阅读说明：本技术 用于沿着虚拟的轨系统自动地导向车辆的方法 (Method for automatically guiding a vehicle along a virtual rail system ) 是由 S·西蒙 于 2018-09-14 设计创作，主要内容包括：本发明涉及一种用于沿着虚拟的轨系统自动地导向车辆的方法,其中,感测地表的特征,所述车辆在所述地表上运动或者将要在所述地表上运动,并且将所述特征转化为至少一个工作签名(A*-O*),其中,检查,所述至少一个工作签名(A*-O*)是否与所述虚拟的轨系统的至少一个参考签名(A-O)一致,其中,所述至少一个参考签名(A-O)配属有在所述虚拟的轨系统上的位置,并且,如果所述至少一个工作签名(A*-O*)和所述至少一个参考签名(A-O)一致,则推断出所述车辆在所述虚拟的轨系统上的位置。(The invention relates to a method for automatically guiding a vehicle along a virtual rail system, wherein a feature of a ground surface on which the vehicle is moving or is about to move is sensed and converted into at least one operating signature (A-O), wherein it is checked whether the at least one operating signature (A-O) corresponds to at least one reference signature (A-O) of the virtual rail system, wherein a position of the vehicle on the virtual rail system is associated with the at least one reference signature (A-O), and the position of the vehicle on the virtual rail system is inferred if the at least one operating signature (A-O) and the at least one reference signature (A-O) correspond.)

1. A method for automatically guiding a vehicle (1) along a virtual rail system (3),

wherein a characteristic of a ground surface (2) on which the vehicle (1) is moving and/or is about to move is sensed,

and converting said features into at least one working signature (A x-O),

wherein it is checked whether the at least one working signature (A X-O) corresponds to at least one reference signature (A-O, S) of the virtual rail system (3), wherein the at least one reference signature (A-O, S) is associated with a position on the virtual rail system (3),

and, if the at least one operating signature (A-O) and the at least one reference signature (A-O, S) are identical, the position of the vehicle (1) on the virtual rail system (3) is deduced.

2. Method according to claim 1, characterized in that a plurality of working signatures (A x O) and a plurality of reference signatures (A O, S) are checked for consistency and the position on the virtual rail system (3) where the highest number of consistency between the working signatures (A x O) and the reference signatures (A O, S) is inferred for the vehicle (1).

3. Method according to claim 2, characterized in that the highest number of correspondences between the working signature (A-O) and the reference signature (A-O, S) is found from at least one histogram of said correspondences.

4. Method according to one of the preceding claims, characterized in that the assignment of the reference signatures (a-O, S) to the position on the virtual rail system (3) is stored in a correspondence table.

5. The method of claim 4, wherein the correspondence table is derived by transmission from at least one transmitter.

6. Method according to one of the preceding claims, characterized in that if the position of the vehicle (1) on the virtual rail system (3) has already been determined, the reference signatures (A-O, S) for comparison when determining the next position are limited to those reference signatures (A-O, S) in position tracking: the reference signature is located within a search area, which is generated by or around a search area surrounding the found position.

7. Method according to one of the preceding claims, characterized in that the reference signature (A-O, S) is updated at least partially by means of the operating signature (A-O) if the position of the vehicle (1) on the virtual rail system (3) has been determined.

8. Method according to one of the preceding claims, characterized in that a control signal for the vehicle (1) is provided in relation to the position of the vehicle (1) on the virtual rail system (3), with which control signal the movement of the vehicle (1) is controlled.

9. Method according to claim 8, characterized in that control signals controlling the steering mechanism of the vehicle (1) are set for a lateral direction relative to the virtual rail system (3) and control signals controlling the drive mechanism of the vehicle (1) are set for a longitudinal direction relative to the virtual rail system (3).

10. Method according to one of the preceding claims, characterized in that a control signal for the vehicle (1) is provided in relation to the position of the vehicle (1) on the virtual rail system (3), with which control signal at least one work implement (21) of the vehicle (1) is controlled.

11. The method according to any one of the preceding claims, wherein the characteristics of the earth's surface are sensed only for a predetermined section and/or layer (31) of the earth's surface.

12. Method according to any one of the preceding claims, characterized in that the illumination device (5) is operated by means of a direction sensor (9) such that the illumination light or illumination color illuminates the illuminated surface from a predetermined direction independently of the orientation of the vehicle (1).

13. A computer program arranged for carrying out each step of the method according to any one of claims 1 to 12.

14. A machine-readable storage medium on which a computer program according to claim 13 is stored.

15. An electronic control unit (20) which is provided for automatically guiding a vehicle along a virtual rail system by means of a method according to one of claims 1 to 12.

16. A vehicle (1) having sensing means for identifying characteristics of a ground surface (2) and being arranged for being automatically guided along a virtual rail system (3) with a method according to any one of claims 1 to 12.

17. Vehicle (1) according to claim 16, characterised in that the sensing device has at least one optical image sensing device (4).

18. Vehicle (1) according to claim 17, characterized in that an illumination device (5) is provided, which illuminates the ground surface (2) sensed by the image sensing device (4).

19. Vehicle (1) according to claim 16, characterized in that the sensing device has at least one touch sensor (23, 26).

20. Vehicle (1) according to claim 16, characterized in that the sensing device has at least one air pulse sensor or air block sensor (32).

21. Vehicle (1) according to claim 16, characterized in that said sensing means have at least one electromagnetic sensor.

22. Vehicle (1) according to claim 16, characterized in that the sensing device has at least one acoustic wave sensor.

23. Vehicle (1) according to any of claims 16 to 22, characterized in that the vehicle (1) has a direction sensor (9).

Technical Field

The invention relates to a method for automatically guiding a vehicle along a virtual rail system. Furthermore, the invention relates to a computer program implementing each step of the method when the computer program runs on a computer; and a machine-readable storage medium storing the computer program; and an electronic controller, which is provided for carrying out the method according to the invention. Finally, the invention relates to a vehicle which is provided for being guided automatically along a virtual rail system.

Background

Nowadays, automatically traveling vehicles or robots are used in a variety of situations, which rely on self-positioning. Typically, in vehicles with image sensing devices, such as cameras, features in the image are detected, which are assigned descriptors and then the real-time features in the image are compared with features of a database and assigned to a location. In the case of the assignment of features and positions, distance measurements are usually also taken into account. The method used in this respect is "Simultaneous Localization and Mapping (SLAM)". In this method, it is provided that the vehicle can construct a map at each location, locate itself within the map, and accurately track its location.

Usually, for the positioning, projecting features (safety features) of the structure are used, which can also be found again at other viewing angles and at other distances and to which descriptors are assigned. As examples, pillars, trees or parts of trees, building parts, walls, corners, etc. are mentioned. Most of these structures are located in the remote field of the vehicle. These structures are generally not durable or cannot be perceived under certain light conditions. These features cannot be found again. It may also happen that: the prominent features that can be found again cannot be sensed by the vehicle, since there is no suitable structure in the environment, for example because the environment lacks structure and the distant field is poorly visible, for example due to low camera height.

Automatically driven vehicles are usually intended to move on a definable route. For example, automatically driven vehicles are intended to be moved on roads (and to parking lots), and automatically driven transport vehicles are intended to be moved on a planned road between a pickup position and a discharge position, for example, in a factory building.

Disclosure of Invention

A method for automatically guiding a vehicle along a virtual rail system is proposed. The term vehicle here includes, in addition to motor vehicles, also commercial vehicles, transport means, mobile robots which are self-propelled, surface conveyers, and also aircraft which move close to the ground, for example unmanned aircraft or landing aircraft.

Sensing characteristics of such a surface: the vehicle moves on the earth's surface and/or will move on the earth's surface in the future. These features are obtained from sensor signals, such as two-dimensional grayscale signals or two-dimensional color image signals of an image sensing unit of the vehicle, for example a camera. The sensor signal may also be constituted by a signal of a one-dimensional image sensing unit when the vehicle is moving in the second direction. The features are intermediate stages of signal extraction with which locations on the earth's surface can be characterized. For this purpose, the observation signal is assigned to the local segment of the location. For example, the convolution or filtering of the segment from which the N-dimensional vector for the feature is known may be performed using one or more wavelets. Preferably, the features are determined in the same manner at each position.

The earth's surface may be various types of man-made or natural earth's surface having distinguishable features that remain at least partially unchanged over a period of time. In particular, the method may be used in the case of a floor having a random pattern. A typical random pattern provides sufficient variation in its surface structure, brightness or color so that distinguishable features can be sensed at different locations. The following ground types are particularly suitable for this:

-bitumen;

-natural or artificial stone floors or stone pavers with random surfaces;

-concrete or screed floors;

hard floors painted with color, wherein so-called color flakes having other colors are sprinkled in the still moist pigment;

-a felt floor;

-a cork floor;

-a carpet;

-a linoleum floor;

-industrially manufactured plastic flooring with a random surface;

-a farmland or farmland ground; or

-grass, lawn or green belts.

The first-mentioned new types of ground are important primarily for industrial and transport robots, while the last-mentioned types of ground, i.e. grass, lawn or green belts in particular, are very important for mowing robots. Here, it should be noted that plants growing on the ground surface, such as grass, are also sensed when sensing the characteristics of the ground surface. If features on all layers of the ground surface are sensed along with the plants, the sensed features of the ground surface change significantly due to growth of the plants within the relevant time segment, for example, between two mowing cycles of the mowing robot.

It may be provided that the characteristics of the earth's surface are sensed only for a predetermined section of the earth's surface, in particular only for a specific layer of the earth's surface. In connection with grass, lawn and green belts, turf, and thus the uppermost ground layer on which plants grow, can advantageously be used for sensing features. Here, in particular ground structures, small stones and/or withered plants can be used as features.

The features of the earth's surface are converted into at least one working signature. The signature is a code of the feature or of the sensor signal belonging to the feature, said code characterizing the position on the surface of the earth. The signature may be electronically stored and processing continued. In converting features into signals, information loss due to encoding typically occurs. In this case, parts of the information that are not necessary for characterizing the position are discarded first. For example, the signature may be constructed as a concatenation of binary coded numbers that embody a vector of features in quantized form. Preferably, the conversion of the features into signatures can be weighted, wherein the weighting can be carried out by the user or, for example, by a neural network in the sense of training. Alternatively, a plurality of working signatures may be constituted by one feature or a signature may be constituted by a plurality of features or a plurality of signatures may be constituted by a plurality of features.

A virtual rail system is a region between at least one starting point and at least one end point, in which region the vehicle moves on a predetermined route (virtual rail). In this case, the vehicle can be moved in both directions on a predetermined virtual rail. For the exemplary case where the vehicle is a transport robot, the virtual rail reflects the transport route for the transport robot. For another exemplary case where the vehicle is a lawn mowing robot, the virtual rail reflects a track: the mowing robot mows grass along the track. Preferably, these tracks extend in serpentine shape and take into account the working width of the mowing tool.

The virtual rail system is stored as a data record of the position of the ground surface on which the vehicle is to be moved. Reference signatures are associated with these positions on the virtual rail system. The reference signature thus constitutes a map of the virtual rail system. In this case, the reference signature can be associated with exactly one position on the virtual rail system, which can be unambiguously determined from a reference signature. The reference signature can also be associated with a plurality of positions on the virtual rail system. The position cannot be derived from only one reference signature but a plurality of reference signatures are required, which are assigned to the same position or to adjacent positions. Preferably, the assignment of the reference signature and the position is stored in a correspondence table, which is designed as a lookup table. The locations on the on-track system are assigned addresses within the correspondence table. The signature is regarded as a number indicating an address of the correspondence table, and is thus used to determine the address in the correspondence table. Thus, storing the signature itself can be omitted. If storage or removal of the reference signature is subsequently referred to in association with the correspondence table, this simplified expression can be understood as: the reference location belonging to the signature is stored or cleared.

The signature need not be univocal, so that multiple locations on the virtual rail system can have the same signature and can therefore refer to the same address in the correspondence table. It is therefore advantageously provided that a plurality of positions can be stored for each reference signature in the correspondence table. It should be noted here that the surface covered by the virtual rail system, and thus its length and its width, has an influence on the probability of such multiple occurrences of the signature in the correspondence table. As explained above, the correspondence table is advantageously set up such that each table area can hold a plurality of locations. In this case, each table area can be fixedly assigned a memory capacity or the entire available memory can be flexibly subdivided into the table areas, for example by means of dynamic lists. As a result, a plurality of locations can be saved in a smaller correspondence table, i.e. in a smaller range of addresses or signatures. The maximum number or average number of positions that can be stored per table area depends on the conversion of the characteristics and the length of the virtual rail system and therefore also on the intended use area of the vehicle.

The virtual rail system can preferably be divided into a plurality of sections. Preferably, each section of the virtual rail system can be assigned a part of the correspondence table. This can also be understood as: each section of the virtual rail system can be assigned its own correspondence table, wherein the different correspondence tables are advantageously compatible with one another in that they have the same address range and are premised on the same type of signature. The electronic control unit of the vehicle can advantageously call up the part of the correspondence table associated with the current section of the virtual rail system on which the vehicle is located or is moving, and the part of the correspondence table associated with the adjacent section of the virtual rail system. The unneeded portion of the correspondence table may then be removed and stored, for example, in a flash memory or "data cloud". As a result, the electronic controller can cope with a smaller memory, which only has to maintain the just mentioned part of the correspondence table.

The reference signature can be derived in the same way as explained for the working signature. In other words, the reference signature constitutes a map of the virtual rail system. To derive the reference signature, one or more of the following methods may be selected.

The vehicle may complete the "learning run" before the vehicle is automatically guided. In learning to travel, the vehicle is controlled or guided by the user or by another vehicle that has already been learned. The vehicle moves along a line which should then constitute a virtual rail system. During the learning mode of travel or at a point in time before the following automatic guidance, the reference signature for the virtual rail system is formed from the features in the manner described above. The reference signature thus obtained can then preferably be stored in the aforementioned correspondence table in that the associated position is stored in the correspondence table. In this way, the vehicle can be automatically guided in each new surroundings following the learning. For the exemplary case where the vehicle is a robot lawnmower, it may be provided that the learning travel is carried out along the boundary of the area to be mowed. In other words, the learning travel is guided along the edge of the grass surface. In this case, the areas that should not be mowed, for example flower beds or channels, can be omitted during driving learning, or can be marked specifically as described again below. By this learning travel, the boundary line conventionally used for guiding the mowing robot can be omitted.

According to one aspect, the correspondence table and thus also the reference signature may be transmitted by at least one transmitter. Thus, the vehicle can be automatically guided in the new surroundings immediately. The transmission takes place wirelessly or by wire and can take place directly between the transmitter and the electronic control unit of the vehicle or be guided by a server, that is to say, in other words, be called up from a "data cloud", wherein the electronic control unit is preferably connected to the receiving device. In this case, the correspondence table can be transmitted in its entirety or only in sections, wherein the section of the correspondence table is associated with the reachable position on the on-rail system. The transmitter may for example be integrated in another vehicle. This is expedient when the vehicle follows another vehicle, for example in a fleet of motor vehicles or trucks on the road or in a fleet of transport fleets as a plurality of mobile transport means. Alternatively, the transmitter may be fixedly arranged. In this case, in particular, a plurality of radio beacons may be provided, which each transmit a part of the correspondence table which is assigned to the following section of the virtual rail system: over which the vehicle can move within the transmission radius of the transmitter.

Alternatively, the ground can also be sensed beforehand, independently of the vehicle, with a sensing mechanism designed for this purpose, for example a ground scanner. Advantageously, the sensor device is advantageously provided for effectively sensing a larger section of the virtual rail system. The reference signature can then be ascertained from the sensed sensor signals, as already explained, by means of the features and stored in the central server. Finally, the reference signature is transmitted to the vehicle, preferably in the form of a correspondence table, as explained above, via a radio connection. In this case, a virtual rail system can be planned at the PC workstation and the desired rail course, curve radius and/or spiral curve can be determined there. In particular, safe distances can be observed in this way. In the case of transport means in the factory, collisions with persons, objects, infrastructure and/or collisions of these transport means with one another can thus be avoided. In road traffic, it can be ensured by this planning that traffic regulations are complied with.

For self-localization, it is checked whether at least one of the working signatures obtained in the current run corresponds to at least one of the reference signatures of the virtual rail system. This is important in particular for this case: in this case, exactly one position is assigned to the reference signature. The position of the vehicle on the virtual rail system is inferred if the at least one operating signature and the at least one reference signature of the virtual rail system coincide. If a unique working signature is compared with a unique reference signature, instead of searching for a consistency as good as possible, as is usually done in connection with similarity or distance measures, the perfect consistency, i.e. identity, of the two signatures is searched. This results in the following advantages: the check for consistency can be carried out with significantly less computational effort than the check for similarity or dissimilarity by means of a similarity measure or a distance measure.

When using the correspondence table already set forth, the working signature is also regarded as a number indicating the address of the correspondence table. Since both the working signature and the reference signature of the virtual track system specify the same address and therefore point to the same table area, both are considered to be corresponding, identical signatures. The position of the vehicle on the virtual rail system is deduced from the table areas pointed to by the two signatures.

The signatures preferably have a length of between 8 and 32 bits, so that a compromise can be found between too short signatures, in which case only a small number of positions can be distinguished, and too long signatures, which lead to large correspondence tables which are premised on a large memory capacity and which additionally increase the probability of errors in consistency. According to one aspect, signatures that are too short can be merged by observing groups with a fixed geometric arrangement, for example two signatures of the same length at two positions staggered with respect to each other. Furthermore, it should be noted that the signature need not be selected to be so long that all positions on the virtual rail system or on the current section of the virtual rail system are unambiguously associated, since, as already mentioned, multiple signatures are allowed to occur.

According to one aspect, it is checked whether the plurality of working signatures are consistent with the plurality of reference signatures. This is particularly important in the case of a reference signature with several positions associated with it. The number of correspondences between the operating signature and the reference signature is counted and the number of correspondences is assigned to the respective position of the vehicle on the virtual rail system. Finally, the position of the vehicle with the highest number of consistencies on the virtual rail system is deduced. Optionally, a weighting of the consistency can be performed here. The following properties are utilized here: the position determined by the mapping table is accumulated according to the correspondence between the working signature and the reference signature, which enables an inference of the actual position on the rail system. In other words, each correspondence between a working signature and a reference signature embodies a determination of the location or locations assigned to the reference signature. Thus, not all features that are converted to a working signature must be consistent with features that have been converted to a reference signature. As a result, measurement uncertainties and inaccuracies can be compensated for on the one hand. On the other hand, changes in the features themselves, for example caused by dirt, damage, wear and/or other effects, may be obscured.

According to one aspect, the highest number of correspondences between the working signature and the reference signature is taken from at least one histogram of correspondences. Each coincidence is associated with a respective histogram bar, which in turn is associated with a position on the rail system. Next, such histogram bars are searched: the histogram bar has the most consistency and the searched position is deduced therefrom. Alternatively, the group with the most consistency in the aggregate of the adjacent histogram bins may be found. If a plurality of correspondence tables are used or the correspondence tables are divided into a plurality of sections, a histogram may be created as explained before for each correspondence table or for each section of the correspondence table and finally the histogram bin with the most consistency across all histograms is found. Alternatively, a plurality of histograms with different local resolutions can be used for the correspondence table or for a part of the correspondence table, wherein if a histogram bin has been found for a histogram with a low resolution, the histogram with the higher resolution for the associated area/length is used in the next step. The histogram may be one-dimensional and/or two-dimensional. For example, in a two-step method, the first histogram is one-dimensional and has a resolution of 1m per histogram bin, while the second histogram is two-dimensional and has a resolution of 1cm × 1cm per histogram bin. This provides the following advantages: the histogram used may on the one hand require little storage capacity and on the other hand the maximum value may simply be found.

If the position of the vehicle on the rail system has already been determined, the following steps can be carried out: if a location Tracking is carried out, also referred to as "Tracking", in which the corresponding next location is determined if the initial location is known, the reference signature used for comparison in determining the next location can be limited to this reference signature: the reference signature is located within a search area that is generated by a search area surrounding the found position or around a search area surrounding the next position. By limiting the reference signatures to those located within the search area, the computational effort and/or storage capacity can be reduced, since the complete virtual track system does not have to be considered.

Furthermore, for the case where the position of the vehicle on the rail system has already been determined, the reference signature can be updated at least partially by means of the operating signature. Thereby, it is possible to adapt the reference signature of features that have changed, e.g. due to aging, wear, contamination, tire wear, repair, etc. Preferably, such updating has already been done when a number of other reference signatures also point to a location. Particularly preferably, the updating is carried out permanently during operation. Advantageously, the reference signature is not replaced at the time of updating, but is saved multiple times. Different short-term conditions, such as a dry state and a wet state of the ground surface, can thus be taken into account. However, in order to clear reference signatures that are in fact outdated due to the aforementioned effects and thus empty the memory, additional information may be set to each reference signature, said additional information being suitable for detecting outdated reference signatures. For example, a counter may be provided as additional information for each position stored in the correspondence table, the counter being incremented when the working signature coincides with the reference signature at the determined position and the determined position coincides sufficiently precisely with the stored position. Finally, during the cleaning process, such positions assigned to the reference signature can be cleaned: the counter for the position is below a threshold. In the remaining positions assigned, the counter can be reset. In this way or in a similar way, a plurality of reference signatures pointing to the same location can also be reduced or merged. The counter may be stored in a correspondence table. In this case, a counter can be associated with each position stored in the correspondence table.

According to one aspect, it is provided that control signals for the vehicle are provided in dependence on the position of the vehicle on the virtual rail system, with which control signals the movement of the vehicle is controlled. The control signal can be used, in particular, to control a drive and/or steering of the vehicle. The vehicle can thus be moved in such a way that it is guided on the virtual rail system. Preferably, a control signal for the transverse direction relative to the virtual rail system is provided, which control signal controls the steering mechanism of the vehicle, and a control signal for the longitudinal direction relative to the virtual rail system is provided, which control signal controls the drive mechanism of the vehicle.

According to one aspect, a control signal for the vehicle is provided as a function of the position of the vehicle on the virtual rail system, with which control signal at least one work implement of the vehicle is controlled. In this way, the work apparatus of the vehicle can be operated in a desired manner at a predetermined position and the provided processing can be carried out or interrupted. A mowing tool of a mowing robot should be used as an example for such a working implement. Areas that should not be mowed, such as flower beds or channels, may be provided. If the robot lawnmower is located outside of these zones, then such control signals are provided: the mowing tool operates based on the control signal. If the mowing robot is traveling in such a position: the position is preferably assigned to such a region via a correspondence table, and a control signal for switching off the mowing tool is provided. Furthermore, the rotational speed or the power of the mowing tool, for example, can be controlled by means of the control signal.

Alternatively, the additional features may be stored in a correspondence table. Among these additional features are:

-information about a curve radius by means of which on the one hand the speed in the curve is adapted and on the other hand the steering mechanism is controlled beforehand;

-a suggested speed and/or speed limit;

-information for switching between correspondence tables or parts of correspondence tables;

-direction selection at an intersection;

-control of a function;

-control of the work implement;

information about the location to which the reference signature is saved or to which it is assigned, such as date/time, dry or moist state, (day) light or darkness, further environmental conditions.

According to a further aspect, the lighting device of the vehicle can be activated by means of the direction sensor in such a way that it illuminates the ground surface in the same way regardless of the surroundings (time of day, weather) so that the illumination light or illumination color illuminates the illuminated surface from a predetermined direction, to be precise regardless of which orientation the vehicle has in relation to its surroundings at the time. With this orientation, the earth's surface is always photographed in the same way. Due to the defined illumination direction, the image content of the image captured by the illumination device is also largely independent of the orientation of the vehicle, since the image can be oriented by means of the direction sensor.

The computer program is provided for carrying out each step of the method, in particular when the computer program is executed on a computer or a controller. It is possible to implement the method in a conventional electronic controller without having to make structural changes thereon. To this end, the computer program is stored on a machine-readable storage medium.

By installing the computer program on a conventional electronic controller, an electronic controller is obtained that is arranged for automatically guiding the vehicle along a virtual rail system. Alternatively, a Field Programmable Gate Array (FPGA) or an Application Specific Integrated Circuit (ASIC) may be provided in order to implement the method for automatically guiding the vehicle along the virtual rail system.

Furthermore, a vehicle is proposed, which has a sensing device for detecting a characteristic of the ground surface and which is provided for being guided automatically along a virtual rail system in the manner set forth above. For this purpose, the vehicle may have an electronic control as described above. In the following, three examples are listed for the vehicle and its field of application. The vehicle may be an industrial robot which moves primarily within an industrial facility and autonomously performs actions therein. Furthermore, the vehicle can be a transport robot which transports the goods autonomously on a predefinable route corresponding to a virtual rail system. Further, the vehicle may be a mowing robot that autonomously mows grass, lawn, or green. However, the invention is not limited to the examples mentioned.

According to one aspect, the sensing device has an optical image sensing device, in particular a camera or a camera system, which captures images of the ground surface. The feature can be sensed directly from the captured image. Thus, the optical image sensing device can be used for all types of earth surfaces where features can be optically distinguished. Examples for this are given in the preceding examples. Advantageously, the sensing region is aimed vertically downwards at the surface.

Preferably, the vehicle has in this case a lighting device which is assigned to the image sensing device and illuminates the area of the ground surface which is sensed by the image sensing device. The lighting device can have a plurality of light sources with different colors. Preferably, the ground is illuminated in different colors from different directions so that the features can be better identified. Preferably, the illumination device can be operated in pulsed excitation to avoid motion blur in the image. Here, the pulse duration of the illumination device and the photographing duration of the image sensing device may be synchronized.

Furthermore, the optical image sensing device offers the advantage that: the layer of the earth's surface determined by the focal point may be selected for observation. Advantageously, for optical image sensing devices, cameras or camera systems with a small depth resolution are used, wherein the layer to be imaged in resolution is located approximately at the height of the feature to be sensed. Interfering objects outside the depth-clearing range are imaged blurred and are therefore not considered as features. Such interfering objects are, for example, plants.

According to one aspect, the sensing device has at least one touch sensor which measures the height and properties, in particular the deflection, of deposits on the earth's surface and the like at each location of a solid earth surface by pressing the sensor in with a force which is dependent on these parameters. The touch sensor can be designed, for example, in the form of a measuring finger. The pin moves up and down several times per second and in this case strikes the solid ground with little force. The stroke and, if necessary, other variables such as damping, delay or generated harmonics vary as a function of height and properties. Alternatively, the touch sensor can be designed, for example, in the form of a measuring wheel with a spring. The measuring wheel rolls on a solid ground surface with little force. The measuring wheel is pressed against the spring in relation to the height and properties of the ground surface. Alternatively, a plurality of touch sensors may be arranged in a line across the width of the lane perpendicular to the direction of motion of the vehicle. Touch sensors are suitable for uneven terrain, in which the height and/or properties vary in relation to the position by orders of magnitude in the region (model or submodel) that is important for forming the signature. As examples of this, grasslands, lawns and green belts are mainly cited but cultivated lands and farmlands are also cited. Because the touch sensor senses a firm ground surface, the height of the turf is measured and plants growing thereon are not perceived.

According to one aspect, the sensing device has at least one air pulse sensor or air block sensor, also referred to as ground effect sensor. The air (or other gas) is discharged from the opening in a direction toward the surface, preferably vertically downward. This can be done both in pulses and continuously. The air escapes in relation to the height and nature of the earth's surface, the height and nature of the sediments on the earth's surface and the like. In connection with this, a reaction force is generated, which is measured by an air pulse sensor or an air blocking sensor by means of a reaction force sensor and from which the height and properties of the ground surface, of the sediments on the ground surface and the like are determined. Alternatively, a plurality of air pulse sensors or air block sensors may be arranged in a line across the width of the roadway perpendicular to the direction of movement. Air pulse sensors or air block sensors are suitable for uneven terrain, in which the height and/or properties vary in relation to the location in the order of magnitude of the region (model or submodel) that is important for forming the signature. As examples of this, grasslands, lawns and green belts are mainly cited but cultivated lands and farmlands are also cited. In the case of grass, lawns and green belts, the openings through which the air is discharged in the direction of the ground surface are preferably arranged at a height at which the air directly hits the turf and does not hit the plants.

Alternatively or additionally, the sensing device may have a further sensor. The sensors described below also measure the height of the earth's surface and generate depth images of the earth's surface, respectively. For example, an acoustic wave sensor, in particular an ultrasonic sensor, can be provided. As a further example, an electromagnetic sensor, such as an ultra-wide band sensor or a radar sensor, may be provided. The latter may have an entry depth into the earth surface in the range of centimeters. In particular, methods of image generation, for example based on sensor arrays or sensor rows, can be used. The sensors mentioned are particularly suitable for uneven ground surfaces, primarily grass, lawn and green belts. The sound or electromagnetic waves penetrate the plants and can take a depth image of the turf.

According to one aspect, the vehicle may have a direction sensor, by means of which the reference signal and the operating signal can be directed. This results in an advantage in terms of consistency of the working signature and the reference signature at self-localization. Furthermore, a rail that is suitable for further travel in the desired direction can be selected in association with the intersecting virtual rail.

According to a further aspect, the direction sensor can be used to activate the lighting device in such a way that the lighting light or the lighting color illuminates the illuminated area from a predetermined direction, to be precise, independently of which orientation the vehicle has in relation to its surroundings at the time.

In addition to the at least one virtual rail, the virtual rail system can have different virtual components, such as switches, points, intersections, parking positions, avoidance positions, etc. The vehicle is thus guided on the virtual rail system in a similar way to the actual rail system, for example for a train.

The virtual rail system itself is at least invisible to humans or indistinguishable from the surrounding environment. It can be provided that visible markings are installed on the virtual rail system in its actual location. The visible marking indicates to the person that a virtual rail system is present in this region and that the vehicle driving on its own is to be considered accordingly. In the case of a self-propelled motor vehicle, this can signal: the driver can switch to automated driving. There are many possibilities to implement such a visible marking. By way of example, among a plurality of further possibilities, especially in the case of carpeted floors, the placement of color dots, the introduction of color patches, bands with other patterns or other colors are also to be mentioned.

At the same time, the formation of a suitable signature can also be assisted by the installation of such visible markings, for example when a color band with a color patch sprinkled in is applied to the surface of the ground which is not textured or is textured in itself at the location where the virtual rail system is to be inserted.

Drawings

Embodiments of the invention are illustrated in the drawings and will be explained in detail in the following description.

FIG. 1 illustrates a cross-sectional view of a vehicle according to one embodiment of the present invention.

Fig. 2 shows an oblique view of a vehicle according to another embodiment of the invention.

Fig. 3 shows a view from below of the vehicle according to fig. 1 on a virtual rail.

Fig. 4 shows a view from below of a vehicle configured as a mowing robot according to another embodiment of the invention.

Fig. 5 shows a schematic view of a touch sensor in the form of a measuring finger.

Fig. 6 shows a schematic view of a touch sensor in the form of a measuring wheel.

Fig. 7 shows a schematic view of an arrangement of a plurality of touch sensors of fig. 5.

Fig. 8 shows a schematic view of an air block sensor.

Fig. 9 shows a schematic view of an arrangement of a plurality of the air blocking sensors of fig. 8.

Fig. 10 shows a schematic cross-sectional view of an optical image sensing device.

Fig. 11 shows a schematic illustration of the captured sensor signal measurement points, the feature formation and the operating signature according to an embodiment of the invention.

FIG. 12 shows a schematic diagram of a virtual rail system, multiple sensing zones, and common feature formation according to one embodiment of the present invention.

Fig. 13 shows a correspondence table according to an embodiment of the present invention.

FIG. 14 shows a schematic diagram of a working signature and a reference signature and their correspondence according to one embodiment of the invention.

Detailed Description

Fig. 1 to 3 show different views of a vehicle 1 according to two exemplary embodiments of the present invention. The vehicle 1 may be used as an industrial robot or as a transport robot, for example. The vehicle 1 moves on the ground 2 along a virtual rail system 3 (not shown in fig. 1). The virtual rail system 3 is, for example, a travel route or a transport route along which the vehicle 1 moves. The vehicles 1 each have an image sensing device 4, for example a camera, and a lighting device 5, which are connected to an electronic control unit 20. The image sensing device 4 may include one or more of the following sensors:

-a monocular image sensor;

a one-dimensional single-row sensor arranged or sensing transversely to the direction of motion of the vehicle 1; deriving a second dimension from the motion of the vehicle 1;

-a conventional image sensor providing a two-dimensional sensor signal, either in grayscale or in color images; ideally, the conventional image sensor operates with a short exposure time and a small f-number (i.e., a large f-number aperture), thereby simultaneously achieving a small motion blur and a sufficient light output during driving;

distance measuring sensors, for example based on ultrasound, radar or light propagation time measurements, or stereo cameras or structured lighting devices with cameras;

sensors for measuring the orientation, for example based on a monocular camera and a polychromatic illumination device, by means of which the surface orientation can be determined and can be provided, for example, as an image of a normal vector.

The image sensing device 4 has a sensing area 6 for features of the earth's surface 2 (not shown in detail). There are many possibilities for the arrangement of the image sensing device 4 or the sensor. In the three-wheeled or four-wheeled vehicle 1 having the knuckle steering mechanism, the image sensing device 4 is arranged centrally on the non-steered shaft. In the case of all-wheel steering, skid steering, or articulated steering, the image sensing device 4 is disposed near the center of the vehicle 1. As a result, the smallest possible misalignment of the sensing region 6 relative to the virtual rail during a steering maneuver is achieved.

Fig. 1 shows an embodiment of the invention, in which the image sensing device 4 is arranged below the vehicle 1. The image sensing device 4 is here offset in the direction of the interior of the vehicle 1, in order to achieve a larger sensing area 6 on the one hand and to protect the image sensing device 4 against dirt, wear and the like on the other hand. Furthermore, different sun positions and/or rain have no direct influence on the image sensing or image sensing device 4. The illumination device 5 is arranged annularly around the image sensing device 4 and illuminates at least the sensing region 6. Furthermore, the vehicle 1 comprises a direction sensor 9, which is also connected to the electronic control unit 20 and with which the orientation of the vehicle 1 relative to the ground surface 2 can be determined. The illumination direction of the illumination device 5 is controlled in relation to the orientation of the vehicle 1 determined by the direction sensor 9.

Fig. 2 shows a further embodiment of the invention, in which the image sensing device 4 is arranged on the front side of the vehicle 1 in the direction of travel. The image sensing device 4 can additionally be used to avoid collisions, wherein the sensing region 7 for avoiding collisions is designed to be larger than the sensing region 6 for the features of the ground surface 2. The illumination device 5 is also arranged on the front side of the vehicle 1 and illuminates at least the sensing area 6 for the feature.

Fig. 3 shows a view of the vehicle of fig. 1 from below, in which the image sensing device 4 is arranged centrally below the vehicle 1, wherein the vehicle 1 moves on a virtual rail system 3. The lighting device 5 comprises a plurality of light emitting modules 8 which emit light with different colors (here four different colors) from different directions. Thus, features can also be distinguished in a weakly textured, but structured surface. The lighting device 5 is rotatable, either mechanically or electronically by using a multicolored light module 8. By using the direction sensor 9 to control the rotation of the lighting device 5, it is ensured that the lighting direction for each color is independent of the current orientation of the vehicle relative to its surroundings or relative to the orientation of the virtual rail system 3 at the location of the vehicle 1.

Furthermore, the illumination device 5 can be operated in pulsed excitation to avoid motion blur in the image. Here, the pulse duration of the illumination device 5 and the photographing duration of the image sensing device 4 are synchronized. Fig. 3 also shows that the virtual rail system 3 has a switch 11 next to at least one virtual rail 10. Also possible are diversions, intersections, T-junctions, parking positions, avoidance positions, etc., which are not shown in detail here. The virtual rail system 3 can be made visible to humans by marking the ground surface 2 with different colors and/or shapes.

Fig. 4 shows a view from below of a vehicle 1 according to another embodiment of the invention, in which the vehicle 1 is configured as a mowing robot. The vehicle 1 has a working implement in the form of a mowing tool 21, which is arranged on the underside of the vehicle 1. The vehicle 1 moves on a virtual rail system 3, wherein the virtual rail system 3 is a track along which the mowing robot mows grass. Furthermore, a sensor device 22 is provided in fig. 4, which is different from the image sensing device 4 illustrated in fig. 1 to 3. As shown in fig. 10 and explained in connection therewith, the optical image sensing device 4 as shown in fig. 1 to 3 can also be used for a lawn mowing robot. The following describes a sensor used in the sensor device 22.

Fig. 5 shows a first embodiment of a touch sensor 23, which is designed in the form of a measuring finger. The pin 24 is arranged in a housing 25 and protrudes from the housing. A drive device, not shown, which moves the pin 24 up and down a plurality of times per second is provided in the housing 25. The end of the pin 24 strikes the solid ground 2 with little force. The course and, if necessary, other variables such as damping, delay or generated harmonics vary in relation to the height and deflection of the earth's surface 2.

Fig. 6 shows a second embodiment of the touch sensor 26 in the form of a measuring wheel. The measuring wheel 27 rolls with little force on the solid ground surface 2. The position of the measuring wheel 27 varies in relation to the height and deflection of the ground surface and the spring 29 is pressed by the suspension rod 28. The height of the ground surface 2 is inferred by the amplitude of the suspension rod 28 at the spring 29, for example by means of a spring travel sensor.

Fig. 7 shows the arrangement of a plurality of touch sensors 23 according to the first embodiment of fig. 5 for use in the case of grass as a ground surface 2. The grass plants 30 grow on the turf 31, which represents the transition to solid ground, wherein the roots of the grass plants 30 are located in the turf 31. The growth of these grass plants 30 causes a change in characteristics within a short time, in particular between two mowing cycles. Thus, these grass plants 30 (typically all plants) should not be perceived when sensing the characteristic. These touch sensors 23 are arranged in a row perpendicular to the direction of movement of the vehicle 1, to be precise in such a way that the pins 24 reach a solid ground or turf 31. Thus, only features of the ground, sediments, quasi-permanent parts such as roots, e.g. stones and grasses (usually all plants) are sensed. The one-dimensional measurement of the sensor device 22 is expanded to a two-dimensional measurement by the movement of the vehicle 1. This arrangement can also be transferred to the touch sensor 26 according to the second embodiment. A plurality of measuring wheels 27 are arranged in a line perpendicular to the direction of movement of the vehicle 1 and roll over the turf 31 in the direction of movement.

Fig. 8 shows one embodiment of an air blocking sensor 32, which may also be used in the sensor device 22. Air is discharged from an opening 33 on the underside of the air blocking sensor 32 vertically downwards in the direction of the ground surface 2 (see fig. 9). This can be done both in pulses and continuously. The air escapes in connection with the height and nature of the ground surface 2, with deposits, such as stones, on the ground surface 2 and the like. In connection with this, a reaction force is generated, which is measured by means of the reaction force sensor 34 and from which the characteristics of the ground surface 2 are determined.

Fig. 9 shows the arrangement of a plurality of air retardation sensors 32 according to the embodiment of fig. 8 for use in the case of grass as a ground surface 2. The air blocking sensors 32 are arranged in a line perpendicular to the direction of movement of the vehicle 1. The openings 33, which thus discharge the air in the direction of the ground surface 2, are arranged at a height lying in the plane of the plants 30, so that the air directly hits the turf 31.

As already explained, in a further embodiment, the sensor device 22 can have a further sensor. For example, an acoustic wave sensor, in particular an ultrasonic sensor, can be provided. As a further example, an electromagnetic sensor, such as an ultra-wideband sensor or a radar sensor, may be provided. The sensors may also be arranged in a line perpendicular to the direction of movement of the vehicle 1 or in an array. The acoustic or electromagnetic waves penetrate the grass plants 30 and a depth image of the turf 31 can be photographed.

Fig. 10 shows a particular development of the optical image sensing device 4 of the first embodiment of the vehicle 1 according to fig. 1 and 3, in the case of use as a mowing robot for grass, lawn or green belts. Because the growth of these grass plants 30 changes characteristics in a short time, particularly between two mowing cycles, these grass plants 30 (typically all plants) should not be perceived when sensing the characteristics. In the case where the optical image sensing device 4 is arranged on the underside of the vehicle 1 and the sensing area 6 of the image sensing device is located vertically downward on the ground surface 2 (see fig. 1), and in the case where the optical image sensing device 4 is arranged on the front side of the vehicle 1 depending on the conditions (see fig. 2), a camera having a small depth clear range may be used. The optical image sensing device 4 is focused at a narrow area 33 around the turf 31. Thus, the grass plants 30 are partially not perceived. The already described lighting device 5 can also be used here.

In fig. 11, it is shown how features are obtained from the two-dimensional sensor signals of the image sensing device 4 and a signature is constructed. The signal measurement points 12 represent the measured values sensed by the image sensing device 4, i.e. for example grey values, distance values or height values, or represent measurement vectors, i.e. for example colour vectors or normal vectors, or represent combinations between measured values and/or measurement vectors. The signal measurement points 12 form the bands shown in fig. 11, which are segments of the virtual rail system 3. The vehicle 1, not shown here, is intended to be moved from the bottom to the top in the direction of the arrow 13. Even if the steps explained next are carried out at each position of the vehicle 1, these steps are described separately from each other. To be able to follow the order of the steps, fig. 11 should be read from bottom to top. The signal measurement points 12 are not usually captured simultaneously, but rather continuously in time with the movement of the vehicle 1. The horizontal and vertical distances of the signal measurement points 12 from each other are depicted as being approximately equal in this embodiment, but may be different in other embodiments. The width of the lane 14 sensed by the image sensing device 4 may be limited by the width of the vehicle 1, or by the width of the image sensing device 4 or the extension of the sensing area 6 for the feature. In fig. 11, the lane 14 comprises 25 signal measurement points 12 in this embodiment. In practice, the number of signal measurement points 12 is significantly higher. Preferably, one signal measurement point 12 corresponds to one image point (pixel).

In a further exemplary embodiment, the features can be obtained in a similar manner from the two-dimensional sensor signal of the sensor device 22 and form a signature. Preferably, one signal measurement point 12 corresponds to one sensor measurement.

In this embodiment, a model 15 is shown, within which features are obtained from the signal measurement points 12. The model 15 extends over a two-dimensional surface which comprises a plurality of signal measurement points 12. The model 15 does not have to cover the entire width of the lane 14. In this embodiment, the model covers just half of the lane 14, but may cover more or less in other embodiments. The mould 15 has here the shape of an octagon, but other shapes are also feasible, such as a circle, an ellipse, an egg shape, a square, a rectangle, a polygon or a line. Within the model 15a plurality of submodels 16 is provided, which are here embodied as 37 circles. Other shapes and/or other numbers are also possible here. These submodels 16 are arranged here without overlapping in the model 15 and fill it to a large extent. In other embodiments, these submodels 16 are arranged overlapping. For each submodel 16, a submodel measurement value or a submodel vector is composed of values and/or vectors of signal measurement points 12 that are at least partially covered by the corresponding submodel 16 or located in the environment of the submodel. For example, a weighted average is determined from the values and/or vectors of the four nearest signal measurement points 12, wherein the weights are selected, for example, in relation to the distance between the signal measurement points 12 and the center of the submodel 16 and the sum of the weights is 1. This step can also be understood as an interpolation step.

Values/vectors which are invalid for the submodels 16 are also provided, which are used in the event of defects in the image sensing unit 4, in particular in the event of defective pixels or unreliable sensor signals, or in the event of a lack of measured values/measured vectors because the model 15 is out of the lane 14. For the invalid value/vector, the weight is chosen to be zero, whereby the invalid value is not considered.

If sub-model measurements or sub-model measurement vectors are available for the model 15, a pre-processing of the sub-model measurements/sub-model measurement vectors, such as a normalization or "equalization" of the data, is implemented in this embodiment. This preprocessing serves to geometrically correct the image (for example with homographic imaging) so that it then corresponds to the imaging of a forward parallel projection of the ground surface 2. This compensates for the optionally present inclination of the vehicle relative to the ground surface 2. If the distance of the image sensing unit 4 relative to the earth's surface 2 is changeable and a non-telecentric light fixture is used, it may be advantageous to scale the image or model 15 of the signal measurement points 12 accordingly in order to compensate for the distance difference and thus ensure that the signature formed is not related to the distance. Various types of height sensing means can be used for determining the distance.

Next, a feature is determined from the preprocessed submodel measurement/submodel measurement vector, which feature is to characterize the position. The features may be, for example, vectors of numbers, wherein each number embodies a sub-feature, for example, the result of convolving or filtering a signal segment with a wavelet. In particular, different wavelets are used here and the segments vary, for example, in their size. Two examples of the many possibilities for the way in which such sub-features can be determined are shown below, which are known in principle to the person skilled in the art:

convolving the image segment of gray values with a wavelet that implements a smoothed second derivative in a predetermined direction (fixed relation to the vehicle coordinate system, either in relation to the direction sensor 9 or not).

-convolving the first color channel with a first wavelet that implements the smoothed first derivative in a first predetermined direction, and convolving the second color channel with a second wavelet that implements the smoothed first derivative in a second predetermined direction, and finally differencing the results of the two convolutions.

The determination of the features (apart from the preprocessing) is not location-dependent, i.e. it can be carried out in the same manner at each position.

5 overlapping models 15a-E are shown, and finally a signature E, L, J, G or D is formed from the belonging features of the models. This may be, for example, a 16-bit wide signature E, L, J, G or D, where 16 binary values may be computed individually, e.g. by 16 different weighted associations of 37 sub-model values/sub-model vectors, followed by threshold decisions separately. From this, 2 expressing different positions by signature is derived¹⁶And (4) possibility. The determination of the weighted correlation and the determination of the associated threshold values are carried out, for example, by the user or by the neural network with the aid of training data and can also be adapted automatically during operation in order to adapt to different earth surfaces 2. It is also possible to determine a plurality of signatures E, L, J, G or D from one feature, or a plurality of signatures E, L, J, G or D from a plurality of features.

Each signature E, L, J, G or D is associated with a location, such as the center of the model 15a-E from which the signature was derived. Since here 5 models 15a-E in parallel are evaluated analytically, as a result 5 signatures E, L, J, G and D are generated, which 5 signatures belong to 5 adjacent positions. The position may be described in coordinates: the coordinate x is transverse to the virtual rail system 3, has a positive and a negative sign and in said coordinate the value zero corresponds to the center of the virtual rail system 3, and the coordinate s extends along the virtual rail system 3 and starts at zero at the start. It should be noted here that the virtual rail system 3 may be curved. The scaling of the two coordinates x and s can be different and be configured in meters, or in relation to the image sensing device 4 (e.g. pixels) or in another embodiment in relation to the sensor device 22. After the vehicle 1 has continued to move for a while, this configuration of the signatures E, L, J, G and D is always implemented again. Thus, a band 17 is generated which consists of a random, usually different, signature, here designated only by S. The area 18 includes signatures that are observed together in order to find a position. These different signatures are discussed in detail in fig. 6.

Fig. 12 shows that, when the feature is sensed in successive measurements (which are designated here by two sensing regions 6a and 6b, 6a being assigned to the first measurement and 6b being assigned to the second measurement), the model 15 is located in the overlap region of the two sensing regions 6a and 6 b. Accordingly, the signal measurement points 12 from both measurements are considered for determining the features. The signature thus found has redundancy and may be of interest when the characteristics change depending on the position of the vehicle 1, even if this is not otherwise the case.

A correspondence table is shown in fig. 13. The reference signatures a-O that have been created during the learning travel of the vehicle 1 as explained in connection with fig. 11 are assigned respective positions. This assignment between the reference signatures a-O and the location is entered into the correspondence table. When the coordinate values (x, s) are assigned to the reference signatures a-O, the same positions also result in the same location parameters (x, s) when the signatures a-O are determined. For this purpose, it may be necessary to take into account the coordinates in the 2-dimensional sensor signal, for example the image coordinates of the center point of the model 15. Each of the positions (x, s) thus formed corresponds to a position on the virtual rail system. Here, the signature is used only once to determine the address in the correspondence table. The location is stored at the address. I.e. the signature itself is not stored. The signature may occur multiple times, wherein the probability of the multiple occurrence of the signature increases with the stored route section (in the correspondence table) and with the width of the strip 17. In the correspondence table, each table area holds a plurality of positions. In this case, a certain amount of memory can be fixedly associated with each table area, or in other embodiments the available memory can be flexibly allocated, for example by means of a dynamic list, below the table area.

Additional features not shown here may be stored in the correspondence table. Among these additional features are:

-information about a curve radius by which on the one hand the speed in the curve is adapted and on the other hand steering is controlled beforehand;

-a suggested speed and/or speed limit;

-information for switching between correspondence tables or parts of correspondence tables;

-direction selection at an intersection;

-control of a function;

saved information about the reference signatures a-O, such as date/time, dry or wet status, (day) light or dark, further environmental conditions.

Fig. 14 schematically illustrates the positioning according to an exemplary embodiment of the method according to the present invention. The reference signatures a-O have been extracted by learning the driving and saved in the correspondence table shown in fig. 13 together with the belonging positions. The reference signatures are shown on their respective positions on the right in fig. 14, wherein these reference signatures are referenced to a reference coordinate system (x, s).

Here, the reference numerals a-O show a total of 15 different signatures. In this example, 100 reference signatures are sensed, in practice, significantly more reference signatures are typically sensed. Some specific reference signatures a-O typically occur multiple times because the cardinality of the amount of possible signatures exceeds the number of different reference signatures a-O.

On the left side, a working signature a-O is shown, which has currently been converted from features captured by the image sensing device 4 of the vehicle 1. Now, the 15 latest recorded operating signatures a-O and their relative operating coordinate system x moving with the vehicle 1 are used_A，s_AThe position of the vehicle 1 is found by the correspondence with the reference signatures a-O by means of the correspondence table. If, for example, the working signature E is looked at, it can be determined that the corresponding reference signature E appears 7 times. Accordingly, 7 different positions are stored in the correspondence table of fig. 13. That is, for the working signature E, there are 7 possibilities to assign the working signature to one reference signature E and thus to one location.These 7 possibilities are marked in fig. 14 with 7 connecting lines. Thus, the assignment is ambiguous, since the most of these 7 possibilities can be correct. To resolve ambiguity, look at additional working signatures a x-O, in particular all 15 working signatures a x-O. For the sake of clarity, not all connecting lines are drawn in fig. 14 for these further working signatures a-O, but only the correct connecting lines. This highlighted correspondence is characterized with respect to all other possibilities in that the working signature and the reference signature are mutually validated. If an error is detected in this assignment, it is provided that the reference signature A-O is updated with the aid of the working signature A-O.

If the coordinate difference is determined according to equation 1 for any of these identified correspondences, the position (x) of the vehicle is consistently found here_F，s_F)

Thus, the position of the vehicle 1 on the virtual rail system 3 is obtained. In this exemplary embodiment, the vehicle 1 is therefore currently located at a distance coordinate of 75, offset by two units to the left from the center of the virtual rail system 3.

In the case of a plurality of possible associations, a group 19 of 9 working signatures a-O, indicated by squares in fig. 14, is highlighted. In this group 19, all 9 working signatures a-O point to the group 19 with adjacently located reference signatures a-O. Not much like consistency can be found for any other location. The amount of coherence can be found using a multi-level histogram. In the following, this should be briefly described:

the first histogram is one-dimensional and has a resolution of 1m per histogram bar, i.e. a coarse resolution. Recording all the journey positions s of the vehicle into the histogram_FThe path positions are generated in the formation of possible correspondences. In this case, each correspondence is entered into a respective histogram bar at a respective position

-then, finding the maximum agreementThe histogram bar of nature. Alternatively, the group with the most consistency of the adjacent histogram bins may also be found. Thereby, along the coordinate s_AIs already precisely known to about 1 m.

These two steps can be repeated a number of times, in particular if the virtual rail system 3 is divided into sections and a separate correspondence table is provided for each section. A histogram is created for each correspondence table and finally such histogram bars are found: the histogram bars have the most consistency overall; the associated histogram can infer the associated correspondence table and therefore the associated section of the virtual rail system 3.

In a next step, the position is precisely determined. For this purpose, a two-dimensional second histogram with a resolution of 1cm × 1cm is used. Only such correspondences participate in the second coordination: the correspondence relation has a correspondence with the histogram bar found for the position in the first histogram or is located in the vicinity on the spot.

-re-finding the histogram bin or group of histogram bins with the most agreement in the second histogram. In the case of a group of histogram bars, an averaging or weighting may additionally be provided. The position coinciding with the histogram bar/bars corresponds to the position of the vehicle 1.

If the position of the vehicle 1 is found, the next position searched for in the process of position Tracking (Tracking) is found based on the position that has been found. Accordingly, the reference signatures A-O are restricted to such search areas: the search area is generated from a search area surrounding the determined position or around a search area surrounding the next position.

Furthermore, control of the vehicle 1 is provided. For control in the transverse direction, the lateral position x is used_FIs known. The steering is carried out in such a way that the lateral position x_FAnd becomes smaller. This is a task known to the person skilled in the art from the adjustment technique. If this variable is positive, for example, it means that the vehicle 1 is located to the right of the center of the virtual rail system 3 in the direction of travel. The steering intervention should go to the leftThe steering intervention is at least selected to be so great that the lateral position x is_FBecomes smaller but not too large in order not to generate excessive wobble.

The direction of travel is first of all involved in the control in the longitudinal direction. The direction of travel has been determined implicitly in the correspondence table: in this embodiment, the vehicle 1 is at a path position s_FIs driven in the direction of increasing value of (c). This knowledge about the direction of travel is used in the location tracking in order to predict the next location and to specify a search area. Likewise, the vehicle may also travel in the opposite direction. If the vehicle is turned around for this purpose, this generally means that the sensor has turned 180 °. This is compensated for by a corresponding rotation of the two-dimensional sensor signal or by appropriately taking the rotation into account when forming the operating signature a-O. The additional features already explained and stored in the correspondence table are taken into account when guiding and controlling the vehicle lane.