EP2047447B1 - Method and device for the fusion of traffic data when information is incomplete - Google Patents

Abstract

A method for merging imprecisely localized traffic reports with precisely localized traffic data includes obtaining a plurality of possible positions (x) of the localized traffic reports having imprecise position indications. The plurality of possible positions is evaluated using overlap functions. Substantially precise positions for the localized traffic reports are defined by solving an extremum problem.

Description

The invention relates to a method and apparatus for merging traffic data with incomplete information, taking into account information from different sources mapped to functions to obtain a result based on calculations with these functions.

Field of the invention:

Generation of traffic information for real-time information or navigation services relies on multiple data sources to achieve the best possible quality. These data sources can be of different nature, such as human observation (police, traffic jam) on the one hand and automatic measurement of traffic data (stationary sensors, floating cars) on the other. As a result, both apparent and actual contradictions occur between individual sources, and certain information elements can only be supplied by one source, others only by the other source. For example, the cause of a traffic disruption is typically the accessible to human observation, while the average velocity is typically determined by an automatic measuring system. This leads to the requirement to associate information from different sources.

In the DE10002918C2 It is suggested that the consideration of different sources should be done with the help of spatial overlap. However, the question remains as to how the data fusion of a source with a high and a source with low spatial accuracy should happen.

From the publication US 2002/0262781 A1 a method is known in which the normalized travel times are taken into account. This reveals the union of NTT values to provide a specific value. Here, a confidence interval is considered. In contrast to the present invention, on the one hand different values are taken into account, on the other hand no overlapping function is used in the prior art and furthermore no optimum search is applied to the overlapping function.

In the following, a corresponding example is shown. The police report "5 km traffic jam between junction 1 and junction 5". Between the two junctions are 30km highway and 3 other junctions - the position of the traffic jam is therefore determined only very inaccurate. At the same time sensors on the motorway section report 3km congestion and 20km free travel, while 7 km are not monitored. What information is to be passed on to the service? Where exactly is the traffic congestion and which route is affected?

Overview of the invention:

The object of the present invention is to overcome the aforementioned disadvantages of the prior art.

The object is achieved by a method and a device having the features of the independent claims.

Specifically, it is a method for local determination of traffic situation data, taking into account a plurality of traffic reports with fuzzy location information, which are to be merged as best as possible. It includes the following steps.

Taking into account all possible positions of the traffic reports with blurred Ortsangaben- This can be done on the basis of mileage.

Evaluate these positions using multiple overlap functions. Such functions will be described below. They can be weighted and summarized to a valuation function. On the basis of these functions and certain marginal parameters, optimal, sharp positions for the traffic reports with fuzzy location information are found by solving an extreme value problem. Here, the standard methods can be used to find the extreme values.

Description of the figures:

Fig. 1

zeigt eine schematische Darstellung einer ungenauen Quelle für Verkehrsinformationen, die eine Störung der Länge L zwischen den Anschlussstellen AS1 und AS4 meldet;

Fig. 2

zeigt eine schematische Darstellung der Datenfusion, wobei oben das Ergebnis der Datenfusion angezeigt wird, in der Mitte die Meldung aus der ungenauen Quelle und die ortsgenauen numerischen Verkehrslagedaten angezeigt werden;

Fig. 3

Verkehrslagedaten und -meldungen auf der A555 von Köln nach Bonn am 10.01.2005.

For a better understanding of the invention, the figures are briefly discussed below

Fig. 1

shows a schematic representation of an inaccurate source of traffic information, which reports a disturbance of length L between the terminals AS1 and AS4;

Fig. 2

shows a schematic representation of the data fusion, with the result of the data fusion is displayed above, in the middle of the message from the inaccurate Source and the local numerical traffic situation data are displayed;

Fig. 3

Traffic situation data and messages on the A555 from Cologne to Bonn on 10.01.2005.

Description of the preferred embodiment:

The present invention report deals with an optimal solution of the data fusion problem with incomplete information.

The general solution to the problem is the situation in Fig.1 considered.

In this Fig. 1 An inaccurate traffic information source reported a length L disturbance between ports AS1 and AS4.

One might assume in this situation that the traffic jam covers the AS2 and AS3 junctions since otherwise it would have been reported between AS1 and AS3 or AS2 and AS4. In fact, this is not always the case. In practice, there are also frequent cases where the distance between AS2 and AS3 is greater than L. Because of such practical problems, all positions from x = location (AS1) to x + L = location (AS4) must first be considered equivalent become.

1. 1. Bestätigung: Für jede mögliche Position x innerhalb des erlaubten Wertebereichs wird eine Funktion b(x) ermittelt, die den Anteil der gemeldeten Störung angibt, der durch die verkehrslagedaten bestätigt wird.
2. 2. Lückenschluss: Für jede mögliche Position x innerhalb des erlaubten Wertebereichs wird eine Funktion n(x) ermittelt, die den Anteil der gemeldeten Störung angibt, der nicht durch vorhandene Verkehrslagedaten widerlegt werden kann (Nichtwissen wegen nicht vorhandener Detektion).
3. 3. Widerlegung: Für jede mögliche Position x innerhalb des erlaubten Wertebereichs wird eine Funktion w(x) ermittelt, die den Anteil der gemeldeten Störung angibt, der durch vorhandene Verkehrslagedaten widerlegt werden kann.

In the next step, the compatibility with the local, numerical traffic situation data is checked for all these possible positions - as far as they are available - as follows:

1. 1. Confirmation: A function b (x) is determined for every possible position x within the permitted value range. indicating the proportion of the reported incident confirmed by the traffic situation data.
2. 2. Gap closure: For each possible position x within the allowed value range, a function n (x) is determined which indicates the proportion of the reported disturbance that can not be refuted by existing traffic situation data (ignorance due to non-existent detection).
3. 3. Refutation: For each possible position x within the allowed range of values, a function w (x) is determined that indicates the proportion of the reported error that can be refuted by existing traffic situation data.

For all x we have the relationship: b ( x ) + w ( x ) + n ( x ) = 1

Here is an example:

The functions b, n and w indicate the degree of spatial overlap. Assuming that a badly locatable message with the length specification "10km" is placed on a test basis at position x and it turns out that 5 of the 10km can not be checked because there are no data that can be located well, 4 of the 10km are in agreement with well localizable data Data are located and 1 of the 10km is at odds with well localizable data. Then n (x) = 0.5, b (x) = 0.4 and w (x) = 0.1. The general condition n (x) + b (x) + w (x) = 1 is thus satisfied, in particular for this x.

$$f_{\mathit{apo}}\left(x\right)=g_b\cdot b\left(x\right)+g_n\cdot n\left(x\right)+g_w\cdot w\left(x\right),$$

$$x_1=\min \left({\mathit{km}}_1,{\mathit{km}}_3-L\right),$$

$$x_2=\min \left({\mathit{km}}_1+L,{\mathit{km}}_3\right)\mathrm{.}$$

Mit

g_b

Gewicht des Kriteriums "Bestätigung"

g_n

Gewicht des Kriteriums "Lückenschluss"

g_w

Gewicht des Kriteriums "Widerlegung"

b(x)

Anteil der gemeldeten Störung, der bei einer angenommenen Position ^x durch Verkehrslagedaten bestätigt wird.

w(x)

Anteil der gemeldeten Störung, der bei einer angenommenen Position x durch Verkehrslagedaten widerlegt wird.

n(x)

Anteil der gemeldeten Störung, der bei einer angenommenen Position x durch Verkehrslagedaten weder bestätigt noch widerlegt wird (Nichtwissen).

x

Mögliche Position für das stromaufwärtige Ende der gemeldeten Störung.

All three criteria are weighted and summarized to the following extreme value problem (the solution denoted by x is searched for): $$\nexists \mathit{x\; \text{'}}\in \left[x_1,x_2\right]:\left\{f_{\mathit{apo}}\left(\mathit{x\; \text{'}}\right)>f_{\mathit{apo}}\left(\mathit{x}\right)\right\}\vee \left\{\left(f_{\mathit{apo}}\left(\mathit{x\; \text{'}}\right)=f_{\mathit{apo}}\left(\mathit{x}\right)\right)\wedge \left(\mathit{x\; \text{'}}>x\right)\right\}.$$

$$f_{\mathit{apo}}\left(x\right)=G_b\cdot b\left(x\right)+G_n\cdot n\left(x\right)+G_w\cdot w\left(x\right).$$

$$x_1=\min \left({\mathit{km}}_1.{\mathit{km}}_3-L\right).$$

$$x_2=\min \left({\mathit{km}}_1+L.{\mathit{km}}_3\right)\mathrm{,}$$

With

g _b

Weight of the criterion "Confirmation"

g _n

Weight of the criterion "gap closure"

g _w

Weight of the criterion "refutation"

b ( x )

Proportion of the reported disturbance confirmed at an assumed position ^x by traffic situation data.

w ( x )

Proportion of reported disturbance that is refuted by traffic situation data at an assumed position x .

n ( x )

Proportion of the reported disturbance that is neither confirmed nor refuted by an assumed position x by traffic data (ignorance).

x

Possible position for the upstream end of the reported fault.

The weighting factors gx can be determined on the basis of a priori knowledge of the quality of a source. For example, disturbances reported by the police as locally blurred are generally credible and attempts should be made to confirm them; however, police reports are usually not timely deregistered. Further criteria for the gx are the (origin) position of disturbances (disturbances arise at bottlenecks, therefore they are preferably positioned as far downstream as possible) and quality requirements for the end product (eg correctness could be more important than completeness, then the confirmation gb would be strong too) weight). By statistically evaluating the distribution among the categories "affirmation,""ignorance," and "refutation," the assumptions made to set the weights may be reviewed and an adjustment may be made.

The extreme value for x can either be found with common methods of optimization calculation ("curve discussion") or by complete calculation of the objective function f _apo (x) with a step size of eg Im, which does not cause any difficulties for today's computers.

The km _i indicate the locations of the connection points. And L the length of the possible error.

As a result, a position x of the traffic disruption is found which - as controlled by the weighting factors - is confirmed as well as possible by the locally accurate numerical traffic situation data or, if this does not succeed to a sufficient extent, at least not contradict them.

After positioning the message from the inaccurate source, fusion with the local numerical traffic data is performed as follows. Wherever the sharply localized message competes with non-knowledge from the traffic assessment, the relevant share of the message is adopted as the end product. In all other places, the numerical traffic situation data is preferred.

The Fig. 2 and Fig. 3 shows a traffic disruption that occurred on 10.01.2005 due to an accident on the A555 from Cologne to Bonn shortly after the junction Bornheim / Alfter (- km 16). From direct observation it was known that the traffic congestion was in the area km 13 - 17. There is no stationary measurement infrastructure in this area.

In the period from 8:03 to 8:28, between Cologne-Godorf and Bonn-Nord, the police reported only one three and then one two-kilometer fault. In the case of a naive, central positioning of the message between Cologne-Godorf and Bonn-Nord (red polygon), this message would completely contradict the numerical traffic situation data (green background) and would be completely rejected by the data fusion.

When positioning with the method proposed here (dark blue polygon), the message is almost completely localized in the area "Ignorance" (light blue background) and thus almost completely taken over by the data fusion (see Fig. 3 ).

The content of the present invention is not limited to the description. Rather, it is determined by the knowledge of the expert in this field. Furthermore, the scope of protection is determined by the claims, the description intends no limitations.

Claims (6)

1. Method for merging imprecisely localized traffic reports with precisely localized traffic data, comprising the following steps:

   - taking into account possible positions of the traffic reports having imprecise position data;

   - evaluating these positions with the aid of the following functions:

   wherein, as part of the evaluation, there is ascertained for each possible position x within the allowed range of values a function b(x) that specifies the portion of the reported traffic disturbance that is confirmed by the traffic situation data,

   and

   there is ascertained for each possible position x within the allowed range of values a function n(x) that specifies the portion of the reported traffic disturbance that cannot be refuted by available traffic situation data,

   and

   there is ascertained for each possible position x within the allowed range of values a function w(x) that specifies the portion of the reported traffic disturbance that can be refuted by available traffic situation data, wherein the relation $$\mathit{b}\left(\mathit{x}\right)\mathit{+}\mathit{w}\left(x\right)\mathit{+}\mathit{n}\left(\mathit{x}\right)\mathit{=}1$$

   

   applies for all x,

   - establishing optimal, precise positions for the traffic reports having imprecise position data by solving an extreme value problem on the basis of the functions.
2. The method according to the preceding claim, wherein a weighting of the three criteria can be performed during the evaluation and is combined into the following extreme value problem: $$\nexists \mathit{xʹ}\in \left[x_1,x_2\right]:\left\{f_{\mathit{apo}}\left(\mathit{xʹ}\right)>f_{\mathit{apo}}\left(\mathit{x}\right)\right\}\vee \left\{\left(f_{\mathit{apo}}\left(\mathit{xʹ}\right)=f_{\mathit{apo}}\left(\mathit{x}\right)\right)\wedge \left(\mathit{xʹ}>x\right)\right\},$$

   

   $$f_{\mathit{apo}}\left(x\right)=g_b\cdot b\left(x\right)+g_n\cdot n\left(x\right)+g_w\cdot w\left(x\right),$$

   

   $$x_1=\min \left({\mathit{km}}_1,{\mathit{km}}_3-L\right),$$

   

   $$x_2=\min \left({\mathit{km}}_1+L,{\mathit{km}}_3\right)\mathrm{.}$$

   

   where g_b is the weight of the criterion "confirmation,"

   g_n is the weight of the criterion "gap-filling,"

   g_w is the weight of the criterion "refutation,"

   b(x) is the portion of the reported disturbance that is confirmed at an assumed position x by traffic situation data,

   w(x) is the portion of the reported disturbance that is refuted at an assumed position x by traffic situation data,

   n(x) is the portion of the reported disturbance that is neither confirmed nor refuted (lack of knowledge) at an assumed position x by traffic situation data,

   x is the possible position of the upstream end of the reported disturbance,

   wherein km_i specifies the location of the junctions and L specifies the length of the possible disturbance.
3. The method according to one or more of the preceding claims, wherein the merging with the precisely localized numeric traffic situation data is carried out after the positioning of the report from an imprecise source.
4. The method according to the preceding claim, wherein at every point where the precisely localized traffic report conflicts with lack of knowledge from the estimated traffic situation, the relevant portion of the report is used as the end product, and at all other points the numerical traffic situation data take precedence.
5. Apparatus for precisely localized determination of traffic situation data taking into account a plurality of traffic reports having imprecise position data, characterized by a device that is designed for the execution of a method according to one or more of the preceding method claims.
6. Data media product characterized by a data structure that is designed such that it brings about the execution of a method according to one or more of the preceding method claims after being loaded onto a computer.

Images