DE102014204933A1 - Method and device for generating a segmentation layer - Google Patents

Abstract

Device (1) and method for generating a segmentation layer (SEG-L) in a driver assistance system of a vehicle (F) comprising: an evaluation unit (1A), which receives sensor data from at least one vehicle sensor (2) mounted on the vehicle (F) ), for evaluating environmental structure information evaluates, and with a segmentation unit (1B), the free areas of an occupancy grid (BG) of an environment of the vehicle (F) by means of the generated environment structure information segmented into passable segments (SEG), each having a segment ID obtained, which are assigned to the corresponding cells of the free areas of the occupancy grid (BG) for generating the segmentation layer (SEG-L).

Description

The invention relates to a method and a device for generating a segmentation layer in a driver assistance system of a vehicle.

In conventional driver assistance systems, the modeling of a vehicle environment of the vehicle takes place by means of model-based tracking for road users and an occupancy grid or occupancy grid for the static environment of the vehicle. The further data processing of the environment data is necessary to connect this information with each other. Furthermore, it is necessary to map further information or data, in particular rules or traffic rules in the system. These rules or traffic rules include, for example, speed limits, lanes separated by markings, prescribed directions of travel, such as turning lanes, or prohibited driving areas, in particular one-way streets. In conventional systems, some of this information is mapped, such as by globally storing a detected traffic rule. However, many of these rules or traffic rules only apply to individual mobile segments, in particular lanes.

It is therefore an object of the present invention to provide a method and an apparatus for generating a segmentation layer in a driver assistance system of a vehicle, which provides an overall environment model of the vehicle.

This object is achieved according to a first aspect of the invention by a method having the features specified in claim 1.

The invention accordingly provides a method for generating a segmentation layer in a driver assistance system of a vehicle with the steps:
Providing an occupancy grid of an environment of the vehicle having cells each containing information about an occupancy of the respective cell, in particular a probability value indicating the probability that the respective cell is occupied, segmenting free areas of the provided occupancy grid which are not occupied cells comprise, by means of detected environment structure information for the formation of passable segments, each receiving an individual segment ID, and
Assigning the segment IDs of the formed segments to the respective unoccupied cells of the provided allocation lattice for generating the segmentation layer.

In one possible embodiment of the method according to the invention, the surroundings structure information is generated by evaluating sensor data supplied by at least one vehicle sensor mounted on the vehicle.

In a further possible embodiment of the method according to the invention, the vehicle sensors have at least one vehicle camera which detects markings, in particular lanes, which lie on a surface located in front of the vehicle as environment structure information.

In a further possible embodiment of the method according to the invention, further vehicle sensors are provided which detect movements of other vehicles which drive ahead of the vehicle as environment structure information which is stored in a movement profile layer.

In a further possible embodiment of the method according to the invention, the vehicle sensors have at least one vehicle camera which detects a color and / or a texture of a surface located in front of the vehicle as environmental structure information.

In a further possible embodiment of the method according to the invention, the vehicle sensors detect a height profile in the surroundings of the vehicle as environment structure information that is stored in a height profile layer.

In a further possible embodiment of the method according to the invention, the segments formed are assigned an attribute list and / or an object list.

In a further possible embodiment of the method according to the invention, the attribute list for each segment ID has segment attributes for an allowable movement of the vehicle within the respective passable segment.

In a further possible embodiment of the method according to the invention, the object list for each movable object which is located in the vicinity of the vehicle, its relative position and / or its relative speed relative to the vehicle and the segment ID of that segment in which the movable object just stops.

In a further possible embodiment of the method according to the invention, the formed passable segments of the generated Segmentation layers accessible lanes of a roadway.

In a further possible embodiment of the method according to the invention, data of the generated segmentation layer are compared with data from at least two further layers, in particular the provided allocation grid and / or a height profile layer and / or a motion profile layer, for mutual integrity checking.

In a further possible embodiment of the method according to the invention, a layer whose data is in contradiction to the data of the other layers is evaluated by the driver assistance system as presumably defective.

The invention further provides, according to a second aspect, an apparatus for generating a segmentation layer in a driver assistance system of a vehicle having the features specified in claim 13.

The invention accordingly provides a device for generating a segmentation layer in a driver assistance system of a vehicle
an evaluation unit, which evaluates sensor data, which originate from at least one vehicle sensor mounted on the vehicle, for generating surroundings structure information and with
a segmentation unit, the free areas of an occupancy grid of an environment of the vehicle by means of the generated environment structure information segmented into drivable segments, each receiving a segment ID, which are assigned to the corresponding cells of the free areas of the occupancy grid for generating the segmentation layer.

In one possible embodiment of the device according to the invention, the segmentation unit links the segments formed with an attribute list and / or an object list,
wherein the attribute list for each segment ID has segment attributes for allowable movement of the vehicle within the respective drivable segment, and
wherein the object list for each movable object located in the vicinity of the vehicle indicates its relative position and / or its relative speed relative to the vehicle and the segment ID of that segment in which the mobile object is currently located.

In another possible embodiment of the device according to the invention, an integrity checking unit compares the data of the segmentation layer generated by the segmentation unit with at least two further layers, in particular the assignment grid and / or a height profile layer and / or a motion profile layer, to check its integrity ,

In the following, possible embodiments of the method according to the invention and of the device according to the invention for generating a segmentation layer in a driver assistance system of a vehicle will be explained in more detail with reference to the attached figures.

Show it:

1 a block diagram of a possible embodiment of the device according to the invention for generating a segmentation layer of a driver assistance system of a vehicle;

2 a flowchart for illustrating an embodiment of a method according to the invention for generating a segmentation layer in a driver assistance system of a vehicle;

3 a schematic representation of an application example to explain the operation of the method according to the invention;

4 an exemplary data structure for explaining the operation of the method according to the invention;

5 a block diagram illustrating a possible embodiment of the inventive device for generating a segmentation layer in a driver assistance system of a vehicle;

6 another application example to explain the operation of the method according to the invention;

7 a diagram illustrating a further embodiment of the method according to the invention;

8th a diagram illustrating a possible security layer, as used in the inventive method;

9 a diagram illustrating another possible security layer, as it can be used in the inventive method;

10 a diagram illustrating a further security layer, as it can be used in the inventive method;

11 a diagram illustrating the performed in a possible embodiment of the method integrity check according to the invention;

12 a schematic diagram for explaining the operation of a specific embodiment of the method according to the invention;

13 a diagram for explaining the performed in a possible embodiment of the method according to the invention integrity check;

14 a diagram for explaining the operation of performed in a possible embodiment of the method integrity check according to the invention;

15 a further diagram for explaining the operation of the performed in a possible embodiment of the method according to the invention integrity check;

16 a further diagram for explaining the operation of the performed in a possible embodiment of the method according to the invention integrity check;

17 a further diagram illustrating an application example of the inventive method;

18 another diagram for explaining an application example of the inventive method;

19 a diagram for explaining an application example of the inventive method;

20 a diagram for explaining an application example of the inventive method;

21 a further diagram for illustrating an application example of the method according to the invention;

22 a block diagram illustrating a possible implementation of the device according to the invention for generating a segmentation layer in a driver assistance system of a vehicle.

How to get in 1 can recognize, has a device according to the invention 1 for generating a segmentation layer SEG-L in the illustrated embodiment an evaluation unit 1A and a segmentation unit 1B. In the 1 illustrated device 1 is preferably used to generate a segmentation layer SEG-L in a driver assistance system of a vehicle. The evaluation unit 1A evaluates sensor data obtained from at least one vehicle sensor mounted on the vehicle 2 originate to generate environmental structure information. The segmentation unit 1B segments free areas of an occupancy grid BG or occupancy grids of an environment of the vehicle by means of those of the evaluation unit 1A supplied environment structure information in passable segments, each receiving a segment ID, which are assigned to the corresponding cells of the free areas of the occupancy grid BG to generate the segmentation layer SEG-L. This in 1 shown occupancy grid BG can be read in a possible embodiment of a data memory. The occupancy grid BG may comprise a plurality of cells each containing a probability value indicating the probability that the respective cell is occupied. The occupancy grid BG can be generated by a data processing unit, which evaluates sensor data to calculate a probability of occupancy of the cells. The segmentation unit 1B adds a segmentation layer SEG-L to the occupancy grid BG or occupancy grid, wherein in this segmentation layer of each cell a unique ID or identification for each detected or detected passable segment can additionally be stored.

2 shows a simple flowchart for illustrating an embodiment of the method according to the invention. This in 2 The method illustrated has a plurality of method steps that can be executed by a control program on a processor, for example. In a first step S1, an occupancy grid BG of an environment of the vehicle is provided, wherein the occupancy grid BG has cells each containing a probability value which indicates the probability that the respective cell is occupied.

In a further step S2, free regions of the allocated occupancy grid BG, which comprise unoccupied cells, are segmented by means of detected environmental structure information to form passable segments, which each receive an individual segment ID.

In a further step S3, the segment IDs of the segments formed are assigned to the respective unoccupied cells of the allocated occupancy grid BG for generating the segmentation layer SEG-L.

The environment structure information used in step S2 is generated in one possible embodiment by evaluating sensor data provided by at least one vehicle sensor mounted on the vehicle. In a possible embodiment, these vehicle sensors have at least one vehicle camera which detects markings, in particular lanes, which lie on a surface located in front of the vehicle as environmental structure information. Furthermore, vehicle sensors may be provided which detect movements of other vehicles that drive ahead of the vehicle as environment structure information that is stored in a movement profile layer. Furthermore, the vehicle sensors of the vehicle may have at least one further vehicle camera, which detects a color and / or a texture of a surface located in front of the vehicle as surroundings structure information, which are evaluated in step S2 for segmenting free areas of the provided occupancy grid BG to form passable segments , Furthermore, in one possible embodiment, vehicle sensors may provide a height profile in the environment of the vehicle as environmental structure information stored in a height profile layer. In one possible embodiment, this environment structure information can also be used to segment the free areas of the allocated occupancy grid BG.

In one possible embodiment of the method according to the invention, the segments formed in step S2 are assigned an attribute list AL and / or an object list OL. In this case, the attribute list AL for each segment ID has segment attributes for a permissible movement of the vehicle within the respective drivable segment. The object list OL indicates, for each moving object located in the vicinity of the vehicle, its relative position and / or its relative speed relative to the vehicle as well as the segment ID of that segment in which the mobile object is currently located.

In one possible embodiment of the method according to the invention, the drivable segments formed in step S2 of the generated segmentation layer SEG-L have traffic lanes on a roadway. In one possible embodiment, the data of the generated segmentation layer SEG-L can be compared with further data from at least two layers L for mutual integrity checking. These further layers L include, for example, the assigned coverage grid BG, an elevation profile layer HP-L or a motion profile layer BP-L. A layer L whose data is in contradiction to the data of the other layers after the integrity check can in one possible embodiment be assessed or detected as suspect by the driver assistance system.

3 shows a diagram for explaining the operation of the method according to the invention, as for example in 2 is shown. One recognizes in 3 a vehicle F, which moves on a two-lane roadway, such as a highway. This in 3 The vehicle shown approaches a construction site in which the vehicles are diverted, for example by means of corresponding road signs or markings, in particular of so-called beacons located at the edge of the roadway. To assist the driver, there are markings or lanes M at the site redirection. The sensors provide the driver assistance system of the vehicle F with sensor data for generating an occupancy grid BG within the range of the sensors, which comprises a multiplicity of cells or data cells, for example some Thousand cells. Each cell represents a surface area with a certain cell size, for example 8 × 8 cm. A cell may contain a probability value P indicating the probability that the area of the respective cell is occupied or not passable. In one possible embodiment, the allocation grid BG is a binary allocation grid with only two logical values, which indicate whether the area associated with the respective cell is free or passable or not free or not passable. For example, receive all free cells whose surfaces are in a passable area, the logical value 0 and all cells whose surfaces are in a non-passable area, the logic value of 1. The occupancy grid BG can on the basis of sensor data, such as radar sensor data or image data, by a data processing unit are generated and temporarily stored in a data memory. The free areas of the provided occupancy grid BG, ie comprising unoccupied or passable cells, are segmented by means of detected environment structure information for the formation of passable segments. The vehicle F has at the in 3 illustrated example of at least one vehicle camera, which detects the mark M, ie the lanes on the front of the vehicle F surface as environment structure information. The sensor data originating from the camera are evaluated to generate the environmental structure information. With the aid of the determined environment structure information, the free areas of the allocated occupancy grid BG, which comprise unoccupied or accessible cells, are then segmented to form passable segments. At the in 3 In the simple example shown, the drivable free area between the boundaries or beacons B is segmented into two segments SEG-A and SEG-B. These two Segments correspond in the illustrated simple example two lanes that are passable by the vehicle F at the site redirection.

4 schematically shows an associated simple data structure with two layers L, namely the provided allocation grid BG and the segmentation layer SEG-L generated therefrom by the method according to the invention. In the schematic representation according to 4 contains the occupancy grid BG 4 × 6 = 24 cells. In a real implementation, the occupancy grid BG has a much higher number of cells, for example a few hundred thousand cells. How to get in 4 can recognize, in the illustrated example, the cells located on the edge are occupied in the direction of travel of the vehicle F or not passable and accordingly have a logic value of 1. The cells located on the road are passable and have a logic value of 0.

In another layer, namely the segmentation layer, the corresponding cells are assigned the segmentation ID of the corresponding segments SEG-A, SEG-B. For this purpose, for the respective cell Z, which is stored for example in a memory with a memory address, in addition to the occupancy value 0 or 1, the assigned segmentation ID A, B of the two determined passable segments are additionally stored. Thus, with the method according to the invention, at least one further level or a further layer, namely the segmentation layer SEG-L, is added to the occupancy grid or occupancy grid BG which stores an occupancy probability or an occupancy value in each cell in an efficient manner ,

5 shows a further embodiment of the device according to the invention 1 for generating a segmentation layer SEG-L. In the illustrated embodiment, the generated segmentation layer SEG-L of an integrity checking unit 1C fed. The integrity check unit 1C compares the data of the segmentation unit 1B generated segmentation layer SEG-L with at least two other layers L to verify its integrity. These further layers L may, for example, be the allocation grid BG itself and / or a height profile layer HP-L and / or a movement profile layer BP-L.

At the in 5 The embodiment illustrated links the segmentation unit 1B the formed segments also with an attribute list AL and / or an object list OL. For each segment ID, the attribute list AL has segment attributes for an allowable movement of the vehicle F within the respective passable segment. The object list OL indicates, in particular, another vehicle, its relative position and / or its relative speed relative to the vehicle F as well as the segment ID of that segment in which the other movable one is located for each movable object which is located in the vicinity of the vehicle F Object just stops.

6 shows an application example for explaining the operation of the method according to the invention and the device according to the invention for generating a segmentation layer.

6 for example, a traffic situation on a road, which has two lanes in both traffic directions. The two directions of travel are separated by a roadway separation T. The roadway separation T forms a special type of roadway marking. The two traffic lanes in both traffic directions are separated from each other by lanes or lane markings M. Four lanes form four mobile segments each having an individual segment ID ID1, ID2, ID3, ID4. In the example shown the 6 is located in the first segment or the first lane, a first object, namely a vehicle F1. This vehicle F1 moves in the illustrated example to a first traffic signal A1, which is switched to green. The lane with the segment ID ID1 is a right turn lane so that the vehicle F1 turns right when the first traffic light A1 is green. The second vehicle F2 located on the second segment moves in the illustrated traffic situation to a second red traffic light A2 and stops in front of it.

6 shows by way of example the attribute list AL, which has for each segment ID ID1, ID2, ID3, ID4 segment attributes for an allowable movement of the vehicle or object within the respective passable segment. For example, the segment with the segment ID ID1 has the segment attributes: right turn lane, traffic light green, maximum allowed travel speed 30 km / h. The second segment with the segment ID ID2 has as segment attributes: straight-line track, traffic light red and maximum permitted speed 50 km / h.

In the 6 Object List OL shown for each moving object that is in the vicinity of a vehicle, its relative position and / or its relative speed relative to the vehicle F and the segment ID of that segment in which the mobile object currently resides. The object list OL indicates whether the vehicle F1 or object 1 is located in the segment with the segment ID ID 1. Furthermore, the object list OL indicates that the object 2 or the vehicle F2 in the segment with the segment ID ID 2 is located. The vehicle F1 moves in the example shown with the relative speed 28 km / h and is located at the relative position x = 30 m, y = -3.5 m. Furthermore, the object list OL indicates that the vehicle F2 is moving at a relative speed of 55 km / h and is at a relative position of x = 30 m and y = +0.24 m.

In one possible embodiment, the segment IDs of the segments can also be used to assign elements from the local grid-based environment model to abstract road profiles in digital maps or navigation maps, as shown schematically in FIG 7 shown. In this embodiment, the grid based environment model represents a highly accurate and up to date section around the vehicle F that fits seamlessly into the more abstract information of the digital navigation map, as in FIG 7 shown.

In particular, the back end can also act as an input sensor. It can for example provide a layer in which the right of way rules are stored. Thus, not only vehicle-mounted sensors are involved in the construction of the grid layer but also non-vehicle-mounted sensors such. the back end. Information from the backend may e.g. Traffic rules, such as Right before links or the like, traffic jam, accident or the like.

Since the ID information, in particular the segmentation IDs, does not change over a wide range and only a small number of bits are required for their representation or storage, for example 4 bits at 16 segments, the generated segmentation layer SEG- L compress using compression methods and save data efficiently. Compared with a representation, for example by means of several spline curves, for which a multiplicity of control points with associated coordinates have to be stored, the additional effort for the storage of the segmentation layer SEG-L is minimal. Due to the grid-based use, even segments with a more complex course, for example multi-lane bypass lanes at construction sites, can be displayed without errors. The direct assignment of the cells to the occupancy grid BG or occupancy grid layer allows easy combining of segments, for example lanes, with data regarding possible obstacles for the vehicle F. A further increase in efficiency is achieved by different in one possible embodiment Resolutions for the segmentation layer SEG-L and the other layer L is selected. For example, the occupancy grid BG may contain cells whose associated areas are 8 × 8 cm cell size, and the segmentation layer SEG-L has cell resolution for areas having a width B and height H of 32 cm each , The limit for a control algorithm, such as lateral guidance, then results from the occupancy / structure layer.

A driver assistance function, such as Lane Keeping / Lane Departure Warning, would keep an EGO vehicle F within the limits of the current segment unless the driver of the vehicle F dictates otherwise. Furthermore, for example, the information for a speed limit display is taken from the attribute list AL for the current segment. The data or information required for the segmentation can be obtained from the allocation grid BG. In addition to occupancy probabilities or occupancy values (free / occupied), additional environment structure information can be used. This environment structure information is obtained, for example, from a texture analysis or recognition of lanes.

If, for example, a representation for a sensor data fusion based on the so-called Dempster-Shafer theory is used, then a particularly helpful extension of the data representation can be made. Instead of two classes for free and occupied, an extension of the Dempster-Shafer theory to three classes, namely free (0), occupied (1) and "structure" (logical 2) made, the occupancy value "2" or " Structure "indicates that the vehicle F can drive in principle on the calculated cell, but it should better be avoided for various reasons. This is the case, for example, when it is a curb or a lane marking. The different sensors can provide evidence or probability values for individual classes or combinations of classes. For example, a radar sensor of the vehicle F can not provide evidence for the class "occupied" and "structure" because it can not reliably distinguish between raised and flat objects.

For example, more information about the Structure class can be provided by lane detection or monotexture analysis. In this case, model-based information or data about lane markings M available as the clothoid estimate can be entered, for example. Alternatively, tracking can also be purely grid-based. Another possibility is to evaluate lanes of road users recorded by means of object list OL. For example, in an area where many road users or moving objects are moving, a traffic lane is likely to be expected (ant-street principle).

Alternatively, a representation using multiple layers, such as a Layers for "occupied" and "free", a layer for "structure" and a layer for "lanes". In addition to the layer for "occupied" and "free", ie in the classic occupancy grid or occupancy grid BG, preferably at least one further layer L is used.

Since there is height information by means of, for example, a stereo camera of the vehicle F, it can distinguish between "structure" in the case of low objects, for example a flat curb, or "occupied" in the case of tall objects. As an alternative to discrete classes, a quasicontinuous modeling between "free" and "occupied" can be made. About a height layer classified as "occupied" cells or those whose overrun is classified as critical, also be given in a memory-saving way, a height information.

Based on the structure class or a low criticality, the segmentation is now performed. The occupancy grid BG provides information about static obstacles. For example, several free lanes are recognized side by side as a large free passable area. Different segmentation methods can be used for the segmentation, for example segmentation by means of simple comparison with threshold values. Furthermore, edge-based or region-based segmentation methods can be used, whereby coherent segments are extracted. For example, k-means clustering techniques can be used, where data is divided into k-segments. The number k and the location of the cluster can be specified for example by data from a road map. Furthermore, the number k and the location of the clusters can be estimated, for example, from the width of the free area. Another possible segmentation method is the so-called region-growing method, which determines for given start cells whether their neighboring cells belong to the corresponding segment.

The generated segmentation layer SEG-L and its high-performance transmission make it possible to provide additional security mechanisms in order, for example, to enable highly automated driving of the vehicle F. The system is monitored, whereby small incorrect interventions are noticed early, without the driver being available as a fallback level.

The procedure is carried out via the segmentation of a surface which is located around the vehicle F around, in various safety layers (Safety Layer), which all represent the drive-over area. For example, three security layers SL can be used.

8th . 9 . 10 schematically represent different layer L or security layers that can be used.

8th schematically represents an occupancy grid BG or operation grid as a safety layer. The layer is oriented to lines and can thus span an area in which a possible passable area, namely the lane, is located. This passable area can then be represented by a few parameters and an ID. For example, instead of grid-based lane detection, it is also possible to use classic lane detection using a camera and a clothoid model.

9 shows another security layer, namely a height profile layer HP-L. This second layer is based on the height profile in the vicinity of the vehicle F. This layer HP-L ends in height jumps such as guide posts, crash barriers or grass patches, etc. Here again a plane is spanned, which represents the drivable area in front of the vehicle F. Based on the presented environment model, the occupancy grid and the occupancy grid and the corresponding classes are evaluated. In addition, an evaluation of the altitude layer HP-L can be made, if available.

This in 10 illustrated third security layer is a motion profile layer BP-L. In this case, movements of other objects or vehicles which precede the vehicle F are detected as environment structure information and stored in the movement profile layer BP-L. This in 10 shown motion profile layer BP-L is thus based on the traffic, which detect the sensors of the vehicle. All surfaces that are run over by a vehicle in front, are considered to be in principle by the own vehicle F also passable. This information or data can for example be extracted from a structure layer.

11 schematically shows a made in a possible embodiment of the method integrity check. In the illustrated example, different layers, in particular security layers SL, are compared with each other, so that their integrity can be checked. The data stored in the various layers or security layers SL are compared with each other to see if they conform or contradict each other. This makes it possible to detect individual sensor or data processing errors in a timely manner and to check whether the respective units of the driver assistance system are still working properly and reliably. Since the detection and processing of single layer L based on different sensors and algorithms, thereby systematic errors within the driver assistance system can be quickly detected.

In one possible embodiment, all layers L, in particular security layer SL, are equal to one another. In an alternative embodiment, certain layers or security layers SL are rated as more reliable than other security layers SL. For example, data information from the security layer SL3, namely the movement profile layer BP-L, for example, that a certain area has been run over by another vehicle, be rated higher than information from the first security layer SL1, for example, a lane detection , As a result, in particular, the function of the integrity check can still remain active if e.g. no lane markings are recognizable on the road, but vehicles drive in front of their own vehicle F and furthermore height elements are recognizable, such as crash barriers LP or sods (Safety Layer SL2).

12 schematically shows the use of various safety or safety layers SL, which differ in their complexity and to which a different security relevance is awarded. In the example shown, the security layer SL3 has the least complexity and the highest granted security relevance. This allows, for example, the following procedure. First, with respect to the security layer SL1, the lane data is recognized, and there is no way to know if that lane data is correct or wrong. In a non-highly automated driving engages in this case, the driver when he realizes that the lane is not correct. In contrast, the driver in a highly automated driving of the vehicle F is not necessarily alert and, for example, while reading a book read.

As in 14 is shown by evaluation of the second level of security SL2 that there is a contradiction to the data of the first level of security SL1. For example, a guardrail LP limits the driving range in the security layer SL2. From the identified contradiction of the data between the first security layer SL1 and the second security layer SL2 can not be derived whether the data of the first security layer SL1 or the second security layer SL2 are also applicable.

As in 15 can be determined by taking into account the third safety layer or safety layer SL3 (motion profile layer), which takes into account the movement of the surrounding objects and thus also spans a plane, at which point the deviation or the error occured. In the illustrated example, for example, the third security layer SL3 or the movement profile layer BP-L confirm the data of the second security layer SL2 and contradict the data of the first security layer SL1, so that the data of the first security layer SLl disregarded, as in 16 shown.

The system thus enables highly automated driving of the vehicle F with the aid of the various layers or safety layers SL and their high-performance representation.

The number of security layers SL used may vary in alternative embodiments.

17 shows a further possible application of the method according to the invention. In this case, another vehicle F moves in a left lane at a relatively high driving speed.

18 shows another case where another vehicle F is approaching at a relatively low speed.

19 FIG. 12 shows a case where a vehicle is approaching at high speed on the ego lane of the own vehicle F, so that it must be assumed that it will probably overtake the own vehicle F.

As a result, the area in which the own vehicle F can drive safely, relatively small, as in 19 shown. This area represents the safety area or the safety dimension of the vehicle F. This safety area can be represented by a few parameters.

In a first embodiment variant, the security area is represented by its extent in the xy direction.

Another possibility is represented by the vehicle characteristics of the vehicle F, such as vehicle angle and acceleration a. The limits of the safety range here are the maximum acceleration a max and the maximum steering wheel angle, which adjusts the vehicle F within the safety range.

20 schematically shows the case that the vehicle F remains within the safety area.

21 shows the case in which the vehicle F does not remain in its safety area.

If the situation occurs that causes the vehicle F to move outside the safety dimension, this can be deactivated. In one possible embodiment, the time until the driver regains control of the vehicle F is bridged by the driver assistance system of the vehicle F in an emergency operating mode, wherein in the emergency operating mode the focus is on the vehicle F always being within the safety dimension or the security area remains.

The check as to whether a function or regulation is within the safety dimension can, in one possible embodiment, be carried out by a so-called observer. This observer checks at least one condition and takes over the emergency operation, if an error occurs. Since this function is critical to safety and the interface can be relatively narrow, this function is also outsourced decentralized in a possible embodiment, for example in an ESP control unit of the vehicle, which can be realized with a high ASIL level.

In one possible embodiment, the individual layers L can be generated at different locations in order to make the entire system robust against failures.

The different layers L can be calculated, for example, in the sensors or in a central fusion unit. The Observer can also be accommodated either in a central fusion unit or in a safety controller.

This will allow the return of a HAF function even if the entire controller on which the HAF function is performed crashes. For example, the Safety Layer SL1 can be based on the previous camera lane detection. Furthermore, the second security layer SL2 and the third security layer SL3 may be based on an occupancy grid merger in a central environment model controller. In one possible embodiment, a plausibility check takes place in a steering controller which receives the data of the first security layer SL1 directly from the camera and the data of the second security layer SL2 and the third security layer SL3 from an environment model controller. Even with a malfunction or a failure of the environment model control unit thus more information is available. A further plausibility check can be carried out, for example, based on the objects tracked by a radar sensor.

22 schematically shows several embodiments, namely the variant A, in which the layers L are calculated by a central merger unit (XPU), and the variant B, in which the calculations of the layer L in the sensors 2A . 2 B . 2C he follows.

In the method according to the invention, all environment information or environment data can be merged with the aid of a segmentation layer SEG-L in a uniform environment model. Even a large number of attributes for the different layers L can be efficiently stored in an attribute list AL, the scope of the attributes benefiting from the freedom of the segmentation layer in terms of geometric models. All information and data are also linked by the grid-based model in a location-based manner.

A segmentation layer SEG-L is added to the allocation grid, the segmentation IDs of which can be assigned to an attribute list AL with valid evaluation rules and to an object list OL in an efficient manner. It is also advantageous to supplement the occupancy grid BG to information or data that promote a segmentation in separate traffic areas, such as lanes. The system provides a holistic redundant environment model for providing a segmentation level SEG-L for an occupancy grid BG.

Claims (15)

 Method for generating a segmentation layer (SEG-L) in a driver assistance system of a vehicle (F) with the steps: (A) providing (S1) an occupancy grid (BG) of an environment of the vehicle (F) having cells, each containing information about an occupancy of the respective cell, in particular a probability value indicating the probability that occupies the respective cell is; (b) segmenting (S2) free areas of the provided occupancy grid (BG) comprising unoccupied cells by means of detected environment structure information to form passable segments, each receiving an individual segment ID; and (c) assigning (S3) the segment IDs of the formed segments to the respective unoccupied cells of the provided allocation lattice (BG) for generating the segmentation layer (SEG-L). The method of claim 1, wherein the environment structure information is generated by evaluating sensor data received from at least one vehicle sensor mounted on the vehicle (F). 2 ) to be delivered. Method according to claim 2, wherein the vehicle sensors ( 2 ) at least one vehicle camera ( 2A ), the markings, in particular lanes, which lie on a front of the vehicle (F) surface, detected as environment structure information. Method according to one of the preceding claims 1 to 3, wherein further vehicle sensors ( 2 ), the motions of other vehicles driving ahead of the vehicle (F) are detected as environmental structure information stored in a motion profile layer (BP-L). Method according to one of the preceding claims 2 to 4, wherein the vehicle sensors ( 2 ) comprise at least one vehicle camera which detects a color and / or a texture of a surface located in front of the vehicle (F) as surroundings structure information. Method according to one of the preceding claims 2 to 5, wherein the vehicle sensors ( 2 ) capture an elevation profile in the environment of the vehicle as environmental structure information stored in a height profile layer (HP-L).  Method according to one of the preceding claims 1 to 6, wherein the segments formed an attribute list (AL) and / or an object list (OL) is assigned.  The method of claim 7, wherein the attribute list (AL) for each segment ID has segment attributes for allowable movement of the vehicle (F) within the respective drivable segment.  Method according to claim 7, wherein the object list (OL) for each movable object located in the vicinity of the vehicle (F) indicates its relative position and / or its relative speed relative to the vehicle (F) as well as the segment ID of that segment, in which the moving object is currently located.  Method according to one of the preceding claims 1 to 9, wherein the formed passable segments of the generated segmentation layer (SEG-L) have drivable lanes of a roadway.  Method according to one of the preceding claims 1 to 10, wherein data of the generated segmentation layer (SEG-L) with data at least two further layers (L), in particular the provided allocation grid (BG) and / or a height profile layer (HP-L ) and / or a motion profile layer (BP-L) for mutual integrity checking.  A method according to claim 11, wherein a layer (L) whose data is in contradiction to the data of the remaining layers is recognized by the driver assistance system as presumably defective. Contraption ( 1 ) for generating a segmentation layer (SEG-L) in a driver assistance system of a vehicle (F) comprising: (a) an evaluation unit ( 1A ), the sensor data which is from at least one of the vehicle (F) mounted vehicle sensor ( 2 ), to evaluate environmental structure information, and (b) a segmentation unit ( 1B ), the free areas of an occupancy grid (BG) of an environment of the vehicle (F) by means of the generated environment structure information segmented into passable segments (SEG), each receiving a segment ID, which the corresponding cells of the free areas of the occupancy grid (BG) Generation of the segmentation layer (SEG-L). Apparatus according to claim 13, wherein the segmentation unit ( 1B ) the formed segments (SEG) are linked to an attribute list (AL) and / or an object list (OL), wherein the attribute list (AL) for each segment ID comprises segment attributes for an allowable movement of the vehicle (F) within the respective drivable segment ( SEG), and wherein the object list (OL) for each movable object located in the vicinity of the vehicle (F), its relative position and / or its relative speed relative to the vehicle (F) and the segment ID of that segment (SEG ) in which the movable object is currently located. Apparatus according to claim 13 or 14, wherein an integrity checking unit ( 1C ) the data of the segmentation unit ( 1B ) generated segmentation layer (SEG-L) with at least two other layers (L), in particular the occupancy grid (BG) and / or a height profile layer (HP-L) and / or a BewegungsprofilLayer (BP-L), for examination of its integrity.

Images