CN105489028A - Supersaturation multi-intersection cooperative control optimization method - Google Patents

Abstract

The invention discloses a supersaturation multi-intersection cooperative control optimization method. The method comprises performing division in a primary channel direction and a secondary channel direction, wherein the primary channel direction is a passage direction of a whole city traffic trip main OD expectation line in an area, concrete calculation involves OD backstepping through the traffic flow of an inlet channel of each intersection, and secondary channels are in the directions of other inlet channels, which are intersected with a primary channel, of the intersections; by taking the quantity of supersaturation interactions and maximum vehicle queuing lengths of the primary channel and the secondary channels of each intersection as condition attributes, respectively taking a green lamp extension node, a green lamp extension phase and green lamp extension time as decision attributes; carrying attribute value fuzzification; carrying out decision table construction and attribute reduction; and carrying out evaluation index calculation.

Description

Supersaturation multi-intersection Collaborative Control optimization method

Technical field

The present invention relates to a kind of supersaturation multi-intersection Collaborative Control optimization method.

Background technology

Along with the fast development of Chinese society economy and the continuous acceleration of urbanization process, the height bringing city is flourishing, but meanwhile urban traffic blocking is increasingly serious, and mass transportation " paralysis " often appears in city peak time especially sooner or later.National urban automobile pollution increases as 15%-20% every year in recent years, and some cities are more up to 30%, and road provisioning resources increasess slowly under limited city space, and traffic imbalance between supply and demand becomes increasingly conspicuous.

Integrative design intersection carries out time devided modulation by signal phase to crossing limited passage resource, if but crossing flows exceed the traffic capacity, then and traditional signal control method effect is undesirable, and now crossing is in hypersaturated state.When crossing multiple in region is all in hypersaturated state, then this region belongs to multi-intersection supersaturation.There are some researches show that supersaturation multi-intersection signal controls compared with unsaturated state, have the feature of himself in control objectives sequence, the optimization coordination of multi-control target, the dynamic change etc. of coordination mode, and often occur wagon flow return overflow or adjacent lane block cause the phenomenons [1-2] such as traffic capacity reduction.The identification of oversaturated intersection by high-resolution intersection signal data, mainly to minimize intersection vehicles queue length, the total traffic capacity of balanced road network queue length or crossing is oversaturated intersection control target signal to the maximum.Controlling model and algorithm mainly contain sets up based on traffic shock wave theory Single Intersection supersaturation cooperation control model, discrete time changing method model, adopts the theoretical dynamic phasing combinational algorithm combined with fuzzy control of the ripple that falls apart.But supersaturation Single Intersection is more suitable for adopting multi-period timing controlled or traffic police manually to command effect to control better than signal also to have part research to think.In supersaturation multi-intersection, TONG etc. incur loss through delay minimum for objective function with vehicle, construct oversaturated intersection group Stochastic Programming Model.Lei Lei etc., based on methods engineering visual angle, establish oversaturated intersection group Controlling model, and devise the derivation algorithm of this model, achieve the optimum control of traffic method.SUN etc. devise the reduced form continuous stream crossing scheme for oversaturated intersection group, and example analysis results shows that the program has the probability of 90% to promote intersection efficiency [14].Rough set is proposed in nineteen eighty-two by Polish mathematician Z.Pawlak, can utilize not too complete, information not too accurately, find the rule that applicable decision-making judges, thus carry out artificial intelligence decision-making.Although developing history is shorter but all achieve great successes at theoretical research or Based Intelligent Control, data mining, fault diagnosis, electric load etc.Rough set is also in starting developing stage in the application of field of traffic, is mainly reflected in the aspect such as identification, passenger traffic volume forecast of the extraction of Regional Road Network transport information, traffic flow congestion status at present.

Mostly launch for crossing unsaturation traffic flow modes in being fruitful, study less for saturated with the integrative design intersection under hypersaturated state, the achievement in research relating to region supersaturation multi-intersection is more limited, not yet by the approach application of fuzzy control in supersaturation multi-intersection control in.Research object is expanded for supersaturation multi-intersection herein, according to its traffic stream characteristics, expect that line angle degree proposes supersaturation main channel control strategy from regional traffic travelling OD, by multi-intersection signal Collaborative Control feasible region Large Copacity express passway, improve the current usefulness of regional traffic entirety.And based on rough set theory, build with oversaturated intersection quantity, direction, main channel vehicle maximum queue length, subchannel direction vehicle maximum queue length etc. for conditional attribute, the supersaturation multi-intersection fuzzy control model being decision attribute with green light extension system, green light prolongation phase place, green extension.

Summary of the invention

For the problems referred to above, the invention provides and a kind ofly solve the supersaturation multi-intersection Collaborative Control optimization method that existing fuzzy intelligent control method is only applicable to the problem of Single Intersection unsaturated state.

For achieving the above object, supersaturation multi-intersection Collaborative Control optimization method of the present invention, comprising:

Carry out direction, main channel, subchannel direction divides, wherein, direction, main channel is that the whole urban transportation main OD that goes on a journey expects that line passes through direction in this region, concrete calculating is undertaken by each crossing inlet road magnitude of traffic flow that OD is counter to be pushed away, and subchannel is then other entrance driveway direction crossing with main channel, crossing;

With oversaturated intersection quantity, the maximum vehicle queue length of each crossing primary and secondary passage for conditional attribute, extend phase place, green extension for decision attribute with green light extension system, green light respectively;

Carry out property value obfuscation;

Carry out decision table structure and attribute reduction;

Carry out evaluation index calculating.

Further, described property value obfuscation specifically comprises

By maximum queue length q in the n-th crossing primary and secondary channel direction _znand q _cnproperty value carry out Fuzzy processing, at q by the subordinate function of linear distribution _znand q _cndomain on definition 7 fuzzy language subsets { corresponding property value is { 0,1,2,3,4,5,6} for very short VS, short S, shorter RS, general M, longer RL, long L, very long VL}.

Respectively by q _znand q _cnthe maximal value of many groups real data and minimum value, discrete by unique step is 7 grades, is designated as: with then q _znand q _cnto the degree of membership computing method belonging to kth level as shown in the formula, the rank corresponding to maximum membership degree is q _znand q _cnproperty value,

${\mathrm{\μ}}_{zn}^k=\left\{\begin{array}{cc}0& q_{zn}\≤q_{zn}^k,q_{zn}\≥q_{zn}^{k+1}\\ \frac{q_{zn}-q_{zn}^k}{q_{zn}^{k+1}-q_{zn}^k}& q_{zn}^k\≤q_{zn}\≤q_{zn}^{k+1}\end{array}\right.$

${\mathrm{\μ}}_{cn}^k=\left\{\begin{array}{cc}0& q_{cn}\≤q_{cn}^k,q_{cn}\≥q_{cn}^{k+1}\\ \frac{q_{cn}-q_{cn}^k}{q_{cn}^{k+1}-q_{cn}^k}& q_{cn}^k\≤q_{cn}\≤q_{cn}^{k+1}\end{array}\right.$

Wherein: q _znit is the actual maximum queue length in the n-th direction, main channel, crossing; q _cnthe actual maximum queue length of the n-th subchannel direction, crossing; for q _znbelong to the degree of membership of kth level; for q _cnbelong to the degree of membership of kth level; for the higher limit of direction, main channel kth level; for the higher limit of subchannel direction kth level.

Crossing quantity in conditional attribute, the property value being N with actual round values;

In decision attribute, which direction, main channel, crossing Green extension green light extension system W refers to, actual conditions are controlled according to region intersection traffic, green light extension system is defined as the long green light time prolongation of direction, oversaturated intersection main channel and direction, main channel, all crossings long green light time all extends two kinds of situations, property value corresponds to W=0 respectively, W=1; Green light extends phase place E and refers to which phase place Green extension of direction, main channel, crossing, and definition green light extends phase place value E=0, refers to main channel craspedodrome phase place Green extension, and subchannel left turn phase green time reduces; E=1, refer to that keep straight in main channel and left turn phase green time all extends, subchannel is kept straight on and left turn phase green time reduces; Green extension G property value is the actual value of time expand.

Further, decision table builds and comprises with attribute reduction:

Decision table builds: conditional attribute and decision attribute data acquisition are carried out respective mode gelatinization process, thus forms the multi-intersection supersaturation optimal control decision table containing 2N+1 conditional attribute, 3 decision attributes, carries out yojan one by one to decision table;

Attribute reduction based on recognizable vector and attribute frequency:

1 decision table T=(U, C ∪ D), | U|=n, the recognizable vector corresponding to decision table T is M=(Cij) n × n, wherein:

Recognizable vector is the matrix about diagonal line symmetry, and diagonal entry is 0, and when the element x i of two in domain U is identical with the decision attribute values corresponding to xj, in recognizable vector, element gets 0; Otherwise in recognizable vector, element value is the different value in the two conditional attribute, simultaneously, number of times p (a) occurred at recognizable vector M by conditional attribute a is with the significance level of characterization attributes a, p (a)=SGF (a, R, D);

Rule Extraction: form and corresponding decision attribute according to the conditional attribute element in yojan set B, extracts multi-intersection supersaturation optimal control decision rule.

Further, step is specifically comprised based on the algorithm of the attribute reduction of recognizable vector and attribute frequency as follows:

Step1: if the conditional attribute value in decision table and decision attribute values exist continuous variable, then carry out sliding-model control.Attribute reduction set

Step2: generate recognizable vector M.

Step3: the core set Core finding out recognizable vector, combinations of attributes number is 1, and upgrades yojan set B=Core.

Step4: occuring simultaneously with B in deletion recognizable vector is not empty element, and from conditional attribute set C, delete element in B, C=C-B.

Step5: number of times p (c) that in design conditions community set C, remaining all elements occurs in recognizable vector M, element corresponding to maximum times is added in yojan community set B, B=B+cq, p (cq)=max{p (c) }.

Step6: if then export yojan set B; Otherwise, return step3.

Further, described evaluation index is:

Support(X _i→D _j)＝|X _i∩D _j|

Accuracy(X _i→D _j)＝|X _i∩D _j|/|X _i|

Coverage(X _i→D _j)＝|X _i∩D _j|/|D _j|

Wherein, || represent set in element number, this element refers to the data item in domain and a line in decision table, by precision in evaluation index lower than 50% rule reject.

Beneficial effect

Supersaturation multi-intersection Collaborative Control optimization method of the present invention possesses following beneficial effect:

The present invention, on rough set knowledge reasoning basis, constructs with multi-intersection status information for conditional attribute, with green light extension system, green light extend phase place, green extension 3 parameters be many decision attributes fuzzy control model of decision attribute.Use the attribute reduction method of recognizable vector and attribute frequency to carry out yojan to model, extract decision rule.Solve the problem that existing fuzzy intelligent control method is only applicable to Single Intersection unsaturated state, meet the needs that regional traffic supersaturation multi-intersection signal works in coordination with coordinated signals.

Accompanying drawing explanation

Fig. 1 is the intersection simulation modeling of the present invention a certain road.

Embodiment

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

Supersaturation multi-intersection Collaborative Control optimization method of the present invention, specifically comprises

Passage divides: supersaturation many intersections cooperate optimization model to be gone on a journey important main channel to get through urban transportation by direction, extended area green light phase place duration, reduces the control strategy of also saturated subchannel green light phase place duration.Wherein, direction, main channel be whole urban transportation go on a journey main OD expect line in this region by direction, specifically calculate and to be undertaken by each crossing inlet road magnitude of traffic flow that OD is counter to be pushed away, subchannel is then other entrance driveway direction crossing with main channel, crossing.Main channel control strategy be by crossing time resource to a direction current tilt, thus active region traffic crucial direction path, the overall traffic in feasible region unimpeded.Main channel control strategy is different from arterial highway green wave band, and green wave band is optimized from the angle of passage rate intersection signal timing, and main channel strategy carries out signal control with the maximum target that turns to of regional traffic overall efficiency (efficiency and ability).

Attributions selection: model with oversaturated intersection quantity, the maximum vehicle queue length of each crossing primary and secondary passage for conditional attribute, extend phase place, green extension for decision attribute with green light extension system, green light respectively, thus build three decision tables of the different decision attribute of identical conditions attribute.Model does not adopt crossing in rush hour each entrance driveway flow sum as conditional attribute, is to consider that each crossing inlet road traffic capacity is inconsistent, if unfounded for conditional attribute will cause signal to control with entrance driveway flow merely.The crossing maximum vehicle queue length of primary and secondary passage is shown below respectively:

q _zn＝max{q _zn1,q _zn2,…,q _znI}；(4)

q _cn＝max{q _cn1,q _cn2,…,q _cnJ}.(5)

Wherein, in region, oversaturated intersection total quantity is N; q _znibe the vehicle queue length in i-th track in the n-th main channel, crossing direction red signal interval, 0≤i≤I; q _cnjbe the vehicle queue length in a jth track in the n-th crossing subchannel direction red signal interval, 0≤j≤J.

The determination of decision attribute values is parameter based on conditional attribute value, minimum for objective function with region multi-intersection total delay, compares thus the optimal value determined at Synchro platform Multi simulation running.Constructed cooperate optimization Controlling model vital role is to obtain the troublesome calculation of an optimal value or the basis of traffic police's artificial experience decision-making from every a line conditional attribute data simulation, extract abstract decision rule by rough set instrument, thus provide decision recommendation for the multi-intersection supersaturation Based Intelligent Control of different cities zones of different.

Property value obfuscation: for avoiding the continuity of property value, by maximum queue length q in the n-th crossing primary and secondary channel direction _znand q _cnproperty value no longer adopt its actual queue length value, but carry out Fuzzy processing, at q by the subordinate function of linear distribution _znand q _cndomain on definition 7 fuzzy language subsets { corresponding property value is { 0,1,2,3,4,5,6} for very short VS, short S, shorter RS, general M, longer RL, long L, very long VL}.

Respectively by q _znand q _cnthe maximal value of many groups real data and minimum value, discrete by unique step is 7 grades, is designated as: with then q _znand q _cnto belonging to the degree of membership computing method of kth level such as formula shown in (6), (7) ^[20], the rank corresponding to maximum membership degree is q _znand q _cnproperty value.

${\mathrm{\μ}}_{zn}^k=\left\{\begin{array}{cc}0& q_{zn}\≤q_{zn}^k,q_{zn}\≥q_{zn}^{k+1}\\ \frac{q_{zn}-q_{zn}^k}{q_{zn}^{k+1}-q_{zn}^k}& q_{zn}^k\≤q_{zn}\≤q_{zn}^{k+1}\end{array}\right.---\left(6\right)$

${\mathrm{\μ}}_{cn}^k=\left\{\begin{array}{cc}0& q_{cn}\≤q_{cn}^k,q_{cn}\≥q_{cn}^{k+1}\\ \frac{q_{cn}-q_{cn}^k}{q_{cn}^{k+1}-q_{cn}^k}& q_{cn}^k\≤q_{cn}\≤q_{cn}^{k+1}\end{array}\right.---\left(7\right)$

Wherein: q _znit is the actual maximum queue length in the n-th direction, main channel, crossing; q _cnthe actual maximum queue length of the n-th subchannel direction, crossing; for q _znbelong to the degree of membership of kth level; for q _cnbelong to the degree of membership of kth level; for the higher limit of direction, main channel kth level; for the higher limit of subchannel direction kth level.

In conditional attribute, crossing quantity has been discrete data, then with the property value that actual round values is N, as: in region, oversaturated intersection quantity is 1, then N=1.

In decision attribute, which direction, main channel, crossing Green extension green light extension system W refers to, actual conditions are controlled according to region intersection traffic, in model, green light extension system is defined as direction, oversaturated intersection main channel long green light time and extends and direction, main channel, all crossings long green light time all extends two kinds of situations, property value corresponds to W=0 respectively, W=1.Green light extends phase place E and refers to which phase place Green extension of direction, main channel, crossing, and definition green light extends phase place value E=0, refers to main channel craspedodrome phase place Green extension, and subchannel left turn phase green time reduces; E=1, refer to that keep straight in main channel and left turn phase green time all extends, subchannel is kept straight on and left turn phase green time reduces.Green extension G property value is the actual value of time expand.

Decision table builds and comprises with attribute reduction: decision table builds, the attribute reduction based on recognizable vector and attribute frequency, Rule Extraction.

Wherein, decision table builds: conditional attribute and decision attribute data acquisition are carried out respective mode gelatinization process, thus the multi-intersection supersaturation optimal control decision table formed containing 2N+1 conditional attribute, 3 decision attributes, because 3 decision attributes cannot simultaneously yojan, need yojan one by one, be actually equivalent to 3 decision tables.

Attribute reduction based on recognizable vector and attribute frequency: suppose 1 decision table T=(U, C ∪ D), | U|=n, the recognizable vector corresponding to decision table T is M=(C _ij) n × n, wherein:

Recognizable vector is the matrix about diagonal line symmetry, and diagonal entry is 0.When the element x of two in domain U _iand x _jwhen corresponding decision attribute values is identical, in recognizable vector, element gets 0; Otherwise element value is the different value in the two conditional attribute in recognizable vector.Meanwhile, number of times p (a) occurred at recognizable vector M by conditional attribute a with the significance level of characterization attributes a, p (a)=SGF (a, R, D).

Algorithm idea based on the attribute reduction of recognizable vector and attribute frequency is that decision table is converted into recognizable vector, to obtain the number of times that all non-core conditional attributes occur at recognizable vector, thus non-core conditional attribute maximum for number of times is included in yojan set, delete all properties combination comprising this attribute ^[21-22].Specific algorithm step is as follows:

Step1: if the conditional attribute value in decision table and decision attribute values exist continuous variable, then carry out sliding-model control.Attribute reduction set

Step2: generate recognizable vector M according to decision table and formula (8).

Step3: the core set Core (combinations of attributes number is 1) finding out recognizable vector, and upgrade yojan set B=Core.

Step4: occuring simultaneously with B in deletion recognizable vector is not empty element, and from conditional attribute set C, delete element in B, C=C-B.

Step5: number of times p (c) that in design conditions community set C, remaining all elements occurs in recognizable vector M, is added in yojan community set B, B=B+c by the element corresponding to maximum times _q, p (c _q)=max{p (c) }.

Step6: if then export yojan set B; Otherwise, return step3.

Rule Extraction: form and corresponding decision attribute according to the conditional attribute element in yojan set B, extracts multi-intersection supersaturation optimal control decision rule.

Evaluation index calculates

It is that inspection rule is accurate, the important evidence of the information amount of containing that evaluation index after Rule Extraction calculates, and adopts regular support, degree of accuracy and coverage three indexs to weigh.Specific targets are ^[20]:

Support(X _i→D _j)＝|X _i∩D _j|(9)

Accuracy(X _i→D _j)＝|X _i∩D _j|/|X _i|(10)

Coverage(X _i→D _j)＝|X _i∩D _j|/|D _j|(11)

In formula: || represent the element number in set, this element refers to the data item in domain and a line in decision table.By precision in evaluation index lower than 50% rule reject, with ensure rule validity.

Case-study

With continuous crossing, 4, a certain road for instance analysis object, changing road is the important major trunk roads running through Jiangbei District east-west direction, and circumferential distribution has the multiple industry situation such as large-scale house, business, office.Wherein crossing 4 is new South Road and Avenue of Stars crossing, and Avenue of Stars is Jiangbei District North and South direction major trunk roads, and sooner or later peak time, these 4 crossings, new South Road are all in hypersaturated state.Because Jiangbei District, Chongqing City main traffic travelling OD expects that line is east-west direction, therefore determine that new South Road is direction, main channel, each road direction vertical is with it subchannel direction.

By evening peak crossing real data continuous observation in many days, and obtain 10 kinds of sight conditional attribute values of 4 crossing supersaturation optimal controls according to formula (4) to formula (7).By the crossing inlet road data on flows input Synchro simulation software in various sight, optimize and draw signal time distributing conception.Artificial many experiments adjustment direction, main channel, crossing green light extension system, green light extend phase place and green extension, and when when total delay, long pointer is optimum, this time experiment parameter value is decision attribute values, and concrete decision table is in table 1.

Table 1 multi-intersection supersaturation optimal control decision table

Take W as decision attribute, according to Decision Table Reduction algorithm, obtain yojan set for { q _z2, q _z3, q _z4.Identical decision rule is merged, and through type (9) to formula (11) calculates each decision rule evaluation index value in table 2.

Table 2 take W as the Decision Table Reduction result of decision attribute

In like manner, be decision attribute with E, obtain yojan set for { q _z1, q _z4.Identical decision rule merged, final yojan the results are shown in Table 3.

Table 3 take E as the Decision Table Reduction result of decision attribute

Be decision attribute with G, obtain yojan set for { q _c1, q _z3, q _z4.Identical decision rule merged, final yojan the results are shown in Table 4.

Table 4 take dg as the Decision Table Reduction result of decision attribute

Extract according to three groups of Decision Table Reduction results and show that the fuzzy control decision rule of 4 oversaturated intersections is in table 5.

Table 5 fuzzy control decision rule

The fuzzy decision rule of his-and-hers watches 5 carries out analysis discussion: (1) when more than direction, main channel queue length rank in crossing 1,2,3 is for general (M), the mode that crossing green light signals extends can expand to other crossing equidirectional green time also corresponding prolongation adjacent to oversaturated intersection from only direction, oversaturated intersection main channel Green extension; Green light extends phase place also can from the control strategy extending main channel craspedodrome phase place green time, minimizing subchannel left turn phase green time expands to main channel craspedodrome and left turn phase green time all extends, subchannel is kept straight on and left turn phase green time reduces, the research conclusion of document [10] that this is abundant and perfect further.(2) the concrete numerical value of direction, main channel, crossing Green extension was advisable at 3 ~ 8 seconds, concrete numerical value and the highest crossing 3,4 direction, the main channel queue length of saturation degree are closely related, and also have direct relation with the subchannel direction queue length of crossing 1.Meanwhile, green extension also extends phase place two decision attributes and there is relevance with green light extension system, green light, and when to show as W, E be 0, green light phase place time expand is shorter, and when W, E are 1, green light phase place time expand is longer; The direction queue length of main channel, crossing is longer, the fuzzy control law that green light phase place time expand is longer.

Supersaturation multi-intersection cooperate optimization model has features such as containing much information, complicacy is high, randomness is strong, and traditional deterministic optimization modelling effect is limited.Although use fuzzy control method to carry out large quantity research to single oversaturated intersection both at home and abroad in recent years, but supersaturation multi-intersection intelligent signal controls not to be simple single fuzzy control model superposition, should from the research of region supersaturation multi-intersection method cooperate optimization angle.Herein on the basis analyzing supersaturation multi-intersection Based Intelligent Control feature, propose main channel control strategy to realize decision-making association between supersaturation multi-intersection, thus improve the current usefulness of region entirety.And build with oversaturated intersection quantity, primary and secondary channel queue length as conditional attribute according to the actual conditions that crossing, region controls, with the rough set fuzzy control model that green light extension system, green light prolongation phase place, green extension are decision attribute, use recognizable vector to carry out yojan with the same terms attribute, different decision attribute to 3 groups of rough sets respectively from the attribute reduction method of attribute frequency, extract decision rule and Calculation Estimation index.

According to the instance analysis of these 4 oversaturated intersections in road, summarize the basic law that supersaturation multi-intersection controls, research conclusion not only can provide decision references for traffic police directs traffic under peak time morning and evening region large area supersaturation traffic behavior, also can be big city peak time regional traffic and works in coordination with the thinking that coordinated signals provides new.But the disposable Algorithm for Reduction of the relevance between many decision attributes on the impact of rough set attribute reduction and many decision attributes need further research, and the concrete value of decision attribute also has room for improvement.

The present invention, multi-intersection main channel Green extension 3-8 effectively can improve regional traffic entirety current usefulness second, time expand, is not only relevant with hypersaturated state vehicle maximum queue length simultaneously, also extend phase place with green light extension system, green light and there is association, this is consistent with the control law that traffic police summarizes the experience.

The present invention be should be understood that; above-described embodiment; further detailed description has been carried out to object of the present invention, technical scheme and beneficial effect; these are only embodiments of the invention; be not intended to limit the present invention, every within spiritual principles of the present invention, done any amendment, equivalent replacement, improvement etc.; all should be included within protection scope of the present invention, the protection domain that protection scope of the present invention should define with claim is as the criterion.

Claims (5)

1. a supersaturation multi-intersection Collaborative Control optimization method, is characterized in that, comprising:

Carry out direction, main channel, subchannel direction divides, wherein, direction, main channel is that the whole urban transportation main OD that goes on a journey expects that line passes through direction in this region, concrete calculating is undertaken by each crossing inlet road magnitude of traffic flow that OD is counter to be pushed away, and subchannel is then other entrance driveway direction crossing with main channel, crossing;

With oversaturated intersection quantity, the maximum vehicle queue length of each crossing primary and secondary passage for conditional attribute, extend phase place, green extension for decision attribute with green light extension system, green light respectively;

Carry out property value obfuscation;

Carry out decision table structure and attribute reduction;

Carry out evaluation index calculating.

2. supersaturation multi-intersection Collaborative Control optimization method according to claim 1, it is characterized in that, described property value obfuscation specifically comprises

By maximum queue length q in the n-th crossing primary and secondary channel direction _znand q _cnproperty value carry out Fuzzy processing, at q by the subordinate function of linear distribution _znand q _cndomain on definition 7 fuzzy language subsets { corresponding property value is { 0,1,2,3,4,5,6} for very short VS, short S, shorter RS, general M, longer RL, long L, very long VL}.

Respectively by q _znand q _cnthe maximal value of many groups real data and minimum value, discrete by unique step is 7 grades, is designated as: with then q _znand q _cnto the degree of membership computing method belonging to kth level as shown in the formula, the rank corresponding to maximum membership degree is q _znand q _cnproperty value,

${\mathrm{\μ}}_{zn}^k=\left\{\begin{array}{cc}0& q_{zn}\≤q_{zn}^k,q_{zn}\≥q_{zn}^{k+1}\\ \frac{q_{zn}-q_{zn}^k}{q_{zn}^{k+1}-q_{zn}^k}& q_{zn}^k\≤q_{zn}\≤q_{zn}^{k+1}\end{array}\right.$

${\mathrm{\μ}}_{cn}^k=\left\{\begin{array}{cc}0& q_{cn}\≤q_{cn}^k,q_{cn}\≥q_{cn}^{k+1}\\ \frac{q_{cn}-q_{cn}^k}{q_{cn}^{k+1}-q_{cn}^k}& q_{cn}^k\≤q_{cn}\≤q_{cn}^{k+1}\end{array}\right.$

Wherein: q _znit is the actual maximum queue length in the n-th direction, main channel, crossing; q _cnthe actual maximum queue length of the n-th subchannel direction, crossing; for q _znbelong to the degree of membership of kth level; for q _cnbelong to the degree of membership of kth level; for the higher limit of direction, main channel kth level; for the higher limit of subchannel direction kth level.

Crossing quantity in conditional attribute, the property value being N with actual round values;

In decision attribute, which direction, main channel, crossing Green extension green light extension system W refers to, actual conditions are controlled according to region intersection traffic, green light extension system is defined as the long green light time prolongation of direction, oversaturated intersection main channel and direction, main channel, all crossings long green light time all extends two kinds of situations, property value corresponds to W=0 respectively, W=1; Green light extends phase place E and refers to which phase place Green extension of direction, main channel, crossing, and definition green light extends phase place value E=0, refers to main channel craspedodrome phase place Green extension, and subchannel left turn phase green time reduces; E=1, refer to that keep straight in main channel and left turn phase green time all extends, subchannel is kept straight on and left turn phase green time reduces; Green extension G property value is the actual value of time expand.

3. supersaturation multi-intersection Collaborative Control optimization method according to claim 1, is characterized in that, decision table builds and comprises with attribute reduction:

Decision table builds: conditional attribute and decision attribute data acquisition are carried out respective mode gelatinization process, thus forms the multi-intersection supersaturation optimal control decision table containing 2N+1 conditional attribute, 3 decision attributes, carries out yojan one by one to decision table;

Attribute reduction based on recognizable vector and attribute frequency:

1 decision table T=(U, C ∪ D), | U|=n, the recognizable vector corresponding to decision table T is M=(Cij) n × n, wherein:

Recognizable vector is the matrix about diagonal line symmetry, and diagonal entry is 0, and when the element x i of two in domain U is identical with the decision attribute values corresponding to xj, in recognizable vector, element gets 0; Otherwise in recognizable vector, element value is the different value in the two conditional attribute, simultaneously, number of times p (a) occurred at recognizable vector M by conditional attribute a is with the significance level of characterization attributes a, p (a)=SGF (a, R, D);

Rule Extraction: form and corresponding decision attribute according to the conditional attribute element in yojan set B, extracts multi-intersection supersaturation optimal control decision rule.

4. supersaturation multi-intersection Collaborative Control optimization method according to claim 3, is characterized in that,

It is as follows that algorithm based on the attribute reduction of recognizable vector and attribute frequency specifically comprises step:

Step1: if the conditional attribute value in decision table and decision attribute values exist continuous variable, then carry out sliding-model control.Attribute reduction set

Step2: generate recognizable vector M;

Step3: the core set Core finding out recognizable vector, combinations of attributes number is 1, and upgrades yojan set B=Core;

Step4: occuring simultaneously with B in deletion recognizable vector is not empty element, M=M-Q, and from conditional attribute set C, delete element in B, C=C-B;

Step5: number of times p (c) that in design conditions community set C, remaining all elements occurs in recognizable vector M, element corresponding to maximum times is added in yojan community set B, B=B+cq, p (cq)=max{p (c) };

Step6: if then export yojan set B; Otherwise, return step3.

5. supersaturation multi-intersection Collaborative Control optimization method according to claim 1, it is characterized in that, described evaluation index is:

Support(X _i→D _j)＝|X _i∩D _j|

Accuracy(X _i→D _j)＝|X _i∩D _j|/|X _i|

Coverage(X _i→D _j)＝|X _i∩D _j|/|D _j|

Wherein, || represent set in element number, this element refers to the data item in domain and a line in decision table, by precision in evaluation index lower than 50% rule reject.