EP2047448B1 - Method and device for generating early warnings signalling traffic collapses at narrow points - Google Patents

Description

Field of the invention:

Traffic information services disseminated through media as diverse as broadcast, mobile, or the Internet are nowadays focused on describing the current traffic situation.

In general, however, the future information, ie. the traffic situation to be expected during the journey, which is more important for individual road users, especially in dynamic navigation. Therefore, there is already a first attempt to make traffic forecasts available online with a horizon of 30 or 60 minutes (www.autobahn.nrw.de). However, this approach has significant disadvantages.

On the one hand, the information is not available while driving and therefore can not be used for navigation purposes.

On the other hand, the prognosis is calculated with a deterministic simulation method and comes to a clear result for the traffic situation in 30 or 60 minutes, which due to the stochastic nature of traffic collapses often incorrect.

A recent research finding corroborates this last aspect ([1] Brilon, W .; Zurlinden, H .: Capacity of roads as a random variable. Road Traffic Technology 4/2004, pp. 164-172 ). The capacity of a road, especially a bottleneck, is not a fixed quantity, beyond which traffic inevitably comes to a standstill, but a random variable, ie. Although the probability of a traffic collapse increases with increasing traffic volume, the critical amount, and thus the time at which an event occurs, remains indeterminate as long as no event occurs. Therefore, a prognosis of the formation of traffic jams is in principle uncertain (if) and fuzzy (when). Intuitively, this is understandable as a collapse of unpredictable events such as abrupt braking or truck overtaking maneuvers can be triggered.

In the EP-A-1657 693 describe systems and methods for generating predictive models based on statistical machine learning techniques that can be used to predict traffic flow and congestion based on abstraction of a traffic system in a set of random variables. These methods include variables that determine the length of time and the location where a traffic congestion occurs, and the time it takes for a congestion to resolve. The monitoring data used include traffic flows and other context data such as time of day, day, week, holidays, school seasons, times of important events such as sporting events, holidays, weather conditions and construction sites.

DE 19954971 A 1 refers to a system for influencing the traffic flow of vehicles. Another part relates to the traffic situation detection means for detecting and outputting traffic-indicator-indicative data computer-aided traffic control center for receiving and evaluating the traffic-situation-indicative data for the purpose of obtaining issued traffic flow influencing data as well as a respective vehicle-side system part which receives and utilizes the traffic flow influencing data.

According to the invention, the traffic control center is set up to determine traffic flow interference influencing traffic flow disturbance amplitude data, which includes acceleration, speed and / or distance-related data for individual vehicles. The respective vehicle-side system part comprises longitudinal movement control means, to which the received traffic flow influencing data are fed and which make the vehicle longitudinal movement influencing control interventions. This approach can be used for. B. to optimize the traffic performance of a road network.

The US 202/120389 A1 discloses a method for predicting traffic conditions, implementing a prediction and assuming traffic disturbance situation in an area where test vehicles are not currently traveling, by sending flow data concerning the time, position, and traveled areas, which are then sent to a central device become. The central device accumulates the flow data in a database of traffic conditions and acceptance means, and also determines the traffic prediction of congestion in the observed areas of the flow vehicles, and can thus predict towards the flow vehicles and also determine information regarding the rearward traffic conditions. Thus, future events and events of the past can be detected.

Overview of the invention:

Thus, the object is to provide a method and a device that find a form of traffic forecast, which the above-mentioned. Overcome disadvantages and combine a number of advantages. The invention is intended to be spread over existing coded traffic message channels to be available while in motion.

Furthermore, it should be suitable for use in dynamic navigation.

Also, it should meet objective quality statements that withstand roadside inspection. Furthermore, the stochastic nature of traffic breakdowns should be taken into account.

Furthermore, it is important for the application, how the stochastic nature of traffic collapses has a concrete effect on quality characteristics and which quality statements can realistically be made. This question is not yet answered for predictions about the emergence of traffic crashes on motorways.

The object is achieved by an invention having the features of the independent claims. Advantageous embodiments are described in the subclaims.

Brief description of the figures:

• Fig. 1 ein Beispiel für eine Engstelle auf einer Netzkante darstellt;
• Fig. 2 die Susammenbruchswahrscheinlichkeit als Funktion der Verkehrsstärke zeigt;
• Fig. 3 zeigt die Abhängigkeit des Verfahrensparameters Warnschwelle von dem Verfahrensparameter Vorhersagezeitraum.;
• Fig. 4 Die Arbeitscharakteristik, bei der auf der x-Achse die Kennzahl FPR (Kosten) und auf der y-Achse die Kennzahl TPR (Nutzen) dargestellt sind.

For a better understanding of the invention reference is made to the figures, wherein

• Fig. 1 an example of a bottleneck on a mesh edge;
• Fig. 2 shows the probability of collapse as a function of traffic intensity;
• Fig. 3 shows the dependence of the parameter Alert Threshold on the method parameter Prediction Period .;
• Fig. 4 The operating characteristic in which the key figure FPR (cost) is displayed on the x-axis and the key figure TPR (benefit) on the y-axis.

Preferred embodiment:

To solve the above problem, it is necessary to understand congestion at bottlenecks. The Fig. 1 shows an example of a bottleneck. On a network edge (between two connection points), the number of tracks is reduced from 3 to 2.

Other examples of bottlenecks may be tributaries at the junctions themselves or at highway junctions, construction sites, lane constrictions, etc. Since the knowledge available in a central office about the presence and type of bottlenecks in a national road network is unlikely to be complete and current at any time, a pragmatic approach is to consider each network edge as a potential bottleneck.

die Wahrscheinlichkeit für Verkehrszusammenbrüche Pbd in Abhängigkeit von der gemessenen, auf die Engstelle zuströmenden Verkehrsmenge gebracht werden. Es versteht sich, dass diese Funktion auch eine andere Form aufweisen kann und nur als Beispiel dient. Grundlage dazu sind aufgezeichnete Messungen der Verkehrsmenge durch Detektoren, z.B. Induktionsschleifen oder Überkopfdetektoren, und Beobachtungen von Verkehrsstaus, die sich tatsächlich ereignet haben. Hierbei bezeichnet A den Flaschenhals, der durch Wegfall einer Spur entsteht, B einen weiteren Flaschenhals, der durch den Zufluss an der stromabwärtigen Anschlussstelle gegeben ist. Die Fig. 1 zeigt die entsprechenden Engstellen.For all assumed bottlenecks, it is now possible, for example, to use the function (Weibull distribution) found to be suitable according to empirical investigations [1]. $$\mathrm{P\; bd}\left(\mathrm{q}\right)=\mathrm{1}-\exp \left(-{\left(\mathrm{q}/\mathrm{b}\right)}^{\mathrm{a}}\right)$$

the probability of traffic collapses Pbd are brought in dependence on the measured traffic volume flowing to the bottleneck. It is understood that this function can also have a different form and serves only as an example. The basis for this is recorded measurements of the volume of traffic by detectors, eg induction loops or overhead detectors, and observations of traffic congestion that actually occurred. Here, A designates the bottleneck that results from omission of a track, B another bottleneck, which is given by the inflow at the downstream connection point. The Fig. 1 shows the corresponding bottlenecks.

In the above. Equation a and b are fit parameters of the Weibull distribution. For bottleneck A q-Q_arr (A) and for bottleneck B q = Q_arr (B), respectively, as long as the bottlenecks are inactive, ie. as long as no traffic disruption has developed. Here, the respective measuring cross sections measure the traffic flow that flows through the bottlenecks, and which leads with increasing probability to traffic collapse.

In this procedure, fit parameters (a and b in the example above) are determined for all bottlenecks in the considered road network. The FIG. 2 shows an example of the result. The function Q_arr is given by the measured values for the traffic flow of the respective measuring cross section. The parameters a and b can be determined by common mathematical fit methods. It is known from the past in which traffic flows (q (t) = Q_arr (A, t)) traffic crashes occurred. The collapse probability at the bottleneck A as a function of the traffic flow q can be determined, for example, with the Kaplan-Meier estimator for the survival probability. The past values determined in this way can be adapted to the Weibull distribution, eg by a least-square-fit method.

Up to this point, all method steps in the preferred embodiment are off-line and merely prepare the generation of forecast messages in real time.

• Zwischen dem Zeitbereich, für den Messwerte extrapoliert werden, und dem Zeitbereich, für den die Ganglinie zum Tragen kommt, wird interpoliert.
• Zwischen dem Zeitbereich, für den Messwerte extrapoliert werden, und dem Zeitbereich, für den die Ganglinie zum Tragen kommt, werden Werte aus der Extrapolation und der Ganglinie gewichtet gemittelt.
• Zwischen dem Zeitbereich, für den Messwerte extrapoliert werden, und dem Zeitbereich, für den die Ganglinie zum Tragen kommt, bestimmen extrapolierte Werte das absolute Niveau und die Ganglinie die Kurvenform.

In order to warn of the disruptions that have not yet occurred after the preparatory steps in real time, it is first necessary to know the current demand. This can be assumed to be known from detector measurements. It is next predicted for an application-dependent forecast horizon. For short forecast horizons of a few minutes linear or quadratic extrapolations are suitable for this, for longer (30 or 60 minutes, as above) one will fall back on learned demand curves. However, alternatives are also conceivable for this approach. For the transition from short-term extrapolation to the use of the hydrograph, several techniques are possible, which are listed below:

• Between the time range, for which measured values are extrapolated, and the time range, for which the hydrograph comes into play, is interpolated.
• Between the time range for which readings are extrapolated and the time range for which the hydrograph takes effect, values from the extrapolation and the hydrograph are weighted weighted.
• Between the time range for which measurements are extrapolated and the time range for which the hydrograph is used, extrapolated values determine the absolute level and the hydrograph determine the waveform.

With the help of the forecasted demand, it is possible to determine from the fit function (for each relevant bottleneck) ( Fig. 2 ) read the collapse probability. If the collapse probability accumulated over the period limited by the forecast horizon exceeds a threshold value (eg 80%, the "warning threshold"), it is appropriate to issue an early warning of a traffic jam.

Since due to the stochastic nature of traffic crashes no absolute quality (before each collapse is warned and warned early), the investigation of the quality aspects is important. Otherwise, there is the danger that no or unrealistic quality statements or even statements are made, which ultimately suffers the acceptance of the information.

With full knowledge of the dependency, as a function of adjustable parameters of an early warning system between i.d.R. competing quality features, in fact, adjustable quality compromises can be specifically aligned with the requirements of particular applications such as traffic information services.

Ereignis e ∈

Bedingung

Z

Störungsmeldungen sind Bestandteil von e.

G

Frühwarnungen sind Bestandteil von e.

Z ∩ G

Störungsmeldungen und Frühwarnungen sind Bestandteil von e.

!Z ∩ G

Ausschließlich Frühwarnungen sind Bestandteil von e.

Z ∩ !G

Ausschließlich Störungsmeldungen sind Bestandteil von e.

To form statistical units to which quality characteristics refer, spatially and temporally closely correlated early warnings (type g) and error messages (type z) are combined into events (e = g∪z) across all types and the following event sets are formed:

Event e ∈

condition

Z

Fault reports are part of e.

G

Early warnings are part of e.

Z ∩ G

Fault reports and early warnings are part of e.

! Z ∩ G

Only early warnings are part of e.

Z ∩! G

Only fault messages are part of e.

The following confusion matrix clarifies the meaning of these event sets. Early Warning: Is there a risk (G) of a collapse? Z Yes No Error message: Has actually one Yes Early warning Z ∩ G omitted early warning Z ∩! G Collapse (Z) occurred? No Blind alarm! Z ∩ G G

For a set of statistical units, quality metrics are defined below, where the width of an event set X with | X | is designated. A compact description of the information quality is the mean feature or share value of all quality indicators in the population. These values can be estimated from a sample of data using statistical methods.

Z-related quality indicators

True Positive Rate: TPR = | Z ∩ G | / | Z |, ie the share of disturbance events covered by an early warning event (early warning rate).

False negative rate: FNR = | Z ∩! G | / | Z |, ie the proportion of disturbance events that are not covered by an early warning event.

The relationship FNR = 1 - TPR applies.

G-related quality indicators

Positive Predictive Value: PPV = | Z ∩ G | / | G |, ie the proportion of early warning events that relate to incident events (relevant early warning events).

False Positive Rate: FPR = | ! Z ∩ G | / | G |, ie the proportion of early warning events that are not related to incident events (blind alarm rate).

The relation FPR = 1 - PPV applies.

Z ∩ G - related quality indicators

Pre-warning time ( VWZ ): The period between the notification of the risk of a traffic collapse and the notification of the traffic collapse.

The adjustable parameters of the early warning system include the prediction period (Δ T ) and the warning threshold ( P _bd ) , ie the threshold for the probability of a collapse, which controls the sign-on and log-off of the early warning.

mit

• P_bd = Warnschwelle der Frühwarnung [%],
• ΔT = gewünschter Vorhersagezeitraum (z.B. 15 Minuten) [min],
• At = Referenzvorhersagezeitraum (z.B. 5 Minuten) [min].

The warning threshold P _bd has a dependency on the prediction time period Δ T , which derives from the condition that early warnings with different-length prediction time periods Δ T are to evaluate the same traffic situations equally: $$P_{\mathit{bd}}\left(\mathrm{\Delta }T\right)=1-{\left(1-P_{\mathit{bd}}\left(\mathrm{\Delta }t\right)\right)}^{\frac{\mathrm{\Delta }T}{\mathrm{\Delta }t}}$$

With

• P _bd = early warning warning threshold [%],
• Δ T = desired forecast period (eg, 15 minutes) [min]
• At = reference forecast period (eg 5 minutes) [min].

Fig. 3 The warning threshold x of an early warning with a prediction period of = 5 minutes corresponds to the warning threshold y of an early warning with a prediction period of = 15 minutes.

To determine the operating characteristics of the early warning system, all settings for the parameters forecast period and warning threshold are played through, the resulting totalities of early warning and disturbance events are determined, the quality indicators are measured and entered into a quality diagram.

The TPR varies between the value one (zero) for the smallest (largest) value of the warning threshold, since a small (large) warning threshold means that virtually every (no) fault is warned. The key figure FPR assumes a value that is specific for the smallest value of the warning threshold is for a contemplated bottleneck and can sometimes be significantly less than one; when the warning threshold is opened, the code FPR shows a falling tendency.

The value assumed by the FPR for the lowest value of the warning threshold is, by definition, the probability that there will be no traffic breakdown at the bottleneck in the critical area of transport demand.

The FIG. 4 shows the working characteristic, in which the key figure FPR (costs) is displayed on the x-axis and the key figure TPR (benefits) on the y-axis. This working characteristic is called Receiver Operating Characteristics (ROC). Each (working) point on the graph of the ROC marks the maximum benefit of the early warning information product on the condition that the cost does not exceed a predetermined value.

An operating point that corresponds to a conservative setting of the early warning system with a high warning threshold (circle), means a small proportion of false-positive classified traffic situations (congestion reported, but no congestion occurs) but at the same time also a low TPR, i. a small proportion of traffic breakdowns that are warned.

The opposite setting of the system is reached by a low selected warning threshold (rectangle) and leads to the fact that virtually before each collapse is warned but also frequent blind alarms occur.

The proposed method solves the problem formulated at the outset.

Early warnings of the type described here can be spread by common media such as the TMC channel. For Early alerts can use existing codes such as "stationary traffic expected" or "queuing traffic expected".

They are suitable for use in dynamic navigation because each early warning naturally carries a weighting factor (the probability of collapse) that the routing algorithm can use along with the forecasting horizon to construct the dynamic cost function.

They are adapted to the local conditions throughout the network and therefore have optimal early warning and blind warning rates as well as early warning times with regular calibration.

They take into account the random nature of traffic crashes and do not pretend to the user that congestion can be accurately predicted to the minutes.

Their quality of information can be assessed objectively on the basis of measurable quality criteria and, based on the working characteristics, can be specifically adapted to the specific quality requirements of an application.

It is understood that the preferred embodiments are not intended to limit the subject of the application. Rather, they serve for understanding. Consequently, the claims serve to determine the scope of protection.

literature

1. [1] Brilon, W .; Zurlinden, H.: Capacity of roads as a random variable. Road Traffic Technology 4/2004, pp. 164-172

Claims (33)

1. Method for generating early warnings of traffic crashes comprising the following steps:

   - dynamically estimating the capacity of bottlenecks in the road network,

   - determining the current demand at bottlenecks,

   - determining the expected demand at bottlenecks for a selectable forecast horizon, based on the current demand at bottlenecks,

   - triggering early warnings as soon as the probability of a traffic breakdown calculated by the expected demand and dynamic capacity exceeds a limit, the warning threshold.
2. The method according to the preceding claim, further comprising:

   estimating the capacity of bottlenecks in the road network, taking into account that the capacity of bottlenecks is a Weibull distributed random variable.
3. The method according to the preceding claim, further comprising that the equation $$\mathrm{Pbd}\left(\mathrm{q}\right)=\mathrm{1}-\exp \left(-{\left(\mathrm{q}/\mathrm{b}\right)}^{\mathrm{a}}\right)$$

   

   
   (the Weibull distribution) calculates the probability Pbd of a traffic breakdown, according to the measured traffic volume, flowing to the bottlenecks, wherein the parameters a and b are determined by common mathematical fit methods from observations or measurements, that have as information, in which traffic flows (q(t)=Q_arr(A,t)) traffic crashes have occurred.
4. The method according to one or more of the preceding claims, further comprising: determining the current demand at bottlenecks by means of current and / or historical data for the traffic volume.
5. The method according to one or more of the preceding claims, further comprising: determining the expected demand for a selectable forecast horizon by means of extrapolation, in particular by means of transition line based extrapolation.
6. The method according to one or more of the preceding claims, further comprising, that between the time domain, for which measured values are extrapolated and the time domain, for which the transition line has effect, interpolation is done.
7. The method according to one or more of the preceding claims, further comprising that within the time domain, for which measured values are extrapolated, and the time domain, for which the transition line has effect, values from the extrapolation and the transition line are weighted averaged.
8. The method according to one or more of the preceding claims, further comprising, that between the time domain,for which measured values are extrapolated, and the time domain, for which the transition line has effect, extrapolated values determine the absolute level and the transition line determines the curve shape.
9. The method according to one or more of the preceding claims, further comprising that each network edge has to be considered as a potential bottleneck in a road network.
10. The method according to one or more of the preceding claims, further comprising that early warnings are spread on common media, such as TMC channel.
11. The method according to one or more of the preceding claims, wherein a method according to one or more of the following process claims 12-17 is used for the formation of statistical units.
12. The method of claim 1, to which quality characteristics are related, to determine traffic conditions, characterized in that
    spatially and temporally closely correlated early warnings (type g) and fault messages (type z) type overlapping are combined to events (e = g∪z), wherein the early warning defines whether the risk of a collapse of traffic exists and the fault message defines whether in fact a collapse has occurred.
13. The method according to the preceding claim, wherein one or more of the following sets of events are formed to form quality indicators, wherein:

    Z: fault messages are part of e,

    G: early warnings are part of e,

    Z ∩ G: fault reports and early warnings are part of e,

    I Z ∩ G: only early warnings are part of e,

    Z ∩ ! G: only error messages are part of e, and wherein taking as Z -related quality indicators the

    true positive rate TPR = | Z n G | / | Z | , and the

    false negative rate : FNR = | Z ∩ !G | / | Z | are used , while always maintaining the relationship FNR = 1 - TPR

    is fulfilled,

    and wherein

    as G -related quality indicators of

    positive predictive value : PPV = | Z ∩ G | / | G | and the false positive rate : FPR = | !Z n G | / | G | are used , while always maintaining the relationship FPR = 1 - PPV applies.
14. The method according to one or more of the preceding two claims, further comprising that the warning level P_bd according to the relation $$P_{\mathit{bd}}\left(\mathrm{\Delta }T\right)=1-{\left(1-P_{\mathit{bd}}\left(\mathrm{\Delta }t\right)\right)}^{\frac{\mathrm{\Delta }T}{\mathrm{\Delta }t}}$$

    

    with P_bd = warning level of early warning [%],

    Δt = desired forecast period (e.g. 15 minutes) [min],

    Δt = reference forecast period [min]

    is tracked with a variation of the forecast period.
15. The method according to one or more of the preceding three claims, further comprising that to determine the operating characteristics of the early warning system, all settings for the parameters of the forecast period and warning threshold are run through, the resulting total of early warning and error events are determined, the quality indicators are measured and entered in a quality chart.
16. The method according to the preceding claim, further comprising that the information quality of early warning is specifically set by recourse to the operating characteristic of the early warning system.
17. Apparatus for computer aided generation of early warnings of traffic crashes, comprising means

    - for dynamic estimation of the capacity of bottlenecks in the road network,

    - for determining the current demand at bottlenecks,

    - for determining the expected demand for a selectable forecast horizon, based on the current demand at bottlenecks,

    - for triggering early warnings as soon as the probability of a traffic breakdown based on expected demand and dynamic capacity exceeds a limit, the warning threshold.
18. The apparatus according to the preceding claim, further comprising means for estimating the capacity of bottlenecks in the road network, taking into account that the capacity of bottlenecks is a Weibull distributed random variable.
19. The apparatus according to the preceding claim, further comprising means that, taking into account the equation $$\mathrm{Pbd}\left(\mathrm{q}\right)=\mathrm{1}-\exp \left(-{\left(\mathrm{q}/\mathrm{b}\right)}^{\mathrm{a}}\right)$$

    

    
    (the Weibull distribution) calculating the probability of a traffic breakdown Pbd, according to the measured traffic volume, flowing to the bottlenecks, wherein the parameters a and b are determined by common mathematical fit methods from observations or measurements, that have the information, in which traffic flows (q(t)=Q_arr(A,t)) traffic crashes have occurred.
20. The apparatus according to one or more of the preceding apparatus claims, further comprising means
    to determine the current demand at bottlenecks by means of current and / or historical data for the traffic volume.
21. The apparatus according to one or more of the preceding apparatus claims, further comprising means for
    determining the expected demand for a selectable forecast horizon by means of extrapolation, in particular by means of transition line based extrapolation.
22. The apparatus according to one or more of the preceding apparatus claims, further comprising means, in which between the time domain, for which the measured values are extrapolated, and the time domain, for which the transition line has effect, interpolation is done.
23. The apparatus according to one or more of the preceding apparatus claims, further comprising means which are designed so that between the time domain, for which measured values are extrapolated, and for the and the time domain, for which the transition line has effect, values from the extrapolation and from transition line are averaged weighted.
24. The apparatus according to one or more of the preceding apparatus claims , further comprising means which are designed so that between the time domain, for which the measured values are extrapolated, and the time domain, for which the transition line has effect, extrapolated values determine the absolute level and the hydrograph determine the curve shape.
25. The apparatus according to one or more of the preceding apparatus claims, further comprising means which consider each network edge in a road network as a potential bottleneck.
26. The apparatus according to one or more of the preceding apparatus claims, further comprising means which spread and / or receive early warnings of common media, such as TMC channel.
27. The apparatus according to one or more of the preceding device claims, comprising a device according to one or more of the following apparatus claims.
28. The apparatus of claim 17 for computer-aided calculation of quality characteristics of traffic conditions, to form statistical units on which quality characteristics are related to, to determine traffic conditions, characterized in that spatially and temporally closely correlated early warnings (type g) and fault messages (type z) are combined for all types to events (e = g∪z), wherein early warning defines whether the risk of a collapse of traffic exists and the fault message defines whether in fact a collapse has occurred.
29. The device according to the preceding claim, wherein means are provided to form one or more of the following sets of events, in order to form the quality metrics, wherein:

    Z: fault messages are part of e,

    G: early warnings are part of e,

    Z ∩ G: fault reports and early warnings are part of e,

    ! Z ∩ G : only early warnings are part of e,

    Z ∩ !G : only fault messages are part of e, and wherein when

    Z -related quality indicators the

    true positive rate TPR = | Z ∩ G | / | Z | , and the

    false negative rate : FNR = | Z n !G | / | Z | are used, while always maintaining the relationship FNR = 1 - TPR

    is fulfilled

    and wherein

    as G -related quality indicators of

    positive predictive value : PPV = | Z ∩ G | / | G | and the false positive rate : FPR = | !Z ∩ G | / | G | are used, while always maintaining the relationship FPR = 1 - PPV applies.
30. The device according to one or more of the preceding two apparatus claims, further comprising means for tracking the warning threshold for variation in the forecast period, according to the relationship $$P_{\mathit{bd}}\left(\mathrm{\Delta }T\right)=1-{\left(1-P_{\mathit{bd}}\left(\mathrm{\Delta }t\right)\right)}^{\frac{\mathrm{\Delta }T}{\mathrm{\Delta }t}}$$

    

    with P_bd = warning level of early warning [%],

    ΔT = desired forecast period (e.g. 15 minutes) [min]

    Δt = = reference forecast period [min].
31. The device according to one or more of the preceding three claims, further comprising means for determining the operating characteristic of the early warning system, and all the settings for the parameters of the forecast period and warning threshold are run through, determining the resulting amount of early warnings and fault events, the quality indicators are measured and entered in a quality graph.
32. The device according to the preceding claims, further comprising that the information quality of early warning is specifically set by recourse to the operating characteristic of the early warning system.
33. Device according to one or more of the preceding device claims, characterized In that a network connection to external sensors exists, of which current traffic data will be obtained.

Images