CN105513359A - Method for estimating city expressway traffic states based on mobile detection of smartphones - Google Patents

Abstract

The invention discloses a method for estimating city expressway traffic states based on mobile detection of smartphones, wherein a city expressway cell transmission model is built firstly, a observing network is built by adopting a smartphone to rapidly detect parameters, then a state space model based on a lighthill-whitham-richards (LWR) traffic flow model is designed, a traffic state and a boundary flux are synchronously estimated by utilizing three-step type recursive filters algorithm, then are coalesced with upstream and downstream subsection boundary flux by adopting a weighted average algorithm, and traffic parameter estimation is upgraded, thereby achieving real-time distributed estimation of city expressway network traffic state. The method for estimating the city expressway traffic states based on the mobile detection of the smartphones can collect average speed information of vehicles on any time and space positions of a loop, enables traffic estimation not to be restrained by the position of a detector, can achieve synchronization estimation of traffic density and boundary flux by designing a state-space model and a three-step recursive filter, and achieves the problem of large range expressway network traffic estimation by being coalesced with subsection boundary flux, reduces model order, and improves efficiency of algorithm.

Description

A kind of urban expressway traffic method for estimating state based on smart mobile phone mobility detect

Technical field

The invention belongs to intelligent transport system field, relate to traffic data collection technology, Cell Transmission Model, single order macroscopic traffic flow and Distribution fusion algorithm.The present invention combines based on the traffic data collection technology of smart mobile phone, single order macroscopic traffic flow, Kalman filtering and Distribution fusion algorithm, realizes city expressway distributed traffic state and estimates in real time.

Background technology

Along with socioeconomic development, the quickening of urbanization, motorize speed, the problems such as traffic congestion, traffic hazard, environmental pollution, energy shortage have become the common issue that countries in the world face.By implementing effective traffic control and road traffic congestion is alleviated in induction, to improve the research of traffic administration service level increasingly mature.And the prerequisite implementing traffic control is estimate the real-time traffic states of road with basis.

Traffic behavior is estimated mainly to be divided into and is estimated based on the traffic behavior estimation of macroscopic traffic flow and the traffic behavior based on data.Traffic flow is regarded as the fluid media (medium) be made up of vehicle by macroscopic traffic flow, concern be the comprehensive average behavior of entire vehicle, mainly pay close attention to average density, the macrovariable that average velocity etc. are average.Estimation problem based on macroscopic traffic flow obtains extensive research in nearly ten years in the past, typically comprises: utilize particle filter come estimating speed expansion random Cell Transmission Model SCTM, utilize EKF to estimate traffic behavior, to estimate based on the traffic behavior of continuous Monte carlo algorithm, describe traffic behavior etc. based on single order switching model.Although the various algorithm based on macroscopic traffic flow is suggested, all there is respective problem, cannot obtain as being subject to constraint, the boundary flux of DETECTION OF TRAFFIC PARAMETERS device position, can not the problems such as different roads being applicable to.Traffic behavior estimation based on traffic data is the important foundation of traffic control and induction.For the Forecasting Methodology based on traffic data, main based on regretional analysis in early days, but Forecasting Methodology is only limited to linear analysis, cannot react uncertainty and the nonlinear characteristic of traffic system.Along with the development of the information processing technology, some methods that are adaptable, that do not have fixed model are applied to traffic data short-term prediction field, as the Forecasting Methodology etc. of artificial neural network, knowledge-based systems, the prediction of traffic data seasonal effect in time series is obtained and develops better.

The real-time estimation developing into traffic behavior of traffic data collection technology provides the foundation traffic data.Existing traffic data collection technology has two classes: fixed detecting device and motion type data acquisition.Fixed detecting device can gather the traffic datas such as flow, speed, density, but it is huge to there is image data amount, and information processing difficulty is comparatively large, cost of investment and maintenance cost higher, the problems such as coverage rate is low.What current portable traffic data collection technology employing was maximum is floating vehicle data acquisition technology, based on GPS positioning system, mobile phone positioning system, vehicle dynamic information is sent to Floating Car information processing centre in real time.But due to Floating Car quantity very few, floating car data is not enough to provide telecommunication flow information and traffic flow density.Along with the raising of mobile phone popularity rate, based on the traffic data collection technology of mobile phone location because cost is low and the features such as efficiency is high, it is real-time to have, wide coverage, day by day come into one's own.

The principal element affecting urban expressway traffic state estimation comprises 2 points, one is boundary flux problem, traffic flow model is dynamically determined jointly by system current traffic condition and boundary flux, and adopting based on the traffic data collection technology of data in mobile phone, the boundary flux on sub-section is normally unknown; Two is system model order problems, and urban freeway network is divided into up to a hundred cellulars usually, and when application Kalman filtering algorithm is estimated, system model order is higher, and calculated amount is comparatively large, is difficult to requirement of real time.

Therefore, the present invention proposes a kind of new Expressway Traffic method for estimating state.First city expressway Cell Transmission Model is set up; Secondly observation grid is built by smart mobile phone velocity measuring parameter; Then design the state-space model based on single order macroscopic view LWR traffic flow model, utilize three-wave-length Recursive Filter Algorithm Using synchronously to estimate traffic behavior and boundary flux; Then adopt average weighted algorithm fusion upper and lower alien section boundary flux, reduce model order, realize the real-time distributed estimation of urban freeway network traffic behavior.

Summary of the invention

The present invention proposes a kind of urban expressway traffic method for estimating state based on smart mobile phone mobility detect newly, observation grid can be built by smart mobile phone velocity measuring parameter, single order macroscopic view LWR traffic flow model and three-wave-length Recursive Filter Algorithm Using synchronously estimate traffic behavior and section boundary flux of coming in and going out, and realize through street distributed traffic state estimation on a large scale by blending algorithm.

The object of the invention is to not increase on the basis of information collecting device cost, improve accuracy and the real-time of urban expressway traffic state estimation.

The technical solution adopted for the present invention to solve the technical problems comprises four parts, this four part is respectively city expressway MCTM model modeling part A, based on the traffic data collection part B of smart mobile phone, the synchronous estimating part C of traffic behavior and boundary flux, through street distributed traffic state estimation part D on a large scale; This four part relations as shown in Figure 1, wherein:

Shown in A, part is city expressway MCTM model.According to MCTM model, through street is divided into some sub-sections, every sub-section is made up of different cellulars, and this road division methods is the basis that traffic data collection and traffic behavior are estimated, reduces model order, shortcut calculation.

Part shown in B is the traffic data collection based on smart mobile phone.This method utilizes the smart mobile phone in vehicle to gather crucial space-time position place vehicular traffic parameter, calculates vehicle travel speed, then calculates vehicle average velocity in cellular, builds observation grid, estimate traffic behavior with average velocity in cellular.

Part shown in C is traffic behavior and the synchronous estimation of boundary flux.This method design based on the state-space model of LWR traffic flow model, and designs a kind of three-wave-length Recursive Filter Algorithm Using, and using boundary flux as Unknown worm, in antithetical phrase section, the traffic behavior of some cellulars and boundary flux are synchronously estimated.

Part shown in D is through street distributed traffic state estimation on a large scale.This method is on the basis that the traffic behavior proposed and boundary flux are synchronously estimated, large-scale through street is divided into some sub-sections, then every sub-road section traffic volume state and boundary flux are estimated, utilize Weighted Average Algorithm, the boundary flux in adjacent two sub-sections is merged, thus estimates the traffic behavior of through street on a large scale.

The city expressway MCTM model of part A carries out discretize to the room and time of through street, the traffic data of part B is gathered by smart mobile phone end, the data of smart mobile phone end collection are carried out traffic behavior and boundary flux as the observational variable of C part to through street and are synchronously estimated, finally draw the distributed traffic state estimation of through street on a large scale of D part.

Compared with prior art, the present invention has following beneficial effect.

1, select smart mobile phone as the medium of traffic data collection.Because smart mobile phone market share approaches 100%, and the standard configuration of smart mobile phone on sale comprises GPS module, and this just lays the foundation for utilizing mobile phone to carry out traffic data collection.The features such as through street vehicle flowrate is huge, and detection time is long cause Information Monitoring huge, and traditional traffic behavior monitoring reflects traffic behavior by fixed detector or monitoring camera-shooting, there is transmission and store data volume huge, reflects the problems such as delayed.And the present invention is by mobile phone traffic data collection, only need gather the vehicle speed information of limited quantity in the corresponding cellular of synchronization, calculate average velocity, as the observational variable of traffic flow model.Possess acquisition cost low, to transmit and to store data volume little, the features such as acquisition range is wide, real-time.Be applicable to the traffic data collection of road network on a large scale, make traffic behavior estimate the constraint of no longer examined device position.

2, the state-space model based on LWR traffic flow model and three-wave-length Recursive Filter Algorithm Using is selected.The existing research estimated about traffic behavior is often carried out when boundary flux is known, but boundary flux required when actual conditions often cannot detect estimation, this has just limited to the estimation problem of traffic behavior.The model that the present invention proposes and filtering algorithm solve the problem brought by boundary flux well.In a model, boundary flux, not as known quantity, but as Unknown worm, is synchronously estimated with traffic density when using wave filter.The present invention emphasizes that a Negotiation speed detects data to monitor the potential application of whole urban expressway traffic system state, does not have traditional data to detect, such as boundary flux.The method has the ability to estimate the traffic parameter in any section of urban traffic network.

3, city expressway is divided into some sub-sections, traffic behavior estimation is carried out in first antithetical phrase section, then carries out information fusion to each sub-section boundary flux.Large scale road network traffic flow model order is high, and algorithm is complicated, and is difficult to working time meet the requirement of traffic control to real-time.The present invention adopts the traffic behavior of 3 or 4 cellulars in state-space model and three-wave-length Recursive Filter Algorithm Using antithetical phrase section and boundary flux synchronously to estimate, again the boundary flux in adjacent two sections is merged, fusion method adopts average weighted method, thus obtains the traffic behavior of complete loop.Be that some sub-sections are estimated by longer pavement section, greatly reduce the order of model, improve efficiency of algorithm.

Accompanying drawing explanation

Fig. 1 is based on the urban expressway traffic state estimation schematic diagram of smart mobile phone mobility detect.

Fig. 2 tri-ring through street cellular divides schematic diagram.

Fig. 3 mobile phone traffic data collection technical limit spacing velocity information schematic diagram.

Fig. 4 vehicle position information sampling schematic diagram.

Fig. 5 data sampling and processing flow chart.

Fig. 6 section is divided into N number of cellular.

Fig. 7 city expressway distributed traffic state estimation frame diagram.

Fig. 8 adjacent sub-section boundary flux merges.

Fig. 9 boundary flux blending algorithm process flow diagram.

Embodiment

Below in conjunction with accompanying drawing, the present invention is further described.

Part A sets up city expressway MCTM model, and (for ring through street, Beijing three), on the basis of original Cell Transmission Model CTM, adopts the Cell Transmission Model MCTM improved to carry out modeling.

CTM is using the vehicle number in each cellular as the state variable of cellular, and MCTM is using the traffic density in cellular as state variable, this improvement eliminates CTM must equal restriction for the length of each cellular, the cellular of road is divided more flexible, urban expressway traffic stream can be described with less cellular, thus greatly reduce the dimension of cellular state variable.

According to MCTM model by room and time discretize.Spatial discretization and cellular division methods, be divided into the cellular that multiple length does not wait, will be divided into constant duration simultaneously the time by through street, as time step, and meet in a time step, vehicle is less than cellular length with the distance that maximal rate travels, i.e. L _i≤ v _fΔ t, L _ifor cellular length, Δ t is the time interval, v _ffor free travel speed.

Through street cellular division methods is: cellular length is L _irice, the time interval is Δ t second, and concrete partiting step is as follows:

(1) city expressway is closed annular, in order to simplified model, only considers the road of one-way traffic, i.e. anticlockwise through street.

(2) according to MCTM model by spatial discretization, through street is divided into length not wait several cellulars.Cellular length meets L _i≤ v _fΔ t, sampling time Δ t (in the present embodiment, Δ t=30 second).Cellular numbering is set to i=1,2 ..., N, N value positive integer.

(3) through street is closed annular, Entrance ramp and exit ramp are main input and output, therefore the general border by cellular is arranged on main entrance or exit ramp place, and the turnover flow of entrance and exit ring road is incorporated to cellular boundary flux, shortcut calculation.

By above cellular division methods, for ring through street, Beijing three, be cellular by this pavement section, division result as shown in Figure 2.Because observational variable of the present invention derives from mobile phone traffic data, therefore cellular divides the constraint of no longer examined device position, and the concrete dividing mode of cellular is flexibly, has different dividing mode according to above-mentioned cellular division methods.

The cellular dividing mode of this method is: each cellular length is L _i(i=1,2 ..., 36), each cellular length value is as follows: (unit is km)

1.27,1.45,0.87,1.20,1.18,1.28,1.20,1.65,2.06,2.11,1.43,1.07,1.24,0.65,1.68,1.38,1.18,2.02,1.22,1.52,0.87,1.69,1.47,1.87,1.58,0.97,1.82,1.82,1.45,1.2,0.92,1.25,0.68,0.91,1.43,1.03

Entrance ramp is numbered OR1-OR11, and exit ramp is numbered FR1-FR10.

Set up traffic flow transmission relation between cellular, the delivery flow rate between cellular is:

$y=min\{n_i,Q_i^O,Q_{i+1}^I,{\mathrm{\δ}}_{i+1}(N_{i+1}-n_{i+1})\}$

In formula: y is the outflow vehicle number of cellular i reality in the Δ t time, it is also the inflow vehicle number of cellular i+1 reality in the Δ t time; n _iand n _i+1be respectively existing vehicle number in cellular i and cellular i+1; the maximum vehicle number that can flow into cellular i+1 being respectively that cellular i in the Δ t time can flow out; N _i+1the maximum vehicle number that can hold in cellular i+1; δ _i+1for the free stream velocity of cellular i+1 and the ratio of backward wave speed, i.e. δ=ω/ν.

Utilize above first intercellular traffic flow transmission relation, emulate with the boundary value of the real-time magnitude of traffic flow as section, the traffic datas such as the delivery flow rate between the vehicle number in the different cellular in this section, cellular can be obtained, consistent with actual traffic data, show that MCTM model can well describe the traffic behavior of through street.

So far, the room and time discretize of through street completes, for traffic data collection, structure observation grid and traffic behavior are estimated to lay a solid foundation.

Part B is the traffic data collection based on smart mobile phone.The present invention is observational variable with car speed, estimates traffic behavior.Vehicle speed information utilizes mobile phone traffic data collection technical limit spacing.

Vehicle position information is obtained by the smart mobile phone in vehicle, location information carries out time-sampling at equal intervals, positional information is sent to background data server by mobile communications network, basic data is arranged and analyzes, calculate vehicle average velocity in vehicle travel speed and cellular, estimate as traffic behavior.As shown in Figure 3.

Mobile phone location adopts A-GPS location technology, obtains vehicle position information, carries out time-sampling at equal intervals to vehicle position information, and Fig. 4 is shown in by vehicle position information sampling schematic diagram.Data sampling and treatment scheme as follows: (see Fig. 5).Gather all positional informations that can provide the vehicle t of traffic data in cellular, after sampling time Δ t second, gather t+1 moment vehicle position information, if tested vehicle t+1 moment and t are in same cellular, vehicle travel speed can calculate with following formula:

v _α＝l _α/Δt

L in formula _αfor the alternate position spike in the sampling period, Δ t is the sampling time.

Otherwise reject this vehicle data.

Obtain β in each cellular _ithe travel speed of car, in each cellular, the average velocity of vehicle can calculate with following formula:

$v_i\left(t\right)=<v_{\mathrm{\&alpha;}}>=\frac{1}{{\mathrm{\&beta;}}_i}\underset{\mathrm{\&alpha;}=1}{\overset{{\mathrm{\&beta;}}_i}{\&Sigma;}}{v}_{\mathrm{\&alpha;}}$

In each cellular, the average velocity of vehicle is used as the observational variable of traffic flow model, can build observation grid.

Traffic data collection method of the present invention may cause in certain cellular because of not having vehicle in certain cellular or do not meet data acquisition conditions cannot collect velocity information, can only obtain the velocity information of part cellular thus.And utilize the synchronous estimation method of traffic density of the present invention and boundary flux (hereafter describing in detail), then do not need the velocity information gathering whole cellular, suppose that N number of cellular section needs the velocity information of M cellular, M is less than N, and only needs the number being greater than Unknown worm.

So far, the observational variable needed for traffic behavior estimation, the velocity information of each cellular can be obtained by said method.

C part is the synchronous estimation of traffic behavior and boundary flux.The present invention's design based on the state-space model of LWR traffic flow model, and designs a kind of three-wave-length Recursive Filter Algorithm Using, and using boundary flux as Unknown worm, in antithetical phrase section, the traffic behavior of some cellulars and boundary flux are synchronously estimated.

First the present invention devises the state-space model of city expressway based on LWR traffic flow model.

The system equation of city expressway state-space model obtains as follows:

Single order macroscopic view LWR traffic flow model, describe locus x by vehicle conservation equation, the traffic density ρ (x, t) at time t place is as follows with magnitude of traffic flow q (x, t) relation:

$\frac{\&part;\mathrm{\&rho;}}{\&part;t}+\frac{\&part;q}{\&part;x}=0$

It is as follows that average overall travel speed meets flow-density relations:

$v=\frac{q}{\mathrm{\&rho;}}$

Application Godunov Finite Element Method, the above-mentioned vehicle conservation equation of numerical solution and flow-density relations.Spatially, section is divided into multiple cellular, and as shown in Figure 6, this border, section input and output flow is unknown, is respectively the vehicle flowrate that vehicle sailed and rolled away from cellular i into respectively.

Suppose that cellular length is L _i, i=1 ..., n, time sampling interval is Δ t, and according to Courant-Friedrichs-Lewy condition, when cellular length meets the following conditions, numerical solution is stablized

L _i≤v _fΔt

By space-time discretize, LWR model can be expressed as difference equation

${\mathrm{\&rho;}}_i(k+1)={\mathrm{\&rho;}}_i\left(k\right)+\frac{\mathrm{\&Delta;}t}{L_i}\left(q_i^{in}\right(k)-q_i^{out}(k\left)\right)$

ρ in formula _ik () represents the traffic density in current time cellular i, that vehicle sails and roll away from the flow of cellular i at [k Δ t, (k+1) Δ t] in the time interval, the rolling flow away from and can sail flow into represent with downstream cellular of upstream cellular:

$q_i^{out}\left(k\right)=q_{i+1}^{in}\left(k\right),i=1,\mathrm{...},N-1$

Flow determined by relation between supply and demand, S _ik () represents downstream cellular supply, D _i-1k () represents upstream cellular demand:

$q_i^{in}\left(k\right)=\min \{D_{i-1}\left(k\right),S_i\left(k\right)\}$

Relation between supply and demand is:

D _i(k)＝v(ρ _i)ρ _i,S _i(k)＝q _max,ρ _i(k)≤ρ _c

D _i(k)＝q _max,S _i(k)＝v(ρ _i)ρ _i,ρ _i(k)＞ρ _c

To N number of cellular, sail into roll discharge relation and flow relation between supply and demand away from according to above-mentioned difference equation, upstream and downstream cellular, the system equation of the state-space model system of city expressway can be summarized as:

ρ ^T(k+1)＝F(ρ ^T(k))+G ^Td(k)

ρ in formula ^t=[ρ ₁... ρ _n] be traffic density in each cellular of current time, be border input and output flows, F represents the funtcional relationship of above-mentioned difference equation Midst density and flow,

$G^T=\left[\begin{array}{cccc}0& 0& \mathrm{...}& -\frac{\mathrm{\&Delta;}t}{L_{n_j,j}}\\ \frac{\mathrm{\&Delta;}t}{L_{1,j}}& 0& \mathrm{...}& 0\end{array}\right].$

The observation equation of city expressway state-space model obtains as follows:

The present invention adopts smart mobile phone to gather traffic data, only needs the velocity information of vehicle in collecting part cellular, and does not need to gather boundary flux information, just can estimate traffic behavior.

According to the speed V (i, t) that part B method gathers, the traffic density of current cellular i can be described.Mobile phone gathers traffic data and provides velocity information set y (k)=[v ₁... v _m] ^t, owing to only needing the velocity information knowing part cellular, suppose that N number of cellular section needs the velocity information of M cellular (M is less than N, and only needs the number being greater than Unknown worm).

Application speed-density equation is as follows:

$v=\left\{\begin{array}{cc}\frac{v_c-v_f}{{\mathrm{\&rho;}}_c}\mathrm{\&rho;}+v_f,& \mathrm{\&rho;}\&le;{\mathrm{\&rho;}}_c\\ \frac{{\mathrm{\&rho;}}_c{v}_c}{{\mathrm{\&rho;}}_J-{\mathrm{\&rho;}}_c}(\frac{{\mathrm{\&rho;}}_J}{\mathrm{\&rho;}}-1),& \mathrm{\&rho;}>{\mathrm{\&rho;}}_c\end{array}\right.$

The non-linear observation equation model setting up city expressway state-space model is as follows:

y(k)＝H(ρ(k))

In formula, H describes above-mentioned speed-density equation relation.

So far, the state-space model devising city expressway based on LWR traffic flow model has been set up, as follows:

ρ ^T(k+1)＝F(ρ ^T(k))+G ^Td(k)

y(k)＝H(ρ(k))

Secondly, the present invention devises a kind of three-wave-length Recursive Filter Algorithm Using, can synchronously estimate traffic behavior and boundary flux by the state-space model of city expressway.

For discrete-time system:

x(t+1)＝A(t)x(t)+Gd(t)

y(t)＝C(t)x(t)

In formula, x (t) is state variable, and d (t) is Unknown worm, and y (t) is observed parameter.

Nonlinear relationship F and H, according to EKF, solves as the matrix A of above-mentioned discrete system and C at the Jacobian matrix of current time, namely by the present invention

$A\left(k\right)=\frac{\&part;F}{\&part;\mathrm{\&rho;}}{|}_{\mathrm{\&rho;}\left(k\right|k)}$

$C\left(k\right)=\frac{\&part;H}{\&part;\mathrm{\&rho;}}{|}_{\mathrm{\&rho;}(k+1|k)}$

The present invention devises a kind of three-wave-length Recursive Filter Algorithm Using and carries out compound Weibull process to system state and Unknown worm and boundary flux, and step is as follows:

The first step: estimated state variable.By the measured value in k moment, estimate traffic subsequent time traffic density ρ (k+1|k).

ρ(k+1|k)＝F(ρ(k|k))

Second step: the unknown boundary flux of pre-estimation

$\hat{d}\left(k\right)=M\left(k\right)\left(y\right(k)-C(k\left)\mathrm{\&rho;}\right(k+1|k\left)\right)$

$\stackrel{\&OverBar;}{\mathrm{\&rho;}}(k+1|k+1)=\mathrm{\&rho;}(k+1|k)+G\hat{d}\left(k\right)$

3rd step: more new state.According to the status predication value ρ (k+1|k) of the first step and the Unknown inputs value of second step obtain optimum state variable and estimate ρ (k+1|k+1).

$\mathrm{\&rho;}(k+1|k+1)=\stackrel{\&OverBar;}{\mathrm{\&rho;}}(k+1|k+1)+K\left(k\right)\left(y\right(k)-C(k\left)\stackrel{\&OverBar;}{\mathrm{\&rho;}}\right(k+1|k+1\left)\right)$

Gain matrix M (k) in formula, K (k) can be calculated by covariance:

P(k+1|k)＝A(k)P(k|k)A ^T(k)+Q(k)

$M\left(k\right)={\left(D^T\right(k\left)\tilde{R}\right(k\left)D\right(k\left)\right)}^{-1}{D}^T\left(k\right){\tilde{R}}^{-1}\left(k\right)$

$K\left(k\right)=P(k+1|k)C\left(k\right){\tilde{R}}^{-1}\left(k\right)$

$\tilde{R}\left(k\right)=C\left(k\right)P(k+1|k)C^T\left(k\right)+R\left(k\right)$

P(k+1|k+1)＝(1-M(k)C(k))P(k+1|k)

In formula, the error co-variance matrix that D (k)=C (k) G, Q (k) is state equation, the error co-variance matrix that R (k) is observation equation.

So far, traffic density and the boundary flux of each cellular synchronously can be estimated by above-mentioned state-space model and filtering algorithm.

D part merges boundary flux, realizes through street distributed traffic state estimation on a large scale.The present invention estimates to realize large scale traffic network, according to smart mobile phone traffic data collection situation, through street is divided into n sub-section, every sub-section comprises 3-4 cellular, by the velocity information of the corresponding cellular of mobile phone traffic data collection technical limit spacing, application C partial state space model and filtering algorithm are synchronously estimated every sub-road section traffic volume state and boundary flux, then the boundary flux in adjacent two sub-sections is merged, realize distributed traffic state estimation, thus obtain the traffic behavior of whole loop, as shown in Figure 7.

The division methods of the through street cellular of the present invention according to part A, for ring through street, Beijing three, becomes 36 cellulars by this pavement section, now choose wherein two adjacent sub-sections and carry out distributions estimation, as shown in Figure 8, cellular 6,7,8 is sub-section one, cellular 9,10,11 is sub-section two, the input flow rate of the cellular 9 that the delivery rate of the cellular 8 that sub-section one is estimated and sub-section two are estimated is same boundary flux, merges this boundary flux.

Fusion method of the present invention adopts average weighted method, and weight is the core of Weighted Average Algorithm.This method utilizes Lagrange multiplier to calculate weighted value.In order to obtain reliable weighted value r _m, suppose following condition:

(1) m kind data source does not have systematic error;

(2) variance of a random variable θ _mknown;

(3) error in different pieces of information source is uncorrelated.

Suppose that the estimated value of all types data is x _m, error variance X (x, the t)=∑ of weighted arithmetic mean _mr _mx _mcan be by provide, wherein ∑ _mr _m=1 must meet.Target is by changing weight r _mor weight vectors make error minimize.Thus the constrained optimization problem obtained below:

Minimize objective function:

Constraint condition: ∑ _mr _m=1

The optimal problem of belt restraining can use method of Lagrange multipliers to solve, and final weighted value is as follows:

$r_m=\frac{\frac{1}{{\mathrm{\&theta;}}_m}}{\underset{m^{\&prime;}}{\mathrm{\&Sigma;}}\frac{1}{{\mathrm{\&theta;}}_{m^{\&prime;}}}}$

The step of boundary flux fuse information is as follows: (algorithm flow chart as shown in Figure 9)

The first step: utilize the filtering algorithm antithetical phrase section one of above-mentioned C part to carry out one-step prediction state variable calculated gains matrix K ₁, M ₁, calculate the output estimation value of cellular 8

Second step: use the same method and obtain the one-step prediction state variable in sub-section two gain matrix K ₁, M ₁, and the input estimated value of cellular 9

3rd step: utilize the weighted value r that above-mentioned Weighted Average Algorithm obtains _m, right with be weighted and on average obtain $\hat{d}=r_1{\hat{d}}_1+r_2{\hat{d}}_2.$

4th step: utilize upgrade state variable estimated value with calculate the state variable ρ in each sub-section more respectively ₁and ρ (k+1|k+1) ₂(k+1|k+1).

Data fusion is carried out to the upper delivery rate in a sub-section and the input flow rate in next son section, reduces evaluated error, and be that some sub-sections are estimated by longer pavement section, greatly reduce the order of model, improve efficiency of algorithm.

C part describing method is used to estimate the traffic behavior in whole sub-section, through street and boundary flux respectively, apply D part describing method again the boundary flux estimated value in adjacent sub-section is merged, renewal traffic parameter is estimated, realizes the distributed traffic state estimation of whole section of through street.

Claims (9)

1. the urban expressway traffic method for estimating state based on smart mobile phone mobility detect, it is characterized in that: this four part is respectively city expressway MCTM model modeling part A, based on the traffic data collection part B of smart mobile phone, the synchronous estimating part C of traffic behavior and boundary flux, through street distributed traffic state estimation part D on a large scale; Wherein,

Shown in A, part is city expressway MCTM model; According to MCTM model, through street is divided into some sub-sections, every sub-section is made up of different cellulars, and this road division methods is the basis that traffic data collection and traffic behavior are estimated, reduces model order, shortcut calculation;

Part shown in B is the traffic data collection based on smart mobile phone; This method utilizes the smart mobile phone in vehicle to gather crucial space-time position place vehicular traffic parameter, calculates vehicle travel speed, then calculates vehicle average velocity in cellular, builds observation grid, estimate traffic behavior with average velocity in cellular;

Part shown in C is traffic behavior and the synchronous estimation of boundary flux; This method design based on the state-space model of LWR traffic flow model, and designs a kind of three-wave-length Recursive Filter Algorithm Using, and using boundary flux as Unknown worm, in antithetical phrase section, the traffic behavior of some cellulars and boundary flux are synchronously estimated;

Part shown in D is through street distributed traffic state estimation on a large scale; This method is on the basis that the traffic behavior proposed and boundary flux are synchronously estimated, large-scale through street is divided into some sub-sections, then every sub-road section traffic volume state and boundary flux are estimated, utilize Weighted Average Algorithm, the boundary flux in adjacent two sub-sections is merged, thus estimates the traffic behavior of through street on a large scale;

The city expressway MCTM model of part A carries out discretize to the room and time of through street, the traffic data of part B is gathered by smart mobile phone end, the data of smart mobile phone end collection are carried out traffic behavior and boundary flux as the observational variable of C part to through street and are synchronously estimated, finally draw the distributed traffic state estimation of through street on a large scale of D part.

2. a kind of urban expressway traffic method for estimating state based on smart mobile phone mobility detect according to claim 1, it is characterized in that: the method for based on MCTM model, city expressway being carried out to cellular division, partiting step comprises:

(1) city expressway is closed annular, in order to simplified model, only considers the road of one-way traffic, i.e. anticlockwise through street;

(2) according to MCTM model by spatial discretization, through street is divided into length not wait several cellulars; Cellular length meets L _i≤ v _fΔ t, sampling time Δ t=30 second; Cellular numbering is set to i=1,2 ..., N;

(3) through street is closed annular, Entrance ramp and exit ramp are main input and output, therefore the general border by cellular is arranged on main entrance or exit ramp place, and the turnover flow of entrance and exit ring road is incorporated to cellular boundary flux, shortcut calculation.

3. the urban expressway traffic method for estimating state based on smart mobile phone mobility detect according to claim 1, is characterized in that: set up traffic flow transmission relation between cellular, the delivery flow rate between cellular is:

$y=min\{n_i,Q_i^O,Q_{i+1}^I,{\mathrm{\&delta;}}_{i+1}(N_{i+1}-n_{i+1})\}$

In formula: y is the outflow vehicle number of cellular i reality in the Δ t time, it is also the inflow vehicle number of cellular i+1 reality in the Δ t time; n _iand n _i+1be respectively existing vehicle number in cellular i and cellular i+1; the maximum vehicle number that can flow into cellular i+1 being respectively that cellular i in the Δ t time can flow out; N _i+1the maximum vehicle number that can hold in cellular i+1; δ _i+1for the free stream velocity of cellular i+1 and the ratio of backward wave speed, i.e. δ=ω/ν;

Utilize above first intercellular traffic flow transmission relation, emulate with the boundary value of the real-time magnitude of traffic flow as section, the traffic datas such as the delivery flow rate between the vehicle number in the different cellular in this section, cellular can be obtained, consistent with actual traffic data, show that MCTM model can well describe the traffic behavior of through street;

So far, the room and time discretize of through street completes, for traffic data collection, structure observation grid and traffic behavior are estimated to lay a solid foundation.

4. the urban expressway traffic method for estimating state based on smart mobile phone mobility detect according to claim 1, is characterized in that: part B is the traffic data collection based on smart mobile phone; This method is observational variable with car speed, estimates traffic behavior; Vehicle speed information utilizes mobile phone traffic data collection technical limit spacing;

Vehicle position information is obtained by the smart mobile phone in vehicle, location information carries out time-sampling at equal intervals, positional information is sent to background data server by mobile communications network, basic data is arranged and analyzes, calculate vehicle average velocity in vehicle travel speed and cellular, estimate as traffic behavior;

Mobile phone location adopts A-GPS location technology, obtains vehicle position information, carries out time-sampling at equal intervals to vehicle position information; Data sampling and treatment scheme as follows: gather all positional informations that the vehicle t of traffic data can be provided in cellular, after sampling time Δ t second, gather t+1 moment vehicle position information, if tested vehicle t+1 moment and t are in same cellular, vehicle travel speed can calculate with following formula:

v _α＝l _α/Δt

L in formula _αfor the alternate position spike in the sampling period, Δ t is the sampling time;

Otherwise reject this vehicle data;

Obtain β in each cellular _ithe travel speed of car, in each cellular, the average velocity of vehicle can calculate with following formula:

$v_i\left(t\right)=<v_{\mathrm{\&alpha;}}>=\frac{1}{{\mathrm{\&beta;}}_i}\underset{\mathrm{\&alpha;}=1}{\overset{{\mathrm{\&beta;}}_i}{\&Sigma;}}{v}_{\mathrm{\&alpha;}}$

In each cellular, the average velocity of vehicle is used as the observational variable of traffic flow model, can build observation grid.

5. the urban expressway traffic method for estimating state based on smart mobile phone mobility detect according to claim 1, it is characterized in that: the traffic data collection method of this method may cause in certain cellular because of not having vehicle in certain cellular or do not meet data acquisition conditions and cannot collect velocity information, can only obtain the velocity information of part cellular thus; And utilize the traffic density of this method and the synchronous estimation method of boundary flux, then do not need the velocity information gathering whole cellular, suppose that N number of cellular section needs the velocity information of M cellular, M is less than N, and only needs the number being greater than Unknown worm;

So far, the observational variable needed for traffic behavior estimation, the velocity information of each cellular can be obtained by said method.

6. the urban expressway traffic method for estimating state based on smart mobile phone mobility detect according to claim 1, is characterized in that: C part is the synchronous estimation of traffic behavior and boundary flux; This method design based on the state-space model of LWR traffic flow model, and designs a kind of three-wave-length Recursive Filter Algorithm Using, and using boundary flux as Unknown worm, in antithetical phrase section, the traffic behavior of some cellulars and boundary flux are synchronously estimated;

First this method devises the state-space model of city expressway based on LWR traffic flow model;

The system equation of city expressway state-space model obtains as follows:

Single order macroscopic view LWR traffic flow model, describe locus x by vehicle conservation equation, the traffic density ρ (x, t) at time t place is as follows with magnitude of traffic flow q (x, t) relation:

$\frac{\&part;\mathrm{\&rho;}}{\&part;t}+\frac{\&part;q}{\&part;x}=0$

It is as follows that average overall travel speed meets flow-density relations:

$v=\frac{q}{\mathrm{\&rho;}}$

Application Godunov Finite Element Method, the above-mentioned vehicle conservation equation of numerical solution and flow-density relations; Spatially, section is divided into multiple cellular, and this border, section input and output flow is unknown, is respectively the vehicle flowrate that vehicle sailed and rolled away from cellular i into respectively;

Suppose that cellular length is L _i, i=1 ..., n, time sampling interval is Δ t, and according to Courant-Friedrichs-Lewy condition, when cellular length meets the following conditions, numerical solution is stablized

L _i≤v _fΔt

By space-time discretize, LWR model can be expressed as difference equation

${\mathrm{\&rho;}}_i(k+1)={\mathrm{\&rho;}}_i\left(k\right)+\frac{\mathrm{\&Delta;}t}{L_i}\left(q_i^{in}\right(k)-q_i^{out}(k\left)\right)$

ρ in formula _ik () represents the traffic density in current time cellular i, that vehicle sails and roll away from the flow of cellular i at [k Δ t, (k+1) Δ t] in the time interval, the rolling flow away from and can sail flow into represent with downstream cellular of upstream cellular:

$q_i^{out}\left(k\right)=q_{i+1}^{in}\left(k\right),i=1,...,N-1$

Flow determined by relation between supply and demand, S _ik () represents downstream cellular supply, D _i-1k () represents upstream cellular demand:

$q_i^{in}\left(k\right)=\min \{D_{i-1}\left(k\right),S_i\left(k\right)\}$

Relation between supply and demand is:

D _i(k)＝v(ρ _i)ρ _i,S _i(k)＝q _max,ρ _i(k)≤ρ _c

D _i(k)＝q _max,S _i(k)＝v(ρ _i)ρ _i,ρ _i(k)＞ρ _c

To N number of cellular, sail into roll discharge relation and flow relation between supply and demand away from according to above-mentioned difference equation, upstream and downstream cellular, the system equation of the state-space model system of city expressway can be summarized as:

ρ ^T(k+1)＝F(ρ ^T(k))+G ^Td(k)

ρ in formula ^t=[ρ ₁... ρ _n] be traffic density in each cellular of current time, $d^T\left(k\right)=\left(q_1^{in}\right(k),q_N^{out}(k\left)\right)$ Be border input and output flows, F represents the funtcional relationship of above-mentioned difference equation Midst density and flow,

$G^T=\left[\begin{array}{cccc}0& 0& \mathrm{...}& -\frac{\mathrm{\&Delta;}t}{L_{n_j,j}}\\ \frac{\mathrm{\&Delta;}t}{L_{1,j}}& 0& \mathrm{...}& 0\end{array}\right];$

The observation equation of city expressway state-space model obtains as follows:

This method adopts smart mobile phone to gather traffic data, only needs the velocity information of vehicle in collecting part cellular, and does not need to gather boundary flux information, just can estimate traffic behavior;

According to the speed V (i, t) that part B method gathers, the traffic density of current cellular i can be described; Mobile phone gathers traffic data and provides velocity information set y (k)=[v ₁... v _m] ^t, owing to only needing the velocity information knowing part cellular, suppose that N number of cellular section needs the velocity information of M cellular (M is less than N, and only needs the number being greater than Unknown worm);

Application speed-density equation is as follows:

$v=\left\{\begin{array}{cc}\frac{v_c-v_f}{{\mathrm{\&rho;}}_c}\mathrm{\&rho;}+v_f,& \mathrm{\&rho;}\&le;{\mathrm{\&rho;}}_c\\ \frac{{\mathrm{\&rho;}}_{cc}v}{{\mathrm{\&rho;}}_J-{\mathrm{\&rho;}}_c}(\frac{{\mathrm{\&rho;}}_J}{\mathrm{\&rho;}}-1),& \mathrm{\&rho;}>{\mathrm{\&rho;}}_c\end{array}\right.$

The non-linear observation equation model setting up city expressway state-space model is as follows:

y(k)＝H(ρ(k))

In formula, H describes above-mentioned speed-density equation relation;

So far, the state-space model devising city expressway based on LWR traffic flow model has been set up, as follows:

ρ ^T(k+1)＝F(ρ ^T(k))+G ^Td(k)

y(k)＝H(ρ(k))

Secondly, this method devises a kind of three-wave-length Recursive Filter Algorithm Using, can synchronously estimate traffic behavior and boundary flux by the state-space model of city expressway;

For discrete-time system:

x(t+1)＝A(t)x(t)+Gd(t)

y(t)＝C(t)x(t)

In formula, x (t) is state variable, and d (t) is Unknown worm, and y (t) is observed parameter;

Nonlinear relationship F and H, according to EKF, solves as the matrix A of above-mentioned discrete system and C at the Jacobian matrix of current time, namely by this method

$A\left(k\right)=\frac{\&part;F}{\&part;\mathrm{\&rho;}}{|}_{\mathrm{\&rho;}\left(k\right|k)}$

$C\left(k\right)=\frac{\&part;H}{\&part;\mathrm{\&rho;}}{|}_{\mathrm{\&rho;}(k+1|k)}$ 。

7. the urban expressway traffic method for estimating state based on smart mobile phone mobility detect according to claim 1, it is characterized in that: this method devises a kind of three-wave-length Recursive Filter Algorithm Using and carries out compound Weibull process to system state and Unknown worm and boundary flux, and step is as follows:

The first step: estimated state variable; By the measured value in k moment, estimate traffic subsequent time traffic density ρ (k+1|k);

ρ(k+1|k)＝F(ρ(k|k))

Second step: the unknown boundary flux of pre-estimation

3rd step: more new state; According to the status predication value ρ (k+1|k) of the first step and the Unknown inputs value of second step obtain optimum state variable and estimate ρ (k+1|k+1);

$\mathrm{\&rho;}(k+1|k+1)=\stackrel{\&OverBar;}{\mathrm{\&rho;}}(k+1|k+1)+K\left(k\right)\left(y\right(k)-C(k\left)\stackrel{\&OverBar;}{\mathrm{\&rho;}}\right(k+1|k+1\left)\right)$

Gain matrix M (k) in formula, K (k) can be calculated by covariance:

P(k+1|k)＝A(k)P(k|k)A ^T(k)+Q(k)

$M\left(k\right)={\left(D^T\right(k\left)\tilde{R}\right(k\left)D\right(k\left)\right)}^{-1}{D}^T\left(k\right){\tilde{R}}^{-1}\left(k\right)$

$K\left(k\right)=P(k+1|k)C\left(k\right){\tilde{R}}^{-1}\left(k\right)$

$\tilde{R}\left(k\right)=C\left(k\right)P(k+1|k)C^T\left(k\right)+R\left(k\right)$

P(k+1|k+1)＝(1-M(k)C(k))P(k+1|k)

In formula, the error co-variance matrix that D (k)=C (k) G, Q (k) is state equation, the error co-variance matrix that R (k) is observation equation;

So far, traffic density and the boundary flux of each cellular synchronously can be estimated by above-mentioned state-space model and filtering algorithm.

8. the urban expressway traffic method for estimating state based on smart mobile phone mobility detect according to claim 1, is characterized in that: D part merges boundary flux, realizes through street distributed traffic state estimation on a large scale; This method is estimated to realize large scale traffic network, according to smart mobile phone traffic data collection situation, through street is divided into n sub-section, every sub-section comprises 3-4 cellular, by the velocity information of the corresponding cellular of mobile phone traffic data collection technical limit spacing, application C partial state space model and filtering algorithm are synchronously estimated every sub-road section traffic volume state and boundary flux, then the boundary flux in adjacent two sub-sections is merged, realize distributed traffic state estimation, thus obtain the traffic behavior of whole loop.

9. the urban expressway traffic method for estimating state based on smart mobile phone mobility detect according to claim 1, is characterized in that: this method fusion method adopts average weighted method, and weight is the core of Weighted Average Algorithm; This method utilizes Lagrange multiplier to calculate weighted value; In order to obtain reliable weighted value r _m, suppose following condition:

(1) m kind data source does not have systematic error;

(2) variance of a random variable θ _mknown;

(3) error in different pieces of information source is uncorrelated;

Suppose that the estimated value of all types data is x _m, error variance X (x, the t)=Σ of weighted arithmetic mean _mr _mx _mcan be by provide, wherein Σ _mr _m=1 must meet; Target is by changing weight r _mor weight vectors make error minimize; Thus the constrained optimization problem obtained below:

Minimize objective function: $\mathrm{\&theta;}\left(\stackrel{\&RightArrow;}{r}\right)=\underset{m}{\&Sigma;}{r}_m^2{\mathrm{\&theta;}}_m$

Constraint condition: Σ _mr _m=1

The optimal problem of belt restraining can use method of Lagrange multipliers to solve, and final weighted value is as follows:

$r_m=\frac{\frac{1}{{\mathrm{\&theta;}}_m}}{\underset{m^{\&prime;}}{\&Sigma;}\frac{1}{{\mathrm{\&theta;}}_{m^{\&prime;}}}}$

The step of boundary flux fuse information is as follows:

The first step: utilize the filtering algorithm antithetical phrase section one of above-mentioned C part to carry out one-step prediction state variable calculated gains matrix K ₁, M ₁, calculate the output estimation value of cellular 8

Second step: use the same method and obtain the one-step prediction state variable in sub-section two gain matrix K ₁, M ₁, and the input estimated value of cellular 9

3rd step: utilize the weighted value r that above-mentioned Weighted Average Algorithm obtains _m, right with be weighted and on average obtain $\hat{d},\hat{d}=r_1{\hat{d}}_1+r_2{\hat{d}}_2;$

4th step: utilize upgrade state variable estimated value with calculate the state variable ρ in each sub-section more respectively ₁and ρ (k+1|k+1) ₂(k+1|k+1);

Data fusion is carried out to the upper delivery rate in a sub-section and the input flow rate in next son section, reduces evaluated error, and be that some sub-sections are estimated by longer pavement section, greatly reduce the order of model, improve efficiency of algorithm;

C part describing method is used to estimate the traffic behavior in whole sub-section, through street and boundary flux respectively, apply D part describing method again the boundary flux estimated value in adjacent sub-section is merged, renewal traffic parameter is estimated, realizes the distributed traffic state estimation of whole section of through street.