CN101038700A - Mixed controlling method of single dot signal controlling crossing - Google Patents

Abstract

A mixed control method for a single-point signal control junction is an automatic selection timing control method for different traffic states or a fuzzy control method with a self-adaptive ability, the mixed control method comprises: at first, collecting real-time traffic information of the junction, gathering the collected data in two classes by a clustering method, in which the two classes represent a peak state and a flat state; selecting typical examples near a centre of the gathering class from each class to establish an example library down-line; determining a current traffic state of the junction by comparing the collected real-time traffic flux, speed, driveway occupied ratio, time headway, and a predictive value of the traffic state in a short time with example values in the example library; selecting the timing control or the fuzzy control according to different combinations between the current traffic state and a current signal control mode; the timing control performing signal planning with an object that motor vehicles travel ability is maximal, by considering a relay constrained condition of motor vehicles and non-motorized vihecles; the fuzzy control using a three-dimensional traffic signal fuzzy controller.

Description

The mixing control method of single point signals control crossing

Technical field

The present invention relates to the traffic intelligent control technology of single point signals control crossing, city, belong to the urban traffic signal technical field of control automatically.

Background technology

In urban road network, congested in traffic or obstruction often occurs in intersection (owing to conflict in the crossing traffic flow, in the ordinary course of things, the not enough normal road section of the handling capacity of intersection 50%), therefore, solving the crowded key of urban highway traffic is to improve the handling capacity of crossing, and common way is: build grade separation, widen the crossing, carry out signal lamp control.Build grade separation and take up an area of greatly, influence urban look, in urban road, seldom adopt; Widen the crossing and also will take bigger land resource, in the city of an inch of land is an inch of gold, the crossing is widened very limited; By contrast, signal lamp control does not increase land resource basically, and can realize the separation of conflict traffic flow, makes traffic flow in order by the crossing, becomes the effective measures that improve intersection capacity.

In China's urban road intersection signal controlling, single point signals control is in leading position, and the non-mixed traffic phenomenon of machine shows the most seriously in China's urban road intersection.Therefore, solve the single point signals control problem of intersection mixed traffic flow, the urban traffic congestion situation that alleviation China is on the rise has important meaning.

Summary of the invention

Technical matters: be controlled to be main this reality with single-point at Chinese mixed traffic flow characteristics and whistle control system, the present invention proposes a kind of traffic signals mixing control method of single cross prong, the fuzzy control that this mixing control method can be selected timing controlled automatically or have adaptive ability at different traffic.

Technical scheme: its core concept of mixing control method of single point signals control of the present invention crossing is to be optimized for prerequisite with the phase place phase sequence, based on clustering method division signals crossing traffic behavior, predicted value according to Real-time Traffic Information (volume of traffic, speed, occupation rate, time headway etc.) and short-term traffic flow, the current traffic behavior of use case reasoning and judging, for the various combination between current traffic behavior and the currently used control pattern signal, the fuzzy control of selecting timing controlled automatically or having adaptive ability.Timing controlled is considered the delay constraint condition of motor vehicle and bicycle, be target to the maximum with the autos only ability and carry out signal timing dial, the thought of PREDICTIVE CONTROL has been merged in fuzzy control, adopt three-dimensional fuzzy controller, its input is the prognosis traffic volume of crossing current phase real-time traffic amount, follow-up phase real-time traffic amount, follow-up phase, be output as the green time of follow-up phase, the used fuzzy rule of fuzzy controller is offline optimization.

1 mixed traffic flows down the timing controlled mode of traffic signals

For traffic flow peak period relatively stably, adopt timing controlled, main target is the traffic capacity maximum that guarantees car lane.Under the given condition of intersection signal phase place phase sequence, consider delay (service level) constraint condition of motor vehicle and bicycle, the Optimization Model of autos only ability maximum is:

${C_p}^{*}=\max C_p=\max \underset{i}{\mathrm{\Σ}}\underset{j}{\mathrm{\Σ}}\underset{k}{\mathrm{\Σ}}{C_p}^{l_{\mathrm{ijk}}}=\max \underset{i}{\mathrm{\Σ}}\underset{j}{\mathrm{\Σ}}\underset{k}{\mathrm{\Σ}}{q_{\mathrm{sC}}}^{l_{\mathrm{ijk}}}\·{\mathrm{\λ}}^{l_i}$

$S.t.d_1=\underset{\mathrm{jk}}{\mathrm{\Σ}}{d}_{\mathrm{jk}}{q}_{\mathrm{jk}}/\underset{\mathrm{jk}}{\mathrm{\Σ}}{q}_{\mathrm{jk}}\≤{d_1}^{\mathrm{Li}}$

${\stackrel{\‾}{d}}_b=\frac{{\∫}_{T-r}^{T+m}((T-t)+\frac{1}{Q}{\∫}_{T-r}^{t}q\left(t\right)\mathrm{dt})q\left(t\right)\mathrm{dt}}{{\∫}_{T-r}^{T+g}q\left(t\right)\mathrm{dt}}\≤{{\stackrel{\‾}{d}}_b}^{\mathrm{Li}}$

t _RWM ^p≤t _Rth ^p

g ^imin≥max(g ^iminb，g ^iminp)， i＝1，2，...，n _ph

${\stackrel{\‾}{T}}^{p_w}=\underset{j=1}{\overset{n_{\mathrm{or}}}{\max }}\left({\stackrel{\‾}{T}}^{p_{w}j}\right)\≤{\stackrel{\‾}{T}}^{p_{\mathrm{th}}}$

In the formula: C _pThe traffic capacity of-signalized intersections car lane (pcu/h);

C _p ^LijkThe traffic capacity (pcu/h) of each each track car lane of signal phase all directions of-Di l cycle;

q _SC ^LijkThe correction saturation volume rate (pcu/h) of each each track motor vehicle of signal phase all directions of-Di l cycle;

λ ^LiThe split of each signal phase motor vehicle of-Di l cycle;

The signal period that l-investigated;

I-phase place sequence number;

J-direction sequence number;

K-track sequence number;

d ₁The average letter control of-crossing per car is incured loss through delay;

d _JkThe average letter control of the per car of j direction k track ,-crossing motor vehicle is incured loss through delay;

q _JkThe peak 15min traffic flow rate of j direction k track ,-crossing motor vehicle;

d ₁ ^LiThe average letter control of motor vehicle per car is incured loss through delay under the service level Li of-crossing;

d _bThe mean delay of-bicycle;

d _b ^LiThe mean delay of bicycle under the-service level Li;

t _RWM ^pThe longest wait red time of-pedestrian's street crossing;

t _Rth ^p-pedestrian's street crossing maximum wait time threshold value;

g ⁱThe green light of-crossing i phase place shows the time;

g ^IminbThe Minimum Green Time of-crossing i phase place bicycle;

g ^Iminp-crossing i phase place pedestrian's Minimum Green Time;

n _PhThe signal phase number of-crossing;

T ^Pw-crossing pedestrian's street crossing the average latency;

T ^Pwj-crossing j direction pedestrian's street crossing the average latency;

T ^Pth-crossing pedestrian's street crossing average latency threshold value;

N _OrThe intersecting roads number of-two crossings.

Consider particle swarm optimization algorithm (particle swarm optimization algorithm, PSO) can be used for finding the solution the complicated optimum problem of non-linear in a large number, non-differentiability and multi-peak, program realizes succinct unusually, need the parameter of adjustment few, being particularly suitable for engineering uses, so the utilization particle swarm optimization algorithm carries out optimizing to the Optimization Model of autos only ability maximum in above-mentioned rush hour, thereby obtains the signal period and the signal timing dial parameter of single cross cross road mouth leggy mixed traffic current control.

Particle swarm optimization algorithm is a kind of evolutionary computing based on the colony intelligence method, and this algorithm comes from the simulation to the birds predation.In PSO, separating of each optimization problem all is a bird in the search volume, is referred to as " particle ".All particles all have an adaptive value and speed to determine direction and the distance that they circle in the air.Particles are just followed current optimal particle and are searched in solution space then.

The PSO algorithm can be described below: suppose to form a group by m particle in the target search space of a D dimension, wherein i particle is expressed as

I=1,2 ..., m, i.e. i the position of particle in D dimension search volume.Will

Bring objective function into and just can calculate its adaptive value, weigh according to the size of adaptive value Quality." circling in the air " speed of i particle also is the vector of a D dimension, is designated as

Remember that the optimal location that i particle searches up to now is p _{I_best}, the optimal location that whole population searches up to now is p _{G_best}At first initialization a group of algorithm random particles finds optimum solution by iteration then.In iteration each time, particle upgrades oneself by following the tracks of above-mentioned two " extreme values ", and formula is as follows:

v _i＝v _i+c ₁r ₁(p _{i_best}-x _i)+c ₂r ₂(p _{g_best}-x _i)

x _i＝x _i+v _i

Wherein, study factor c ₁And c ₂It is non-negative constant; r ₁And r ₂It is the random number between [0,1].

Stopping criterion for iteration is set according to particular problem, and is generally constant as the condition that finishes search with optimum solution after maximum iteration time or N iteration of particle.

PSO is similar with genetic algorithm, but does not have the intersection and the mutation operation of genetic algorithm, follows optimum particle in solution space and searches for but particle is potential separating.Therefore, compare, it is advantageous that the simple deep intelligent background that realizes easily having again simultaneously with genetic algorithm.

The real-time fuzzy control in 2 crossings

For the flat peak period that the traffic flow fluctuation is big, randomness is stronger, with the volume of traffic of crossing real-time change and the predicted value of short-term traffic flow is foundation, according to fuzzy control theory, carry out the real-time adaptive optimized distribution of signal green time resource, the operational use time resource.

If current phase is respectively q and q with the follow-up amount of real-time traffic mutually _s, follow-up phase prognosis traffic volume is p _s, the green light duration is e, then e=fc (q, q _s, p _s).f _cBe three-dimensional fuzzy controller to be designed, q, q _sAnd p _sBe input variable of fuzzy controller, e is its output.

Control procedure is as follows: with accurate input quantity (q, q _s, p _s) carry out fuzzy quantization and handle and to become fuzzy quantity, activate the corresponding fuzzy control rule in the rule base, carry out fuzzy reasoning and obtain fuzzy control quantity, handle through sharpening again and transfer accurate amount e, control signal driving circuit to.

This traffic signals fuzzy controller one of is imported the predicted value of short-term traffic flow as controller, has merged the thought of PREDICTIVE CONTROL; Adopt genetic algorithm offline optimization fuzzy rule base in addition, further improved the fuzzy control performance of crossing.

● wavelet analysis-neural network short-term traffic flow forecast model

Traffic volume forecast value p _sWhether accurate performance for fuzzy controller significant effects is arranged.Be used for the traffic volume forecast that the intersection signal timing is optimized, predicted time is very short at interval, the uncertainty of the magnitude of traffic flow is very strong, the present invention uses the compound forecast model of wavelet analysis-neural network that the crossing short-term traffic flow is predicted, resulting Intersection Traffic Volume predicted value is served signal timing dial directly as the input of traffic signals fuzzy controller.

Wavelet analysis-neural net prediction method at first carries out the decomposition of multiresolution to traffic data with wavelet function, the low frequency signal and the high-frequency interferencing signal of expressing traffic flow essential change trend are separated, then the undesired signal of baseband signal and different resolution is set up neural network prediction model, last extrapolatedly predict the outcome and synthesize, thereby obtain predicting the outcome of the volume of traffic.Step is as follows:

(1) wavelet decomposition.Select for use a certain wavelet function that volume of traffic time series is carried out the decomposition of multiresolution, establishing decomposition scale is N, decomposes the back coefficient of dissociation and is made up of two parts: the vectorial d of high frequency coefficient under the low frequency coefficient vector aN of yardstick N and N the different scale _N, d _N-1... d ₁The N size is relevant with sampling time interval, and the short more N of sampling interval should be big more.Because signals sampling is short more at interval, randomness is strong more, and high frequency noise jamming component is just many more, therefore also just needs more multi-layered time decomposition, could be the trend signal from extracting the high-frequency interferencing signal layer by layer.

(2) wavelet reconstruction.With the wavelet function of appointment respectively to the low frequency part a of signal _NWith HFS d _N, d _N-1... d ₁Carry out multiple dimensioned reconstruct, obtain N+1 time series v _N, w _N, w _N-1... w ₁, v wherein _NBe low frequency signal, reflection traffic flow essential change trend, w _i(i=1,2 ..., N) be high-frequency interferencing signal.

(3) time series of an above-mentioned N+1 reconstruct is set up neural network prediction model respectively.

(4) prediction.Use neural network model to predict, obtain N+1 and predict the outcome

(i=1,2, ..., N).

(5) synthetic.Individual the predicting the outcome of above-mentioned N+1 added up, and acquisition predicts the outcome corresponding to raw traffic capacity, promptly

$\tilde{x}={\tilde{v}}_N+\underset{i}{\overset{N}{\mathrm{\Σ}}}{\tilde{w}}_i$

● based on the Optimization of Fuzzy-control Rules of genetic algorithm

The core of fuzzy controller is a fuzzy control rule, generally provides one according to existing experience and investigation directly perceived and guarantees safety and roughly rational language control law, adjusts according to the behavior quality of system again.If variable q, q _s, p _sFive fuzzy subsets of definition on its domain, e defines seven fuzzy subsets on domain, and control law has the 5*5*5=125 bar at most, thereby rule base has 1257 kinds of possible combinations.Concerning the deviser, accurately make the consequent of strictly all rules, its difficulty is well imagined.Because genetic algorithm (Genetic Algorithm, GA) adopt the mode of population to search for, once can provide a plurality of more excellent rule bases for your guidance,, use genetic algorithm offline optimization fuzzy rule base in order further to improve the fuzzy control performance of crossing.

Article one, the gene of the consequent homologue of rule is represented with one 1～7 decimal integer, and the consequent of strictly all rules is expressed as a chromosome in rule base, represents with 125 decimal numbers.

The target of optimizing is to find one group of rule to make average vehicle delay J minimum.If initially waiting for vehicle is q ₀, the vehicle delay summation of i cycle j phase place is d _Ij, the arrival vehicle is c _Ij, objective function is the average vehicle delay in M cycle, then

$J=\underset{i=1}{\overset{M}{\mathrm{\Σ}}}\underset{j=1}{\overset{4}{\mathrm{\Σ}}}{d}_{\mathrm{ij}}/(q_0+\underset{i=1}{\overset{M}{\mathrm{\Σ}}}\underset{j=1}{\overset{4}{\mathrm{\Σ}}}{c}_{\mathrm{ij}})$

Algorithm with the genetic algorithm optimization fuzzy control rule is as follows:

(1) initialization population;

(2) chromosome is decoded as rule, calculates each chromosomal adaptation value;

(3), then finish if restrain or evolutionary generation reaches preset value;

(4) use selection, intersection, mutation operator to produce a new generation, change (2);

The adaptation value fit of chromosome c (c)=J _Max-J _c, J _MaxBe the present age or each maximal value up to the present for objective function in the population.Select to adopt ratio back-and-forth method based on adaptation value.For guaranteeing convergence, also use strategy according to qualifications, be about to the best chromosome of the previous generation and remain in the current new colony.Intersect and adopt evenly hybridization, produce at random and chromosome equilong binary hybridization template, the corresponding position of 0 expression does not exchange, 1 expression exchange.According to template two parents are implemented hybridization then, produce two offsprings.Evenly hybridization can search the pattern that point type hybridization can't search, and properly is used for less population size.And the pattern that point transposition searches is fewer, and at population size hour, its search capability will be subjected to certain influence.

Mutation operator uses heuristic variation, the individuality that adaptive value is big is searched in more among a small circle, and the little individuality of adaptive value is searched in a big way.If chromosome p=is (v ₁, v ₂..., v _n), k gene v _kSelected variation, the chromosome p '=(v after the variation ₁..., v _k' ..., v _n),

$v_k^{\&prime;}=\left\{\begin{array}{cc}{v}_k+\mathrm{\&Delta;}(\mathrm{fit}\left(p\right),7-v_k)& r\&GreaterEqual;0\\ v_k-\mathrm{\&Delta;}(\mathrm{fit}\left(p\right),v_k-1)& r<0\end{array}\right.$

$\mathrm{\&Delta;}(f,y)=y*(1-{\mathrm{\&delta;}}^{{(1-\frac{f}{f_{\max }})}^2})$

Wherein r is a random number, and δ is the random number on [0,1] interval, f _MaxMaximum adaptation value for current colony.The function Δ (f, codomain y) is [0, y], when f increased, (f y) leveled off to 0 probability and increases Δ, and promptly when f increased, mutation operation was to v _kInfluence reduce.Ding Yi mutation operator will be protected preferably and separate like this, and search is carried out in its less field, and the chromosome low to adaptive value, the field of search is bigger.Make variation to adjust the region of search like this, thereby can improve the ability of search more significantly according to the quality adaptation ground of separating.

In order to prevent precocious convergence phenomenon, adopt following method: if during evolution, the optimum solution N continuous then adopts following formula to increase the variation probability to increase individual adaptive value for not improving:

p _m＝p _m*1.1

The genetic algorithm calculated amount is big, and convergence is slow, obviously is unsuitable for the online rule optimization that carries out.But the employing offline optimization, when transportation condition changed, the rule after the optimization may lose original advantage, adopted following two methods to address this problem:

(1), optimizes by GA and to obtain a plurality of rule bases according to the traffic stream characteristics of different crossings or different period (conversion in season, festivals or holidays are with at ordinary times, traffic flow peak and Ping Feng).Can be in the practical application or the different period at different crossings, enable suitable rule base, better to be met the control effect of real-time traffic stream.

(2) according to the traffic stream characteristics at crossing, regularly or irregularly obtain up-to-date rule base, compare with existing rule base, when the difference of two rule bases reaches the threshold values of setting by GA optimization, replace original rule base with up-to-date rule base, otherwise still use the meta-rule storehouse.If d is the difference of two rule bases, d _iBe the distance of i bar rule, the distance between promptly same rule (former piece the is identical) consequent, as hamming distance, Euclidean distance etc., λ is a threshold values, order

$d=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{d}_i{w}_i$

Wherein, N is the rule sum, w _iBe the weight of rule, the degree commonly used of expression rule, the rule that more often is used, weight are big more, and its value can calculate according to the historical information that rule is used.When d＞λ, replace the meta-rule storehouse, otherwise do not change rule base.

Beneficial effect: a kind of mixing control method of single point signals control crossing.This method is according to Real-time Traffic Information and short-term traffic flow predicted value, the current traffic behavior in use case reasoning and judging crossing.Automatically the fuzzy control of selecting timing controlled or having adaptive ability at different traffic.The advantage of doing like this is that the crossing operational mode is not regularly to transform according to built-in timetable, but is determined by current traffic flow operation characteristic, avoids the traffic flow operation characteristic to change and problem that built-in timetable does not match with it and brings.

For peak, crossing state, adopt timing controlled.Under the given condition of intersection signal phase place phase sequence, consider delay (service level) constraint condition of motor vehicle and bicycle, autos only ability maximum is set up Optimization Model, utilize particle swarm optimization algorithm that this model is carried out optimizing, obtain the signal period and the signal timing dial parameter of single cross cross road mouth leggy mixed traffic current control.Its advantage is, the timing model is at the traffic characteristics of China, considered the delay of motor vehicle and bicycle simultaneously, can preferentially guarantee under the state of peak, motor vehicle is in the traffic capacity of crossing, taken into account simultaneously the interests of bicycle again, reduced because bicycle waits the long traffic hazard incidence of making a dash across the red light and causing.In addition, the peak state adopts timing controlled down, and the traffic disturbance that can avoid adaptive control frequent changes signal period, split or phase differential to cause causes congested in traffic the obstruction, increases time delays.

For the flat peak of crossing traffic state, adopt fuzzy control.Having merged the thought of PREDICTIVE CONTROL in the fuzzy control, is input with the volume of traffic of crossing real-time change and the predicted value of short-term traffic flow, carries out the real-time adaptive optimized distribution of signal green time resource.Adopt genetic algorithm offline optimization fuzzy rule base, further improve the fuzzy control performance of crossing.Its advantage is, traffic flow randomness is big under the flat peak state, uncertainty is stronger, fuzzy control makes the crossing traffic signal timing dial adapt to the dynamic need of the magnitude of traffic flow, avoid red light to wait and un-reasonable phenomenon that green light passes through to no vehicle, can effectively reduce the waste of green light resource to vehicle queue.

Description of drawings

Fig. 1 is the process flow diagram of single cross prong mixing control method.Wherein solid line is represented control stream, and dotted line is represented data stream.

Fig. 2 is the structural representation of three-dimensional traffic signal ambiguity controller.Wherein have: current phase real-time traffic amount q, follow-up phase real-time traffic amount q _s, follow-up phase prognosis traffic volume p _s, follow-up phase green light duration e.

Embodiment

As shown in Figure 1, the implementation process of each several part is as follows.

Real-time information collection: utilization ground induction coil or video capture device obtain crossing real-time traffic stream information, comprise the magnitude of traffic flow, speed, lane occupancy ratio, time headway, and sampling step length can be made as 30 seconds or 1 minute.

Case library: the most basic most important knowledge base in the case library case-based reasoning.A continuous week or longer time are gathered crossing real-time traffic stream information, adopt the cluster analysis based on distance, as K-Mean Method or K-center method, are two classes with above-mentioned data clusters, and promptly the crossing running status is divided into peace peak, peak two classes.It should be noted that need be with working day and nonworkdays data difference cluster.Select prominent example (near the example of cluster centre) off-line to set up case library, with the crossroad is example, each example attribute comprises the magnitude of traffic flow, speed, lane occupancy ratio, the time headway of four entrance driveway in crossing, whether add working day and traffic behavior classification, totally 18 attributes, each connection attribute all standard turn to number between 0 and 1, discrete attribute is represented with 0 or 1, as attribute " whether working day " be with 1 expression, 0 expression is not, " traffic behavior classification " with 1 expression peak, 0 represents flat peak.Consider recall precision and accuracy, working day, every class was stored about 500 examples, and the every class of nonworkdays is stored about 200 examples.

Case retrieval: according to the traffic flow character data of being gathered and working day whether, in case library, search similar example, adopt the similarity degree between the euclidean distance metric example, get a most similar n example, wherein n is taken as odd number, as get n=5, adopt the majority voting method to determine the running status that the crossing is current.

The crossing pattern is judged: four kinds of possibilities are arranged between current running status in the crossing that case retrieval obtains and the current control mode:

If current traffic behavior is peak and current control mode for regularly: adopt the timing controlled mode peak period, both are consistent, and control mode need not to change.But, the signal timing dial parameter of Optimization Model gained of section autos only ability maximum might be different with the timing scheme of current employing according to rush hour, when the signal period of calculating gained and original signal between the cycle during (both differences can according to the setting of crossing actual conditions), enable new signal time distributing conception greater than 10 seconds.

If current traffic behavior be peak and current control mode for fuzzy: both conflicts, check continuous 5 times case retrieval result recently, if wherein belong to the peak state at least 4 times, determine that then the current running status in crossing is the peak period, control mode is switched to timing mode from fuzzy control, otherwise it is constant to keep former fuzzy control mode.Why select 5 times, purpose is to make the judgement of the current running status in crossing more reliable, thereby avoids frequently changing between two kinds of control modes.

If current traffic behavior is regularly for the current control mode in flat peak: both conflicts, check continuous 5 times case retrieval result recently, if wherein belong to flat peak state at least 4 times, determine that then the current running status in crossing is the flat peak phase, control mode is switched to the Fuzzy Control Model with adaptive ability from timing controlled, otherwise it is constant to keep when original control mode.

If current traffic behavior is fuzzy for the current control mode in flat peak: flat peak adopts fuzzy control method, and both are consistent, control mode need not to change, and continues the use fuzzy control scheme.

Timing controlled: the Optimization Model of setting up section autos only ability maximum in rush hour, utilize particle swarm optimization algorithm that above-mentioned model is carried out optimizing, obtain the signal timing dial parameter of single cross cross road mouth leggy mixed traffic current control, comprise the green duration of signal period, each phase place.Be that stability considers, rush hour section change the signal timing dial parameter less as far as possible, only between the cycle during, just enable new signal timing dial parameter greater than 10 seconds at signal period of current calculating gained and original signal.

Traffic volume forecast: utilization wavelet analysis-neural network compound forecast model predicts the crossing short-term traffic flow, obtains the predicted value of the volume of traffic by the nearest 4 times real-time traffic amount of follow-up phase place, promptly

x(t)＝f(x(t-4)，x(t-3)，x(t-2)，x(t-1))

X (t) is directly as one of input of traffic signals fuzzy controller.

Fuzzy control: the current phase real-time traffic amount that will gather, follow-up phase real-time traffic amount, follow-up phase prognosis traffic volume are sent into fuzzy controller, determine the green light duration of follow-up phase thus follow-up phase place to be carried out timing according to the output of fuzzy controller.Controller is a three-dimensional fuzzy controller, and fuzzy rule has used the genetic algorithm offline optimization.

The workflow of the mixing control method of single point signals control crossing is:

Step 1. off-line is set up case library.Gather the crossing Real-time Traffic Information, comprise the magnitude of traffic flow, speed, occupation rate, time headway, the crossing running status is divided into peace peak, peak two classes, using clustering method that above-mentioned institute image data is gathered is two classes, selects prominent example (near the example of cluster centre) off-line to set up case library from each class.

Step 2. is gathered the crossing Real-time Traffic Information, comprises the magnitude of traffic flow, speed, occupation rate, time headway, and according to these traffic flow operation characteristics, the current running status in crossing is determined in the use case reasoning.

Step 3. is determined the crossing control model.Traffic behavior is if timing controlled is then selected on the peak, otherwise the fuzzy control that employing has adaptive ability.Fuzzy rule is offline optimization.

Step 4. repeating step 2.

Claims (6)

1. the mixing control method of single point signals control crossing is characterized in that this mixing control method is:

1.) at first gather the crossing Real-time Traffic Information, using clustering method that above-mentioned institute image data is gathered is two classes, represents state peace peak, the peak state of crossing respectively,

2.) from each class, select to set up case library near the prominent example off-line of cluster centre,

3.), determine the current traffic behavior in crossing according to the predicted value and the contrast of the exemplary values in the case library of the magnitude of traffic flow, speed, lane occupancy ratio, time headway and the short-term traffic flow gathered in real time;

4., select timing controlled or fuzzy control) according to the various combination between current traffic behavior and the currently used control pattern signal; Timing controlled is considered the delay constraint condition of motor vehicle and bicycle, is target to the maximum with the autos only ability and carries out signal timing dial; Three-dimensional traffic signal ambiguity controller is adopted in fuzzy control, its input is crossing current phase real-time traffic amount, follow-up phase real-time traffic amount, follow-up phase prognosis traffic volume, be output as the green time of follow-up phase, wherein the used fuzzy rule of fuzzy controller offline optimization.

2. control the mixing control method of crossing by the described single point signals of claim 1, it is characterized in that off-line sets up the method for case library and be, a period of time is gathered crossing real-time traffic stream information continuously, comprise the magnitude of traffic flow, speed, lane occupancy ratio, time headway, the clustering method of adopting K-Mean Method or K-center method is to above-mentioned data clusters, the number of cluster is 2, the running status that is the crossing is divided into peace peak, peak two classes, with working day and nonworkdays data difference cluster; Select to set up case library near the example of cluster centre, each example attribute comprises the magnitude of traffic flow, speed, lane occupancy ratio, time headway of crossing all directions, whether working day and traffic behavior classification, each attribute all standard turn to number between 0 and 1.

3. control the mixing control method of crossing by the described single point signals of claim 1, it is characterized in that determining that by case retrieval the method for the current traffic behavior in crossing is, according to the magnitude of traffic flow of being gathered, speed, lane occupancy ratio, time headway and working day or nonworkdays attribute, in case library, search similar example, adopt the similarity degree between the euclidean distance metric example, get a most similar n example, wherein n is an odd number, adopts the majority voting method to determine the current running status in crossing.

4. by the mixing control method of the described single point signals control of claim 1 crossing, it is characterized in that selecting the method for timing controlled or fuzzy control as follows according to current traffic behavior in crossing and currently used control pattern signal,

If the current running status in crossing be peak and current control mode for regularly, adopt the timing controlled mode peak period, both are consistent, control mode need not to change; But, the signal timing dial parameter of Optimization Model gained of section autos only ability maximum might be different with the timing scheme of current employing according to rush hour, when the signal period of calculating gained and original signal between the cycle during, enable the new signal period and the signal time distributing conception of each phase place greater than 10 seconds;

If the current running status in crossing is that peak and current control mode are for fuzzy, both conflicts, check continuous 5 times case retrieval result recently, if wherein belong to the peak state at least 4 times, determine that then the current running status in crossing is the peak period, control mode is switched to timing mode from fuzzy control, otherwise it is constant to keep former fuzzy control mode;

If the current running status in crossing is regularly for the current control mode in flat peak, both conflicts, check continuous 5 times case retrieval result recently, if wherein belong to flat peak state at least 4 times, determine that then the current running status in crossing is the flat peak phase, control mode is switched to the Fuzzy Control Model with adaptive ability from timing controlled, otherwise it is constant to keep when original control mode;

If the current running status in crossing is fuzzy for the current control mode in flat peak, flat peak state adopts fuzzy control method, and both are consistent, and control mode need not to change, and continues the use fuzzy control scheme.

5. control the mixing control method of crossings by claim 1 or 4 described single point signals, it is characterized in that described time-controlled method is, under the given condition of intersection signal phase place phase sequence, consider the delay constraint condition of motor vehicle and bicycle, autos only ability maximum is set up Optimization Model, utilize particle swarm optimization algorithm that this model is carried out optimizing, obtain the signal timing dial parameter of signal period and each phase place, drive traffic lights thus.

6. control the mixing control method of crossing by the described single point signals of claim 1, the compound forecast model of prognosis traffic volume utilization wavelet analysis-neural network that it is characterized in that follow-up phase place is predicted, at first traffic data is carried out the decomposition of multiresolution with wavelet function, the low frequency signal and the high-frequency interferencing signal of expressing traffic flow essential change trend are separated, then the undesired signal of baseband signal and different resolution is set up neural network prediction model, last extrapolatedly predict the outcome and synthesize, thereby obtain predicting the outcome of the volume of traffic; This traffic volume forecast value is directly sent into an input end of three-dimensional traffic signal ambiguity controller.

Images