DE102011115354A1 - Method for absolute position determination for navigating trailer on base of e.g. factory, involves moving vehicle at maximum speed based on calculated rotational position - Google Patents

Abstract

The method involves admitting the overall pattern (3') for the entire potential driving portion (4) in a database. The bottom image (7) of the base (2) is received based on the position relative to the vehicle through the camera. The pattern of the bottom image is stored in the database, and is compared with the overall pattern of samples (3). The rotational position of the vehicle in the overall pattern is calculated automatically. The vehicle is moved at the maximum speed based on the calculated rotational position.

Description

I. Field of application

The invention relates to a navigation method for autonomous, in particular in a limited environment, unmanned vehicles.

II. Technical background

For navigation, known solutions often use optical, non-contact methods.

A well-known navigation method for accuracy requirements in the range of a few meters is navigation by means of GPS, which, however, can not achieve greater accuracies and is not applicable in closed buildings because of the shielding of the communication from the GPS satellite.

For the navigation of vehicles with higher accuracy, especially within closed buildings, for example in factory buildings, hospitals, etc., and especially when the vehicles are to dock with it automatically to other vehicles or trailers, other methods are known.

From the US 6,868,307 For example, a navigation method for a floor surface cleaning robot is known.

This has at least one camera, which is directed down to the ground, alternatively an additional camera, which is directed upwards against the ceiling.

The downward-facing camera captures ground images from which special features of the optically uneven ground design are extracted and used to determine the direction of travel and the distance traveled since the last image as the vehicle continues, for which a large number of images, for example, 1500 per second, are taken ,

Here, therefore, a relative navigation compared to the last previous floor image is performed, which has the disadvantage that at each step existing errors are cumulated, so that such a relative navigation method - depending on the required accuracy - after some time possibly no usable positioning accuracy provides more.

Since the direction of travel and the route can only be meaningfully used for the navigation, if initially a starting point is known, a position determination is carried out in advance. This is done by means of the ceiling pictures, which receives the upward-facing camera.

In this case, eye-catching elements are used on the ceiling for orientation, such as ceiling lights. It is unclear whether only switched-on ceiling lights are usable here.

On the one hand, this has the disadvantage that here a high amount of stray light can enter the camera pointing upwards, on the other hand, that straight ceiling lights are often present in a regular arrangement, whereby a clear position determination is difficult or completely impossible.

Only when additional features that are not regularly available are added, for example smoke detectors or the like, can an unambiguous position determination be possible at all.

Another disadvantage of this approach is that separate databases of images must be kept for both the floor images and the ceiling pictures, apart from the expense of the two cameras.

According to the EP 0 366 350 navigates an unmanned vehicle by means of a single, directed against the ceiling camera by comparing the image recorded there with corresponding images from the database, which were taken as in the US script by means of a previous learning drive.

Here, ceiling lamps are used as distinguishing features on the ceiling, however, have the aforementioned disadvantages.

As with all navigation methods based on Deckenbilden here is the disadvantage that the distance of the ceiling camera from the ceiling is not always exactly the same and thus the size of the ceiling camera recorded ceiling elements does not always correspond to their actual size, so that an accurate navigation impossible is.

From the DE 10 2004 025 149 a navigation system for vehicles is known in which as a sensor chip that of an optical computer mouse is used. As a result, the movement of the vehicle over ground is determined.

As explained, however, the errors occurring at each relative offset between the preceding and next image are cumulated, which results in an excessively distorted position determination even after a short journey even if one starts from a defined starting point.

III. Presentation of the invention

a) Technical task

It is therefore the object of the invention to provide an optical, automatic navigation system for autonomous, preferably in a limited environment, unmanned and especially also on and uncoupling vehicles, which is able to the vehicle very accurately, especially in the accuracy range less than 1 cm, with the least possible technical effort.

b) Solution of the task

This object is solved by the features of claim 1. Advantageous embodiments will be apparent from the dependent claims.

For the purposes of the present application, some terms are defined below:

Original pattern:

The original patterns present on the ground, be it the streaks of a linoleum floor or the different deposits in a concrete or natural stone or the light and dark, alternating fibers in a chipboard.

Template feature:

The definable optical details of such an original pattern, such as the dark fibers versus the majority of the light fibers or overlying long fibers versus the remaining shorter fibers in a chipboard.

Reduced pattern:

The pattern reduced to the definable optical details.

Total sample:

The pattern extending over the total area.

Extracted characteristic:

A unique geometrical position determined from a pattern feature, generally a point, for example the center or centroid of a two-dimensional pattern feature, and / or the -. B. defined in one direction - tip of a fiber or the end of a streak.

Extracted pattern:

The absolute position and / or the relative position of all extracted from the pattern features features each other, so usually a dot pattern.

Feature group:

A group of extracted features, usually a group of points, created according to given grouping criteria, eg. B. number of features = points per group and defined by the relative position of these extracted features within the feature group to each other (relative position data), and possibly additional, in particular non-geometric, statements such. B. that two features of a group are part of the same wood fiber, such as both ends.

Triangle:

Example of geometric rendering of a group of three points. The group can also consist of more than 3 characteristics or points. Mirrored triangles are not identical because of the mismatched sense of rotation.

Database:

• a) dessen Form und Größe (Relativlage-Daten) eindeutig definiert, indem mindestens die geometrischen Angaben (z. B. Strecken-Längen und/oder Winkel), die die Relativlage der Merkmale der Merkmals-Gruppe relativ zueinander eindeutig beschreiben
• b) als auch deren absolute Lage (Position und Orientierung) in Form der Absolutlage-Daten innerhalb der Gesamtfläche eindeutig definiert.

The extracted pattern is stored in the database by having for each feature group and / or for each extracted feature a record that

• a) its shape and size (relative position data) uniquely defined by at least the geometric information (eg, track lengths and / or angles) that clearly describe the relative position of the features of the feature group relative to each other
• b) as well as their absolute position (position and orientation) in the form of the absolute position data within the total area clearly defined.

Location or position:

As a rule, always includes the rotational position of this object considered in the plan view.

The most important process for the automatic navigation of a vehicle is the determination of the position, in this case the position of the vehicle on the ground, on which the visible patterns consisting of individual pattern features are located.

First of all, the pattern existing over the entire possible traversing surface of the vehicle, that is to say the overall pattern, must be known and stored in searchable form, for example in a database.

To determine the position of the vehicle, the vehicle picks up a ground image by means of one or more cameras directed against the ground. The camera is at least during the preparation of the recording, so the bottom image, in a defined, known relative position to the vehicle and in particular fixedly mounted in the vehicle in a known relative position, so that the position of the bottom image on the optical sensor, such as a CCD Sensor of the camera, the position of the vehicle relative to this ground image in terms of position and rotational position of the vehicle can be closed.

In addition, however, it must be determined where the part of the pattern taken up by the bottom image is located within the overall pattern.

A theoretical possibility would be to save the overall pattern in its original form, and to search by means of EDP support the area which coincides identically with the portion of the original pattern taken up as a bottom image.

• – dabei unterschiedlichen Lichtverhältnissen
• – demzufolge oft auch unterschiedlichen Blenden und anderen Einstellungen,
• – unterschiedlichem Verschmutzungsgrad des Bodens und damit des Musters usw.

unterscheiden, so dass häufig die mit dem Bodenbild identisch übereinstimmende Stelle des Gesamtmusters nicht sicher ermittelt werden kann.In practice, this approach, however, almost never to success, since the captured when capturing the original overall pattern recordings with the later when taking a floor image extracts produced mostly due to

• - while different light conditions
• - as a result often different apertures and other settings,
• - Different degree of contamination of the soil and thus the pattern, etc.

different, so that often can not be determined with certainty identical to the bottom image identical location of the overall pattern.

Another disadvantage of this approach is the high memory requirements, if the original pattern with all the details should be stored in total.

Therefore, instead of the original pattern with all its features, only the extracted pattern from the extracted features, mostly individual points, is used for later comparison with the floor images.

• – entweder für jeden Ort innerhalb des potentiellen Fahrbereiches des Fahrzeuges, also innerhalb des Gesamtmusters, nur einmal vorhanden sind
• – oder die innerhalb des Gesamtmusters mehrfach vorhanden sind, jedoch mit so großem Abstand, dass zwei bei maximaler Fahrgeschwindigkeit des Fahrzeuges hintereinander aufgenommene Bodenbilder einen kleineren Abstand besitzen als der Abstand zwischen zwei Wiederholungen des Musters auf dem Gesamtmuster.

This procedure is only feasible with the further restriction that only those are selected as the original pattern and the pattern derived therefrom

• - Either for each location within the potential driving range of the vehicle, so within the overall pattern, are present only once
• - or which are multiple in the overall pattern, but with such a large distance that two at a maximum speed of the vehicle consecutively recorded floor images have a smaller distance than the distance between two repetitions of the pattern on the overall pattern.

For the absolute determination of the position and orientation you usually have not only the bottom image available, but in addition a rough determination, in which area the vehicle could be, starting z. B. from the last calculated location of the vehicle and the travel time traveled since then, direction of travel and driving speed, even if you do not know details such as the slip and the exact direction of travel exactly.

For this reason, an absolute position and position determination of the vehicle is again preferably carried out with each floor image of the vehicle produced, absolute absolute determination of the position in absolute values within the total area, that is to say the entire potential driving area and thus of the overall pattern.

If the ground on which the vehicle is moving is composed of individual plates, the coarse determination of the vehicle position can also serve by registering the boundaries between individual floor slabs run over since the last position determination (for which, however, it is necessary that the ground be completely covered by floor images is scanned when driving the vehicle), together with or as an alternative to the direction of travel and the driving speed and / or driving time, so that the coarse determination in particular results in the residence of the vehicle within a certain floor slab.

Starting from the rough determination of the vehicle position, a possible positional range of the vehicle is defined independently of this coarse determination, and for the fine determination of the vehicle position, the pattern, in particular the abstracted pattern, from the bottom image exclusively with the stored patterns, in particular the abstracted patterns compared to this possible position range, whereby the computational effort is minimized considerably.

The memory and computational effort and especially the accuracy of results and reproducibility are inventively further improved by the fact that from the patterns - ie both the overall pattern of the soil and the not the original pattern features, ie definable optical details of such a pattern (for example, the dark fibers in a particle board versus the majority of the light fibers or the like), captured and stored, but instead from these definable optical features of the original pattern extracted features and patterns are created, which are thus to represent the most unambiguous geometric position within the overall pattern and also within the individual pattern feature. As a rule, the extracted features are individual points.

Thus, of flat pattern features such as the fiber of a chipboard, automatically a defined point, z. Their center or centroid, and that point is used as the extracted feature.

In the next step, groups of either the original pattern features, in particular the extracted features, of the overall pattern, become groups. Although only the relative position data (and possibly the further, eg non-geometric statements) of the pattern features or extracted features within a group is used for comparisons later (and optionally the other, eg non-geometric Statements), but including the absolute position of this group in the overall pattern including position and rotational position about the vertical axis (orientation) stored in the database, in particular stored as a record

For example, the grouping criteria include the number of features within a group, for example, where three features each form a group.

Another grouping criterion may be that the maximum extent of a group should be smaller than the diameter of a floor image taken by the vehicle.

On the other hand, as a group formation criterion for the distance of the features within a group, a minimum distance can also be specified below which a group formation is not carried out. Such a minimum distance can be set in relation to the diameter of a bottom image, in particular amount to at least one fifth of the image diameter of a bottom image.

In general, it can not be ruled out that a feature can be a member of several groups.

In this way, all possible group formations are performed, but each group is stored only once.

For example, four visible features within a floor image of these four z. B. extracted features or points a maximum of four different triangles are formed, each representing a group, if a group should include three features.

The relative position of the features within a group (relative position data) is used on ascertained and stored information for each group, that is to say the minimum number of geometrical indications necessary for unambiguously defining the group, with which the geometric relative position of the features within the group are described can. These are z. B. in a triangle three necessary information, eg. As the side lengths and / or angle of the triangle.

• – der Abstand der beiden am weitesten voneinander entfernten Merkmale der Gruppe als erste geometrische Angabe ermittelt und gespeichert wird,
• – von einem dieser beiden Merkmale Verbindungslinien zu allen verbleibenden Merkmalen der Gruppe gelegt und deren Länge und Winkel zur ersten Verbindungslinie bestimmt und abgespeichert wird zusammen mit der Richtung, in der der Winkel von der ersten Verbindungslinie aus aufgetragen wird, oder indem ein festgelegter Drehsinn des Winkels, z. B. der Uhrzeigersinn, vorgegeben wird.

Here, specific specifications can be made by, for example, per group

• The distance between the two furthest apart features of the group is determined and stored as the first geometric indication,
• - By one of these two characteristics connecting lines to all remaining features of the group laid and whose length and angle to the first connecting line is determined and stored together with the direction in which the angle is applied from the first connecting line, or by a fixed sense of rotation of the angle , z. B. clockwise, is specified.

• – der Winkel der Verbindungslinie der beiden am weitesten voneinander entfernten Merkmale der Gruppe, also Abstandslinie bei einem Bodenbild, relativ zur Längsachse des Fahrzeuges.

Further, for each group of the floor image, the quasi-absolute position data is determined and stored describing the location of that group within the vehicle, e.g. B.

• - The angle of the connecting line of the two furthest apart features of the group, ie distance line in a floor image, relative to the longitudinal axis of the vehicle.

For determining the position, all groups of features are then first created from the currently recorded floor image, by means of which the position determination is to take place, according to the valid group formation criteria that contains this floor image and their relative position data together with their quasi-absolute position data, the position play this group to the vehicle, determined and possibly cached.

Then, for each group contained in the bottom image, a comparison with the database and all the relative position data stored therein is obtained Groups performed. Each match of the group in the floor image with one of the groups in the database (within the scope of a predetermined, in particular adjustable, tolerance range) represents a possible position of the vehicle.

It will often happen that the relative position data of one of the groups contained in the ground image even affects a plurality of relative position data stored in the database even in the limited scope of the database by the rough determination of the position and / or the "possible location area" of the vehicle fits groups.

Each match is - preferably stored only by depositing the corresponding position and rotational position of the vehicle - as a so-called theory entry.

From the list of theoretical entries is then selected as the actual position including rotational position of the vehicle (which is calculated from the relation of the absolute position data and the quasi-absolute position data) that theory entry, preferably without further measures, the most mentions in the comparison of all pattern in the bottom image with the patterns stored in the database.

For the bottom image, however, the position and rotational position of the individual groups relative to the surface of the optical sensor must be determined and cached, but this usually happens because only from this information, the relative position and direction of rotation of the group contained in ground image to a reference on Vehicle and a reference direction of the vehicle can be determined.

The number of most mentions should be much higher, at least twice, better than three times as high as the second most mentions. Furthermore, the number of most mentions should correspond at most to the number of groups within the floor image.

The determined in this way absolute position of the vehicle on the floor surface, ie within the overall pattern will not vote exactly, just because of the predetermined tolerance range when comparing the groups of the floor image with the groups from the overall picture, which is necessary here always slight differences due to other lighting conditions, current pollution and the like.

Therefore, a fine positioning is preferably carried out as a further step after the fine positioning, but possibly not always, but z. B. only if a coupling of the vehicle z. B. to a trailer or another vehicle or a fixed position, or if particularly difficult maneuvers are to be mastered:
For this purpose, a particularly suitable group from the current ground image for the position determination of the vehicle - that is, for the calculation of the quasi-absolute position data of this group from the ground image with the absolute position data of the corresponding group from the entire pattern - used.

A particularly suitable group is, for example, a triangle, which occupies the greatest possible extent in comparison to the total area of the floor image, since then position inaccuracies of the vertices of the triangle comparatively little effect on the center or center of gravity or another defined point of the triangle, from these three vertices is calculated.

However, as a particularly suitable group can also z. B. a group are used, in which the position of its vertices is subject to particularly low uncertainties.

Especially for this Feinstpositionierung as well as possibly for the previous step of merely fine positioning, so the comparison of the floor image with the database, could bring in terms of the size of the vehicle as large ground images advantages:
Preference is therefore proposed not only use a ground camera, but several ground cameras, namely z. B. with such a number as corresponds to the number of features of a group, so for example, three floor cameras in groups of three.

These three floor cameras are then placed within the vehicle in a triangle, but preferably not in the form of an isosceles triangle. The floor images of each individual camera are small compared to the distance of the floor cameras, so that the floor images preferably do not overlap each other. Nevertheless, it can now by means of appropriate grouping criteria groups, ie z. As triangles are determined, each corner is located in each one of the three floor images of the three floor cameras.

With regard to the finest positioning, when selecting such a large triangle, the positional blurring of its corner points in comparison to its center or center of gravity only plays a very minor role.

At least after Feinstpositionierung so that the absolute position of the vehicle within the entire travel surface, ie within the overall pattern, very accurately known.

This positioning of the vehicle is repeated at short intervals and proceeds in particular in real time during the procedure of the vehicle.

To facilitate the evaluation, the camera is always guided at the same distance from the ground. If the chassis of the vehicle has a sprung suspension, the suspension can change the distance of the camera to the ground when it is on the chassis. In this case, the camera is preferably mounted directly on the chassis or the movements of the suspension are scanned in real time and scaled according to these distance changes the current floor image.

The overall pattern is z. B. recorded by a variety of floor images is created on a learning trip, which partially overlap, and then these floor images are computer-technically set to the overall pattern together. Preferably, this learning run is performed with the same vehicle and in particular the same camera setting, as it is then used for the position determination. This saves conversions such as scaling due to other camera distance, different viewing angle of the camera and the like.

Such a learning journey is carried out in particular with a known route and possibly also known intermediate positions.

In order to reduce the evaluation effort for determining the position, ground images are produced in the moving vehicle at such a time interval that the individual floor images do not overlap with respect to their position on the ground, ie have no connection to one another.

The position and extent of obstacles within the potential driving range should also preferably be predefined for the navigation system before navigation begins.

If the overall pattern contains a pattern that repeats periodically, when driving the vehicle, preferably a repeat counter is counted, which counts the overtravelled repetitions of the periodically repeating pattern, this repeat counter being used, in particular for the rough determination of the position of the vehicle. The prerequisite for this is that all the periodically repeating patterns which are run over by the vehicle are also recognized, ie wherever a crossed-over periodic pattern is present, a bottom image is also produced.

• – entweder die eingescannten Platten in einer vorgegebenen Orientierung und Position im Fahrbereich einzubauen
• – oder die Platten willkürlich im Untergrund zu verbauen und anschließend ihre Lage wiederum mittels einer Lernfahrt des Fahrzeuges zu bestimmen.

Another possibility in a composite of individual plates driving range is the individual plates before their obstruction in the driving range individually with respect to their pattern, eg. B. read in a stationary, large scanner and then

• - Install either the scanned plates in a given orientation and position in the driving area
• - Or install the plates arbitrarily in the underground and then determine their location again by means of a learning drive of the vehicle.

As background materials and samples there are very different materials in question:
A typical application is commercially available chipboard as a substrate, where on the surface of the composition of the surface of individual fibers is visible. The majority of the fibers are light fibers, but there are a small number of dark fibers that can serve as definable extracted pattern features. The same applies to fibers whose length is above a predetermined maximum value, and whose center, center of gravity or a specific one of their tips can serve as an extracted feature.

When using a linoleum floor as a base, the streaks of the randomly arising marbling patterns can be used as a pattern in the production of the linoleum floor.

In the case of a laminate floor in which the individual panels of the laminate floor are imprinted with an imitation wood grain and only a limited number of printing templates are used in the production of the laminate panels, the pattern present on the panels is thus repeated, preferably periodically, in the case of a laid laminate floor at least if the procedure is followed.

If the ground is a wooden floorboard, the grown wood grain usually represents a unique pattern on each board, alternately bright and dark areas of which may serve as extracted features, in particular the line-shaped boundaries between these levels and darker areas. Also, a raw concrete floor and contained therein visible, different, in their relative arrangement each individual inclusions such as large and small pebbles, chippings and the like, can serve as a pattern, with as definable pattern features again inclusions, z. B. above or below a certain size limit or brightness limit lying, can serve.

c) embodiments

Embodiments according to the invention are described in more detail below by way of example. Show it:

1a : the original pattern on the ground,

1b the pattern reduced to the optically definable pattern features,

1c : the extracted pattern as it arises when each pattern feature is reduced to a point-shaped extracted feature, such as the center of each pattern feature,

2 an extracted pattern, as photographed and extracted in the form of a bottom image from the camera of a vehicle from the pattern of the background,

3 : the process of determining the position,

4a , b: the vehicle which travels on the ground and whose position is to be determined, and

5 : the procedure for determining the position.

1a shows a pattern 13 as it is on the ground 2 is present if this example z. B. of individual chipboard, composed of individual floor panels 12 , consists. The illustrated area may be part of a floor panel 12 So in the form of about a chipboard 16 , his, or even the overall pattern 13 ' , so the entire potential driving range 4 , represent.

The original pattern is formed by a variety of differently shaped and in their coloring also different fibers, which each have a planar structure with appropriate magnification.

As definable extracted pattern features 3a In this case, the dark fibers lying below a brightness limit value can - as shown in FIG 1b - used, omitting the pattern of all other parts of the original pattern 3 . 13 result in the following werter is worked.

In the presentation of 1c is each of these pattern features 3a , b, c on a single point, the so-called extracted feature 15a , b, c reduced, z. B. the center 19 of the respective, surface pattern feature 3a , b.

In 1c is also the bottom picture 7 as it regards its location from the pattern 3 photographed (see 1a ). This then becomes limited to the optically definable pattern features 3a , b, c, d and their transformation into extracted features 15a , b a section of the in 1c illustrated extracted pattern 103 created, its location within the pattern 3 or extracted pattern 103 However, it is not known, but should be determined, because this situation, the location of the vehicle 1 represents where the ground image 7 receiving camera - see 4 - is mounted. The following figures show how this position determination works:
As 2 shows, first all the extracted features 15a -I within the recorded floor image 7 determined.

Subsequently, these extracted features become 15a -I feature groups 20a , b created according to the given group formation criteria.

• – Die Gruppe muss sich vollständig innerhalb des Bodenbildes 7 befinden, mit der Folge, dass der Maximalabstand zwischen zwei extrahierten Merkmalen 15a, b nicht größer als der Durchmesser 22 des in diesem Fall runden Bodenbildes 7 sein darf,
• – eine Gruppe umfasst drei extrahierte Merkmale, die nicht auf einer Linie liegen und somit geometrisch betrachtet ein Dreieck bilden,
• – ein extrahiertes Merkmal, z. B. 15a kann Bestandteil mehrerer Gruppen sein,
• – der minimalste Abstand zwischen den zu einer Gruppe verbundenen extrahierten Merkmalen muss mindestens ein Fünftel des Durchmessers 22 des Bodenbildes 7 betragen.

It is assumed for the present example that the grouping criteria include:

• - The group must be completely within the floor image 7 with the result being that the maximum distance between two extracted features 15a , b not larger than the diameter 22 of the in this case round bottom picture 7 may be,
• A group comprises three extracted features which do not lie on a line and thus geometrically form a triangle,
• An extracted feature, e.g. B. 15a can be part of several groups,
• - The minimum distance between the extracted features connected to a group must be at least one fifth of the diameter 22 of the floor image 7 be.

It is assumed that with these group formation criteria from the in 2 a total of nine extracted features 15a -I a total of 12 groups 20a -L form.

• – Zunächst wird von den beiden am weitesten voneinander entfernten Merkmalen 15a, e der Gruppe, deren Abstand 21, also die Länge dieser Verbindungslinie 24a, ermittelt und gespeichert,
• – Anschließend wird von den verbleibenden Merkmalen der Gruppe, in diesem Fall nur noch dem Merkmal 15b, zum einen der beiden Merkmale 15a, b, z. B. dem weiter entfernten Merkmal 15a, eine Verbindungslinie 24b gelegt und deren Länge ermittelt und abgespeichert sowie deren Winkel 23b zur Verbindungslinie 24a zwischen den ersten beiden Merkmalen 15a, e ermittelt und abgespeichert. Das Merkmal 15a bildet dann den Scheitel dieser beiden Verbindungslinien. Dabei wird dieser Winkel 23b in einem festgelegten Winkel-Drehsinn, hier dem Uhrzeigersinn, beginnend von der ersten Verbindungslinie 24a aus, gemessen.

Each of these groups 20a , b, that is geometrically considered each of the triangles is now determined in terms of its unique geometric parameters, but usually not stored, but immediately compared with the database, as explained below. In the case of a triangle, three details (relative position data) are necessary for the shape and size, for example the lengths of the three sides of the triangle, or the length of two triangle sides and the size of the intermediate angle between them or the size of the three angles in the triangle, and for the quasi-absolute position relative to the vehicle further details:
One possibility, the clearly defining geometric parameters within a group z. B. 20a (Relative position data), looks like this:

• - First, of the two most distant characteristics 15a , e of the group whose distance 21 , ie the length of this connecting line 24a , determined and stored,
• - Subsequently, of the remaining characteristics of the group, in this case only the feature 15b , for one of the two characteristics 15a , b, z. B. the more distant feature 15a , a connecting line 24b placed and their length determined and stored and their angle 23b to the connecting line 24a between the first two features 15a , e determined and stored. The feature 15a then forms the apex of these two connecting lines. Thereby this angle becomes 23b in a fixed angular direction of rotation, here in the clockwise direction, starting from the first connecting line 24a measured.

This determines the shape and size of the triangle.

In addition, the angle becomes 23a this connecting line 24 through these two features 15a , e for z. B. longitudinal center axis 10 of the vehicle 1 determined and stored, measured from that of 15a to 15e pointing vector to the lying on the longitudinal central axis, forward-facing longitudinal mean vector. This gives quasi-absolute position data, since this angle 23a together with the known camera position in the vehicle clearly the absolute position of the triangle 15a , b, e within the vehicle.

In the same way, according to the same group formation criteria, is for the overall pattern 3 ' So for the entire potential driving range 4 of the underground 2 on which the vehicle is 1 the formation of all possible groups has been made and in turn each individual group in terms of shape and size defining geometric information (relative position data) as well as with regard to the position of the group in the overall pattern 3 ' (Absolute position data) as a record D1, D2 ... in a database 26 deposited, which accordingly can contain millions of records.

For determining the position from the current floor image 7 will now according to 3a - shown for only two triangles A, B in the ground image 7 - each feature group 20a , b, as triangles A, B thus their relative position data, each with all relative position data records D1, D2 ... the database 26 , ie all triangles a1, a2 ... b1, b2 in the overall pattern, and comparing each possible match, ie the congruence of the triangles except for the set tolerance range, as a so-called theory entry Th1, Th2, as in 3b shown.

Every correspondence, ie every theory entry, represents a possible position of this group z. B. 20a and therefore absolute positioning of the vehicle 1 on the overall pattern 3 ' So the underground 2 , which can be calculated by additional comparison of the absolute position data of the relevant triangle from the database with the cached quasi-absolute position data of the corresponding triangle from the bottom image. Where "position" means the position including the rotational position of the vehicle and the corresponding group 20a relative to the overall pattern 3 and thus to the underground 2 ,

If one of the group 20a of the floor image 7 corresponding record exists multiple times in the database, z. B. because the corresponding pattern in the overall pattern 3 ' repeated several times, can a group 20a also give several theoretical entries.

When checking the individual groups 20a , b of the ground image 7 successively a certain theory entry (= absolute positioning of the vehicle 1 ) are mentioned several times, especially the theory entry, the actual position of the ground image 7 and thus the vehicle 1 reproduces.

This actual position and orientation will be after comparing all the groups 20a , b with the entire database 26 determined by listing which of the theory entries Th1, Th2 as often in the comparison of the floor image 7 has appeared with the database, and the theory entry that occurred with the greatest frequency, in this case Th2, is considered the actual position and orientation of the floor image 7 and thus the vehicle 1 assumed, whereby the fine position determination is completed.

To the computational effort for the comparison of all groups 20a , b of a floor picture 7 with the database 26 To reduce, first, a rough determination of the position of the vehicle 1 carried out and based on this coarse determination a possible location area 14 be determined, on which the vehicle 1 can be maximum.

The comparison according to 3 will not work with the entire database 26 , special only still performed with the records that are within the possible position range 14 are located.

5 shows how starting from the last exactly carried out position determination Pos 1 of the possible position range 14 can be determined for the now to be carried out position determination Pos 2:
If from position 1 the vehicle 1 With a known speed in a known direction over a known journey time was moved and these values were stored, it can be the distance S1 determine the length and direction and from this the current coarse position of the vehicle 1 , However, this does not coincide with the real position, ie the fine position of the vehicle, in the rule, since important parameters for the calculation such as slip of the drive wheels, errors in the direction of travel and the registered speed etc to a deviation of the fine position of lead the coarse position.

Starting from the determined coarse position, a possible position range can now be determined 14 be determined, for example, a circle around the coarse position of the vehicle with a radius of z. B. 30% of the calculated distance traveled since the last position Pos 1.

Now it is determined for the fine position according to the procedures according to 1 - 3 , but limited in the database comparison to the part of the database, the possible location area 14 equivalent.

In this way, the vehicle is approaching 1 the goal, wherein the temporal and / or distance distance of the individual floor images 7 and thus the exactly sequentially determined positions Pos 1, Pos 2 of the vehicle 1 can decrease, the more the vehicle 1 approaching the desired destination.

For the journey to the destination, the navigation system should also be aware of the location of obstacles, such as those in 5 are shown.

4 shows in side view and top view the vehicle 1 and those attached to it, against the ground 2 directed camera 6 ,

This is on the one hand to that vehicle 1 whose position determination is to be performed on the ground. At the same time it can be the vehicle 1 which is used to learn the overall pattern during a learning trip 3 first of all to determine and subsequently in the database 26 save.

This vehicle 1 owns in the supervision of the 4b looks about a triangular basic shape, in which at each of the beveled corners a wheel 8a -C sits, which by a separate engine 9a , b, c, which preferably comprises a revolution counter.

Due to a special design of the wheels 8a , b, c, in whose circumference each spherical rollers are mounted in two radial planes distributed over the circumference, whose axes of rotation form tangents to a circle about the axis of rotation of the wheel, such wheels 8a , B are also moved in the direction of their axis of rotation. By individual control of the motors 9a , b, c by an unillustrated control can thus such a vehicle 1 Turn both on the spot and start in any direction you want without turning it.

In the center of the vehicle 1 is one against the underground 2 directed camera 6 mounted, preferably in an elevated area of the floor of the vehicle 1 is housed and a floor picture 7 to create with the largest possible diameter. The location of the camera 6 to the vehicle 1 is both in position and in terms of rotational position to the longitudinal axis 10 known.

LIST OF REFERENCE NUMBERS

1

vehicle

2

underground

3

template

3a, b, c

pattern feature

3 '

overall pattern

4

potential driving range

5

Database

6

camera

7

Floor picture, picture

8a, b

wheel

9a, b

engine

10

Direction of travel, longitudinal axis

11

border

12

baseplate

13

Pattern in the ground image

14

possible location area

15a, b

extracted feature

16

chipboard

17

fiber

18

antenna

19

Focus

20a, b

Feature group

21

distance

22

diameter

23a, b

angle

24a, b

connecting line

25

theory entry

26

Database

103

extracted pattern

113

extracted pattern

D1, D2

record

Th1, Th2

theory entry

Pos 1, Pos 2

Vehicle positions

S1

driving route

QUOTES INCLUDE IN THE DESCRIPTION

This list of the documents listed by the applicant has been generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Cited patent literature

• US 6868307 [0005]
• EP 0366350 [0014]
• DE 102004025149 [0017]

Claims (19)

Absolute position determination method for navigating a vehicle ( 1 ) on a surface ( 2 ), in particular for the purpose of coupling a vehicle ( 1 ) on a trailer, with the help of visible, especially visible, pattern characteristics ( 3a , b) existing patterns ( 3 ), where - the overall pattern ( 3 ' ) for the entire potential driving range ( 4 ), in particular in a database ( 5 ), - the vehicle ( 1 ) with one against the ground ( 2 ), with their relative position to the vehicle ( 1 ) defined camera ( 6 ) Floor pictures ( 7 ), - at least one of the 7 ) ( 13 ) in the database ( 5 ), in the overall pattern ( 3 ' ), patterns ( 3 ) and - if they match, the absolute position and rotational position of the vehicle ( 1 ) on the overall pattern ( 3 ' ) is calculated automatically, characterized in that as a total pattern ( 3 ' ) those are chosen where: - the patterns ( 3 ) for each location within the potential driving range ( 4 ) are unique or, within the potential driving range ( 4 ) Template ( 3 ) are present several times, in particular periodically, and the distance between two repetitions of the same pattern ( 3 ), in particular the period length, but at least twice the distance from the moving vehicle ( 1 ) recorded floor images ( 7 ) corresponds to when the vehicle ( 1 ) is moved at a fixed, in particular the maximum, speed. Method according to claim 1, characterized in that with each of the vehicle ( 1 ) prepared floor image ( 7 ) an absolute position determination of the vehicle ( 1 ) is carried out. (Rough determination) Method according to one of the preceding claims, characterized in that - for determining the position first a coarse determination and then a fine determination is performed and in particular - for coarse determination and fine determination the same, in particular in a database ( 26 ) deposited patterns ( 3 ), in particular extracted features ( 15a , b) or extracted pattern ( 103 . 113 ) being used as a coarse determination the fine determination on the basis of the last floor image ( 7 ) taking into account the driving time, driving speed and direction of travel of the vehicle ( 1 ) can be used. Method according to one of the preceding claims, characterized in that - in the composition of the substrate ( 2 ) from individual floor slabs ( 12 ) as a rough determination the last fine determination plus the crossed over since then borders between individual floor slabs ( 12 ), so that the rough determination in particular the stay of the vehicle ( 1 ) within a particular floor slab ( 12 ) and / or - based on the position of the coarse determination a possible position range ( 14 ) of the vehicle and for fine determination the patterns ( 13 ) from the recorded soil image ( 7 ) exclusively with the deposited patterns ( 3 ) from the possible positional area ( 14 ). (Pattern generation) Method according to one of the preceding claims, characterized in that - from the patterns ( 3 . 13 ) individual, defined extracted features ( 15a , b), in particular defined points, and from the arrangement of these extracted features ( 15a, b ) extracted patterns to each other ( 103 . 113 ) are created and stored in particular and in particular - of each planar pattern feature, eg. As the fiber of a chipboard, automatically a defined point, z. B. the midpoint ( 19 ) and then as an extracted feature ( 15 ) is used. Method according to one of the preceding claims, characterized in that - groups ( 20a , b) pattern characteristics ( 3a , b), in particular extracted features ( 15a , b), according to predetermined group formation criteria, in particular the number of characteristics ( 15a , b) in a group ( 20a ) and the particular extracted features ( 15a , b) the group ( 20a ) are used in their relative position to each other (relative position data) for comparisons and in particular stored together with their absolute position data in the database and in particular - a group-education criterion is that - either the maximum extent of the group ( 20a ) is smaller than the diameter ( 22 ) of a single floor image ( 7 ) Or any feature, e.g. B. every corner of the polygon, in another of the simultaneously recorded by multiple cameras bottom images. Method according to one of the preceding claims, characterized in that for the Distance of the features ( 15a , b) within a group ( 20a , b), in particular the largest distance ( 21 ) of two specific characteristics ( 15 ) within a group, a minimum distance is specified, below which a group ( 20a , b) is not defined and the minimum distance in relation to the diameter ( 22 ) of the image taken by the camera ( 7 ), in particular at least 1/5 of the image diameter ( 22 ) is. Method according to one of the preceding claims, characterized in that - each extracted feature ( 15a , b) Member of several groups ( 20a , b) characteristics ( 15a , b) and in particular neighboring groups ( 20a , b) can overlap one another and / or - in the case of a learning journey before the start of navigation with in particular known travel route and in particular also known intermediate positions, the patterns ( 3 ) of the underground ( 2 ) in the form of z. B. individual floor images ( 7 ) and a computer-aided design ( 3 ' ) of the potential driving range ( 4 ). Method according to one of the preceding claims, characterized in that the patterns ( 3 ) of the potential driving range ( 4 ) are determined by their location by 12 ) existing potential driving range ( 4 ) before creating the driving range, the patterns ( 3 ) of the plates ( 12 ) individually before obstructing in the underground ( 2 ) are scanned individually as a whole and then - either in a predetermined orientation and position in the driving range ( 4 ) - or the location of the individual installed plates ( 12 ) by means of a learning run of the vehicle ( 1 ) is determined. (Base materials) Method according to one of the preceding claims, characterized in that as a pattern ( 3 ) - when using chipboard ( 16 ) the dark fibers ( 17 ) in the surface of the otherwise bright chipboard and / or the tips of fibers ( 17 ) whose length is above a predetermined minimum value, or - if a linoleum floor is used whose random marbling pattern is used, or - if a laminate floor is used, its periodically repeating pattern, which is unique in the period ( 3 ) is used, or - when using a raw concrete floor whose different, unique in their relative arrangement, inclusions are used, or - using a wooden plank floor whose grown wood grain is used as always unique or at best, periodically repeating similar pattern. (Storage of a group in the database) Method according to one of the preceding claims, characterized in that - the relative position of the extracted features ( 15a , b) within a group ( 20a , b) (relative position data) in the form is determined and stored, that for the clear definition of the group ( 20a , b) necessary, least number of geometrical indications such. As lengths or angles and optionally non-geometric information such as belonging to the same fiber, the group ( 20a , b) is determined and stored, and in particular the relative position data are determined in such a way that - of the two furthest removed features ( 15a , e) the group ( 20a ) the distance ( 21 ) of the two characteristics, - of the two most distant characteristics ( 15a , d) the group ( 20a ) determines the distance between these two features, and - the vertex angle at the vertex ( 5 ) between these two distances is determined as a relative position. Method according to one of the preceding claims, characterized in that the absolute position of each group (eg. 20a , b) is determined and stored within the overall pattern, in particular by - the vertex ( 5 ) with regard to its position and - the distance ( 21 ) between the two farthest features ( 15a , b) the group ( 20a ) is determined and stored in terms of its orientation in the basic pattern. (Theory entries) Method according to one of the preceding claims, characterized in that - from the current floor image ( 7 ) all the pattern features ( 3a , b), in particular all extracted features ( 15a , b), and all ensuing groups ( 20a , b) features are created according to the group-educational criteria, - each feature group ( 20a , b) with those in the database ( 5 ), in particular for the possible location ( 14 ), deposited patterns ( 3 ), ie records (D1, D2) is compared and - for each match as a theory entry (Th1, Th2) of a possibly matching pattern ( 3 ) from the database ( 5 ) and the resulting current position and orientation of the vehicle ( 1 ) is cached. Method according to one of the preceding claims, characterized in that for the conformity of a group from the floor image with a group from the database, a tolerance range, in particular the position of the individual extracted features of the group, is specified, which is adjustable in particular. (Sorting theory entries) Method according to one of the preceding claims, characterized in that over the, in particular all, patterns of a floor image ( 7 ), and the most frequently occurring theory entry (Th2) as the actual position of the vehicle (Th1, Th2) according to the frequency of their occurrence 1 ) is set. (Fine positioning) Method according to one of the preceding claims, characterized in that after determining the most frequently occurring theory entry as the actual position of the vehicle, in particular for coupling to another vehicle or a trailer or a fixed position, a Feinstpositionierung is performed by a group specifically suitable for this purpose (eg 20b ) from the ground image ( 7 ) is selected on the basis of which data the position of the vehicle ( 1 ) is recalculated again, and in particular as a specifically suitable group ( 20b ) the one with the greatest geometric distance of its features, in particular extracted features, is selected. Method according to one of the preceding claims, characterized in that - the group selected specifically comprises a plurality of spaced-apart floor cameras ( 6 ), separate floor images ( 7 ) and in particular each feature is arranged in one of these floor images and / or - the camera ( 6 ) always at the same distance to the ground ( 2 ), in particular the camera ( 6 ) at a relative to the chassis of the vehicle ( 1 ) sprung suspension is mounted on the chassis, or the camera ( 6 ) at the known distance to the ground ( 2 ), which can also change, is mounted. Method according to one of the preceding claims, characterized in that - the floor images ( 7 ) are produced at such a time interval that they are separated from the ground ( 2 ) photographed floor pictures ( 7 ) have no connection to one another and / or - when using repetitive, in particular periodic, patterns ( 3 ) while driving the vehicle ( 1 ) a repetition counter is stored, which is used in particular for the rough determination of the position. Method according to one of the preceding claims, characterized in that - the photographing of the substrate ( 2 ) during the learning trip to create the overall picture ( 3 ' ) using the same camera ( 6 ) and the same vehicle ( 1 ) is carried out, with which then the position determination and in particular the navigation is performed or - when photographing with another camera, in particular an external scanning of the bottom plates ( 12 ), before using the characteristics ( 15a , b) and feature groups ( 20a , b) from the database ( 5 ) a scaling to that of the camera ( 6 ) of the vehicle ( 1 ) generated groups ( 20 ) of characteristics ( 15 ) he follows.

Images