CN102760182B - Population travel demand calculating method based on compressive sensing technology - Google Patents

Abstract

The invention discloses a population travel demand calculating method based on a compressive sensing technology. With the population travel demand calculating method based on the compressive sensing technology, according to the sparsity of travel demands among various activity sites, the travel demand is calculated by restoring sparse vectors through the compressive sensing technology. The population travel demand calculating method based on the compressive sensing technology mainly comprises the following steps: for a travel demand vector whose sparsity does not meet a restoring requirement, training a linear transformation base through an own travel vector of a simulating system so as to make an original travel vector meet an appointed sparsity under the linear transformation base; designing an optimal observation matrix by minimizing coherence of an observation matrix and a sparsity transformation base, and further selecting an observation road section set; and according to traffic observed at an actual observation road section, restoring the original travel vector from a constraint equation set by a minimized L1 norm method in the compressive sensing technology. Compared with the prior art, the population travel demand calculating method based on the compressive sensing technology has the advantage that the travel demands among various sites in actual traffic can be calculated in real time and more accurately, so that in an on-line state, the simulating system can be consistent with an actual traffic condition more reliably.

Description

Population trip requirements computing method based on compressed sensing technology

Technical field

The invention belongs to traffic simulation field, be specifically related to a kind of population trip requirements computing method based on compressed sensing technology.

Background technology

Computer Simulation has become an important means of analysis city traffic.It is advantageous that with the various complicated traffic behaviors of cheap cost simulation and check feasibility and the validity etc. of control algolithm, for the optimum management of urban transportation provides decision references.Calculate population trip requirements and can predict comparatively exactly the trip production in actual traffic system, make emulation approach reality, and then improve the confidence level of traffic simulation.

Current trip requirements computing method mainly contain three kinds, one is that road network is divided into multiple regions, obtain resident's go off daily data by traffic study, then in analogue system, set each initiation region (Origin) and stop the trip requirements amount between region (Destination), carrying out emulation thereby produce row.Trip requirements amount between each region can be write as matrix form, is called OD matrix or OD table in traffic engineering, and as shown in table 1, wherein generating capacity and traffic attraction are the summations of corresponding row or column.The way of this direct surveys OD matrix need to be provided a large amount of questionnaires, expends a large amount of manpower and financial resources, and obtain data accuracy can not ensure, therefore less employing in practice.To be that OD is counter push away the second,, partly or entirely on section, the first-class pick-up unit of shooting being installed, obtains the volume of traffic in observation section etc., then by certain mathematical model Extrapolation OD matrix.The information that this way needs is less, workable, but because being far smaller than, detected amount of traffic information calculates the needed quantity of information of trip requirements, therefore the mathematical model set up is generally a underdetermined system of equations, can only determine unique solution by adding artificially certain assumed condition.According to the difference of subsidiary condition, can be divided into Statistical Estimation Method and equalization two classes.Statistical Estimation Method is by setting up and optimizing probability objective function, and the estimated value using the OD matrix of maximum probability as current trip requirements amount, typically has Maximum Entropy estimation, maximum likelihood estimation, Bayesian Estimation etc.; Balanced rule adopts double-deck law of planning to ask for OD matrix according to Wardrop user equilibrium principle.The third is that first two method is combined with, and is first calculated and is drawn OD matrix by road section traffic volume amount, then calculate result using traffic study data as priori correction.This method remains so that OD is counter and pushes away as main, adopts at present often.

Just calculate result due to Statistical Estimation Method using the solution of certain posterior probability maximum in Feasible Solution Region as OD matrix, therefore might not reflect current trip requirements.On the other hand, actual traffic is often again in transient, and this makes double-deck law of planning need to calculate the parameter of traffic while reaching stable state to determine current OD, thereby causes calculated amount to strengthen, real-time on-line simulation deleterious.

Table 1OD matrix

Summary of the invention

For solving the problems such as existing traffic simulating system is not high to the approximation accuracy of trip requirements, calculating real-time is poor, the present invention has provided a kind of population trip requirements computing method based on compressed sensing technology, detect data by road section traffic volume amount, calculate the trip requirements amount between each playground, so that analogue system can be approached actual traffic state better.

A kind of population trip requirements computing method based on compressed sensing technology that the present invention proposes, is characterized in that, the method comprises the following steps:

Step 1, based on analogue system, the travel amount V in multi collect simulation time section between each place ^*as training sample data collection;

Step 2, using the training sample data collection collecting in described step 1 as input, the trip requirements vector that training obtains between place is specified the sparse transform-based L of degree of rarefication, so that the coordinate of the trip requirements vector V between described place under sparse transform-based L represents that W has appointment degree of rarefication;

Step 3, m is counted in the observation section of specifying according to user, calculates one by one every measurement row vector P that section is corresponding in analogue system road network _iwith the degree of coherence of described sparse transform-based L, choose m the measurement row vector composition measurement matrix P with minimum degree of coherence, using a corresponding m section as observation section collection X, and will measure matrix P, population constraint matrix Q, R composition observing matrix K;

Step 4, is divided into multiple sense cycle by simulation time section, gets a sense cycle as the current detection cycle, the actual traffic amount in acquisition step 3 in definite m observation section obtain observing the volume of traffic of section i wherein α is average carrying number, 1≤i≤m; In gathering simulation system, population when this current sense cycle initial sum stops in each place, obtains total observation vector B simultaneously;

Step 5, according to linear transformation V=LW, is rewritten as B=KLW by observation equation B=KV; Calculate the optimum solution of W according to the total observation vector B, the observing matrix K that calculate, sparse transform-based L and described observation equation, i.e. W ^#;

Step 6, calculates the trip requirements amount V:V=LW between each place in the current detection cycle ^#;

Step 7, judges in simulation time section whether have next sense cycle, if so, returns to step 4, enters the calculating of next sense cycle; Otherwise, calculate and finish.

Advantage of the present invention and good effect are: total simulation time section is divided into multiple less sense cycle, within each cycle, ignore the less trip component between place, thereby ensure analogue system in operation overall process all the time by the actual information detecting as much as possible for the accurate Calculation component of mainly going on a journey, avoid deviation and the excessive problem of double-deck law of planning calculated amount on Statistical Estimation Method probability meaning, realized dynamic simulation and calculate.Meanwhile, the method does not need extensive traffic study data, the volume of traffic that only utilizes the actual section of part to detect, convenient operation.

Brief description of the drawings

Fig. 1 is the population trip requirements computing method process flow diagram that the present invention is based on compressed sensing technology.

Fig. 2 is reported in Tianhe district of Guangzhou map road network.

The road network that the embodiment that Fig. 3 adopts for the present invention uses.

Embodiment

For making the object, technical solutions and advantages of the present invention clearer, below in conjunction with specific embodiment, and with reference to accompanying drawing, the present invention is described in more detail.

Consider traffic network G (N, A, PL), wherein, N represents junction node, and A represents section, and PL represents the playground in road network.If be an observation section collection, and | X|=m is observation section number, | PL|=n is playground number.At certain sense cycle t=[0, T] in, order

$U=\left[\begin{array}{cccccc}{u}_{11}& u_{12}& .& .& .& u_{1n}\\ u_{21}& u_{22}& .& .& .& u_{2n}\\ .& .& & & & .\\ .& .& & & & .\\ .& .& & & & .\\ u_{n1}& u_{n2}& .& .& .& u_{\mathrm{nn}}\end{array}\right]$

For travel amount matrix, wherein, u _ijthe travel amount (i.e. by place i trip number to place j) of representative from place i to place j.Write U as vector form:

[v _1?v ₂?…?v _N] ^T＝[u ₁₁?…?u _1n?u ₂₁?…?u _2n?u ₃₁?…?u _nn] ^T………(1)

Wherein, v _jbe j trip component, N=n × n.Definition for the volume of traffic of observation section i, wherein for the volume of traffic (by the vehicle number in this section) that actual section i detects, α is the average carrying number of vehicle, unit behaviour/, need to obtain by the investigation of traffic control department statistics.With p _ijrepresent corresponding trip component v _jallocation proportion on the i of section, obtains by traffic study.According to flow conserva-tion principle, can obtain measuring system of equations:

Definition [p _i1p _i2p _iN]=P _ifor measurement row vector corresponding to section i.

In the time that sense cycle starts and finish, in analogue system, the population equation of constraint group in each place is:

$\left[\begin{array}{c}{y}_1\\ .\\ .\\ .\\ y_n\\ z_1\\ .\\ .\\ .\\ z_n\end{array}\right]=\left[\begin{array}{cccccc}{q}_{11}& q_{12}& .& .& .& q_{1N}\\ .& .& & & & .\\ .& .& & & & .\\ .& .& & & & .\\ q_{n1}& q_{n2}& .& .& .& q_{\mathrm{nN}}\\ r_{11}& r_{12}& .& .& .& r_{1N}\\ .& .& & & & .\\ .& .& & & & .\\ .& .& & & & .\\ r_{n1}& r_{n2}& .& .& .& r_{\mathrm{nN}}\end{array}\right]\·\left[\begin{array}{c}{v}_1\\ v_2\\ .\\ .\\ .\\ .\\ .\\ .\\ v_{N-1}\\ v_N\end{array}\right]$

Wherein, y _ithe population of place i in the time of t=0, z _ibe place i in the population in t=T moment, in analogue system, all can detect;

Two equations simultaneousnesses obtain observation equation above, by its brief note are:

$\left[\begin{array}{c}X\\ Y\\ Z\end{array}\right]=\begin{array}{c}\left[\begin{array}{c}P\\ Q\\ R\end{array}\right]\·\end{array}V$

Wherein, Q, R are all called population constraint matrix.

Above formula further can be write a Chinese character in simplified form into:

B＝K·V………………………………(3)

Wherein, B ties up total observation vector (known) for (m+2n), and K is defined as observing matrix (known), and V is defined as trip requirements vector, by the column vector of travel amount matrix U represents (as (1) formula is defined, waiting to ask).In reality, observation section number is far smaller than playground number, i.e. m < < n.Therefore (3) formula is an Indeterminate Equation Group, and without unique solution, its free variable number is (n ²-2n-m).

From the character of the angle analysis trip requirements vector V based on activity trip.The population demand that has stemmed from current active and started the next item down activity of going on a journey, and adjacent two movable types are not identical in most cases.For example, initial active when agency's (Agent, i.e. visual human in analogue system) daystart is predetermined to be the rest of being in conventionally, and Section 2 activity subsequently can not remain rest.Otherwise, can be regarded as same activity, without changing playground and then producing trip.Therefore, in road network, belong between the place of same Activity Type and can produce hardly travel amount.In mathematical model, be reflected as most of component in trip requirements vector V and be approximately zero, V has sparse property.Therefore can consider to adopt compressed sensing technology to solve (3) formula.

Compressed sensing technology, in the time solving the sparse solution of the underdetermined system of equations, needs observing matrix K to meet two conditions:

1) the vectorial V that is s for degree of rarefication (being to only have at the most s component non-zero in V), any 3s of observing matrix K is listed as equal linear independence.When this condition cannot meet if s is large, seek a Sparse transform-based

$L=\left[\begin{array}{cccccc}{l}_{11}& l_{12}& .& .& .& l_{1N}\\ l_{21}& l_{22}& .& .& .& l_{2N}\\ .& .& & & & .\\ .& .& & & & .\\ .& .& & & & .\\ l_{N1}& l_{N2}& .& .& .& l_{\mathrm{NN}}\end{array}\right]$

After making to be linear transformation V=LW, the coordinate under L represents that W reaches enough sparse.

2) degree of coherence of observing matrix K and sparse transform-based L reaches minimum.The degree of coherence of definition K and L is wherein K _i, L _jthe i j capable and L that is respectively K is listed as.Minimizing degree of coherence can realize by choosing suitable observation section collection, below will introduce in detail.

Fig. 1 is the process flow diagram that the present invention is based on the population trip requirements computing method of compressed sensing technology,

Concrete steps are as follows:

Step 1, based on analogue system, the travel amount V in multi collect simulation time section between each place ^*as training sample data collection;

For example, in simulation time section (morning, 6:00 was to 12:00), repeatedly operating simulation system, in direct gathering simulation system, 6:00 is to the travel amount V between the each place of 12:00 ^*as the sample data collection of training, described travel amount V ^*be expressed as [v ₁ ^*v ₂ ^*v _n ^*] ^t, wherein, v _j ^*be j trip component, with the v in (1) formula _jmeaning is identical, and difference is only refer to the trip component in analogue system, can arrive by direct-detection, and v _jrefer to the trip component in reality, cannot direct-detection.This step is mainly to do the preparation in data for the sparse transform-based of follow-up training.

Step 2, using the training sample data collection collecting in described step 1 as input, the trip requirements vector V that training obtains between place specifies the sparse transform-based L (N × N dimension) of degree of rarefication, has appointment degree of rarefication so that the coordinate of the trip requirements vector V between described place under sparse transform-based L represents W (W that between described place, trip requirements vector V obtains after linear transformation V=LW in other words).Conventionally degree of rarefication is appointed as (m+2n), in W, is only had at the most (m+2n) individual element non-vanishing.

In this step, be specially for the training of described sparse transform-based: adopt online dictionary learning algorithm (Online Dictionary Learning) solving-optimizing problem:

$\underset{L,W}{\arg \min }{\left|\right|V^{*}-L\&CenterDot;W\left|\right|}_{l_2}+\mathrm{\&lambda;}{\left|\right|W\left|\right|}_{l_1},$

Wherein, L and W are variable to be optimized, V ^*certain the travel amount sample data collecting for step 1, for asking for the l of W ₁norm, for asking for V ^*the l of-LW ₂norm.

The basic ideas of above-mentioned optimization problem are to optimize another on the basis of fixing a variable, so iterate until restrain, and concrete steps are:

Step 201, initialization intermediary matrix variables A ₀=B ₀=0 (N × N dimension), makes iterations t=1, L=E, and wherein, E is unit matrix;

Step 202, concentrates and takes out sample from described training sample data

Step 203, adopts minimum angular convolution to return (Least Angle Regression) algorithm to calculate W _t:

$W_t=\underset{W\&Element;R^N}{\arg \min }\frac{1}{2}{\left|\right|V_t^{*}-L\&CenterDot;W\left|\right|}_{l_2}+\mathrm{\&lambda;}{\left|\right|W\left|\right|}_{l_1},$

Wherein, λ > 0 is regularization parameter, is generally determined by experience.λ is larger, considers that the degree of rarefication of separating is more when representing optimized, considers that mapping fault is fewer;

Step 204, utilizes sample with the W obtaining _tmiddle matrix variables A and B are upgraded:

$A_t=A_{t-1}+W_t\&CenterDot;W_t^T,$ $B_t=B_{t-1}+V_t^{*}\&CenterDot;W_t^T;$

Step 205, utilizes the intermediary matrix variables A after upgrading _tand B _tupgrade each row in L:

$\frac{1}{A_{\mathrm{jj}}}(b_j-L\&CenterDot;a_j)+l_j\&RightArrow;u_j,$ $\frac{1}{\max ({\left|\right|u_j\left|\right|}_2,1)}\&CenterDot;u_j\&RightArrow;l_j,$

Wherein, a _j, b _j, l _jbe respectively A _t, B _t, L j column vector, A _jjfor A _tin the element of the capable j of j row, u _jfor intermediate variable (1≤j≤N);

Step 206, whether training of judgement sample data is concentrated also has new data untreated, if so, makes t=t+1, returns to step 202; If not, the sparse transform-based L now obtaining is exported as training result.

Step 3: m is counted in the observation section of specifying according to user, calculates one by one every measurement row vector P that section is corresponding in analogue system road network _iwith the degree of coherence of described sparse transform-based L, choose m the measurement row vector composition measurement matrix P with minimum degree of coherence, using a corresponding m section as observation section collection X, and will measure matrix P, population constraint matrix Q, R composition observing matrix K.

This step is measured matrix P and is selected observation section collection X according to following steps design:

Step 301, if m > is c, wherein c is section number total in analogue system road network, makes X=φ, P=φ, (φ is empty set), calculating stops; Otherwise go to step 302;

Step 302, calculates corresponding measurement row vector and the degree of coherence of sparse transform-based L to each the section i in emulation road network, is specially:

Step 30201, by measurement row vector P corresponding to traffic allocation proportion structure section i _i;

Step 30202, calculates P _idegree of coherence μ with L _i(P _i, L):

${\mathrm{\&mu;}}_i(P_i,L)=\sqrt{N}\&CenterDot;\underset{1\&le;j\&le;N}{\max }|<P_i,l_j>|,$

Wherein, < P _i, l _j> represents P _iwith the inner product that the j of L is listed as, N is the dimension of sparse transform-based L;

Step 30203, preserves P _iwith corresponding μ _i(P _i, L);

Step 303, chooses m row vector composition of degree of coherence minimum and measures matrix P, the m bar section composition observation section collection X that a described m row vector is corresponding;

The measurement matrix P obtaining and population constraint matrix Q, R combination are obtained to observing matrix K, $K=\left[\begin{array}{c}P\\ Q\\ R\end{array}\right].$

Step 4, simulation time section is divided into multiple sense cycle (for example, to time period 6:00 to 12:00, taking 15 minutes as sense cycle, being divided into altogether 24 sense cycle), this step is by simulation time section section, thereby reaches the object of dynamic calculation trip requirements.

Get a sense cycle as current detection cycle (generally calculating successively according to time sequencing), the actual traffic amount in acquisition step 3 in definite m observation section obtain observing the volume of traffic of section i wherein α is average carrying number, 1≤i≤m; In gathering simulation system, population when this current sense cycle initial sum stops in each place, obtains total observation vector B simultaneously.

In prior art, generally adopt ground induction coil, infrared ray, the first-class several different methods of video camera to gather actual traffic amount, repeat no more concrete scheme here.If there is historical traffic data to be used.

Step 5, according to linear transformation V=LW, is rewritten as B=KLW by described observation equation B=KV, and wherein, W is the coordinate under base L to be asked.This step is in fact that V is compressed, then the coordinate of investigating under described sparse transform-based L represents W.

According to the total observation vector B calculating, (m+2n is capable, 1 row), (m+2n is capable for observing matrix K, N row), sparse transform-based L (N is capable, N row) and described observation equation calculate the optimum solution (N is capable, and 1 is listed as) of W.

The concrete method that minimizes L1 norm that adopts of this step is recovered W from described observation equation, and the optimum solution of definite mathematical programming problem is below (with W ^#represent optimum solution)

$W^{\#}=\underset{B=K\&CenterDot;L\&CenterDot;W}{\arg \min }{\left|\right|W\left|\right|}_{l_1}.$

The classic algorithm that minimizes L1 norm is a lot, repeats no more herein.

Step 6, calculates the trip requirements amount V:V=LW between each place in the current detection cycle ^#, wherein before L, determine W ^#be the optimum solution of trying to achieve in step 5, V is the trip requirements amount between each place in the current detection cycle;

Step 7, judges in simulation time section whether have next sense cycle, if so, returns to step 4, enters the calculating of next sense cycle; Otherwise, calculate and stop, exit this emulation.

For verifying the feasibility of trip requirements computing method provided by the present invention, choose near the road network in Milky Way sports center, reported in Tianhe district of Guangzhou as emulation experiment road network (as shown in Figure 2), adopt Institute of Automation, CAS complication system to manage the TransWorld traffic simulation software of researching and developing with control National Key Laboratory as emulation platform (interface as shown in Figure 3).Extract 137, main section, 89, crossing (containing empty crossing), 72 of playgrounds, simulation time section is set to 6:00 to 12:00 in the morning, chooses 71, observation section.In the situation that not adopting above-mentioned computing method, trip requirements relative error is 38.6%, and after adopting said method, trip requirements relative error is down to 26.1%.The accuracy that the visible trip requirements computing method based on compressed sensing technology proposed by the invention can significantly improve analogue system approaches actual traffic.

Above-described specific embodiment; object of the present invention, technical scheme and beneficial effect are further described; institute is understood that; the foregoing is only specific embodiments of the invention; be not limited to the present invention; within the spirit and principles in the present invention all, any amendment of making, be equal to replacement, improvement etc., within all should being included in protection scope of the present invention.

Claims (10)

1. the population trip requirements computing method based on compressed sensing technology, is characterized in that, the method comprises the following steps:

Step 1, based on analogue system, the travel amount V in multi collect simulation time section between each place ^*as training sample data collection;

Step 2, using the training sample data collection collecting in described step 1 as input, the trip requirements vector that training obtains between place is specified the sparse transform-based L of degree of rarefication, so that the coordinate of the trip requirements vector V between described place under sparse transform-based L represents that W has appointment degree of rarefication;

Step 3, m is counted in the observation section of specifying according to user, calculates one by one every measurement row vector P that section is corresponding in analogue system road network _iwith the degree of coherence of described sparse transform-based L, choose m the measurement row vector composition measurement matrix P with minimum degree of coherence, using a corresponding m section as observation section collection X, and will measure matrix P, population constraint matrix Q, R composition observing matrix K;

Step 4, is divided into multiple sense cycle by simulation time section, gets a sense cycle as the current detection cycle, the actual traffic amount in acquisition step 3 in definite m observation section obtain observing the volume of traffic of section i wherein α is average carrying number, 1≤i≤m; In gathering simulation system, population when this current sense cycle initial sum stops in each place, obtains total observation vector B simultaneously;

Step 5, according to linear transformation V=LW, is rewritten as B=KLW by observation equation B=KV; Calculate the optimum solution of W according to the total observation vector B, the observing matrix K that calculate, sparse transform-based L and described observation equation, i.e. W ^#;

Step 6, calculates the trip requirements amount V:V=LW between each place in the current detection cycle ^#;

Step 7, judges in simulation time section whether have next sense cycle, if so, returns to step 4, enters the calculating of next sense cycle; Otherwise, calculate and finish.

2. method according to claim 1, is characterized in that, in described step 2, described appointment degree of rarefication is m+2n, in W, only has at the most m+2n element non-vanishing, and n is playground number.

3. method according to claim 1, is characterized in that, in described step 2, for the training of described sparse transform-based L is further: adopt online dictionary learning Algorithm for Solving following formula:

$\underset{L,W}{\arg \min }{\left|\right|V^{*}-L\&CenterDot;W\left|\right|}_{l_2}+\mathrm{\&lambda;}{\left|\right|W\left|\right|}_{l_1},$

Wherein, L and W are variable to be optimized, V ^*certain the travel amount sample data collecting for step 1, represent to ask for the l of W ₁norm, represent to ask for V ^*the l of-LW ₂norm, λ >0 is regularization parameter.

4. method according to claim 3, is characterized in that, solution procedure is further comprising the steps:

Step 201, the intermediary matrix variables A of initialization N × N dimension ₀=B ₀=0, make iterations t=1, L=E, wherein, E is unit matrix;

Step 202, concentrates and takes out sample from described training sample data

Step 203, adopts minimum angle regression algorithm to calculate W _t:

$W_t=\underset{W\&Element;R^N}{\arg \min }\frac{1}{2}{\left|\right|V_t^{*}-L\&CenterDot;W\left|\right|}_{l_2}+\mathrm{\&lambda;}{\left|\right|W\left|\right|}_{l_1},$

Wherein, λ >0 is regularization parameter;

Step 204, utilizes sample with the W obtaining _tmiddle matrix variables A and B are upgraded:

A _t＝A _t-1+W _t·W _t ^T,B _t＝B _t-1+V _t ^*·W _t ^T；

Step 205, utilizes the intermediary matrix variables A after upgrading _tand B _tupgrade each row in L:

$\frac{1}{A_{\mathrm{jj}}}(b_j-L\&CenterDot;a_j)+l_j\&RightArrow;u_j,\frac{1}{\max ({\left|\right|u_j\left|\right|}_2,1)}\&CenterDot;u_j\&RightArrow;l_j,$

Wherein, a _j, b _j, l _jbe respectively A _t, B _t, L j column vector, A _jjfor A _tin the element of the capable j of j row, u _jfor intermediate variable, 1≤j≤N;

Step 206, whether training of judgement sample data is concentrated also has new data untreated, if so, makes t=t+1, returns to step 202; If not, the sparse transform-based L now obtaining is exported as training result.

5. method according to claim 1, is characterized in that, calculates every measurement row vector P that section is corresponding in analogue system road network in described step 3 _ibefore further comprising the steps of with the degree of coherence of described sparse transform-based L:

Judge whether m is greater than c, wherein c is section number total in analogue system road network, if make X=φ, and P=φ, φ is empty set, calculating stops, otherwise calculates every measurement row vector P that section is corresponding in analogue system road network _idegree of coherence with described sparse transform-based L.

6. method according to claim 2, is characterized in that,

Travel amount V described in described step 1 ^*be expressed as ［ v ₁ ^*v ₂ ^*v _n ^*? ^t, wherein, v _j ^*for V ^*j trip component, N is the dimension of sparse transform-based L;

In described step 3, measure row vector P _ibe expressed as:

P _i＝[p _i1?p _i2?…?p _iN]，

Wherein, p _ijrepresent corresponding trip component v _jallocation proportion on the i of section;

Population constraint matrix Q, R are expressed as:

Wherein,

Described observing matrix K is expressed as:

$K=\left[\begin{array}{c}P\\ Q\\ R\end{array}\right].$

7. method according to claim 1, is characterized in that, calculates every measurement row vector P that section is corresponding in analogue system road network in described step 3 _ifurther comprising the steps with the degree of coherence of described sparse transform-based L:

Step 30201, by measurement row vector P corresponding to traffic allocation proportion structure section i _i;

Step 30202, calculates P _idegree of coherence μ with L _i(P _i, L):

${\mathrm{\&mu;}}_i(P_i,L)=\sqrt{N}\&CenterDot;\underset{1\&le;j\&le;N}{\max }|<P_j,l_j>|,$

Wherein, <P _i, l _j> represents P _iwith the inner product that the j of L is listed as, N is the dimension of sparse transform-based L.

8. method according to claim 7, is characterized in that, further comprises and preserve P after described step 30202 _iwith corresponding μ _i(P _i, L) step.

9. method according to claim 1, is characterized in that, in described step 4, total observation vector B is expressed as:

$B=\left[\begin{array}{c}X\\ Y\\ Z\end{array}\right],$

Wherein, $X=\left[\begin{array}{c}{x}_1\\ x_2\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ x_m\end{array}\right],Y=\left[\begin{array}{c}{y}_1\\ y_2\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ y_n\end{array}\right],Z=\left[\begin{array}{c}{z}_1\\ z_2\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ z_n\end{array}\right],$ Y _ithe population of place i in the time of t=0, z _ibe place i in the population in t=T moment, n is playground number.

10. method according to claim 1, is characterized in that, described step 5 further comprises the optimum solution W that adopts the method that minimizes L1 norm to calculate W from described observation equation ^#.