DE102016212774B4 - Method and device for generating a map of the surroundings and for locating a vehicle - Google Patents

Abstract

Method for generating an environment map for the purpose of locating a vehicle (4, 5, 25) in its environment, with the steps:
- cyclic creation of images of the surroundings of the vehicle (4, 5, 25) using at least one image-generating device (30),
- Extracting landmarks (27, 28, 29) from the cyclic images of the image-generating device and creating a measurement data set for each image,
- Determining the vehicle position and the vehicle's own movement,
- Creating a map of the surroundings from the measurement data set, with only a specified field of view (26) being taken into account within the measurement data set, as well as the vehicle position and the vehicle's own movement, characterized in that the specified field of view (26) for evaluating the cyclic measurement data sets is a function of the current weather condition .

Description

The invention relates to a method for generating an environment map for the purpose of locating a vehicle in its environment according to the preamble of claim 1, a corresponding device according to the preamble of claim 6, a method for locating an autonomous vehicle in its environment according to the preamble of claim 9 and a corresponding device with the features of claim 10.

Automated or autonomous driving requires a highly accurate localization of the vehicle in its environment. Since automated vehicles do not have to move in "good-natured" environments but in the real world and are therefore directly exposed to atmospheric disturbances such as seasons, weather changes and light effects, the localization methods must be robust and localization must be guaranteed in every situation, since these atmospheric disturbances can affect the localization.

When talking about localization in the context of automated driving, localization consists of a two-part approach. In the first step, a map is generated from observations made during one or more measurement runs. In the next step, an attempt is made to locate oneself on the previously created maps using observations from a subsequent trip. In both steps, the observations are crucial for the localization. However, these observations depend on the sensors used and the situation surrounding the vehicle. For example, observations from a laser scanner behave differently when it is raining than when it is dry. A camera also provides different observations in rain, wet and dry conditions.

• - Sammlung von Beobachtungen in unterschiedlichen Situationen, wie beispielsweise bei Trockenheit, Regen, Nässe, etc., zur Erstellung von Karten, die weitestgehend unabhängig von atmosphärischen Gegebenheit sind. Live-Beobachtungen können so allerdings nicht beeinflusst werden.
• - Anpassung der Sensor-Detektion der Beobachtungen. Dabei wird der Detektionsalgorithmus für jede atmosphärische Störung trainiert, um beispielsweise die Detektion von Bildobjekten in den jeweiligen Situationen zu verbessern.

So far, attempts have been made to solve the above problem using the following approaches:

• - Collection of observations in different situations, such as dry, rainy, wet, etc., to create maps that are largely independent of atmospheric conditions. However, live observations cannot be influenced in this way.
• - Adjustment of the sensor detection of the observations. The detection algorithm is trained for each atmospheric disturbance in order to improve the detection of image objects in the respective situations, for example.

• - Werden die Karten aus unterschiedlichen Beobachtungen in unterschiedlichen Situationen erstellt, so bedeutet dies nicht, dass während der Lokalisierung in einer bestimmten Situation, wie beispielsweise Regen, die Beobachtungen gut zur Karte korrespondieren. Eine Robustheit der Lokalisierung gegen atmosphärische Störungen ist deshalb nicht zwingend gegeben.
• - Durch die reine Anpassung der Sensor-Detektion der Beobachtungen werden die Ausgaben des Sensors zwar direkt für jede Situation verbessert, ob dies jedoch direkt eine Verbesserung der Lokalisierung mit sich bringt ist bislang offen.

These approaches have two inherent disadvantages:

• - If the maps are created from different observations in different situations, this does not mean that the observations correspond well to the map during the localization in a certain situation, such as rain. A robustness of the localization against atmospheric disturbances is therefore not necessarily given.
• - By simply adapting the sensor detection of the observations, the outputs of the sensor are improved directly for every situation, but whether this directly results in an improvement in the localization is still open.

Overall, this means that the adjustments made to the map and observations without considering the localization results were not necessarily chosen in a sensible manner for localization optimization.

The pamphlet EP 0 913 751 B1 relates to an autonomous vehicle and a method for controlling an autonomous vehicle, comprising an input unit for one or more travel orders, a device for route planning, comprising at least one position detection device and a digital road map, a device for generating a route, an arrangement of sensors, each comprising at least one distance sensor for detecting objects and characteristics of the path, a collision avoidance device, a device for status data recognition of the vehicle, a vehicle control device and a device for controlling the vehicle actuators as a function of the signals generated by the vehicle control device. The arrangement of sensors includes at least the following sensors: two substantially horizontally aligned distance sensors arranged in the front area of the vehicle, at least one distance sensor arranged in the front area of the vehicle, at least one distance sensor arranged in the rear area of the vehicle, at least one on the roof of the Vehicle arranged, directed towards the roadway distance sensor with trackable scanning angle, arranged on the respective sides of the vehicle ultrasonic sensors or microwave radar sensors and at least one camera each arranged in the front and rear area of the vehicle.

The pamphlet DE 10 2008 034 923 A1 discloses a method for controlling the climatic conditions of a building containing a heating and/or air conditioning system with at least one sensor arranged outside the building and used to record the temperature, using weather data distributed via a data channel of a forecast general weather situation including the location of the building and using the physical variable detected by the sensor by calculating a forecast of the smallest weather situation relating to the location of the building, the forecast being used to control the heating or air conditioning system.

The pamphlet DE 10 2007 003 014 A1 relates to a vehicle position determination device for a vehicle, comprising a sensor unit for detecting an unevenness in the roadway on which the vehicle is driving or in the area surrounding the vehicle, a sensor unit for detecting a roadway marking in the area surrounding the vehicle and/or a sensor unit for detecting a roadway roofing underneath and for providing at least one corresponding sensor signal and an evaluation unit with a digital map on which the positions of bumps in the roadway, roadway markings and/or roadway roofing are stored, the evaluation unit being designed to determine a limited spatial area as a function of the spatial area and the at least one sensor signal Determine area for a more accurate vehicle position.

Fully autonomous vehicles require precise localization of the vehicle in its environment, i.e. on the street or lane, which may not be sufficiently guaranteed due to weather conditions such as rain, snow, fog, etc.

The invention is therefore based on the object of improving the generation of an environment map for the purpose of locating a vehicle in its environment and the localization of a vehicle in its environment.

This object is achieved by a method for generating an environment map with the features of claim 1, by a corresponding device with the features of claim 6, by a method for locating a vehicle with the features of claim 9 and by a corresponding device with the features of Claim 10 solved. Preferred embodiments of the invention are the subject matter of the dependent claims.

• - zyklisches Erstellen von Bilder des Umfelds des Fahrzeugs mittels mindestens einer bilderzeugenden Einrichtung,
• - Extrahieren von Landmarken aus den zyklischen Bildern der bilderzeugenden Einrichtung und Erstellen eines Messdatensatzes für jedes Bild,
• - Ermitteln der Fahrzeugposition und der Fahrzeugeigenbewegung,
• - Erstellen einer Umfeldkarte aus dem Messdatensatz sowie der Fahrzeugposition und der Fahrzeugeigenbewegung, wobei innerhalb des Messdatensatzes nur ein vorgegebenes Sichtfeld berücksichtigt wird, wobei das vorgegebene Sichtfeld zur Auswertung der zyklischen Messdatensätze eine Funktion der aktuellen Wetterbedingung ist.

The method according to the invention for generating an environment map for the purpose of locating a vehicle in its environment comprises the following steps:

• - cyclic creation of images of the environment of the vehicle by means of at least one image-generating device,
• - Extraction of landmarks from the cyclic images of the imaging device and creation of a measurement data set for each image,
• - Determining the vehicle position and the vehicle's own movement,
• - Creation of an environment map from the measurement data set and the vehicle position and the vehicle's own movement, wherein within the measurement data set only a predetermined field of view is taken into account, the predetermined field of view for evaluating the cyclic measurement data sets is a function of the current weather condition.

The environment map is optimized through the coupling between the field of view and the prevailing weather conditions.

Preferably, the predetermined field of view is parameterized by a set of parameters, the set of parameters being a function of the weather condition. The parameterization of the field of view enables the field of view to be described and manipulated in the algorithm run.

More preferably, the weather condition is classified into a predetermined number of weather classes. In other words, the currently prevailing weather condition is assigned to a class, such as "dry", "wet" or "rain", whereby these three classes are only to be understood as examples, since a continuously closed description of weather conditions is hardly possible.

More preferably, the landmarks are formed by the lane markings, the environment by the lane in front of the vehicle and the field of vision by a trapezoid on the lane, the trapezoid only being used within the measurement data set and being “imaginary” so to speak.

The specified field of view for a weather condition is preferably determined by a specified number of measurement runs for this weather condition, with the field of view being optimized by localizing the vehicle for each measurement run and comparing each localization determined in this way with a reference localization. By carrying out a large number of measurement runs for different weather conditions and different parameter sets for the field of view, optimal parameter values for the individual weather classes can be determined, which are then used to determine and optimize the local environment map.

• - eine bilderzeugende Einrichtung zur Erzeugung zyklischer Bilder des Umfelds des Fahrzeugs,
• - eine Einrichtung zum Extrahieren von Landmarken aus den zyklischen Bildern der bilderzeugenden Einrichtung und zum Erstellen eines Messdatensatzes für jedes Bild,
• - eine Einrichtung zum Ermitteln der Fahrzeugposition und der Fahrzeugeigenbewegung,
• - eine Einrichtung zum Erstellen einer Umfeldkarte aus dem Messdatensatz sowie der Fahrzeugposition und der Fahrzeugeigenbewegung, wobei innerhalb des Messdatensatzes nur ein vorgegebenes Sichtfeld berücksichtigt wird, und
• - eine Einrichtung zur Bestimmung der aktuellen Wetterbedingung, wobei das vorgegebene Sichtfeld eine Funktion der aktuellen Wetterbedingung ist.

The device according to the invention for generating a map of the surroundings for the purpose of locating a vehicle in its surroundings, the device being set up and designed to carry out the method explained above

• - an image-generating device for generating cyclic images of the surroundings of the vehicle,
• - a device for extracting landmarks from the cyclic images of the image-generating device and for creating a measurement data set for each image,
• - a device for determining the vehicle position and the vehicle's own movement,
• - a device for creating a map of the surroundings from the measurement dataset and the vehicle position and the vehicle's own movement, with only a predetermined field of view being taken into account within the measurement dataset, and
• - means for determining the current weather condition, the predetermined field of view being a function of the current weather condition.

More preferably, the device for determining the current weather condition classifies the weather condition into a predetermined number of weather classes.

Precisely one field of view is preferably stored for each weather class, which is determined by measuring runs for the respective weather class.

• - Erzeugen einer Umfeldkarte des aktuellen Umfelds des Fahrzeugs, wobei die Umfeldkarte mit dem oben erläuterten Verfahren zur Erzeugung einer Umfeldkarte erzeugt wird,
• - Bereitstellen einer digitalen Karte des aktuellen Umfelds des Fahrzeugs, wobei die digitale Karte die Landmarken des aktuellen Umfelds aufweist, und
• - Fusionieren der Umfeldkarte und der digitalen Karte zum Erzeugen einer Lokalisierung des Fahrzeugs in seinem Umfeld.

The method according to the invention for locating a vehicle in its surroundings carries out the following steps:

• - Generating an environment map of the current environment of the vehicle, the environment map being generated using the above-explained method for generating an environment map,
• - Providing a digital map of the current surroundings of the vehicle, the digital map having the landmarks of the current surroundings, and
• - Merging the environment map and the digital map to generate a localization of the vehicle in its environment.

• - eine Einrichtung zum Erzeugen einer Umfeldkarte des aktuellen Umfelds des Fahrzeugs, wobei eine oben erläuterte Vorrichtung zur Erzeugung einer Umfeldkarte verwendet wird,
• - eine Einrichtung zum Bereitstellen einer digitalen Karte des aktuellen Umfelds des Fahrzeugs, wobei die digitale Karte die Landmarken des aktuellen Umfelds aufweist, und
• - eine Einrichtung zum Fusionieren der Umfeldkarte und der digitalen Karte zum Erzeugen einer Lokalisierung des Fahrzeugs in seinem Umfeld,

The device according to the invention for locating a vehicle in its surroundings, wherein the device is set up and designed for carrying out the method for locating a vehicle explained above

• - a device for generating an environment map of the current environment of the vehicle, wherein a device explained above for generating an environment map is used,
• - A device for providing a digital map of the current surroundings of the vehicle, the digital map having the landmarks of the current surroundings, and
• - a device for merging the environment map and the digital map to generate a localization of the vehicle in its environment,

The landmark-based localization is thus expanded to include an adaptive adaptation to the external atmospheric conditions. The method explained above describes how parameters for localization in certain weather situations, such as fog, rain and drought, can be determined.

• - Durchführen von Messfahrten für die jeweiligen Klassen, wie beispielsweise Regen, Nässe und Trockenheit. Die Messfahrten sollen alle Daten für die Lokalisierung enthalten und zudem muss eine Referenzeinheit zur Referenzlokalisierung mitaufgezeichnet werden.
• - Einteilen der Fahrten in die jeweiligen Klassen - diese Datengrundlage kann als Trainingsset bezeichnet werden.
• - Innerhalb jeder Klasse werden jetzt die Parameter optimiert. Hier sei dies am Beispiel des Sichtbereichs der Kamera erläutert. Es werden die minimalen und maximalen Parameterwerte sowie die Schrittweite festgelegt. Für jede Fahrt des Trainingssets werden nun alle Parameterkombinationen durchgetestet. Dabei können unterschiedliche zusätzliche Bedingungen an die Parameterkombinationen festgelegt werden. Als Güte-Maß zu Bewertung einer Parameterkombination wird die Referenzlokalisierung herangezogen. Über ein festgelegtes Distanzmaß, wie beispielsweise das euklidisch Abstandsmaß, kann so der Fehler zwischen Referenzlokalisierung und Parameterkombination ermittelt werden. Nachdem alle Fahrten mit allen Kombinationen durchgetestet wurden, wird die beste Fahrt herangezogen. Da Witterung nicht diskret einteilbar ist, wird versucht das Rauschen der besten Fahrt zu bestimmen. Dies geschieht, indem die Abstände der Lokalisierung zur Referenzlokalisierung der besten Fahrt diskretisiert werden. Über diese diskreten Abstände wird die Standardabweichung gebildet. Alle weiteren getesteten Fahrten werden nur weiter betrachtet, wenn sie innerhalb von 2 Standardabweichungen zur besten Fahrt der Klasse liegen. Anschließend können die Parameterwerte aller betrachteten Fahrten gemittelt werden. Sie gelten als optimale Parameterwerte für die Lokalisierung der jeweiligen Klasse.

To determine these parameters, proceed as follows:

• - Carrying out measurement runs for the respective classes, such as rain, wet and dry. The measurement runs should contain all the data for the localization and a reference unit for the reference localization must also be recorded.
• - Dividing the trips into the respective classes - this database can be referred to as a training set.
• - Within each class the parameters are now optimized. This is explained here using the example of the field of view of the camera. The minimum and maximum parameter values and the increment are specified. All parameter combinations are now tested for each ride in the training set. Various additional conditions can be specified for the parameter combinations. The reference localization is used as a quality measure for evaluating a parameter combination. The error between the reference localization and the parameter combination can be determined using a specified distance measure, such as the Euclidean distance measure. After all rides with all combinations have been tested, the best ride is used. Since the weather cannot be classified discretely, an attempt is made to determine the noise of the best trip. This is done by discretizing the distances between the localization and the reference localization of the best trip. The standard deviation is formed using these discrete distances. All other tested rides are only considered further if they are within 2 standard deviations of the best ride in class. The parameter values of all the trips considered can then be averaged. They are considered the optimal parameter values for localizing the respective class.

• - Es wird nicht nur ein Parameterset für alle Wetterbedingungen verwendet, sondern für jede Witterung wird ein spezielles Parameterset ermittelt und eingesetzt,
• - durch die ermittelten witterungsspezifischen Parametersets kann die Lokalisierung an die jeweilige Situation adaptiv auf Robustheit angepasst werden, um Fehllokalisierungen zu vermieden.

The method presented offers a robust precise localization with the following events:

• - Not only one parameter set is used for all weather conditions, but a special parameter set is determined and used for each weather condition,
• - Using the determined weather-specific parameter sets, the localization can be adaptively adapted to the respective situation for robustness in order to avoid incorrect localizations.

• 1 eine schematische Darstellung der Lokalisierungsproblematik anhand einer Verkehrssituation,
• 2 ein Sichtfeld einer Kamera eines Fahrzeugs,
• 3 Definition des Sichtfeldes,
• 4 unterschiedliche Sichtfelder als Funktion der Klasse der Umgebungsbedingung, und
• 5 eine schematische Darstellung der Vorrichtung zur präzisen Lokalisierung.

A preferred embodiment of the invention is explained below with reference to the drawings. while showing

• 1 a schematic representation of the localization problem based on a traffic situation,
• 2 a field of view of a camera of a vehicle,
• 3 definition of the field of view,
• 4 different fields of view as a function of the environmental condition class, and
• 5 a schematic representation of the device for precise localization.

As mentioned above, an autonomously driving vehicle needs to be localized as precisely as possible in its traffic environment. In other words, the position of the vehicle on the road or lane must be known as precisely as possible, i.e. within a few centimetres.

So shows 1 a traffic situation at an intersection 1, which is formed by the two streets 2 and 3. A first vehicle 4 is moving straight ahead on one street 3 via the intersection 1 and a second vehicle 5 on the other street 2 wants to turn left and crosses the path of the vehicle 4 driving straight ahead. Both vehicles 4, 5 have a satellite-based positioning system, such as a GPS system.

The inaccuracies caused by GPS positioning are in 1 shown as circles of uncertainty 6, 8 around the respective vehicles 4, 5. Thus, with GPS, the first vehicle 4 can only be localized within the GPS circle of uncertainty 6 and the second vehicle 5 within the GPS circle of uncertainty 8 . The GPS uncertainty circles 6, 8 in the localization of the two vehicles 4, 5 lead to overlapping GPS uncertainty areas 7 and 9 due to the respective vehicle movement, the overlapping area shown hatched forming the GPS collision area 10. In other words, due to the inaccuracies in the GPS localizations of the two vehicles 4, 5, there is a large possible GPS collision area 10, which is insufficient for the safety of autonomously moving vehicles 4, 5.

A precise localization of the two vehicles 4, 5 on the roads 2, 3 leads to significantly smaller circles of uncertainty 11 and 13 around the vehicles 4, 5, so that the areas of uncertainty 12 and 14 caused by the vehicle movements with regard to the precise localization are in a significantly smaller Intersect collision area 15, which is represented by two dashed lines. It is therefore necessary for autonomous driving to provide a precise localization of the vehicles 4, 5 in the area, here the intersection area 1.

For precise localization, the position of the vehicle on the road is therefore determined, for example using the road markings and other features, using a suitable sensor system, with a monocular camera being used here in the example. The results of the evaluation of the sensor data are merged with a digital map and the GPS position in order to arrive at a precise localization. Since automatic vehicles, i.e. autonomously driving vehicles, move in the real world and have to deal with the real world events, the precise localization based on lane markings is affected by weather conditions. So it is obvious that the visibility of the road markings is different for different weather conditions such as dryness, rain, snow or fog. In other words, the precise localization using road markings or surrounding objects such as traffic lights etc. recorded in the digital map depends on the weather conditions.

2 shows the view from a vehicle 25 onto a roadway 20 with different lanes 21, 22, 23, 24, the trapezoidal field of view 26 of a camera (not shown) of the vehicle 25 being the area of the camera that is used to detect markings on the roadway 20 serves. In 2 the already recognized markings 27, 28, 29 are shown as black bars, which have occurred or are currently occurring within the field of vision 26 moving across the roadway 20 due to the forward movement of the vehicle.

In 3 will that be in the 2 Perspectively shown trapezoidal field of view 26 formally defined more precisely. In this case, the field of view 26 of a camera 30 arranged on the front side of a vehicle is defined by the aperture angles α and β, it being assumed for further discussion that α=β applies. If the field of view 26 is not arranged symmetrically in front of the vehicle, α≠β applies. The total height h of the in 3 The triangle shown consists of a fixed height h _F and the height of the trapezoid h _T , ie: $$\text{H}={\text{H}}_{\text{f}}+{\text{H}}_{\text{T}}.$$

wobei für α = β gilt w₁ = w₂. Es können daher die Höhe h_T und die Breite w des Trapezes 26 variiert und als veränderliche Parameter des Sichtfeldes 26 der Kamera 30 betrachtet werden.The fixed height h _F is due to the mounting of the camera 30 on the vehicle and is therefore predetermined. The height h _T of the trapezium 26 and thus its upper width w, which is made up of the left width w ₁ and the right width w ₂ , are variable. Consequently: $$\text{w}={\text{w}}_{\text{1}}+{\text{w}}_2,$$

where for α = β applies w ₁ = w ₂ . Therefore, the height h _T and the width w of the trapezium 26 can be varied and considered to be variable parameters of the field of view 26 of the camera 30 .

4 shows the result for fields of vision 26 in front of a vehicle 25 as a function of the weather class, with the three weather classes “dry” (dry), “wet” (wet) and “rain” (rain) being used in the present example. It is obvious that further weather classes can be defined, for example the "rain" class could be differentiated into "light rain", "medium rain" and "heavy rain". Furthermore, classes such as "fog" and "snowfall" could also be defined, which can be differentiated into further subclasses.

It turns out that for each of the weather classes "dry", "wet" and "rain" 26 different parameters are used to form the respective fields of vision for optimal recognition of the lane markings. These different parameters for the height h, dfie composed of the trapezium height h _T and the fixed height h _F , and the width w of the field of view 26 lead to different trapeziums, namely the trapezium T _R for the weather class "rain", the trapezium T _D for the "Dry" weather class and the Trapez Tw for the "Wet" weather class. In the example, the three weather classes are self-explanatory. "Dry" stands for a dry road, "wet" for a wet road without rain and "rain" for a road with current rain. The weather classes can be derived, for example, from the Derive data from a suitable sensor, for example a rain sensor in combination with other sensors.

To determine the parameters, training drives are carried out under the classified weather conditions and the road markings are determined, with the parameters height of the trapezoid h _T and width w of the trapezium being varied within specified limits to evaluate the measurements. In other words, the road markings are measured in the specified weather class for a specified number of training drives, the measurement data are evaluated with specified sets of parameters for the field of vision and the localization of the vehicle is determined for the respective set of parameters and compared with reference values. A localization quality for the parameter set used for the respective weather class can then be determined from the deviation of the localization and the reference values, it being possible, for example, to use the absolute Euclidean distance with subsequent averaging. Average values of the localization quantities for the various sets of parameters can then be determined from the large number of training drives, so that an optimal set of parameters can be determined for a given weather class.

The following Table 1 shows an example of the parameter values “Width w” and “Total height h” determined by training drives for the weather classes “Dry”, “Rain” and “Wet” used here: Table 1 Class Width w (m) height h (m) Area (m ² ) Rain 7.3 16.1 99.82 Wet 10 16.4 123.82 Dry 9.7 15.5 113.96

5 shows a schematic representation of the device for the precise localization of a vehicle in its environment, ie on the roadway. A mono camera 30 creates cyclic images of the roadway or the environment in front of the moving vehicle. The image data from the camera 30 are fed to a device 31 for extracting the lane markings, which extracts the lane markings from the camera images and forwards them as measurement data. The road markings are usually extracted from the camera images using suitable image processing methods. These measurement data are supplied to an evaluation unit 34, which filters and stores the data to reduce the measurement noise. A current map of the surroundings of the vehicle is created from the data, which is continuously supplemented with the new data, ie the new data is tracked and merged with the previous data, ie the surroundings map.

In order to be able to create the environment map in the evaluation device 34, the vehicle position and the vehicle's own movement are determined from the data from a sensor system 32, which includes various vehicle sensors such as ESP and ABS for supplying odometric data and a GPS device, by means of a downstream device 33. This information is fed to the evaluation device 34 to create the environment map. Furthermore, the information from the device 33 for determining the vehicle's own movement is supplied to an electronic horizon 36 which, based on the vehicle position, provides a digital map of the corresponding surroundings of the vehicle, the digital map containing the lane markings. The electronic map can be requested and made available by the electronic horizon 36 via a mobile radio interface.

In order to be able to optimally localize and track the road markings in the evaluation unit 34, so that an optimal creation of the environment map of the evaluation unit 34 is possible, the evaluation unit must know the weather conditions, i.e. the weather class, so that, depending on the weather class, the corresponding parameter set for the field of vision of the Camera can be selected for evaluation. The weather class is determined via a corresponding device 35, which uses various sensors present in the vehicle to clarify what weather conditions prevail in the area surrounding the vehicle. Information from the rain sensor, the air conditioning system, the time, the lighting system of the vehicle, or another suitable environment sensor system can be used for this purpose. The prevailing weather is determined from the information and classified into weather classes. The result of the classification, i.e. the weather class, is fed to the evaluation device 34 for use in locating the road markings to create the environment map.

The environment map created by the evaluation device 34 and the digital map provided by the electronic horizon 36 are merged in a fusion device 37 and a precise localization of the vehicle in its environment is carried out. To do this, the lane markings recognized on the map of the surroundings are compared with the lane markings on the digital map and the position of the vehicle is determined again. The localization of the vehicle determined in this way is supplied to an output device 38 for further use.

Reference List

1

Crossing

2

Street

3

Street

4

first vehicle

5

second vehicle

6

GPS uncertainty circle first vehicle

7

GPS uncertainty area first vehicle

8th

GPS uncertainty circle second vehicle

9

GPS uncertainty area second vehicle

10

GPS collision area

11

circle of uncertainty precise localization of first vehicle

12

Uncertainty area precise localization of first vehicle

13

circle of uncertainty precise localization of second vehicle

14

Uncertainty area precise localization of second vehicle

15

Collision area precise localization

20

roadway

21

lane

22

lane

23

lane

24

lane

25

vehicle

26

field of view camera

27

recognized mark

28

recognized mark

29

recognized mark

30

camera

31

Extraction of the marks

32

Vehicle sensors, GPS

33

Determination of the absolute vehicle position and vehicle movement

34

Evaluation of the road markings, creation of a measured environment map

35

Environment sensors to determine the weather class

36

Electronic horizon (digital environment map)

37

Fusion of digital map and environment map

38

Precise localization output

H

overall height

hF

fixed height

hT

Height trapezoid

w

Overall width trapezoid

w1

left latitude trapezium

w2

right latitude trapezium

a

angle

β

angle

f

vehicle

TR

Trapeze for class "Rain" (Rain)

TD

Trapeze for class "Dry" (Dry)

TW

Wet Class Trapeze

Claims (10)

Method for generating an environment map for the purpose of locating a vehicle (4, 5, 25) in its environment, with the steps: - cyclic creation of images of the environment of the vehicle (4, 5, 25) using at least one image-generating device (30), - Extracting landmarks (27, 28, 29) from the cyclic images of the image-generating device and creating a measurement data set for each image, - determining the vehicle position and the vehicle's own movement, - creating an environment map from the measurement data set, with only a predetermined field of view within the measurement data set (26) is taken into account, as well as the vehicle position and the vehicle's own movement, characterized in that the specified field of view (26) for evaluating the cyclic measurement data records is a function of the current weather conditions. procedure after claim 1 , characterized in that the predetermined field of view (26) is parameterized by a set of parameters (h, w), the set of parameters being a function of the weather condition. procedure after claim 2 , characterized in that the weather condition is classified into a predetermined number of weather classes. Method according to one of the preceding claims, characterized in that the landmarks are defined by the lane markings (27, 28, 29), the surroundings by the lane in front of the vehicle (4, 5, 25) and the field of vision (26) by a trapezoid (26 ) are formed on the lane (20). Method according to one of the preceding claims, characterized in that the specified field of view for a weather condition is determined by a specified number of measurement runs for the respective weather condition, the optimization of the field of view (26) being carried out by localizing the vehicle (4, 5, 25 ) for each measurement run and comparison of each localization determined in this way with a reference localization. Device for generating an environment map for the purpose of locating a vehicle (4, 5, 25) in its environment, the device being set up and designed for carrying out the method according to one of the preceding claims, having: - an image-generating device (30) for generation cyclic images of the surroundings of the vehicle (4, 5, 25), - a device (31) for extracting landmarks (27, 28, 29) from the cyclic images of the image-generating device (30) and creating a measurement data set for each image, - a device (33) for determining the vehicle position and the vehicle's own movement, - a device (34) for creating a map of the surroundings from the measurement data record, wherein only a predetermined field of view (26) is taken into account within the measurement data record, as well as the vehicle position and the vehicle's own movement, characterized by - means (35) for determining the current weather condition, the predetermined field of view (26) being a function of the current weather condition. device after claim 6 , characterized in that the device (35) for determining the current weather condition carries out a classification of the weather condition into a predetermined number of weather classes. device after claim 7 , characterized in that exactly one field of view (26) is stored for each weather class, which is determined by measuring runs for the respective weather class. Method for locating a vehicle (4, 5, 25) in its environment with the steps: - Generating an environment map of the current environment of the vehicle (4, 5, 25), - Providing a digital map of the current environment of the vehicle, the digital Map has the landmarks (27, 28, 29) of the current environment, - merging the environment map and the digital map to generate a localization of the vehicle (4, 5, 25) in its environment, characterized in that the environment map with a method one of the Claims 1 until 5 is produced. Device for locating a vehicle (4, 5, 25) in its environment, the device for carrying out the method claim 9 is set up and designed, having - a device (34) for generating an environment map of the current environment of the vehicle (4, 5, 25), - a device (36) for providing a digital map of the current environment of the vehicle (4, 5, 25), the digital map having the landmarks (27, 28, 29) of the current environment, - a device (37) for merging the environment map and the digital map to generate a localization of the vehicle (4, 5, 25) in its Environment, characterized by a device for generating an environment map according to one of Claims 6 until 8th .

Images