DE19841457B4 - Method for determining a traffic-dependent signal program for signal groups of traffic signal systems - Google Patents

Abstract

Method for determining a traffic-dependent signal program (SP) for signal groups of traffic signals, the signal program (SP) being characterized by a sequence of release times (tg) for the different signal groups,
characterized by the combination of the following steps:
a) m signal programs (SP) of the first order are formed, each with a time history consisting of sections of randomly selected release times (tg),
b) a total of k second-order derived signal programs (SP) are formed from two signal programs (SP) of the first order by mutually exchanging the sections (A1... A3, B1... B3) of the respective time courses (tg),
c) the m + k signal programs (SP) of the first and the second order are each assigned a quality value (Q (tg)) which takes into account currently determined traffic data,
d) from the m + k signal programs, the m signal programs with the best quality values (Q (tg)) are selected, interpreted as first-order signal programs (SP) and the steps b) to d) are repeated cyclically until the quality value (Q (tg)) of the ...

Description

method to determine a traffic-dependent Signal program for Signal groups of traffic signal systems The invention relates to to a method for determining a traffic-dependent signal program for signal groups of traffic lights.

Traffic-dependent traffic control nowadays takes place, for example, via traffic lights, variable message signs, changeable Parking space information signs or radio announcements. The traffic-dependent data are over Traffic detectors such as induction loops, radar or infrared detectors won.

For example be in light signal systems whose phase sequences or signal sequences determined by predetermined signal programs. The signal programs are depending on the selected Control method variable. So are in traffic-dependent control procedures For example, the duration of each phase, the order the individual phases and the number of different phases (if required) changed in the signal program. For daytime-dependent control procedures be set at fixed times of the day (for example at rush hour) different signal programs and thus different phase sequences connected.

For the election of the signal program or for the choice of the phase duration, the sequence of phases or the number the phases are evaluated traffic technical parameters, the help be determined by traffic detectors. In the case of a signal program with fixed times, the parameters are in the draft processing Implemented off-line. In the case of signal program adaptation or signal program formation the parameters continuously processed, with the possibility a changing influence between traffic flow and signal control. The current signal programs are doing according to a predetermined Control logic based on updated characteristics online calculated and evaluated. With the help of traffic detectors will be in the spatial Surroundings of traffic signal systems Occupancy values detected, off those characteristics of the Traffic flow are derived. As parameters serve, for example, the Waiting time of the vehicles in front of the traffic signal system, the traffic jam length the traffic signal system, the traffic volume, i. Vehicles per cross section, the speed of travel, a registration (request) by pedestrians, cyclists and / or vehicles, the occupancy rate, the traffic density, the degree of utilization and the burden quotient. In a traffic-dependent signal program selection become the processed characteristics of the Traffic flow in the control logic for selecting the signal programs with Condition equations and thresholds linked.

In the guidelines for Traffic light systems (RiLSA) - Traffic lights for the Road Traffic - Edition 1992, published by the Research Association for Road and Rail Transportation, Working Group Traffic Management and Traffic Safety, Pages 41 to 44) are described methods using Traffic detectors time gaps, Occupancy degrees or ram lengths measure in the driveways to a crossroads and from this a release time adjustment determine.

Out Fellendorf, Wiedemann, "Reasonable calculation signal controlled nodes ", Traffic engineering, Issue 2, 1990 is for the determination of release times a so-called blocking group method known. For complex problems (consideration of more than, for example 32 signal groups), however, a large amount of time is needed to solve the problem. Besides, they are the conventional ones Procedure very inflexible in relation to the change of the quality functions necessary for the evaluation.

From the JP 08171694 A A method is known for determining a traffic-dependent signal program for signal groups of traffic signals using an optimization process. Initially, signal programs with random traffic signal control patterns are formed. Subsequently, an interchange operation is applied to these light signal control patterns. Furthermore, the signal programs are assigned a quality value. Finally, by cyclically repeating this process, optimized control programs are determined.

From the publication Joshi, Rathi, Tew: "An improved response surface methodology algorithm with an application to traffic signal optimization for urban networks", IEEE, Proceedings of the 1995 Winter Simulation Conference, 03.-06.12.1995, p 1104-1109 discloses a method for determining a traffic dependent signal program using signal programs with randomly generated traffic signal control patterns In a first step, a first order regression model is fitted to the averaged response over some restricted areas of the factor space steepest gradient using the first order model. In a third step, the steepest slope is searched along the path until no further improvement in the middle response can be observed. In a fourth step, the most advantageous answer is estimated on the basis of the first three steps. In a fifth step, steps 1 to 4 are repeated over a new area. If a curvature is detected, a finer search must be performed using a second order experimental design. Finally, the optimal solution is reported.

The Wong, Law, Tong: "Determination of Optimal Successor Function in Phase-based Control Using Neural Network ", IEEE, Proceedings of the 1996 Intelligent Vehicles Symposium, 19.-20.09.1996, Pp. 120-125 is a method for determining a traffic-dependent signal program known, which first Signal programs with random generated light signal control patterns used. In a light signal controlled Crossing with eight access roads, the traffic is through, for example controlled nine signal phases with 14 pairs of incompatible phases. The number of neurons in the input and output layers are Accordingly, eight or 14. The number of neurons in the hidden For example, layer is assumed to be 20. There will be 400 quantities generated by random flow patterns, with the delay-minimizing, optimal successor function is determined using a Software packages SIGSIGN. The first 200 data sets become training of the neural network while the remaining 200 Sets of data for an effectiveness test the method of reproducing the optimal successor function be used.

It It is therefore the object of the invention to provide a method for determining a traffic-adapted Signal program to specify which flexible to different optimization criteria adaptable is and for complex Traffic light systems is suitable.

The Task is achieved by a method of the type mentioned above with the features of the claim 1 solved.

According to claim 1, a method is applied which is based on the foundations of evolution based. Signal programs of the first order become signal programs assembled second order, and then all signal programs via a predetermined quality function a quality value assigned. The best signal programs will be back afterwards for the Generating new signal programs used, which are re-evaluated. This cycle continues until the quality does not improve more improved. With the best signal program then the traffic signal system operated. Such a determination of a traffic-adapted signal program has the advantage that the Requirements for the quality function are low, as they are not difficult in the context of optimization subject to mathematical operations.

By the reduction of the signal programs according to claim 2 becomes the required computing time reduced without the Results are worse.

By the random factor, as he when generating signal programs second Order according to claim 3 taken into account will be optimized adaptation strategies even for complex Won questions.

The special formation of the second-order signal programs according to claim 4 gives a fast distribution of high quality values evaluated properties among the examined signal programs.

According to claim 5 and 6 become the minimum release time and the minimum wait time, respectively used as easy-to-implement quality features.

In an advantageous embodiment of the method according to claim 7 become fixed boundary conditions for the signal programs specified, for example, minimum release times for individual traffic to obtain.

Based of exemplary embodiments the invention will be explained in more detail with the aid of the figures of the drawing.

there demonstrate

1 a structogram of the sequence of the method according to the invention,

2 a scheme for the formation of second-order signal programs from two first-order signal programs.

A signal program SP is to be calculated which is optimal with respect to certain criteria. The signal program SP consists of a circulation time Tu and a fixed number n of signal groups SGi or their release time start tgbi and release time end tgei (modulo the circulation time Tu):
SP = SP (Tu, tgb1, tge1, ..., tgbn, tgen).

• • Umlaufzeit Tu
• • Zwischenzeiten tzz,
• • Versatzzeiten tvb und tve,
• • Mindestgrünzeiten tmin,
• • Verkehrsbelastungen B,
• • Sättigungsverkehrsstärken S,
• • Auslastungsgraden A,
• • Zeitmehrbedarfswerten tmb..

The calculation depends on the following output data:

• • circulation time Tu
• • split tzz,
• • Offset times tvb and tve,
• • minimum green times tmin,
• • traffic loads B,
• Saturation traffic intensities S,
• • utilization levels A,
• • Time extra requirements tmb ..

The Data of the traffic load B, the saturation traffic S, of the Degree of utilization A and the additional time value tmb determine the required release time length tgerf. So you only indirectly enter into the calculation.

The Offsets tvb and tve are always all effective and accurate respected. The associated signal groups are usually in the same phase.

is because of too hard boundary conditions, the calculation of an optimal solution impossible, so the calculation method should provide a solution, which is a Optimum possible approaching. Are the preconditions contradictory, this is how the process should break off and give no result at all.

It should over it out possible be any, but consistent, subsets of a signal program pretend to complement her allow. Specifications can both fixed release windows and release start / release times be individual signal groups. The procedure should then try the remaining part to be set according to the required Strategy to complement optimally.

• 1. "Maximale Freigabezeit" In einem ersten Schritt werden die erforderlichen Freigabezeiten aufgrund von Belastungen und Auslastungsgraden berechnet. Liegen keine Belastungen vor, so wird die erforderliche Freigabezeit gleich der minimalen Freigabezeit gesetzt. Die Freigabezeiten der einzelnen Signalgruppen werden nun so berechnet, daß ihre Verhältnisse untereinander möglichst genau den gegenseitigen Verhältnissen der zugehörigen erforderlichen Freigabezeiten entsprechen. Ein solchermaßen berechnetes Signalprogramm wird in der Regel keine oder nur geringe Reservezeiten aufweisen!
• 2. "Minimale Wartezeit" Die Freigabezeiten des Signalprogramms werden so bemessen, daß die Wartezeiten minimiert werden. Diese Wartezeitminimierung läßt sich aber nur auf Signalgruppen anwenden, für die Belastungen und Auslastungsgrade vorgegeben werden. Alle anderen Signalgruppen werden mit Mindestfreigabezeit eingepaßt.

Various strategies are available, for example:

• 1. "Maximum release time" In a first step, the necessary release times are calculated on the basis of loads and utilization levels. If there are no loads, the required release time is set equal to the minimum release time. The release times of the individual signal groups are now calculated so that their relationships with each other as closely as possible correspond to the mutual ratios of the associated required release times. As a rule, a signal program calculated in this way will have little or no reserve time!
• 2. "Minimum waiting time" The release times of the signal program are calculated so that the waiting times are minimized. However, this minimization of waiting time can only be applied to signal groups for which loads and load factors are specified. All other signal groups are fitted with minimum release time.

In principle, a signal program of the circulation time Tu is defined by the release time start and release time end points of all its signal groups. A signal program without reserve times is already completely defined by its release time start points. The missing endpoints can be calculated from the start points by using the split times: tgj: = min {t = || tGBj - tGBi ||; SGj is hostile to SGi}, 1≤j≤n.
with the distance || t2 - t2 || : = [(t2 + Tu) - t1 - tzz (SG1, SG2)] mod Tu.

This the process takes advantage of the number of optimized ones To reduce parameters and hence the rate of convergence to increase. A signal program thus consists of a set of release time start points, which can vary over a time interval of length Tu. With This parameter set contains the basic structure of the signal program (Split times, phase split). He is one, so to speak reduced version of the signal program, as it has the facilities or possibilities a signal program includes. A complete signal program, however Still has to meet some other structural features. The release times of individual signal groups may not be maximal and the start / end points depend on the release times u.U. above Offset times from each other.

The sequence of the calculation takes place as follows (schematically in 1 shown):
The method starts with m first order signal programs, where m = n * 3 with n = number of signal groups. Each of these signal programs is generated purely at random; ie its n release time start points are determined with a random number generator. From the resulting first-order reduced signal programs, the complete first-order signal programs are developed. These complete signal programs are evaluated via the quality function. The evaluation of the complete signal programs and the formation of new signal programs with reduced signal programs leads to an acceleration of the procedure.

• 1. "Maximale Freigabezeit". Jeder Signalgruppe Sgi wird ein Gewicht Gi = tgerfi/tgerfsum zugeordnet (1≤j≤n, ΣGi=1). Entsprechend dieser Gewichtung wird die Grünzeitbemessung vorgenommen, indem folgende Zielfunktion Q angewendet wird:
  

Possible quality functions can be used:

• 1. "Maximum release time". Each signal group Sgi is assigned a weight Gi = tgerfi / tgerfsum (1≤j≤n, ΣGi = 1). According to this weighting, the green time measurement is performed by applying the following objective function Q:
  

die Summe all dieser Intervalle. Der erste Summand Q1 der Funktion ist für die relative Gewichtung der Freigabezeitanteile untereinander zuständig, wogegen Q2 dafür sorgt, daß die Freigabezeitsumme über alle Signalgruppen SG maximiert wird.Here are tgi the calculated release time intervals and

the sum of all these intervals. The first summand Q1 of the function is responsible for the relative weighting of the release time shares among each other, whereas Q2 ensures that the release time sum is maximized across all signal groups SG.

The Constant α sets determine what influence the have two summands Q1 and Q2 on the overall result: for small α, the green times preferably exactly want to keep the relative weights with each other, even if this green spaces be given away; large α, on the other hand rather bring the exhausting all green times at the expense of relative weighting. With increasing α, reserve times become increasingly unlikely.

zeigen, wie α dimensioniert werden muß, damit die beiden Summanden Q1 und Q2 auf die Gesamtqualität Q etwa gleichen Einfluß haben!The partial derivatives

show how α must be dimensioned so that the two summands Q1 and Q2 have approximately the same influence on the overall quality Q!

folgt über einen Vergleich von Q1 und Q2:

• 2. "Minimierung Wartezeit". Die mittlere Wartezeit einer Signalgruppe berechnet sich nach Ning Wu („Wartezeiten an festzeitgesteuerten Lichtsignalanlagen unter zeitlich veränderlichen (instationären) Verkehrsbedingungen", Straßenverkehrstechnik, Heft 3, 1992) zu:
  

wobei

• NGE = mittlerer Stau bei Grün-Ende,
• λi = tgi/Tu, Freigabezeitanteil,
• qi = (Si/3600)·λi = (Si/3600)·(tgi/Tu), Leistungsfähigkeit
  
• und xi = Bi/(Si·tgi), mittlerer Auslastungsgrad.

Out

follows through a comparison of Q1 and Q2:

• 2. "minimize waiting time". The mean waiting time of a signal group is calculated according to Ning Wu ("Waiting times on fixed-time traffic light systems under time-variable (transient) traffic conditions", Straßenverkehrstechnik, Issue 3, 1992) to:
  

in which

• NGE = medium traffic jam at green end,
• λi = tgi / Tu, release time share,
• qi = (Si / 3600) · λi = (Si / 3600) · (tgi / Tu), performance
  
• and xi = Bi / (Si · tgi), average duty cycle.

The quality function is the sum of the waiting times across all signal groups:

The next steps are repeated for each cycle:
Further k (for example, k = m) second-order signal programs are formed by randomly selecting two first-order signal programs each and combining them in the following manner, as shown in FIG 2 is shown.

Randomly, two numbers i and j (i <j) are chosen between 1 and n (= number of signal groups). For one of the second-order signal programs SP ( 3 ) the release times 1 to i (A1) and j + 1 to n (A3) of the first ( 1 ) of the two signal programs SP of the first order, and the release times i + 1 to j (B2) of the second ( 2 ). The second signal program of second order ( 4 ) is formed from the complementary release times (B1, A2, B3).

By this type of formation of second-order signal programs successful features exponentially fast.

One certain percentage of the second-order signal programs additionally changed i.e. that one or several of their release time start points are moved randomly.

This is done by a certain percentage of the generated signal programs second Order one or more of the release time start points arbitrarily moved. In general, the so changed Signal programs of poor quality; in rare cases can such a change but also to a sudden improvement of the signal program to lead.

The Second-order signal programs now likewise become complete signal programs trained and evaluated.

In the next step, the k worst signal programs become the meanwhile m + k signal program me (first and second order) eliminated. The number of signal programs is thus decimated (by half at m = k) and, as before, again counts m individuals and the process begins again.

The Signal programs with the highest quality value will over observed several repetitions. Its quality value remains over at least 30 repetitions constant, it is assumed that the associated signal program optimal. The procedure is then aborted and the complete signal program with the highest quality value is considered a solution hergenommen. If the quality value has changed (improved), so the calculation is continued by k new signal programs SP are generated.

is no consistency of quality also over Repeat the procedure is aborted after 1000 passes.

Claims (7)

Method for determining a traffic-dependent signal program (SP) for Signal groups of traffic signal installations, the signal program (SP) by a sequence of enable times (tg) for the different signal groups is characterized characterized by the combination of the following Steps: a) there are m signal programs (SP) of the first order each with a time history consisting of sections of randomly selected release times (tg) formed, b) there will be two signal programs each (SP) first order by mutual exchange of the sections (A1 ... A3, B1 ... B3) of the respective time courses (tg) from a total of k derived second-order signal programs (SP), c) the m + k signal programs (SP) of the first and the second order each becomes a quality value (Q (tg)), which takes into account currently determined traffic data, d) From the m + k signal programs, the m signal programs with the best quality values (Q (tg)) selected, interpreted as signal programs (SP) of the first order and the steps b) to d) cyclically repeated until the quality value (Q (tg)) the highest evaluated signal program (SP) via a given number of repetitions no longer changes, and e) with the highest signal program (SP), the traffic light system is operated. Method for determining a traffic-dependent signal program (SP) for Signal groups of traffic signal systems according to claim 1, thereby in that before forming the second-order signal programs (SP) from the signal programs (SP) first-order versions reduced to only the release start times (tgb) over time, formed, that from these reduced versions reduced signal programs (SP) second order be formed and that out The reduced versions then formed complete signal programs which are the quality value is assigned. Method for determining a traffic-dependent signal program (SP) for Signal groups of traffic signal systems according to one of claims 1 or 2, characterized in that a Part of the signal programs (SP) before the evaluation at least partially fortuitously changed becomes. Method for determining a traffic-dependent signal program (SP) for signal groups of traffic signal systems according to one of Claims 1, 2 or 3, characterized in that the time profiles of the two signal programs (SP) of the first order (SP) 1 . 2 ) are divided into three sections (A1 ... A3, B1 ... B3) and the signal program of the second order ( 3 ) results from the first (A1) and the third section (A3) of the one of the two signal programs (SP) of the first order and the second section (B2) of the other of the two signal programs (SP) of the first order. Method for determining a traffic-dependent signal program (SP) for Signal groups of traffic signal systems according to one of claims 1 to 4, characterized in that the quality value (Q (tg)) is determined by the maximum release time. Method for determining a traffic-dependent signal program (SP) for Signal groups of traffic signal systems according to one of claims 1 to 4, characterized in that the quality value (Q (tg)) is used by the minimum waiting time. Method for determining a traffic-dependent signal program (SP) for Signal groups of traffic signal systems according to one of claims 1 to 5, characterized in that fixed Boundary conditions for the Time course of the signal programs (SP) can be specified.