EP1758065A2 - Method for forecasting a traffic condition in a road network and traffic management centre - Google Patents

Abstract

The method involves determining traffic volumes with respect to a time point at supporting points (S1, S2, S3) arranged over a road network, where the traffic volumes are combined at a data set. Other two data sets are formed in a similar way and characterize the traffic volumes in the supporting points with respect to other time points. The data sets are combined to classes, which characterize similar traffic situations in the road network. The classes are used for forecasting a traffic condition and/or detecting an accident. Independent claims are also included for the following: (1) a traffic management centre, in which a method for forecasting a traffic condition in a road network is implemented (2) a method for controlling a traffic in the road network.

Description

The invention relates to a method for forecasting a future traffic condition characterizing a road network with a plurality of network nodes and links and / or for detecting an accident in the road network. The invention also relates to a traffic management center.

Out DE 101 46 398 A1 A method is known for controlling traffic light signaling at a node of a road traffic network in which traffic conditions already acquired at an earlier point in time at the node are clustered into classes, a characteristic traffic state being determined as a representative for each class of traffic states. When a current traffic status occurs, a defined metric determines the characteristic traffic state closest to the current traffic state and a signal program for the traffic signal transmitter assigned to this characteristic traffic state is executed. The clustering limits the variety of traffic conditions occurring at the node to a reasonably limited number of typically occurring traffic conditions, the characteristic traffic conditions. This in turn restricts the number of signal programs to be stored, which are adapted to the characteristic traffic conditions. If the current traffic status is also included in the clustering for future evaluations, this results in a method for controlling a light signal transmitter at a node, which learns dynamically from traffic state to traffic state in order always to select the best possible signal program in future traffic states.

For a forecast of traffic conditions or traffic conditions in a road network, it is also known on a single road so-called hydrographs, which are time series of traffic loads one day, usually in equidistant sections of 15 min., First - for reasons of data reduction - by interpolation using predetermined Gaussian curves to filter and the result of the filtering, so-called amplitude vectors, then by a Classify clustering into classes of similar amplitude vectors.

The invention has for its object to provide a method for forecasting future traffic conditions and / or for detecting an accident, with which the precision or quality of the prognosis is improved compared to known methods.

• zur mehreren über das Straßennetz verteilten Stützstellen ein Verkehrsaufkommen bezüglich eines zurückliegenden Zeitpunktes ermittelt wird,
• die Verkehrsaufkommen zu einem Datensatz zusammengefasst werden,
• in gleicher Weise weitere Datensätze gebildet werden, die das Verkehrsaufkommen an den Stützstellen bezüglich weiterer zurückliegender Zeitpunkte beschreiben,
• die Datensätze zu Klassen zusammengefasst werden, die einander ähnliche Verkehrssituationen im Straßennetz beschreiben, und
• die Klassen zur Prognose des künftigen Verkehrszustandes bzw. zur Erkennung des Störfalles verwendet werden.

This object is achieved according to a first variant of the invention in that

• for the several nodes distributed over the road network, a traffic volume with respect to a previous point in time is determined,
• the traffic is combined into one record,
• in the same way, further data sets are formed, which describe the traffic volume at the interpolation points with respect to other past times,
• the data sets are grouped together to describe similar traffic situations in the road network, and
• the classes are used to predict the future traffic condition or to detect the incident.

In this method, clustering of traffic conditions deposited in the data sets takes place not only locally at a single intersection but also at interpolation points distributed over the road network.

• zu mehreren über das Straßennetz verteilten Stützstellen ein Verkehrsverlauf gebildet wird, der jeweils eine Folge von Werten bezüglich des Verkehrsaufkommens an dieser Stützstelle innerhalb eines zurückliegenden Zeitintervalls enthält,
• die Verkehrsverläufe zu einem Datensatz zusammengefasst werden,
• in gleicher Weise weitere Datensätze gebildet werden, die das Verkehrsaufkommen an den Stützstellen bezüglich weiterer zurückliegender Zeitintervalle beschreiben,
• die Datensätze zu Klassen zusammengefasst werden, die einander ähnliche Verkehrssituationsverläufe im Straßennetz beschreiben, und
• die Klassen zur Prognose des künftigen Verkehrszustandes bzw. zur Erkennung des Störfalles verwendet werden.

According to a second variant, the invention solves the stated object in that

• a traffic pattern is formed for a plurality of nodes distributed over the road network, each of which contains a sequence of values relating to the traffic volume at this node within a past time interval,
• the traffic courses are combined to a data record,
• in the same way, further data sets are formed, which describe the traffic volume at the interpolation points with respect to further past time intervals,
• the datasets are combined into classes describing similar traffic situations in the road network, and
• the classes are used to predict the future traffic condition or to detect the incident.

In this variant, the clustering also takes place based on data sets, in particular vectors, which are based on many nodes distributed over the road network. In contrast to the first variant, the data sets or vectors in the second variant do not only consist of individual values of the interpolation points, but of local traffic patterns, for example hydrographs, which concern the individual interpolation points.

Common to both variants of the invention is the advantage that by including a large number of nodes distributed over the road network, dependencies or correlations of the traffic loads between the individual nodes are included in the classification, which are then available later for use - for example in the form of class representatives - for forecasting purposes stand. The prediction methods according to the invention with their special clustering method include the dependencies or correlations of the traffic conditions of individual links or roads of the road network into the evaluation. Another advantage is that the methods for controlling failed street-side detectors are suitable, for example by Substitute values are formed for the classes or clusters or for class representatives formed for this purpose.

An accident detection can in both cases be done in particular by comparing the current traffic volume or a current traffic flow with classes or their representatives, and concluding that an accident is the result of the absence of a similar accident-free class.

Thus, in the second variant of the invention, for example, the data sets or vectors used for clustering contain not only time profiles of the traffic volume or - possibly in amplitude representation interpolated - hydrographs of a single measuring point in the road network, but it is the hydrographs or time courses of all or many measuring points in the road network in a vector or Record included. Preferably, the data set or vector for each link or street contains exactly one hydrograph. In particular, many data sets or vectors are formed over several days, which are then used for clustering.

The traffic courses or hydrographs are, in particular, daily diurnal lines encompassing one day interval.

In the context of the invention, a link is, for example, a road, a tunnel, a bridge or the like, that is to say a trafficable connection for motor vehicles between two network nodes, which is understood to mean, in particular, an intersection, a junction, a driveway or the like.

According to a preferred embodiment of the second variant of the invention, the traffic courses before being combined into the data records are interpolated by a plurality of curve courses, in particular by Gaussian bell curves. This method is for a single measuring point in the beginning described technical article. Preferably, the amplitudes, and preferably only these, of the Gaussian bell curves are used to form the data sets. The following preferred embodiments relate to both variants of the invention:

The values of the traffic volume flowing into the data sets, for example a traffic flow measured in number of motor vehicles per unit of time, are measured in particular by a plurality of roadside detectors.

Essential to the invention is that a plurality of nodes distributed over the road network are used to construct the datasets or the vector. The support points are in particular distributed over the road network in such a way that the traffic volume at a plurality of links and / or network nodes in the road network is determined.

In particular, the interpolation points have a distance of at least 1 km, in particular of at least 3 km, relative to one another.

In a most preferred embodiment, for each class, a characteristic class representative is used, which is used for the prediction.

For example, the class representatives or prototypes are the center of all the amplitude vectors aggregated in the cluster or class. A good cluster ring is given if each amplitude vector is as close as possible to the prototype of the cluster to which it is assigned. For example, the distance between any amplitude vector and the prototype can be defined by a weighted maximum norm as a metric. For a given set of prototypes, the class or cluster division that assigns each amplitude vector to the nearest prototype is optimal. As part of an optimization calculation, a cluster division can be calculated, in which alternately the Cluster centers and cluster divisions are recalculated. Once a local optimum is reached, class scheduling no longer changes in successive iterations so that the procedure can be terminated. Details of this type of cluster method are described in the above-mentioned technical article.

For a short-term or long-term prognosis, according to a preferred embodiment, a current traffic situation is compared with the characteristic class representatives or prototypes. For example, the forecasting process selects from a larger set of class representatives a class representative well suited to the current situation and bases the prediction on them.

Preferably, a metric is defined for comparing the data records with one another and / or for comparing the data records with a current traffic situation, in particular a standard. For example, the metric is a measure of the mutual location or spacing of the dataset vectors.

The predicted traffic condition can be displayed on a display device in graphical representation.

When classifying or clustering, disjoint classes are preferably used.

In the context of the invention is also a method for influencing the traffic in a road network, wherein by means of one of the above methods, a prognosis of a future traffic condition is determined and using the predicted traffic condition driving a roadside signaling device - by an operator or automatically - is made.

The method according to the invention is preferably implemented in a traffic management center.

FIG 1

ein Straßennetz in schematischer Darstellung, für das eine Prognose erstellt werden soll,

FIG 2

den schematischen Ablauf eines ersten Ausführungsbeispiels eines Verfahrens nach der Erfindung, und

FIG 3

den schematischen Ablauf eines zweiten Ausführungsbeispiels eines Verfahrens nach der Erfindung.

Two embodiments of the method according to the invention will be explained in more detail with reference to Figures 1 to 3. Show it:

FIG. 1

a road network in a schematic representation, for which a prognosis is to be created,

FIG. 2

the schematic sequence of a first embodiment of a method according to the invention, and

FIG. 3

the schematic sequence of a second embodiment of a method according to the invention.

1 shows a road network 1 in which a plurality of network nodes K1, K2, K3 are interconnected by roads or links L1, L2, L3. For monitoring and generating traffic forecasts for the traffic in the road network 1, a traffic management center 3 is provided, in which a display device 5 for displaying current and predicted traffic conditions of the entire road network 1 in graphic or topological form is formed. The traffic management center 3 is prepared for remotely influencing signaling devices 7 present at the nodes K1, K2, K3, of which only one is indicated schematically. In the road traffic network 1 several measuring points or detectors M1, M2, M3 are installed, each of which measures a local traffic flow.

In the first exemplary embodiment illustrated in FIG. 2 relating to the method of the second embodiment variant according to the invention, the measured values of the detectors M1, M2, M3 are typically at n = 10... 100 different support points S1, S2, S3 (see FIG Road Traffic Network 1 measured many daily diurnal or traffic patterns 12,13,14,15,16,17,18, 19,20. The schematically indicated measured value extrapolation 11 expresses that the measuring points do not have to be identical to the supporting points S1, S2, S3. Rather, in the measured value extrapolation 11, a Extrapolation take place at locations where there are no detectors. For this purpose, corresponding methods are known, for example a so-called current route determination with a subsequent allocation-based dynamic measured value propagation.

Since these are daily gaits recorded on several days d ₁ , d ₂ , d ₃ ..., the corresponding time interval Δt is 24 hours. The many hundreds of individual hydrographs 12 to 20 of the individual links L1, L2, L3 are first interpolated with a small number N of curves 23, 24, 25 (typically N = 4... 6) to determine the number of parameters for describing the To reduce hydrographs 12 to 20 drastically. The Gaussian bell curves 23, 24, 25 used for interpolation are shown for illustrative reasons only for one of the traffic courses 12 (day d ₃ ). The hydrographs 12 to 20 each represent a plot of the traffic volume or traffic flow V over the time t. The result of the interpolation are the amplitudes a ₁ , a ₂ , a ₃ ,... Of the Gaussian bell curves 23 to 25 used with the same width:

wobei $${\mathrm{G}}_{\mathrm{ti}}\mathrm{=}\exp \left(\mathrm{-}\frac{{\left(\mathrm{t}\mathrm{-}\mathrm{ti}\right)}^2}{\mathrm{2}{\mathrm{\sigma }}^{\mathrm{2}}}\right)$$

und $${\mathrm{t}}_{\mathrm{i}}\mathrm{=}\left(\mathrm{i}\mathrm{-}\mathrm{1}\right)\frac{\mathrm{\Delta t}}{\mathrm{N}\mathrm{-}\mathrm{1}}$$

$$\mathrm{\sigma }\mathrm{=}\frac{\mathrm{\Delta t}}{\mathrm{2}\left(\mathrm{N}\mathrm{-}\mathrm{1}\right)}$$

Specifically, the determination of the amplitudes a _i (i = 1.2...) Takes place by means of linear regression. The hydrographs can be written as: $${\mathrm{q}}_{\mathrm{t}}\mathrm{=}{\displaystyle \underset{\mathrm{i}=1}{\overset{\mathrm{N}}{\Sigma }}}{\mathrm{G}}_{\mathrm{ti}}{\mathrm{a}}_{\mathrm{i}}$$

in which $${\mathrm{G}}_{\mathrm{ti}}\mathrm{=}\exp \left(\mathrm{-}\frac{{\left(\mathrm{t}\mathrm{-}\mathrm{<figure-callout\; id="2"\; label="ti"\; filenames="imgf0003.png"\; state="\{\{state\}\}">ti</figure-callout>}\right)}^2}{\mathrm{2}{\mathrm{<figure-callout\; id="2"\; label="\sigma "\; filenames="imgf0003.png"\; state="\{\{state\}\}">\sigma </figure-callout>}}^{\mathrm{2}}}\right)$$

and $${\mathrm{t}}_{\mathrm{i}}\mathrm{=}\left(\mathrm{i}\mathrm{-}\mathrm{1}\right)\frac{\mathrm{.delta.t}}{\mathrm{N}\mathrm{-}\mathrm{1}}$$

$$\mathrm{\sigma }\mathrm{=}\frac{\mathrm{.<figure-callout\; id="2"\; label="delta.t"\; filenames="imgf0003.png"\; state="\{\{state\}\}">delta.t</figure-callout>}}{\mathrm{2}\left(\mathrm{N}\mathrm{-}\mathrm{1}\right)}$$

$$\tilde{\mathrm{G}}\mathrm{=}{\left({\mathrm{G}}^{\mathrm{T}}•\mathrm{G}\right)}^{\mathrm{-}\mathrm{1}}{\mathrm{G}}^{\mathrm{T}}$$

wobei G^T die Transponierte und G̃ die Pseudoinverse der Matrix G ist.The linear regression yields: $${\mathrm{a}}_{\mathrm{i}}\mathrm{=}{\displaystyle \underset{\mathrm{t}}{\Sigma }}{\tilde{\mathrm{G}}}_{\mathrm{it}}\cdot {\mathrm{q}}_{\mathrm{t}}$$

$$\stackrel{\mathrm{~}}{\mathrm{G}}\mathrm{=}{\left({\mathrm{G}}^{\mathrm{T}}•\mathrm{G}\right)}^{\mathrm{-}\mathrm{1}}{\mathrm{G}}^{\mathrm{T}}$$

where G ^{T is} the transpose and G is the pseudoinverse of the matrix G.

zusammengefasst. Die Doppel-Indizes der Amplituden a stehen dann für einen bestimmten Tag und eine bestimmte Glockenkurve; b und c sind im Beispiel für die Amplituden von den anderen Links verwendet.In a next step, an entirety of the interpolated hydrographs 12 to 20 of all links L1, L2, L3 of the road network 1, in the embodiment illustrated here, their Gauss bell interpolation values, ie amplitudes a ₁ , a ₂ , a ₃ , ... to a single Feature vector or dataset ${\overrightarrow{\mathrm{a}}}_{\mathrm{d}\mathrm{1}}$

summarized. The double indices of the amplitudes a then stand for a certain day and a specific bell curve; b and c are used in the example for the amplitudes of the other links.

wird der entstandene Merkmalsvektor oder Datensatz ${\overrightarrow{\mathrm{a}}}_{\mathrm{d}\mathrm{1}},$

der genau eine Verkehrssituation des Straßennetzes 1 beschreibt, in einem Clusterverfahren verarbeitet. Als Clusterverfahren kann beispielsweise das im Folgenden dargestellte Verfahren mit den in FIG 2 angegebenen Vektoren ${\overrightarrow{\mathrm{a}}}_{\mathrm{d}\mathrm{1}},{\overrightarrow{\mathrm{a}}}_{\mathrm{d}2},{\overrightarrow{\mathrm{a}}}_{\mathrm{d}3},\mathrm{...}$

angewendet werden. Die Clusterung oder Klassenbildung ist in FIG 2 noch schematisch angedeutet. Die Vektoren sind dort mit ihren Endpunkten symbolisiert, die real in einem N x n-dimensionalen Raum liegen. Die Vektoren ${\overrightarrow{\mathrm{a}}}_{\mathrm{d}\mathrm{1}},{\overrightarrow{\mathrm{a}}}_{\mathrm{d}2},{\overrightarrow{\mathrm{a}}}_{\mathrm{d}3},\mathrm{...}$

werden zu Clustern oder Klassen C1,C2,C3 zusammengefasst, die einander ähnliche Verkehrssituationsverläufe im Straßennetz 1 beschreiben. Zu jeder Klasse C1,C2,C3 wird ein charakteristischer Klassenrepräsentant R1,R2,R3, ... ermittelt, der für die jeweilige Klasse typisch oder aussagekräftig ist. Mit anderen Worten: Innerhalb jeder Klasse wird ein Prototyp festgelegt, der als Klassenrepräsentant für spätere Langzeitverkehrsprognosen verwendbar ist. Zum Vergleich der Datensätze untereinander, also für das Clustering, wird als Metrik D eine gewichtete Maximumsnorm verwendet.Together with other feature vectors or datasets created in the same way ${\overrightarrow{\mathrm{a}}}_{\mathrm{d}\mathrm{2}}.{\overrightarrow{\mathrm{a}}}_{\mathrm{d}\mathrm{3}}$

becomes the resulting feature vector or record ${\overrightarrow{\mathrm{a}}}_{\mathrm{d}\mathrm{1}}.$

which describes exactly one traffic situation of the road network 1, processed in a cluster procedure. As a cluster method, for example, the method shown below with the vectors shown in FIG 2 ${\overrightarrow{\mathrm{a}}}_{\mathrm{d}\mathrm{1}}.{\overrightarrow{\mathrm{a}}}_{\mathrm{d}2}.{\overrightarrow{\mathrm{a}}}_{\mathrm{d}3}.\mathrm{...}$

be applied. The clustering or class formation is indicated schematically in FIG. The vectors are symbolized there with their endpoints, which lie real in a N x n-dimensional space. The vectors ${\overrightarrow{\mathrm{a}}}_{\mathrm{d}\mathrm{1}}.{\overrightarrow{\mathrm{a}}}_{\mathrm{d}2}.{\overrightarrow{\mathrm{a}}}_{\mathrm{d}3}.\mathrm{...}$

are grouped into clusters or classes C1, C2, C3 which describe similar traffic situations in the road network 1. For each class C1, C2, C3 a characteristic class representative R1, R2, R3, ... is determined which is typical or meaningful for the respective class. In other words, within each class, a prototype is defined as the class representative is usable for later long-term traffic forecasts. To compare the data sets with one another, ie for clustering, the metric D is a weighted maximum standard.

The exemplary clustering method can be formulated mathematically as follows:

A division of d _max amplitude vectors into c clusters can be represented by a cxd _max -dimensional matrix whose elements are defined as follows: $${\mathrm{U}}_{\mathrm{id}}\mathrm{:}=\mathrm{\{}\begin{array}{l}\mathrm{1}\mathrm{.}\mathrm{if}{\overrightarrow{\mathrm{a}}}_{\mathrm{d}}\mathrm{the\; i}\mathrm{-}\mathrm{cluster\; is\; assigned}\\ \mathrm{0}\mathrm{.}\mathrm{otherwise}\mathrm{,}\end{array}$$

Each cluster i = 1, ..., c is assigned a prototype or representative R _i , which results as the midpoint of all the amplitude vectors aggregated in the cluster: $${\overrightarrow{\mathrm{R}}}_{\mathrm{i}}\mathrm{:}\mathrm{=}\frac{{\displaystyle \underset{\mathrm{d}\mathrm{=}{\mathrm{d}}_{\mathrm{1}}}{\overset{{\mathrm{d}}_{\mathrm{Max}}}{\Sigma }}}{\mathrm{U}}_{\mathrm{id}}{\overrightarrow{\mathbf{a}}}_{\mathrm{d}}}{{\displaystyle \underset{\mathrm{d}\mathrm{=}{\mathrm{d}}_{\mathrm{1}}}{\overset{{\mathrm{d}}_{\mathrm{Max}}}{\Sigma }}}{\mathrm{U}}_{\mathrm{id}}}$$

Good clustering is given when each amplitude vector is as close as possible to the prototype of the cluster to which it is assigned. The distance D between the d-th amplitude vector and the i-th prototype R _i is defined, for example, by the weighted maximum norm δ _id . In the simplest case, the weights are all the same.

For a given set of prototypes, the clustering that maps each amplitude vector to the nearest prototype is optimal: $${\mathrm{U}}_{\mathrm{id}}\mathrm{:}=\mathrm{\{}\begin{array}{l}\mathrm{1}\mathrm{.}\mathrm{if}{\delta }_{\mathrm{id}}={\min }_{\mathrm{j}\mathrm{=}\mathrm{1}\mathrm{.}\mathrm{...}\mathrm{c}}\left\{{\delta }_{\mathrm{jd}}\right\}\\ \mathrm{0}\mathrm{.}\mathrm{otherwise}\mathrm{,}\end{array}$$

A (locally) optimized cluster division can now be calculated by recalculating the cluster centers using G1.8 and the cluster divisions using G1.9, starting from any initial cluster division. The process is aborted as soon as the cluster division barely changes in successive iterations.

enthält demzufolge keinen Verkehrsverlauf, sondern jeweils einen einzelnen Messwert des Verkehrsflusses V zu den n Stützstellen (n-dimensionaler Raum).3, instead of the traffic courses 12 to 15 at different times t ₁ , t ₂ , t _3, the traffic volume V, ie in each case a single value, at the individual support points S1, S2, S3 determined. Every record or vector ${\overrightarrow{\mathrm{a}}}_{\mathrm{t}\mathrm{1}}\mathrm{.}{\overrightarrow{\mathrm{a}}}_{\mathrm{t}\mathrm{2}}\mathrm{.}{\overrightarrow{\mathrm{a}}}_{\mathrm{t}\mathrm{3}}\mathrm{.}\mathrm{...}$

contains therefore no traffic course, but in each case a single measured value of the traffic flow V to the n support points (n-dimensional space).

werden in analoger Weise zu Klassen C1,C2,C3, ... geclustert, was schematisch dargestellt ist.The corresponding vectors ${\overrightarrow{\mathrm{a}}}_{\mathrm{t}\mathrm{1}}\mathrm{.}{\overrightarrow{\mathrm{a}}}_{\mathrm{t}\mathrm{2}}\mathrm{.}{\overrightarrow{\mathrm{a}}}_{\mathrm{t}\mathrm{3}}\mathrm{.}\mathrm{...}$

are clustered in an analogous manner to classes C1, C2, C3,..., which is shown schematically.

Claims (15)

Method for forecasting a future traffic condition characterizing a road network (1) having a plurality of network nodes (K1, K2, K3) and links (L1, L2, L3,...) And / or detecting a fault in the road network (1), wherein a traffic volume (V1 _t1 , V2 _t1 , V3 _t1 ,...) with respect to a past point in time (t ₁ ) is determined for several support points (S1, S2, S3,...) distributed over the road network (1), - The traffic (V1 _t1 , V2 _t1 , V3 _t1 , ...) to a record $\left({\overrightarrow{a}}_{t1}\right)$

be summarized - in the same way more records $\left({\overrightarrow{a}}_{\mathit{t}2},,{\overrightarrow{a}}_{\mathit{t}3},...\right)$

are formed, which describe the traffic volume (V1 _t2 , V2 _t2 ,...) at the interpolation points (S1, S2, S3,...) with respect to further previous times (t ₂ , t ₃ ), - the records $\left({\overrightarrow{a}}_{\mathit{t}1},,{\overrightarrow{a}}_{\mathit{t}2},,{\overrightarrow{a}}_{\mathit{t}3},...\right)$

to classes (C1, C2, C3, ...) describing similar traffic situations in the road network (1), - The classes (C1, C2, C3) are used to predict the future traffic condition or to detect the incident. Method for forecasting a future traffic condition characterizing a road network (1) with several network nodes (K1, K2, K3, ...) and links (L1, L2, L3,...) And / or detecting an accident in the road network ( 1), where a traffic course (14,17,20,...) is formed for a plurality of support points (S1, S2, S3,...) distributed over the road network (1), each of which has a sequence of values relating to the traffic volume (V) at this interpolation point (S1, S2, S3, ...) within a past time interval (Δt), - The traffic patterns (14,17,20, ...) to a record $\left({\overrightarrow{a}}_{d1}\right)$

be summarized - in the same way more records $\left({\overrightarrow{a}}_{d2},,{\overrightarrow{a}}_{d3},...\right)$

be formed, the traffic (V) at the reference points (S1, S2, S3, ...) with regard to further time intervals (Δt), - the records $\left({\overrightarrow{a}}_{\mathrm{d}\mathrm{1}},,{\overrightarrow{a}}_{\mathrm{d}\mathrm{2}},,{\overrightarrow{a}}_{\mathrm{d}\mathrm{3}},\mathrm{...}\right)$

to classes (C1, C2, C3) which describe similar traffic situations in the road network (1), - The classes (C1, C2, C3, ...) are used to predict the future traffic condition or to detect the incident. Method according to claim 2,
wherein as traffic patterns (12-20) hydrographs, in particular one daily interval comprehensive daily snaps are used. Method according to claim 2 or 3,
where the traffic patterns (12-20) before the summary to the records $\left({\overrightarrow{a}}_{d1},,{\overrightarrow{a}}_{d2},,{\overrightarrow{a}}_{d3},...\right)$

be interpolated by several curves (23,24,25, ...), in particular by Gaussian bell curves. Method according to claim 4,
where the amplitudes (a, b, c, ...) and preferably only these, the Gaussian bell curves to form the data sets $\left({\overrightarrow{a}}_{d1}.{\overrightarrow{a}}_{d2}.{\overrightarrow{a}}_{d3}....;{\overrightarrow{a}}_{\mathit{t}1}.{\overrightarrow{a}}_{\mathit{t}2}.{\overrightarrow{a}}_{\mathit{t}3}....\right)$

be used. Method according to one of the preceding claims,
wherein the traffic volume (V) is locally measured by a plurality of roadside detectors (M1, M2, M3, ...). Method according to one of the preceding claims,
wherein the interpolation points (S1, S2, S3, ...) are distributed over the road network (1) in such a way that the traffic volume (V) on a plurality of links (L1, L2, L3, ...) and / or Network node (K1, K2, K3, ...) in the road network (1) is determined. Method according to one of the preceding claims,
wherein the support points (S1, S2, S3, ...) have a distance of at least 1 km, in particular of at least 3 km, relative to one another. Method according to one of the preceding claims,
wherein for each class (C1, C2, C3, ...) a characteristic class representative (R1, R2, R3, ...) is determined which is used for the prediction. Method according to claim 9,
wherein a current traffic situation is compared with the characteristic class representatives (R1, R2, R3,...) for the prognosis. Method according to one of the preceding claims,
comparing the records $\left({\overrightarrow{a}}_{d1}.{\overrightarrow{a}}_{d2}.{\overrightarrow{a}}_{d3}....;{\overrightarrow{a}}_{\mathit{t}1}.{\overrightarrow{a}}_{\mathit{t}2}.{\overrightarrow{a}}_{\mathit{t}3}....\right)$

with each other and / or to compare the records $\left({\overrightarrow{a}}_{d1}.{\overrightarrow{a}}_{d2}.{\overrightarrow{a}}_{d3}....;{\overrightarrow{a}}_{\mathit{t}1}.{\overrightarrow{a}}_{\mathit{t}2}.{\overrightarrow{a}}_{\mathit{t}3}....\right)$

with a current traffic situation a metric (D) is defined, in particular a norm. Method according to one of the preceding claims,
wherein the predicted traffic condition is displayed in graphical form on a display device (5). Method according to one of the preceding claims,
where the classes are formed disjointly. Method for influencing the traffic in a road network, wherein by means of a method according to one of the preceding claims, a prognosis of a future traffic condition is determined and, using the predicted traffic condition, an activation of a roadside signaling device (7) is undertaken. Traffic management center (3), in which a method according to one of the preceding method claims is implemented.

Images