopts this traffic parameter design event data recognition methods of flow.

(2) design of the automatic detection algorithm of traffic events

The mutual cross connection of SCATS covering path is network configuration, if adopt the double sections Algorithm for Traffic Incidents Detection, be difficult to determine that the cross section, upstream in highway section takes place concrete incident, and consider and have only this traffic parameter of flow to use, therefore, the present invention adopts the standard deviation method that event data is estimated, the arithmetic mean of the traffic parameter data (refering in particular to flow among the present invention) in n sampling period is as the predicted value of traffic parameter at moment t before this algorithm utilization moment t, measure the change degree of traffic parameter data with the standard normal deviation again with respect to its former mean value, when it surpasses pre-set threshold, then think sporadic traffic congestion has taken place.

If the actual value of t traffic parameter is x (t) constantly, the traffic parameter actual value in n sampling period is x (t-n) before the t constantly, x (t-n+1) ..., x (t-1), then discrimination formula is

Wherein

In the formula,

Current predicted value for traffic parameter; S is the standard deviation of preceding n sampling period traffic parameter; K is a decision-making value; SND is a normal deviate.

Fig. 4 has provided the flow process that the standard deviation method is crowded and differentiated.The traffic parameter of being imported can be the magnitude of traffic flow, occupation rate and speed etc.In this algorithm, the time window width n of moving average calculation has a significant impact the crowded effect tool of differentiating.The n value is too big, makes that interior traffic behavior of time window width n period is comparatively average, might cause the omission of traffic events; The n value is too little, and the normal fluctuation of traffic parameter also may be judged by accident to traffic events takes place.Therefore, the n value can not too greatly can not be too little, generally gets 5-10.

In order to improve the precision of event detection, can detect incident generation continuation and detect.The concrete grammar that continuation detects is as follows.

When SND surpasses institute for the first time and formulates threshold k, should moment data markers be " traffic event data may take place "; When SND surpasses institute for the second time and formulates threshold k, should moment data markers be " traffic event data may take place " equally; When SND surpasses institute for the third time and formulates threshold k, should moment data markers be " generation traffic event data ", otherwise data not carried out any mark.

3) quality assessment of SCATS coil measured data and control method

In SCATS, though the sampling period difference of actual traffic supplemental characteristic, because SCATS takes continuous small step to adjust the signal period apart from mode, a new cycle was compared with last one-period, its length variations is limited in ± 6s in, the duration difference of adjacent periods is little.Therefore, in order to improve the effect of SCATS dynamic traffic data processing, in the first step quality of data of the present invention evaluation and control, ignore this difference, it be considered as horizontal time Series Processing, the evaluation of finishing missing data and control and with the functions such as detection of traffic events.

(1) identification of obliterated data and control method

Traffic parameter virtual time sequence is to make up under the prerequisite that the traffic parameter time series data does not have to lack, and therefore, the data basis of the evaluation of obliterated data and control is a traffic parameter elapsed time sequence data.In SCATS, having only flow is the traffic parameter of inductive coil actual measurement.And average headway is the phase place duration of SCATS system file output and the ratio of flow, if data on flows is lost or be 0, then the average headway output valve is the phase place duration.Therefore, the evaluation of obliterated data and control method research are only at this traffic parameter of flow.

The SCATS system file is worth for exporting 0 the processing mode of flow loss of data, can not illustrate it must is loss of data but flow is 0 value, also might be normal data.In order to distinguish the data qualification of flow 0 value, the present invention proposes the approximate recognition methods of following horizontal time series segmentation moving average.

According to the moving law of traffic flow as can be known, in the road that SCATS covers, peak period, flow was that the possibility of 0 value is less, and morning low ebb flow be that the possibility of 0 value is bigger.Laterally time series segmentation moving average recognition methods just be based on traffic flow this moving law proposed, its basic thought is, if SCATS system file delivery rate is 0 value, then calculate the average θ and the mean square deviation δ of the preceding n bar data of these data, calculate (θ+2 δ) and (θ-2 δ), if 0 value can be similar in [θ-2 δ, θ+2 δ] is interval and be considered as normal data, otherwise is obliterated data.Because the traffic parameter data can not be negative value, thus when (θ-2 δ) less than 0 the time, then make (θ-2 δ) to equal 0.The computing formula of average and variance is as follows:

${\mathrm{\θ}}_i=\frac{1}{n}\underset{i-1}{\overset{i-n}{\mathrm{\Σ}}}{q}_i^j---\left(2\right)$

${\mathrm{\δ}}_i=\sqrt{\frac{\underset{i-1}{\overset{i-n}{\mathrm{\Σ}}}{\left(q_i^j\right)}^2-{{\mathrm{n\θ}}_i}^2}{n-1}}---\left(3\right)$

In the formula, θ _iAverage for i n bar data before the time interval; δ _iMean square deviation for i n bar data before the time interval; q _i ^jFlow for i time interval j measuring station; N is the number of the selected data of moving average.

Because the variation of same place adjacent moment traffic parameter is little, so the present invention adopts the mean value θ of segment data when adjacent _iObliterated data is repaired θ _iCalculating as the formula (2).

The horizontal time series segmentation method of foundation is lost after identification and the control flow, should calculate again for the pairing average headway of traffic loss data.The computing method of average headway are the ratio of the flow after phase place duration and the reparation.

(2) identification of event data and control

The utilization designed traffic events of the present invention automatic detection algorithm can identify the actual measurement traffic data of the traffic events influence that is subjected to happening suddenly, because these data have embodied the true running status of traffic flow, therefore do not need it is repaired control, its subsequent processes is identical with the appropriate section of through street system.

4) construction method of SCATS traffic parameter virtual time sequence

In order to access the fixing SCATS covering path traffic parameter time series of sampling interval, the present invention has made up a virtual sampling interval that time span is τ, and this virtual sampling interval τ is inserted time shaft according to method of interpolation.So just form situation as shown in Figure 4, the sampling interval of inductive coil reality is still the signal period of its affiliated teleseme, but the traffic parameter of each virtual sampling interval can obtain by each traffic parameter that collects in the actual samples interval is calculated, just by certain computation rule with the actual samples data map on virtual sampling interval, so just formed the fixed sample traffic parameter virtual time sequence of traffic flow continuously at interval.

That shown in Figure 5 is τ＜c _MinThe time situation, it is that the relation of virtual sampling interval and actual samples compartment can be slightly different that τ gets different value, but can not surpass two kinds of situations shown in Fig. 5 on its intension.Wherein first kind of situation is simpler relatively, and whole virtual sampling interval is positioned at an actual samples at interval; Second kind of situation is complicated a little, and virtual sampling interval is crossed over two actual samples at interval, and the ratio in shared each cycle constantly changes with actual conditions.Each effective traffic parameter mapping relations from the actual samples interval to virtual sampling interval in both cases are discussed respectively below.

(1) construction method of magnitude of traffic flow virtual time sequence

For of the mapping of more real realization reduced discharge, introduce the average magnitude of traffic flow from the actual samples interval to virtual sampling interval

This notion.The average magnitude of traffic flow of definition is the ratio of reduced discharge and signal period in this project, and unit is :/second.Formula is as follows:

${\stackrel{\‾}{q}}_i=\frac{q_i^s}{C_i}---\left(4\right)$

In the formula:

---actual samples is the average magnitude of traffic flow of i (cycle i) at interval;

q _i ^s---actual samples is the reduced discharge of i (cycle i) at interval;

C _i---actual samples is the cycle duration of i (cycle i) at interval.

Reduced discharge from actual samples to the mapping relations of virtual sampling interval is at interval:

$q_k^{\mathrm{\τ}}=\underset{n=i}{\overset{i+N}{\mathrm{\Σ}}}{\stackrel{\‾}{q}}_i\×t_i---\left(5\right)$

In the formula:

q _k ^τ---the flow of virtual sampling interval k;

---actual samples is the average magnitude of traffic flow of i (cycle i) at interval;

τ---virtual sampling interval;

t _i---virtual sampling interval is positioned at the cycle duration of signal period i;

N---virtual sampling interval is crossed over actual samples number at interval.

(2) construction method of average headway virtual time sequence

Average headway from actual samples to the mapping relations of virtual sampling interval is at interval:

$h_k^{\mathrm{\τ}}=\frac{\underset{n=i}{\overset{i+N}{\mathrm{\Σ}}}{h}_i\×t_i\×{\stackrel{\‾}{q}}_i}{\underset{n=i}{\overset{i+N}{\mathrm{\Σ}}}{t}_i\×{\stackrel{\‾}{q}}_i}---\left(6\right)$

In the formula:

h _k ^τ---the average headway of virtual sampling interval k;

h _i---actual samples is the average headway of i (cycle i) at interval.

The notion by virtual sampling interval and the construction method of traffic parameter virtual time sequence just can obtain the SCATS covering path traffic parameter time series data under the virtual sampling interval.And different virtual sampling interval values can construct different SCATS covering path traffic parameter virtual time sequences, and then can have influence on the result of traffic state judging and traffic parameter short-term prediction.

When virtual sampling interval value hour, traffic state judging and traffic parameter short-term prediction have good real-time performance.But the value of virtual sampling interval is too small, and the time series of traffic flow parameter data can have very big random fluctuation, and the stability of traffic state judging and traffic parameter short-term prediction also can be relatively poor.

When virtual sampling interval value was big, the traffic flow parameter data were comparatively level and smooth, and its time sequence and traffic state judging result have stability preferably.But the value of virtual sampling interval is excessive, and unusual traffic behavior might on average be become the normal traffic state, and its discriminant accuracy and real-time can reduce.

Therefore, need to determine a rational virtual sampling interval duration.

The definition of data extraction point time is the time of each end cycle among the present invention, the decision point time is the time that each virtual sampling interval finishes, definition of data extracts point and postpones to be the mistiming of decision point time prior to the data extract point time that decision point postpones to be the mistiming of data extract point time prior to the decision point time.Like this, the delay of each decision point just equals the delay of data extract point and decision point delay sum of this decision point.As Figure 6 and Figure 7.

Fig. 6 is the one dimension time diagram.Shown in the figure τ＜C _MinSituation.Then the delay of each decision point is calculated as:

D ₁＝D _1，1＝t ₁-p ₁ (7)

D ₂＝D′ _2，2+D _2，2＝p ₂-t ₁+t ₂-p ₂ (8)

D ₃＝D _3，2＝t ₂-p ₃ (9)

D ₄＝D′ _4，2+D _4，3＝p ₄-t ₂+t ₃-p ₄ (10)

Fig. 7 is τ＞C _MaxSituation.Then the delay of each decision point is calculated as:

D ₁＝D′ _1，1+D′ _1，2+D _1，3＝p ₁-t ₁+p ₁-t ₂+t ₃-p ₁ (11)

D ₂＝D′ _2，3+D′ _2，4+D _2.5＝p ₂-t ₃+p ₂-t ₄+t ₅-p ₂ (12)

Virtual sampling interval τ or greater than the signal period, or, promptly put p smaller or equal to the signal period _iAt t _iBefore, or (may cross over several signal periods) thereafter or overlap.But no matter the delay factor of which kind of relation calculates and is included in the computing method of above-mentioned two figure.With all decision point p _iDelay calculate summation and just obtain delay factor under this virtual sampling interval τ.

If virtual sampling interval τ is very big, may comprise several signal periods, its decision point postpones will be very big, if opposite virtual sampling interval τ is very little, may a signal period comprise several virtual sampling interval, and its data extract point postpones will be very big.Therefore, decision point delay and data extract point postpone sum of the two for falling the trend that afterwards rises earlier.Like this, each virtual sampling interval τ correspondence a delay factor, and the virtual sampling interval of delay factor minimum is exactly the τ value that the method is determined.

By the delay factor algorithm, can obtain the virtual sampling interval of each crossing delay factor minimum, and the virtual sampling interval that different crossings obtains is different.And the virtual sampling interval of a unification of whole SCATS system's needs is carried out the information issue.The present invention proposes to adopt the system delay factor two factors of stability minimum and the virtual time sequence data to determine the system virtualization sampling interval.

(1) system delay factor minimum

Adopt definite thought of the virtual sampling interval in crossing,, can obtain the virtual sampling interval of the delay factor minimum of total system the delay factor summation of each crossing under each virtual sampling interval correspondence.

(2) stability of virtual time sequence data

Traffic state judging and traffic parameter short-term prediction need have the traffic data of good stability, and the data in same place can compare with historical data, and the data morning of different location, evening peak value also can compare.And the integral multiple in the virtual sampling interval number of winning the confidence cycle and virtual sampling interval long enough can obtain stability virtual time sequence data preferably.

But people relatively accept traditionally as information issuing time of 5,10,15 minutes etc. 5 minutes integral multiples at interval, therefore, determine also to should be taken into account when road network is unified sampling interval the mental habit of traveler, the integral multiple of preferably choosing 5 minutes is as the sampling interval duration.

5) traffic parameter virtual time sequence quality is estimated and control method

Quality assessment of traffic parameter virtual time sequence data and control mainly are identification and the control to misdata.The present invention is divided into misdata evaluation and control method based on horizontal time series with based on two kinds of methods of vertical time series.

(1) based on vertical seasonal effect in time series method

At the design of virtual traffic parameter time series data, the accumulation of virtual traffic parameter time series data can form historical data based on vertical seasonal effect in time series method.The traffic parameter time series data has the theoretical foundation that secular trend are designs of vertical time series recognition methods.When finding by a large amount of actual measurement traffic data statistics, same place is little in the traffic parameter data variation of different operating day or nonworkdays synchronization, the traffic parameter data value fluctuates up and down at a certain numerical value, can adopt mathematical statistic method, formulate the one day 24 hours continuous intervals that change of different data values constantly, to estimate and the control misdata.Concrete grammar is shown in formula (13), (14), (15):

[θ-3δ，θ+3δ] (13)

$\mathrm{\θ}=\frac{1}{n}\underset{i-1}{\overset{i-n}{\mathrm{\Σ}}}{q}_i^j\left(t\right)---\left(14\right)$

$\mathrm{\δ}=\sqrt{\frac{\underset{i-1}{\overset{i-n}{\mathrm{\Σ}}}{q}_i^j\left(t\right)-n{\mathrm{\θ}}^2}{n-1}}---\left(15\right)$

In the formula, θ is the average of n day data before vertical time series; δ is the mean square deviation of preceding n day data; q _i ^j(t) be i days t of j measuring station flows constantly; N is the selected fate of moving average, generally chooses n=3-8.

If the traffic parameter data are surveyed constantly in the interval range that formula (13) calculates in i days, t, then the detected data of this moment detecting device are normal data, otherwise are misdata.

Control method to misdata is the average of output historical data, suc as formula (14).

(2) based on horizontal seasonal effect in time series method

Design at traffic parameter actual measurement traffic parameter time series data based on horizontal seasonal effect in time series method.According to the moving law of traffic flow as can be known, traffic flow is the process of a gradual change, and the traffic parameter of adjacent moment changes and be little.Can adopt mathematical statistic method, formulate the one day 24 hours continuous valid intervals that change of different data values constantly, to estimate and the control misdata.Its method is shown in formula (16), (17), (18):

[θ-3δ，θ+3δ] (16)

$\mathrm{\θ}=\frac{1}{n}\underset{i-1}{\overset{i-n}{\mathrm{\Σ}}}{q}_i^j---\left(17\right)$

$\mathrm{\δ}=\sqrt{\frac{\underset{i-1}{\overset{i-n}{\mathrm{\Σ}}}{\left(q_i^j\right)}^2-n{\mathrm{\θ}}^2}{n-1}}---\left(18\right)$

In the formula, θ is the average of n bar data before the horizontal time series on the same day; δ is the mean square deviation of preceding n bar data; q _i ^jFlow for j measuring station i bar data; N is the number of the selected data of moving average, all exports identical average in order to prevent 0 continuous value, and the n value should be formulated as the case may be.

If i bar actual measurement traffic parameter data are in the interval range that formula (16) calculates, then the detected data of this moment detecting device are normal data, otherwise are misdata.

Control method to misdata is the output moving average, suc as formula (17).

(3) two kinds of method applicability analyses

Based on horizontal seasonal effect in time series method is that therefore, the variation tendency of valid interval can lag behind the variation tendency of measured data with the average of preceding n bar data and the valid interval of variance structure, promptly can the generation time drift based on horizontal seasonal effect in time series method.And when measured discharge from morning low ebb rise to this time period of morning peak, flow has the process of sudden change because the time delay of the variation tendency of valid interval, therefore, the data of sudden change might be mistaken as misdata.But do not required historical data based on horizontal seasonal effect in time series method, only can finish with the data on the same day.

Based on vertical seasonal effect in time series method is with the average of n day data before vertical time series and the valid interval of variance structure, its theoretical foundation is the secular trend of traffic parameter time series data, therefore, the valid interval of this kind method construct is comparatively stable, but off-line is finished, and can not produce horizontal seasonal effect in time series time drift.But the historical data that needs 3-8 week based on vertical seasonal effect in time series method just can be finished, and if historical data when very close, measured data has misjudged possibility.

In sum, when the historical data of abundance, be better than generally based on horizontal seasonal effect in time series method based on vertical seasonal effect in time series method; But effective historical data is not enough or as the national legal festivals and holidays that no historical datas such as " May Day ", " 11 " and the Ching Ming Festival can be used, then need employing based on horizontal seasonal effect in time series method.

6) based on the ACI algorithm design of SCATS coil data

(1) track traffic behavior evaluation method

When the transport need amount was low, the magnitude of traffic flow on the road was less, and all vehicles all can be with higher speed by stop line in the green time, and time headway is bigger, and traffic this moment is in smooth and easy state; Along with the increase of transport need amount, the magnitude of traffic flow on the road constantly increases, and traffic flow speed and time headway descend gradually simultaneously, when the transport need amount reaches road passage capability, the magnitude of traffic flow on the road reaches maximal value, and time headway reaches minimum value, and traffic this moment is in the retardance state; When the transport need amount surpasses the traffic capacity of road, to form queuing before the SCATS covering path stop line, the vehicle that comes the front in the green time still may be that the upper limit is passed through stop line with the traffic capacity of crossing, but can't identify this traffic behavior based on the inductive coil data.

By analyzing flow and the variation tendency of average headway under smooth and easy and retardance state, the present invention adopts flow, average headway combination as the congested in traffic foundation of discerning, and designs following congested in traffic index:

${\mathrm{\γ}}_t^i=\frac{k*q_t^j}{h_t^i}---\left(19\right)$

In the formula, γ _t ⁱThe congested in traffic index in the i elementary cell t time interval, q _t ⁱBe the magnitude of traffic flow in the i elementary cell t time interval, h _t ⁱBe corresponding average headway.Along with the increase (not comprising congestion status) of congested in traffic degree, this index has the monotone increasing characteristic, and k is crowded index sampling factor, and its value should be formulated according to concrete condition.

If boundary threshold value smooth and easy and retardance is γ ₁, then

1) when

The time, highway section i is in smooth and easy state at t in the time interval;

2) when

The time, highway section i is in the retardance state at t in the time interval.

In order to eliminate the influence of fluctuation immediately, improve congested in traffic state recognition result's stability, can carry out 2 times or 3 as required and continue identification, the congested in traffic state that occurs is continuously offered traffic administration person and traffic trip person as final recognition result.

For the traffic parameter data of place, zone-to-zone travel supplemental characteristics such as average stroke time, average travel speed can embody the traffic flow operation conditions in the road network better.Determine boundary threshold gamma smooth and easy and retardance ₁The time, should set up the incidence relation of zone-to-zone travel parameter and congested in traffic index, utilization zone-to-zone travel supplemental characteristic is discerned the congested in traffic state in highway section, and the counter based on this traffic congestion that pushes away based on place traffic parameter data measures threshold gamma ₁

(2) road section traffic volume method for evaluating state

By above-mentioned track traffic behavior evaluation method, can estimate out the traffic behavior in each track on the highway section.And for the traffic behavior in cross section, highway section, originally research and propose according to the track on the different signal phase combination highway sections, and provide corresponding combination track traffic behavior, promptly special-purpose left-hand rotation or exclusive right-turn lane provide its traffic behavior separately, craspedodrome and straight right lane belong to same phase place, and the combination track provides its traffic behavior.

Craspedodrome and straight right lane for after the combination as long as there is a track to be in smooth and easy state, then makes up the track and just are in smooth and easy state.Suppose that there is m track in certain combination track, concrete evaluation method is as follows:

(1) if the traffic behavior in m track all is retardance in the combination track, then makes up the track traffic behavior and be retardance;

(2) if it is smooth and easy in m track the traffic behavior in the individual track of j (j 〉=1) being arranged in the combination track, then the traffic behavior in this combination track is smooth and easy.

(3) if the traffic behavior in m track all is smooth and easy in the combination track, it is smooth and easy then making up the track traffic behavior.

Concrete computing formula is as follows:

T ^j＝min{t ₁，t ₁，...，t _m} (20)

In the formula:

T ^j---the traffic behavior in combination track;

t _i---track traffic behavior in the combination track;

M---the contained number of track-lines in combination track.

7) multistep forecasting method in short-term of traffic parameter virtual time sequence

The vertical time series of traffic parameter has good regularities, so the secular trend of traffic parameter can be estimated by methods such as the method for moving average, exponential smoothing, the summation autoregression methods of moving average, consider the simplicity of calculating, the present invention adopts the method for moving average to predict, the method for moving average comprises the simple method of moving average and two kinds of methods of the method for weighted moving average.The simple method of moving average is applicable to that the basic trend of target of prediction is a situation about fluctuating up and down in certain level, waits weight average to each data; And method of weighted mean is that forecasting institute is treated with a certain discrimination with data, and important predicted data is given bigger weight, with the importance degree of reflection different pieces of information when predicting.Based on two kinds of moving average model(MA model)s, make up traffic parameter data secular trend forecast model respectively suc as formula shown in (21) and the formula (22):

${\hat{x}}_{\mathrm{jk}}^p\left(t\right)=\frac{1}{n}\underset{i=j-n}{\overset{j-1}{\mathrm{\Σ}}}{x}_{\mathrm{jk}}^p\left(t\right)---\left(21\right)$

${\hat{x}}_{\mathrm{jk}}^p\left(t\right)=\underset{i=j-n}{\overset{j-1}{\mathrm{\Σ}}}{a}_i{x}_{\mathrm{jk}}^p---\left(22\right)$

In the formula:

---in j week, the prediction traffic parameter of the normal traffic state in k days t moment p places can be flow, average headway;

x _Jk ^p(t)---in i week, the historical traffic parameter in k days t moment p places can be flow, average headway;

T---detecting device is uploaded the constantly theoretical of data;

J---the j week of vertical time series data;

P---detecting device position numbering;

N---choose vertical time series data number;

a _i---the weight coefficient of contained i the data of moving average.

Above-mentioned model can once be finished predictions in 24 hours of each traffic parameter under the normal traffic state, and can constantly adjust the historical data of forecast model according to the passing of time, to embody the differentiation of the vertical time series secular trend of traffic parameter.

According to aforementioned thought, design secular trend forecast model solution procedure is as follows:

(1) acquisition is with the historical time sequence data of phase same date on the same day;

(2) carrying out data according to need predicted time yardstick synthesizes;

(3) selection needs the time scale of prediction;

(4) historical data of carrying out under the select time yardstick of front is upgraded;

(5) obtain the prediction vertical time series of traffic parameter data constantly;

(6) determine corresponding prediction weight, weight is fine when adopting the simple method of moving average to predict determines, does not do here and gives unnecessary details, and when adopting the method for weighted moving average to predict, utilizes historical optimal weights method to determine.Concrete grammar is: the optimal weight when obtaining to utilize the previous historical time sequence of vertical time series forecasting by processed offline deposits it in historical traffic parameter database with the traffic parameter historical data, and brings in constant renewal in.During prediction current time traffic parameter data, read the prediction weight of the optimal weight of last historical time current time as present moment;

(7) utilize model to predict accordingly, judge whether to finish whole predictions then, carry out next step,, return step (5) if do not have if finish;

(8) judge whether to finish all time scale predictions, if finish, then prediction of output result if do not finish, then returns step (3).

The flow process of short time yardstick traffic parameter secular trend forecast model as shown in Figure 8.

The present invention selects for use the self-adaptation exponential smoothing to make up traffic parameter intermediate trend forecast model.Segment data is predicted when utilizing data that existing exponential smoothing can only utilize the real-time period on the same day that detecting device obtains to next, therefore can't realize the multi-step prediction of day part.In order to realize multi-step prediction, now exponential smoothing is improved, realize multi-step prediction to the traffic parameter intermediate trend.Concrete model is as follows:

It is as follows respectively to go on foot predictor formula after self-adaptation exponential smoothing forecast model reached current second step of period, wherein i 〉=2.

${\hat{y}}_{t+i}^d={\mathrm{\α}}_{t+i-1}{\stackrel{\‾}{y}}_{t+i-1}^d+(1-{\mathrm{\α}}_{t+i-1}){\hat{y}}_{t+i-1}^d---\left(23\right)$

${\mathrm{\α}}_{t+i-1}=\left|\frac{E_{t+i-1}}{A_{t+i-1}}\right|---\left(24\right)$

$E_{t+i-1}={\mathrm{re}}_{t+i-1}^d+(1-r)E_{t+i-2}---\left(25\right)$

$A_{t+i-1}=r\left|e_{t+i-1}^d\right|+(1-r)A_{t+i-2}---\left(26\right)$

$e_{t+i-1}^d={\mathrm{\α}}_{d-1}^{\′}{e}_{t+i-1}^{d-1}+...+{\mathrm{\α}}_{d-i}^{\′}{e}_{t+i-1}^{d-i}+...+{\mathrm{\α}}_{d-N}^{\′}{e}_{t+i-1}^{d-N}---\left(27\right)$

${\stackrel{\‾}{y}}_{t+i-1}^d={{\mathrm{\α}}_{d-1}^{\′}y}_{t+i-1}^{d-1}+...+{\mathrm{\α}}_{d-i}^{\′}{y}_{t+i-1}^{d-i}+...+{\mathrm{\α}}_{d-N}^{\′}{y}_{t+i-1}^{d-N}---\left(28\right)$

${\mathrm{\α}}_{d-i}=\frac{L_{d-i}}{L_{d-i}+...+L_{d-i}...L_{d-n}}---\left(29\right)$

In the formula:

y _T+i ^dCurrent d days t+i period traffic parameters of-detecting device predicted value can be flow, average headway;

e _T+i-1 ^d, e _T+i-1 ^D-i-detecting device d, d-i days t+i-1 period first step self-adaptation exponential smoothing predicated errors;

α _T+i-1, α ' _D-i-weight coefficient;

y _T+i-1 ^D-iThe actual traffic parameter historical data of d-i days t+i-1 periods of-detecting device;

L _D-i-similarity calculates over d-i days the inverse of accumulative total similarity distance.

According to preceding surface analysis, design traffic parameter intermediate trend multi-step prediction model solution step is as follows:

(1) judges that current traffic flow is the intermediate trend data of non-events affecting data;

(2) utilize and to have obtained data and carry out one-step prediction, calculate measurable step number;

(3) judge whether to finish this period multi-step prediction,, carry out the multi-step prediction of next period,, enter next step if do not have if finish;

(4) carry out similar time series search, gauge index smoothing prediction error and signal errors;

(5) the replacement smoothing factor makes i=i+1, returns step (3).

By the intermediate trend multistep forecasting method as can be known, when obtaining the traffic parameter real data of two periods, can't obtain predicated error, can't utilize the self-adaptation EXSMOOTH to predict, can choose fixing α numerical value and calculate this moment, and α can be taken as 0.3.When obtaining the traffic parameter real data of the 3rd period, will obtain a predicated error, the error tracking signal of this moment is very unstable, α _tTo equal 1, be that fixed value is predicted with α still.When obtaining the traffic parameter real data of the 4th period, this moment α _tCan adjust according to error, and can utilize self-adaptation exponential smoothing to improve one's methods and predict.Can finish the calculating of predicted value with vertically moving consensus forecast when exponential smoothing can't realize predicting, the prediction step number of self-adaptation exponential smoothing prediction is determined by the measurable step number of time series.The prediction flow process as shown in Figure 9.

By moving average as can be known, the moving average forecast model goes to predict the traffic parameter data of next period with the traffic parameter time series data of adjacent period.This meets the prediction principle of short-term trend of the times traffic parameter data, can utilize moving average to finish prediction with short-term trend of the times time series data, the data of being utilized during the moving average prediction are with the most contiguous data of incident period of right time, to improve the order of accuarcy of prediction.

Because the different period short-term trend of the times data that incident causes have nothing in common with each other, the weight of weight moving average is difficult to determine, the traffic parameter data fluctuate around certain fixing horizontal usually in the influence time scope of incident simultaneously, can realize its prediction with simple moving average Forecasting Methodology, the self-adaptation exponential model has the very strong tracking power that gets to data simultaneously, and the first step when predicting the outcome its first step as the incident of generation predicts the outcome.The duration of incident is normally uncertain, so the step number of short-term trend prediction is generally one-step prediction, for the prediction effect to difference prediction step number compares, and the design multistep forecasting method.

Day part traffic parameter predictor formula was shown in (30) after second step reached:

${\hat{X}}_{t+i}^d=\frac{X_{t+i-1}^d+X_{t+i-2}^d+...+X_{t+i-n}^d}{N}---\left(30\right)$

In the formula:

-Di d days t+i period detecting device traffic parameter predicted values, X can be flow, average headway;

The prediction step number of N-short-term trend of the times prediction;

X _T+i-n ^d-Di d days t+i-n period detecting device actual traffic parameter X numerical value.

According to preceding surface analysis, under the solution procedure of design traffic parameter trend of the times forecast model:

(1) data of acquisition event data mark;

(2) the average weight coefficient is respectively predicted in input;

(3) utilize model to predict, judge whether to finish prediction, if finish, prediction of output result if do not finish, replaces actual value with predicted value, carries out next step prediction, predicts that specifically flow process as shown in figure 10.

8) congested in traffic range of scatter and duration method of estimation

Congested in traffic regularity by its generation can be divided into often sends out that sexual intercourse is logical can be divided into generally that normal to send out sexual intercourse logical crowded and sporadic congested in traffic two kinds.The normal crowded reasons such as local traffic capacity reduction that generally are the defective on the excessive or road geometric condition causes by transport need of property of sending out cause, normal sending out can be predicted logical crowding of sexual intercourse, it so just can be by repeatedly observing the when and where of predicting its generation often in certain periodic generation in place, highway section.Sporadic crowding is that the result causes crowding because temporary transient one or more track of blocking of the incident on the road causes that road section capacity descends.Incident comprises that traffic hazard, vehicle cast anchor, goods is scattered etc.Because the generation of these incidents is contingency very, does not have regularity, so it is predicted very difficulty.Congested in traffic range of scatter method of estimation must be set up based on traffic parameter data multi-step prediction technology, and it is very difficult to sporadic logical crowded prediction, so this research is primarily aimed at the normal logical crowded design space of the sexual intercourse range of scatter method of estimation of sending out, for sporadic crowded only to its one step of trend of the times prediction prediction, to eliminate the time delay that traffic parameter is gathered, it is not carried out the space range of scatter and estimate.

For the ease of research, the congested in traffic space of this research definition is diffused as a plurality of highway sections continuous in the space the crowded state of the property sent out often takes place simultaneously.The flow process of its scope method of estimation is as follows:

(1) highway section is numbered, under swim over to the upstream and progressively increase in proper order and be numbered;

(2) detect the traffic behavior in each highway section in real time, be in blocked state, then register highway section numbering P in case detect any highway section ₀With current time t ₀

(3) make j=1;

(3) detect highway section P ₀Whether+j congestion state can occur in its measurable step number, if "No" is not then exported highway section P ₀+ j is crowded range of scatter border, upstream, if be "Yes", then makes j=j+1, carries out step (3) again;

(4) detect highway section P ₀Whether-j congestion state can occur in its measurable step number, if "No" is not then exported highway section P ₀-j is crowded range of scatter border, downstream, if be "Yes", then makes j=j+1, carries out step (4) again.

Idiographic flow is illustrated in fig. 11 shown below.

The logical crowded duration algorithm for estimating cardinal principle of sexual intercourse is often sent out in single highway section, reads the real time data that the traffic detecting device collects, and estimates t congested in traffic degree constantly according to the traffic behavior evaluation algorithms.If t is constantly crowded through being evaluated as, then read this measurable step number n of these its data of detection constantly, read the data of multi-step prediction in short-term of this detecting device, continue to judge that next is the t+ δ degree of crowding constantly constantly, if t+ δ also is crowded through estimating constantly, then define t and take place the moment for crowded, be designated as t ₁Each real time data or multi-step prediction data in short-term constantly is crowded through estimating up to a certain moment t after continuing to read, and next is also crowded constantly or reach the maximum predicted step number, then defines this constantly for the crowded finish time, is designated as t ₂t ₂And t ₁Difference be exactly the current crowded duration.Continue to read and judge next real time data or multi-step prediction data constantly in short-term, judge whether to have once more crowded the generation.

It is as follows that the concrete job step of the logical crowded duration algorithm for estimating of sexual intercourse is often sent out in single highway section:

1) reads t real-time/predicted data constantly, enter 2);

2) calculate congested in traffic index, and compare, judge whether traffic behavior is A, if "Yes" then enters 3 with threshold value), "No" then enters 4);

3) make t=t+ δ, enter 1);

4) make t=t+ δ, read t real-time/predicted data constantly, calculate congested in traffic index once more, and compare with threshold value, judge whether traffic behavior is A, why judge interference once more for the data that prevent to fluctuate, if "Yes" then enters 3), "No" then enters 5);

5) be judged as crowded the generation, make t ₁=t enters 6);

6) make t=t+ δ, enter 7);

7) read t real-time/predicted data constantly, calculate congested in traffic index, and compare, judge whether traffic behavior is A, if "Yes" then enters 8 with threshold value), "No" then enters 6);

8) make t=t+ δ, read t real-time/predicted data constantly, calculate congested in traffic index, and compare with threshold value, judge whether traffic behavior is A, why judge it also is interference once more for the data that prevent to fluctuate, if "Yes" then enters 9), "No" then enters 6);

9) be judged as crowded the end, make t ₂=t,, T=t then ₂-t ₁I.e. the crowded duration, enter 3).

Crowded duration calculation process as shown in figure 12.

Claims (7)

1. online quantitative evaluation of traffic behavior and Forecasting Methodology based on SCATS coil data is characterized in that comprising the steps:

(1) the SCATS coil is gathered traffic parameter automatically and is uploaded to traffic information center by the network connection;

(2) traffic information center carries out quality assessment and control by computing machine to SCATS coil measured data, and comprise traffic events is detected automatically, and to vehicle supervision department's issue testing result, for it provides stronger decision-making foundation;

(3) traffic information center by computing machine to quality assessment with control after SCATS coil measured data carry out traffic parameter virtual time sequence construct;

(4) traffic information center carries out quality assessment and control by computing machine to the virtual time sequence data of SCATS traffic parameter;

(5) traffic information center carries out traffic congestion identification automatically by computing machine to the virtual time sequence data of SCATS traffic parameter, and differentiates the result to vehicle supervision department's issue, for it provides stronger decision-making foundation;

(6) traffic information center carries out multi-step prediction in short-term by computing machine to the virtual time sequence data of SCATS traffic parameter;

(7) traffic information center carries out congested in traffic range of scatter and duration prediction on the basis of multi-step prediction by computing machine in short-term to SCATS traffic parameter virtual time sequence data.

2. online quantitative evaluation according to claim 1 and Forecasting Methodology, it is characterized in that: analysis is meant based on the discernible traffic behavior type of SCATS coil data, definition and classification according to traffic behavior in the continuous traffic flow, in conjunction with SCATS inductive coil installation position and dynamic traffic data characteristics, draw and to discern smooth and easy state and retardance state based on SCATS coil data and can't discern the conclusion of congestion status.

3. online quantitative evaluation according to claim 1 and Forecasting Methodology, it is characterized in that: measured data quality assessment of SCATS coil and control are meant, the near-sighted recognition methods of horizontal time series segmentation moving average is adopted in the identification of obliterated data, adopts horizontal time series segmentation moving average to repair to the control of obliterated data; The designed traffic event automatic detection method based on SCATS coil data is adopted in the identification of event data.

4. online quantitative evaluation according to claim 1 and Forecasting Methodology is characterized in that the construction method of SCATS traffic parameter virtual time sequence comprises:

(1) SCATS traffic parameter actual measurement sampling interval is to the mapping relations analysis of virtual sampling interval;

(2) definite method of the virtual sampling interval of single crossing;

(3) definite method of the unified virtual sampling interval of road network.

5. online quantitative evaluation according to claim 1 and Forecasting Methodology, it is characterized in that: the congested in traffic automatic identifying method based on SCATS traffic parameter time series data is meant, according to above-mentioned draw based on SCATS coil data can discern smooth and easy state and the retardance state conclusion, by analyzing flow and the variation tendency of average headway under smooth and easy and retardance state, adopt flow, the average headway combination is as the foundation of congested in traffic identification, propose according to the track on the different signal phase combination highway sections, and provide corresponding combination track traffic behavior.

6. online quantitative evaluation according to claim 1 and Forecasting Methodology, it is characterized in that: the multistep forecasting method in short-term of SCATS traffic parameter virtual time sequence is meant, specificity analysis based on traffic parameter virtual time sequence, carry out traffic parameter secular trend multi-step prediction, traffic parameter intermediate trend multi-step prediction, the prediction of the traffic parameter short-term trend of the times, for the traffic behavior prediction provides the data basis.

7. online quantitative evaluation according to claim 1 and Forecasting Methodology is characterized in that: congested in traffic range of scatter and duration method of estimation comprise the crowded space of single road section traffic volume range of scatter Study on estimation method, the crowded duration Study on estimation method of single road section traffic volume and the crowded range of scatter Study on estimation method of regional road grid traffic.

Images