CN103366557B - Traffic congestion evaluation method based on congestion index - Google Patents

Abstract

The invention discloses a kind of traffic congestion evaluation method based on congestion index, comprise the steps: a, according to the free stream velocity in each section in the speed data calculating road network of road network vehicle equipment offer；B, calculate index hourage in every section；C, exponent data hourage to whole sections carry out cluster analysis and obtain number of categories；D, exponent data hourage to whole sections carry out cluster analysis, it is thus achieved that exponential quantity and class center exponential quantity hourage between the exponential quantity maximum hourage of each class, minimum traveltimes；E, each class is asked the average of its sample gradient-norm, distribute all kinds of corresponding congestion index scopes by the rule being directly proportional with gradient-norm；The congestion index that each apoplexy due to endogenous wind exponential number hourage is respectively mapped to such correspondence by f, definition membership function is interval。The inventive method can be adaptive to rain, snow, mist road passage capability change make decisions blocking up, science provide jam level divide。

Description

Traffic congestion evaluation method based on congestion index

Technical field

The present invention relates to traffic congestion assessment technique field, a kind of specifically can in the traffic congestion evaluation method based on congestion index that traffic congestion is carried out grade separation of science。

Background technology

Traffic congestion is the concentrated reflection that the mouth skewness that causes of economic development imbalance, the limited and city layout of urban transportation supply such as do not mate at the various social contradictioies with economic development, is a global problem。In order to deeply understand the essence of traffic congestion comprehensively, there is provided for every aspect work such as traffic administration, planning, policy appearances and support and correct guidance Public Traveling, conscientiously increasingly serious present situation of blocking up is alleviated, need a set of assessment indicator system that can conscientiously reflect road congestion conditions, wherein just include traffic congestion index。Existing evaluation index of blocking up, first threshold value of blocking up it is manually set, using it as the standard judging whether section blocks up, judge whether to block up to each section, then by the total kilometrage mileage ratio of blocking up divided by road network total kilometrage calculating road network being judged as the section blocked up, and mileage ratio conversion of blocking up is index。And then the mileage ratio cut partition that will block up be unimpeded, substantially unimpeded, slightly block up, moderate is blocked up, five grades of heavy congestion,。The method have drawbacks in that 1, to be manually set threshold value of blocking up too strong to the subjectivity judged that blocks up；2, being manually set jam level, the subjectivity that jam level is divided is too strong, lacks Scientific Meaning；3, the travel speed caused that cannot be distinguished by reducing due to travel speed that the change of the road passage capability such as rain, snow, mist causes and block up due to vehicle more reduce between difference。

Because the defect that above-mentioned existing traffic congestion evaluation method exists, the present inventor's actively in addition research and innovation, energy science to founding a kind of novelty provides the traffic congestion evaluation method based on congestion index of the classification of severity that blocks up, to solve the deficiency that prior art exists。

Summary of the invention

In order to solve the above-mentioned problems in the prior art, the invention provides one to be adaptive to the change of the road passage capability such as rain, snow, mist and make decisions blocking up, that compares science provides the traffic congestion evaluation method based on congestion index that jam level divides simultaneously。

In order to solve above-mentioned technical problem, present invention employs following technical scheme:

Based on the traffic congestion evaluation method of congestion index, comprise the steps:

1) free stream velocity in each section in obtaining 24 hours, described free stream velocity is the travel speed of vehicle in the complete unimpeded situation of the less road of the volume of traffic；

2) calculating index hourage (TTI) in each section, described hourage, index was that vehicle travels the ratio of the time travelling same distance needs with free stream velocity with present speed；

3) according to hourage exponential quantity size the data in whole for road network sections a certain moment are carried out cluster analysis, the principle that in exponent data hourage of system-wide net is big as far as possible according to class inherited, class, difference is the least possible determines optimal classes；

4) on the basis determine optimal classes, index hourage in all sections of road network is clustered, and obtain the maximum TTI value of each class, minimum TTI Zhi Helei center TTI value；

5) for each class is asked it comprise the average of sample gradient-norm, all kinds of mapping range in congestion index are directly proportional to its gradient-norm, and congestion index scope 0-10 is distributed to each class；

6) step 5 give each hourage index class be assigned with on the basis that congestion index is interval, membership function is designed in each class, it is converted into congestion index according to the numerical value of certain rule each apoplexy due to endogenous wind index hourage, and blocks up weight situation with congestion index correspondence from small to large。

As preferably, described step 1) in, 24 hours are divided into y equal interval, each section is calculated the average speed of vehicle in each interval, more respectively to each section before 24 hours ask speed maximum in y interior sample the average of 15% sample as the free stream velocity in this section。

As preferably, described step 2) in, in the way of Floating Car, obtain relevant speed, and then obtain index hourage of respective stretch, adopt formula as follows:

$\begin{array}{c}TTI=\underset{i=1}{\overset{n}{\Σ}}{t}_i/\underset{i=1}{\overset{n}{\Σ}}{T}_i\\ =\underset{i=1}{\overset{n}{\Σ}}{t}_i/\underset{i=1}{\overset{n}{\Σ}}\left(\frac{l_i}{v_f}\right)\end{array}$

Wherein TTI is road trip time index, and n is Floating Car number, t in section_iFor the real travel time of in section i-th Floating Car, T_iIt is that i-th Floating Car travels required time, l with free stream velocity_iIt is i-th Floating Car operating range in section, v_fFor section free stream velocity。

As preferably, described step 3) in, it is determined that the step of optimal classes is as follows:

If index hourage in i-th section is sample a_i, then

$d\left(i\right)={\mathrm{mean}}_{a_j\∈K\left(i\right)}\left(d_{i,j}\right)$

Wherein d (i) is sample a_iWith its belonging to the average distance of other samples, i ≠ j and a in class_j∈ K (i), K (i) are sample a_iThe class of ownership, d_i,jFor sample a_iWith a_jSpacing；

$b\left(i\right)={\min }_{K^{\′}\≠K\left(i\right)}\left({\mathrm{mean}}_{a_j\∈K^{\′}}\right(d_{i,j}\left)\right)$

Wherein b (i) is sample a_iThe minima of sample mean distance in other class, K ' is the class different from class K (i), d_i,jFor the sample a in class K (i)_iTo the middle sample a of class K '_jDistance；

$Sil\left(i\right)=\frac{b\left(i\right)-d\left(i\right)}{max\{d\left(i\right),b\left(i\right)\}}$

Sil (i) is sample a_iClass in Separatory measure amount, Sil (i) ∈ [1 ,-1] between compactness and class, the more big clustering result quality of value is more good；

It is the average Silhouette measure value P that the cluster result in m situation calculates all samples for number of categories_m

$P_m=\underset{i=1}{\overset{N}{\Σ}}Sil\left(i\right)/N$

Wherein N is sample size, then optimal classes m is for making P_mObtain number of categories during maximum；

$P_M={\max }_{m=1,2,...,K_{max}}\left(P_m\right),$

P_mBy traveling through the institute likely value m=1,2 of number of categories m ..., K_maxObtain。

As preferably, described step 4) in index hourage in all sections of road network carried out cluster include following iterative processing steps:

(1) M is arbitrarily selected as initial cluster center from n sample；

(2) calculate the distance of each sample and each class central point, and according to the principle of minimum range, sample is partitioned into a certain class；

(3) calculate the center of each class, the central point of class be its arrive affiliated apoplexy due to endogenous wind other the minimum point of distance sum a little；

(4) judge that whether each center and the last each center clustered of current cluster be consistent, such as inconsistent return step 2, as consistent finishing iteration has clustered。

As preferably, described step 6) in membership function be S function that the change of class center is comparatively sensitive

$f\left(x\right)=\left\{\begin{array}{cc}{D}_i+4{\left(\frac{x-P_i}{P_{i+1}-P_i}\right)}^2,& P_i<x\&le;\frac{P_{i+1}+P_i}{2}\\ D_i+2\left(1-2{\left(\frac{x-P_{i+1}}{P_{i+1}-P_i}\right)}^2\right),& \frac{P_{i+1}+P_i}{2}<x\&le;P_{i+1}\end{array}\right.$

Wherein i=1,2 ..., M, P_iFor exponential lower bound value hourage of cluster Hou Anlei center exponential number hourage sequence the i-th class from small to large, P_M+1It is the upper dividing value of M class, D_iFor the lower bound that the i-th class distribution congestion index in step 5 is interval。

Compared with prior art, the beneficial effects of the present invention is:

1, the present invention utilizes the statistical property of traffic data itself to give the scientific method dividing jam level, it is to avoid artificially arrange subjectivity and the blindness of jam level, the perfect stage theory that blocks up；

2, free stream velocity changes along with the Changes in weather of every day simultaneously, this method calculates index hourage according to the free stream velocity of every day, and then the congestion index drawn also is able to be adaptive to the road passage capability changes such as the natural weathers such as rain, snow, the mist impact on traffic, it is to avoid existing fixed speed threshold value judgement method of blocking up does not adapt to the defect of road passage capability change。

Accompanying drawing explanation

Fig. 1 is the traffic congestion evaluation method that proposes of the present invention and the contrast to Beijing's whole day traffic noise prediction on July 12nd, 2012 of the existing method。

Fig. 2 is the traffic congestion evaluation method that proposes of the present invention and the contrast to December in 2012 Beijing's whole day traffic noise prediction on the 28th of the existing method。

Detailed description of the invention

Below in conjunction with the drawings and specific embodiments, the present invention is described in further detail, but not as a limitation of the invention。

Based on the traffic congestion evaluation method of congestion index, comprise the steps:

1, the free stream velocity in each section in obtaining 24 hours

Free stream velocity is the travel speed of vehicle in the complete unimpeded situation of the less road of the volume of traffic, it is possible to ignore traffic density to its impact, but it is by the impact of the road passage capabilities such as natural conditions and restricted driving mark such as rain, snow, mist,

Free stream velocity is the travel speed of vehicle in the complete unimpeded situation of the less road of the volume of traffic, traffic density can be ignored on its impact, but it is by the impact of the road passage capabilities such as natural conditions and restricted driving mark such as rain, snow, mist, and each section has different free stream velocities。Such as with 15 minutes for interval, whole day is 96 time periods altogether, first each section is calculated the average speed of each time period wherein vehicle, then before asking speed maximum in 96 samples of whole day in each section respectively the average of 15% sample as the free stream velocity in this section。

2, index hourage in each section is calculated

It is defined as the ratio that same distance vehicle travels the time needed and requires time for free stream velocity traveling with present speed index hourage (TravelTimeIndex is abbreviated as TTI)。

$\begin{array}{c}TTI=\underset{i=1}{\overset{n}{\&Sigma;}}{t}_i/\underset{i=1}{\overset{n}{\&Sigma;}}{T}_i\\ =\underset{i=1}{\overset{n}{\&Sigma;}}{t}_i/\underset{i=1}{\overset{n}{\&Sigma;}}\left(\frac{l_i}{v_f}\right)\end{array}$

Wherein TTI is road trip time index, and n is Floating Car number, t in section_iFor the real travel time of in section i-th Floating Car, T_iIt is that i-th Floating Car travels required time, l with free stream velocity_iIt is i-th Floating Car operating range in section, v_fFor section free stream velocity。Namely all floating vehicle travelling time sums and travel the same ratio (its operating range sum is divided by free stream velocity) apart from required time under free stream velocity in same link。The present embodiment is to obtain relevant speed in the way of Floating Car, and then obtains index hourage of respective stretch。The corresponding speed on a certain section can certainly be obtained by additive method, and and then obtain index hourage of respective stretch。

3, the best index classification number hourage is calculated

Hourage, the more big travel speed of exponential quantity was more low, the TTI data in whole for road network sections a certain moment are carried out cluster analysis by the size according to TTI value, first the principle that in the TTI data of system-wide net are big as far as possible according to class inherited, class, difference is the least possible determines optimal classes, and algorithm is as follows:

If the TTI in i-th section is sample a_i

$d\left(i\right)={\mathrm{mean}}_{a_j\&Element;K\left(i\right)}\left(d_{i,j}\right)$

Wherein d (i) is sample a_iWith its belonging to the average distance of other samples, i ≠ j and a in class_j∈ K (i), K (i) are sample a_iThe class of ownership, d_i,jFor sample a_iWith a_jSpacing。

$b\left(i\right)={\min }_{K^{\&prime;}\&NotEqual;K\left(i\right)}\left({\mathrm{mean}}_{a_j\&Element;K^{\&prime;}}\right(d_{i,j}\left)\right)$

Wherein b (i) is sample a_iThe minima of sample mean distance in other class, K ' is the class different from class K (i), d_i,jFor the sample a in class K (i)_iTo the middle sample a of class K '_jDistance。

$Sil\left(i\right)=\frac{b\left(i\right)-d\left(i\right)}{max\{d\left(i\right),b\left(i\right)\}}$

Sil (i) is sample a_iClass in Separatory measure amount, Sil (i) ∈ [1 ,-1] between compactness and class, the more big clustering result quality of value is more good。

It is the average Silhouette measure value P that the cluster result in m situation calculates all samples for number of categories_m

$P_m=\underset{i=1}{\overset{N}{\&Sigma;}}Sil\left(i\right)/N$

Wherein N is sample size。Then optimal classes M is for making P_mObtain number of categories during maximum。

$P_M={\max }_{m=1,2,...,K_{max}}\left(P_m\right),$

P_mBy traveling through the institute likely value m=1,2 of number of categories m ..., K_maxObtain。

4, road network index hourage is classified

Determining, index hourage in all sections of road network is clustered by the basis of optimal classes m, it is thus achieved that exponential quantity and class center exponential quantity hourage between the exponential quantity maximum hourage of each class, minimum traveltimes。

The iterative processing steps that the hourage in road network all sections, index clustered is as follows:

(1) arbitrarily select m (identical with the optimal classes obtained upper step) individual as initial cluster center from n sample；

(2) calculate the distance of each sample and each class central point, and according to the principle of minimum range, sample is partitioned into a certain class；

(3) calculate the center of each class, the central point of class be its arrive affiliated apoplexy due to endogenous wind other the minimum point of distance sum a little；

(4) judge that whether each center and the last each center clustered of current cluster be consistent, such as inconsistent return step 2, as consistent finishing iteration has clustered。

5, congestion index is distributed according to each class sample gradient-norm interval

For each class is asked it comprise the average of sample gradient-norm, all kinds of mapping range on index are directly proportional to its gradient-norm, and congestion index scope 0-10 is distributed to each class。

${\mathrm{Mean}}_{a_i\&Element;K\left(i\right)}\left(\mathrm{mod}\right(Grad\left(a_i\right)\left)\right)\&Proportional;S_i$

Wherein K (i) is sample a_iThe class of ownership, Grad (a_i) for sample a_iThe Grad at place, mod is delivery, and ∝ represents and is proportional to, S_iFor the congestion index scope that class K (i) is corresponding。

Less congestion index interval is given compared with the class of minizone for comprising the TTI changes such as night, interval giving bigger congestion index interval for what comprise that the TTI values such as early evening peak change greatly, the congestion index so drawn is more sensitive to the time period that morning, the traffic such as evening peak changed greatly。

6, index hourage is converted into congestion index by design membership function in each class

Step 5 is assigned with congestion index interval to the class of each TTI, it is necessary to designs membership function in each class, can be converted into congestion index according to certain each apoplexy due to endogenous wind TTI numerical value of rule。

Such as design the S function that the change of class center is comparatively sensitive

$f\left(x\right)=\left\{\begin{array}{cc}{D}_i+4{\left(\frac{x-P_i}{P_{i+1}-P_i}\right)}^2,& P_i<x\&le;\frac{P_{i+1}+P_i}{2}\\ D_i+2\left(1-2{\left(\frac{x-P_{i+1}}{P_{i+1}-P_i}\right)}^2\right),& \frac{P_{i+1}+P_i}{2}<x\&le;P_{i+1}\end{array}\right.$

Wherein i=1,2, M, P_iFor the TTI floor value of cluster Hou Anlei center TTI numerical ordering the i-th class from small to large, P_M+1It is the upper dividing value of M class, D_iFor the lower bound that the i-th class distribution congestion index in step 5 is interval。

The inventive method is used for Beijing be verified。Wherein

Adopt system-wide net interval 15 minutes April in 2013, Beijing hourage exponent data as cluster sample, cluster result be hourage index maximum in being divided into the Silhouette measure value 0.7613 of four classes to be likely to for all classification, four classes are optimal classification number。

According to the following table 1 below of parameter after four class clusters。

Table 1

	One class	Two classes	Three classes	Four classes
					TTI class center	1.0488	1.3577	1.6039	1.88
Class sample size	946	763	760	378
					The TTI upper bound	1.2025	1.4802	1.7392	2.3975
TTI lower bound	0.91423	1.2056	1.4813	1.7432
					Gradient-norm average	0.021974	0.034364	0.040973	0.055681
Congestion index is interval	0-1.3	1.3-3.3	3.3-6.0	6.0-10

What the existing method based on mileage ratio of blocking up was artificial is divided into five class congestion index interval by congestion index, and 0-2 is unimpeded, and 2-4 is substantially unimpeded, and 4-6 is for slightly to block up, and 6-8 is that moderate is blocked up, and 8-10 is heavy congestion。It is divided into four class congestion index interval according to the result congestion index of the present invention, wherein 0-1.3 is unimpeded, 1.3-3.3 is substantially unimpeded, 3.3-6.0 is for slightly to block up, 6.0-10 is heavy congestion, and the present invention compares science the order of severity of blocking up is classified according to the statistical property of traffic data self。

Draw according to the present invention day congestion index change contrast with the existing method based on mileage ratio of blocking up, comparing result is shown in that Fig. 1 and Fig. 2, Fig. 1 are the traffic congestion evaluation method that proposes of the present invention and the contrast to Beijing's whole day traffic noise prediction on July 12nd, 2012 of the existing method。Fig. 2 is the traffic congestion that proposes of the present invention and the contrast to December in 2012 Beijing's whole day traffic noise prediction on the 28th of the existing method。

As can be seen from the figure 1, this method is at the comparatively smooth traffic behavior reflected more really in free stream situation at night at night；2, this method in snowfall situation early the time of blocking up of evening peak longer, compare and tally with the actual situation；3, the present invention has observed the little peak of secondary that 8: 1 evening occurred on ordinary days, more meets reality than existing method。

Claims (5)

1. based on the traffic congestion evaluation method of congestion index, it is characterised in that comprise the steps:

1) free stream velocity in each section in obtaining 24 hours, described free stream velocity is the travel speed of vehicle in the complete unimpeded situation of the less road of the volume of traffic；

2) calculating index TTI hourage in each section, described hourage, index was that vehicle travels the ratio of the time travelling same distance needs with free stream velocity with present speed；

3) according to hourage exponential quantity size the data in whole for road network sections a certain moment are carried out cluster analysis, the principle that in exponent data hourage of system-wide net is big as far as possible according to class inherited, class, difference is the least possible determines optimal classes；

4) on the basis determine optimal classes, index hourage in all sections of road network is clustered, and obtain the maximum TTI value of each class, minimum TTI Zhi Helei center TTI value；

5) for each class is asked it comprise the average of sample gradient-norm, all kinds of mapping range in congestion index are directly proportional to its gradient-norm, and congestion index scope 0-10 is distributed to each class；

6) step 5 give each hourage index class be assigned with on the basis that congestion index is interval, membership function is designed in each class, it is converted into congestion index according to the numerical value of certain rule each apoplexy due to endogenous wind index hourage, and blocks up weight situation with congestion index correspondence from small to large；

Wherein, in step 3, it is determined that the step of optimal classes is as follows:

If index hourage in i-th section is sample a_i, then

$d\left(i\right)={\mathrm{mean}}_{a_j\&Element;K\left(i\right)}\left(d_{i,j}\right)$

Wherein d (i) is sample a_iWith its belonging to the average distance of other samples, i ≠ j and a in class_j∈ K (i), K (i) are sample a_iThe class of ownership, d_i,jFor sample a_iWith a_jSpacing；

$b\left(i\right)={\min }_{K^{\&prime;}\&NotEqual;K\left(i\right)}\left({\mathrm{mean}}_{a_j\&Element;K^{\&prime;}}\right(d_{i,j}\left)\right)$

Wherein b (i) is sample a_iThe minima of sample mean distance in other class, K ' is the class different from class K (i), d_i,jFor the sample a in class K (i)_iTo the middle sample a of class K '_jDistance；

$Sil\left(i\right)=\frac{b\left(i\right)-d\left(i\right)}{max\{d\left(i\right),b\left(i\right)\}}$

Sil (i) is sample a_iClass in Separatory measure amount, Sil (i) ∈ [1 ,-1] between compactness and class, the more big clustering result quality of value is more good；

It is the average Silhouette measure value P that the cluster result in m situation calculates all samples for number of categories_m

$P_m=\underset{i=1}{\overset{N}{\&Sigma;}}Sil\left(i\right)/N$

Wherein N is sample size, then optimal classes m is for making P_mObtain number of categories during maximum；

$P_M={\max }_{m=1,2,...,K_{max}}\left(P_m\right),$

P_mBy traveling through the institute likely value m=1,2 of number of categories m ..., K_maxObtain。

2. the traffic congestion evaluation method based on congestion index according to claim 1, it is characterized in that, described step 1) in, 24 hours are divided into y equal interval, each section is calculated the average speed of vehicle in each interval, more respectively to each section before 24 hours ask speed maximum in y interior sample the average of 15% sample as the free stream velocity in this section。

3. the traffic congestion evaluation method based on congestion index according to claim 1, it is characterised in that described step 2) in, in the way of Floating Car, obtain relevant speed, and then obtain index hourage of respective stretch, adopt formula as follows:

$\begin{array}{c}TTI=\underset{i=1}{\overset{n}{\&Sigma;}}{t}_i/\underset{i=1}{\overset{n}{\&Sigma;}}{T}_i\\ =\underset{i=1}{\overset{n}{\&Sigma;}}{t}_i/\underset{i=1}{\overset{n}{\&Sigma;}}\left(\frac{l_i}{v_f}\right)\end{array}$

Wherein TTI is road trip time index, and n is Floating Car number, t in section_iFor the real travel time of in section i-th Floating Car, T_iIt is that i-th Floating Car travels required time, l with free stream velocity_iIt is i-th Floating Car operating range in section, v_fFor section free stream velocity。

4. the traffic congestion evaluation method based on congestion index according to claim 1, it is characterised in that described step 4) in index hourage in all sections of road network carried out cluster include following iterative processing steps:

(1) M is arbitrarily selected as initial cluster center from n sample；

(2) calculate the distance of each sample and each class central point, and according to the principle of minimum range, sample is partitioned into a certain class；

(3) calculate the center of each class, the central point of class be its arrive affiliated apoplexy due to endogenous wind other the minimum point of distance sum a little；

(4) judge that whether each center and the last each center clustered of current cluster be consistent, such as inconsistent return step 2, as consistent finishing iteration has clustered。

5. the traffic congestion evaluation method based on congestion index according to claim 1, it is characterised in that described step 6) in membership function be S function that the change of class center is comparatively sensitive

$f\left(x\right)=\left\{\begin{array}{cc}{D}_i+4{\left(\frac{x-P_i}{P_{i+1}-P_i}\right)}^2,& P_i<x\&le;\frac{P_{i+1}+P_i}{2}\\ D_i+2(1-2{\left(\frac{x-P_{i+1}}{P_{i+1}-P_i}\right)}^2),& \frac{P_{i+1}+P_i}{2}<x\&le;P_{i+1}\end{array}\right.$

Wherein i=1,2 ..., M, P_iFor the TTI floor value of cluster Hou Anlei center TTI numerical ordering the i-th class from small to large, P_M+1It is the upper dividing value of M class, D_iFor the lower bound that the i-th class distribution congestion index in step 5 is interval。