DE102013003944A1 - Method and device for providing driver assistance functionality - Google Patents

Abstract

A predefined catalog (300) of road scenarios (200) is loaded and sensor data is obtained. The catalog (300) comprises several road scenarios, each describing a road situation by at least one parameterized environment feature. The sensor data each describe at least one measured environmental feature. From a comparison of the respective at least one measured environmental feature with the at least one parameterized environmental feature, a probability value for the road scenarios is obtained in each case. Based on the likelihood values, a road scenario is selected on the basis of which driver assistance functionality is provided.

Description

Various embodiments relate to a method for providing driver assistance functionality and a driver assistance system. In particular, various embodiments relate to techniques that evaluate road scenarios loaded from a predefined catalog based on sensor data having a probability value and use for driver assistance functionality.

Techniques are known which provide driver assistance functionality to a vehicle in response to sensed sensor data describing environmental features in an environment. For example, autonomous lateral and longitudinal guidance of a vehicle requires a sensor-based perception of the vehicle environment that provides a jerk-free and accurate model of road geometry within a range, typically up to 200 m.

Often, a variety of sensors are available, so it may be desirable to merge the appropriate sensor data appropriately. Since the sensors sometimes extract very different sensor information from the vehicle environment, a module for merging the various sensor inputs (fusion) must be able to place the sensor information in a suitable geometric as well as semantic context. Various techniques for fusion of sensor data are known, see for example DE 102 54 806 A1 ,

However, such techniques have certain limitations. Frequently, the fusion of sensor data may require comparatively high computing capacity. As a result, corresponding systems and techniques can be expensive and / or time-consuming. Moreover, such techniques can conventionally fuse the sensor data of the multiple sensors only in a very narrowly defined framework. For example, it may not be possible, or only to a limited extent, to react flexibly to the particular situation in which the vehicle is located.

Therefore, there is a need for improved techniques for providing driver assistance functionality based on sensor data. In particular, there is a need for such techniques which fuse the sensor data in a simple, precise and situation-specific manner.

This object is solved by the features of the independent claims. The dependent claims define embodiments.

According to a first aspect, the invention relates to methods for providing driver assistance functionality in a vehicle. The method comprises loading a predefined catalog of road scenarios from a database, the catalog comprising a plurality of road scenarios each describing a road situation by at least one parameterized environment feature. The method further includes obtaining sensor data from at least one sensor of the vehicle, each describing at least one measured environmental feature in an environment of the vehicle. The method further comprises, for at least some of the road scenarios of the predefined catalog: evaluating the respective road scenario by assigning a probability value to the respective road scenario obtained from at least a comparison of the at least one measured environmental feature with the at least one parameterized environmental feature of the respective road scenario. The method further comprises selecting a road scenario depending on the associated probability values of the at least some road scenarios and providing the driver assistance functionality depending on the selected road scenario and / or depending on the measured environmental characteristics.

The driver assistance functionality may assist the driver of the vehicle while driving. This can be done by providing information via a user interface and / or by interfering with the operation of the vehicle. Typical driver assistance functions relate z. B. a precise location of the vehicle, z. B. lane accuracy or with even greater accuracy, or contextual infotainment. It is possible that the method further comprises predictively determining a traffic lane for the vehicle depending on the selected road scenario and on the measured environmental characteristics. Providing the driver assistance functionality may then be based on the anticipated lane. Are z. For example, if the lanes are known in advance, various techniques of autonomous vehicle operation can be implemented. The lanes can z. B. be provided as a clothoid. It would also be possible to provide a complete model of the vehicle environment, such as in a sensor road graph.

The road scenarios may relate to typical road situations that may be classified within the catalog to a greater or lesser degree. A particular road situation may then be comprehensively characterized by the at least one parameterized environment feature. For example, the environmental characteristics may concern: number of lanes; Width of one Lane; Hierarchy class of the road, z. B. highway, highway, city streets; Green stripes on the right and / or green stripes on the left; Guardrail right and / or guardrail left; Stripe to the right and / or stripe to the left; Acceleration track on the left and / or acceleration track on the right; Road markings; delineators; Street signs; other road users; Column track; Peripheral development; Grid card; Map of a navigation system.

The catalog may contain typified road scenarios, which are to be expected or possible within a valid set of rules. Typically, such determination and typing can be done on a legal basis. For example, the road scenarios may be dictated by guidelines for creating roads that may e.g. B. in Germany by the Federal Agency for Transport, Construction, Urban Planning and Spatial Planning (BMVBS) are determined. They can vary for different places. So z. B. certain criteria for roads defined and binding, such as lane width, hard shoulder, sign arrangement, etc .. Corresponding parameters may, for. B. different for different hierarchy classes of each street.

The comparison does not necessarily take into account all the measured environmental characteristics. Only individual measured environmental features can be taken into account. For example, a large (low) probability value may be obtained if the comparison yields a large (small) match between the measured environmental characteristics of the sensor data and the parameterized environmental characteristics. Match can be z. For example, one or more of the following mean: Same environmental characteristics present; several same environmental features in the same geometric relationship available; comparable environmental features available.

The probability value can be z. B. vary between 0 and 1, where 0 represents a vanishing probability and 1 full security.

The sensor data itself may be associated with a confidence level. The sensor can z. For example, by means of the confidence level, indicate how reliable the corresponding sensor data are, ie how probable a measured environmental feature actually is. It is also possible that the probability value is still obtained from the confidence level of the sensor data. This can allow a particularly accurate determination of the probability value.

The selected street scenario can be used in a simple B. the road scenario with the highest probability. This street scenario can z. B. describe the actual environment of the vehicle particularly well.

Then it may be possible to provide the driver assistance functionality both based on the selected road scenario and based on the measured sensor data. This can provide the driver assistance functionality with high accuracy, low susceptibility to errors and particularly far-sighted.

Based on the selected road scenario, it may be possible to put the measured environmental characteristics of the sensor data in context. The context can be particularly plausible because the street scenars are predefined. In other words, the road scenario may allow the different sensor data to be fused with relatively high accuracy. For example, in particular, sensor data from different sensors may also be fused by the techniques described herein.

The sensor data may be obtained from one or more sensors of the vehicle. For example, the sensors may be selected from the group comprising the following elements: optical camera, infrared camera, radar system, navigation system, distance sensors. The sensor data can also come only from a sensor and z. B. describe each a single or a few environmental feature. Therefore, obtaining sensor data in general may be described as: various inputs into the system implementing the techniques according to the present aspect.

It would be possible for the method to further include: determining the sensor data based on sensor information using techniques selected from the group consisting of image recognition, image segmentation, filtering. Such techniques, which enable the extraction of the environmental features from the sensor information or raw data, are basically known to the person skilled in the art.

It would be possible that the comparison for assigning the probability value further takes into account a type of the respective environmental feature. So it may be z. For example, certain types of environment features, e.g. As sods or crash barriers (lane markers and signs), in particular sensor data are particularly well or poorly reproduced, z. B. in sensor data from a radar sensor (a optical sensor). In particular, it may therefore be unnecessary to compare all sensor data with all parameterized environmental features. This can be a particularly accurate and fast method allows.

In various reference implementations, a single fixed, greatly simplified road scenario can be used. These simplifications typically include assumptions regarding the course of the road, for example, in a motorway situation, the assumption that the lanes are parallel. As soon as such assumptions are no longer met, these approaches usually fail. In a highway situation this is already regularly the case when an acceleration lane, z. As a departure strip to be modeled, as this is not parallel to the main traffic direction.

From the above, it can be seen that with respect to the catalog of road scenarios at least two effects can be achieved: on the one hand a particularly efficient, flexible and targeted fusion of sensor data, and on the other hand a particularly accurate and forward-looking provision of driver assistance functionality by taking into account the selected road scenario.

It is possible that the catalog includes a search tree for the multiple road scenarios that includes links between the multiple road scenarios. The probability value of the respective road scenario can furthermore be obtained from traversing the search tree structure up to the respective road scenario.

By presenting the catalog in the form of a search tree structure, a particularly efficient and fast search of the various road scenarios can be made possible. The selected street scenario, which describes the actual vehicle environment particularly well, can be found quickly.

There are different approaches to organize the search tree so that the effects described above become possible. In different scenarios, different ordering schemes can be used.

For example, the various road scenarios of the catalog may be classified in terms of a level of detail and / or a number of the associated at least one environmental feature. The catalog of road scenarios may include links between the various road scenarios, the links linking together such two road scenarios describing congruent road situations with a different level of detail and / or a different number of associated environmental features.

The level of detail can z. B. defined by an abstraction level, which relates to the description of the parameterized environmental features. Are z. For example, if the road scenarios are only roughly outlined and very broadly specified (accurately described and comprehensively specified), the level of detail may be small (large).

It should be understood that one and the same environment feature z. B. can be reproduced with a large or small degree of detail. For example, a pavement marker may optionally be parameterized as solid or dashed, thick or thin, at a certain distance from adjacent landmarks, with particular color, and so forth. All such properties can increase the level of detail. The preceding list is purely illustrative and not restrictive.

The provision of the links can, for. B. particularly efficient by using graphene structures: z. For example, the various road scenarios may be the nodes of the graph, and the links may be the edges of the graph.

By linking the road scenarios, a particularly efficient evaluation can be achieved by assigning the probability values. So it may be z. For example, it may be possible to use the links to calculate probability values for directly and / or indirectly linked road scenarios. The rating can be done quickly and accurately.

For example, the likelihood value with which the respective road scenario is evaluated can still be obtained from probability values of road scenarios associated with the respective road scenario; z. B. preferably from probability values of such road scenarios, which have a lower level of detail and / or a small number of associated environmental features.

If z. For example, if a particular road scenario evaluates a particularly large (low) probability value, road scenarios associated with the particular road scenario, and therefore describing congruent road situations, may correspondingly have a large (low) probability value. In a simple example, it may be possible to assume that one with a high Probably evaluated road scenario, which is comparatively detailed and provided with a large number of environmental characteristics from a particular set, the linked road scenarios with lower level of detail and fewer features from the particular set also have a high probability value.

By means of such techniques it may be possible to evaluate some or all of the road scenarios by assigning each one probability value. On the basis of the probability values, a specific road scenario can then be selected in order subsequently to provide the driver assistance functionality based thereon. Other criteria may also be considered when selecting the road scenario.

Selecting the road scenario may further take into account the level of detail and / or the number of associated at least one environmental feature for the different road scenarios.

In general, such road scenarios, which have a larger number of environmental features z. B. describe with greater degree of detail tend to have a lower probability. This may be the case because, inherently for such road scenarios, comparing with the measured environmental features opens up more sources of potential mismatch - e.g. B. in comparison to a street scenario that only a single environmental feature, eg. B. specified with a low degree of detail. Therefore, it may be desirable to include the level of detail and / or the number of associated at least one environment feature in the selection.

For example, the method may further include: weighting the assigned likelihood values for the road scenarios using a weighting factor that takes into account the level of detail and / or the number of associated at least one environmental feature. Then z. B. that road scenario can be selected which - on the one hand - the highest possible degree of detail and / or the largest possible number of associated at least one environmental feature, and - on the other hand - the largest possible associated probability value.

In this regard, z. For example, the method comprises performing a threshold comparison between the associated probability value and a predetermined probability threshold. The probability threshold may, for. B. depending on the level of detail and / or the number of associated at least one environment feature.

As a result, the effect of a particularly accurate selection of the road scenario can be achieved, for. B. as that road scenario, which describes the actual vehicle environment particularly accurate and with many environmental features, but still with a sufficiently high probability value.

It is possible that the various road scenarios of the catalog are classified in terms of a number of the associated at least one environment feature, the links between the various road scenarios linking such two road scenarios, of which a higher hierarchy of the two road scenarios includes all environmental features of a low hierarchy of the two road scenarios.

In other words, the catalog can be hierarchically structured. Higher hierarchical road scenarios may include more environmental features than subhierarchical ones. Such a classification would also be possible with respect to the search tree as described above.

It may be dispensable that the road scenarios are also classified in terms of level of detail. However, it is also possible that the road scenarios are classified only in terms of level of detail. It would be possible that higher hierarchical street scenarios have a lower or greater degree of detail than low hierarchies.

In such a case of the hierarchical structuring, a particularly simple provision of the driver assistance functionality can take place, because the selection of the road scenario can be carried out quickly, eg. By means of hypothesis generation (passing through the catalog of low-hierarchy to higher-hierarchical road scenarios) and hypothesis verification (going from higher-hierarchical to lower-hierarchical road scenarios). Congruent road situations can be inherently grouped by hierarchical linking.

It is possible that the evaluation of the respective road scenario is performed stepwise for the obtained sensor data. Preferably, the sensor data can be obtained from at least two sensors.

For example, it may be possible to evaluate with first sensor data, e.g. From a first sensor. A probability value may then be obtained by comparing the first sensor data with a suitable road scenario to be obtained. Subsequently, in a second process stage, it may be possible to evaluate a second road scenario based on second sensor data, e.g. B. originate from a second sensor. For example, the second road scenario may be linked to the first road scenario. For example, the second street scenario may be higher-hierarchical or lower-hierarchical than the first street scenario. Corresponding techniques can also be applied if the different sensor data come from one and the same sensor, but z. B. describe different environmental characteristics.

Such techniques, which progressively through the catalog of road scenarios or the search tree structure to account for the various sensor data, a particularly efficient fusion of the sensor data can be achieved. The catalog of road scenarios sets out the framework by which the merger will be carried out. This allows the merger to be targeted, flexible and efficient.

It is possible that the plurality of road scenarios included in the catalog continue to describe the characteristic road situation by at least one geometric relationship between at least some of the parameterized environmental characteristics. It is also possible that the likelihood value with which the respective road scenario is evaluated is further obtained from a comparison of at least one geometric relationship between the plurality of measured environmental features with the at least one geometric relationship between the respective parameterized environmental features.

In other words, the comparison may relate to the geometric relationships between the plurality of measured environmental features with the geometric relationships between a plurality of parameterized environmental features of the respective road scenario.

If the road scenario includes approximately two lane markings, then the comparison can be made, for example. B. relate to a distance as a geometric relationship between the two lane markers and / or a thickness of the two lane markings.

In other words, it may be possible not only to isolate the various environmental features in isolation and individually, but also to consider a geometric relationship of the various environmental features to one another.

This may make it possible to determine the probability value particularly accurately. It can be checked more closely, to what extent a certain street scenario agrees with the measured environmental parameters.

It is possible that the at least one geometric relationship between the parameterized environmental features of the plurality of road scenarios is further characterized by a predefined statistical distribution describing an expected actual variation of the at least one geometric relationship. It is also possible that this statistical distribution is taken into account in obtaining the probability value from the comparison between the at least one geometric relationship.

For example, the statistical distribution may be described by a normal distribution. In this way an actual fluctuation in the geometric relations of real road situations can be mapped. In addition, a tolerance range can be taken into account when comparing the geometric relationships. Such can z. As inaccuracies, variations, noise and errors in the sensor data are taken into account.

In another aspect, the invention features a driver assistance system in a vehicle that includes a sensor interface configured to obtain sensor data from at least one sensor of the vehicle, each describing at least one measured environmental feature in an environment of the vehicle. The driver assistance system further comprises a computer unit which is set up to carry out the following steps: loading a predefined catalog of road scenarios from a database, the catalog comprising a plurality of road scenarios each describing a road situation by at least one parameterized environment feature; for at least some of the road scenarios of the predefined catalog: evaluating the respective road scenario by assigning a probability value to the respective road scenario obtained from at least a comparison of the at least one measured environmental feature with the at least one parametric environmental feature of the respective road scenario; Selecting a road scenario depending on the associated probability values of the at least some road scenarios; and providing the driver assistance functionality as a function of the selected road scenario and / or as a function of the measured environmental characteristics.

The driver assistance system may be configured to perform the method for providing driver assistance functionality according to another aspect of the invention.

For such a driver assistance system, effects can be achieved which are comparable to the effects that can be achieved for the method for providing driver assistance functionality according to a further aspect of the invention.

In another aspect, the invention relates to a vehicle that includes the driver assistance system according to another aspect of the present invention.

For such a vehicle, effects can be achieved that are comparable to the effects that can be achieved for the method for providing driver assistance functionality according to another aspect of the invention.

The features and features set out above, which are described below, can be used not only in the corresponding combination explicitly set out, but also in further combination or isolation without departing from the scope of the present invention.

The above-described characteristics, features, and advantages of this invention, as well as the manner in which they will be achieved, will become clearer and more clearly understood in connection with the following description of the embodiments, which will be described in detail in conjunction with the drawings.

1 is a plan view of a typical road situation in the vicinity of a vehicle, wherein lanes of the vehicle are further illustrated.

2 FIG. 10 is a schematic view of a driver assistance system according to various embodiments. FIG.

3 illustrates an exemplary road scenario described by various parameterized environmental features.

4 provides a search tree for the street scenario of 3 represents.

5 FIG. 3 is a flowchart of a method for providing driver assistance functionality according to various embodiments.

6 FIG. 11 is a flow chart showing further details of the flowchart of FIG 5 illustrated.

Hereinafter, the present invention will be described with reference to preferred embodiments with reference to the drawings. In the figures, like reference characters designate the same or similar elements. The figures are schematic representations of various embodiments of the invention. Elements shown in the figures are not necessarily drawn to scale. Rather, the various elements shown in the figures are reproduced in such a way that their function and general purpose will be understood by those skilled in the art. Connections and couplings between functional units and elements illustrated in the figures may also be implemented as an indirect connection or coupling. A connection or coupling may be implemented by wire or wireless. Functional units can be implemented as hardware, software or a combination of hardware and software.

Hereinafter, techniques for providing driver assistance functionality in a vehicle will be explained with reference to the figures. Specifically, the techniques include selecting a particular road scenario from a predefined catalog of road scenarios based on received sensor data. The sensor data can z. B. originate from a sensor and describe several environmental features of the vehicle and / or they can come from several sensors. It is then possible to provide the driver assistance functionality based on the selected road scenario and based on the sensor data.

Typically, such techniques that consider multiple sensor data are also referred to as fusion. The fusion typically allows to provide a unified and comprehensive data basis from a plurality of sensor data available in a vehicle, by means of which the driver assistance functionality can be provided with great accuracy.

The driver assistance functionality may be selected from the group comprising the elements: predictively determining a lane, lane assistance, corner assistance, distance assistance, infotainment, autonomous driving. Basically, the expert is aware of performing the driver assistance, which is why no further details need to be mentioned here. According to various embodiments, it is possible to provide the data basis, which the driver assistance functionality relies on, in a particularly comprehensive and accurate manner.

Such driver assistance functionalities are typically based on the most accurate knowledge possible of an environment of the vehicle. In 1 is an environment of a vehicle 100 shown. In 1 there is a passenger car 100 on a three-lane highway. This street scenario is characterized by different environmental characteristics. The environmental features include non-continuous or dashed lane markings 210-1a . 210-1b . 210-1c , as well as continuous lane markings 210-2a . 210-2b , In addition, the environmental features include a left side of the car 100 located guardrail 213 and right side of the car 100 located streaks with turf 211 and an acceleration track 214 ,

In 2 is a driver assistance system 110 shown, which is set up to the environment of the car 100 As described above, to recognize particularly accurate and provide the driver assistance functionality. The driver assistance system 110 includes for this purpose a computer unit 111 , which is set up for performing various tasks, which z. B. include: processing of sensor data, providing driver assistance functionality, eg. B. by the reproduction of information by means of a user interface 112 , Loading a catalog of road scenarios from a database 113 ,

The driver assistance system 110 includes three sensor interfaces 120-1 . 120-2 . 120-3 , the sensor data from each of an optical sensor, for. B. a forward facing camera, from a radar device, and provide a navigation system.

For example, by various techniques described herein sensor data from the three sensor interfaces 120-1 . 120-2 . 210-3 be merged and, together with the catalog of road scenarios, used to the environment of the car 100 to analyze very precisely. Then it may be possible to anticipate lanes 250-1 . 250-2 . 250-3 . 250-4 to determine (cf. 1 ). Driver assistance functionality, such as autonomous driving or driving assistance, may be based on the lanes 250-1 . 250-2 . 250-3 . 250-4 to be provided.

The merging takes place taking into account the catalog of road scenarios. The catalog includes road scenarios that are actually in an environment of the car 100 may occur. The different road scenarios are described by parameterized environmental parameters.

In 3 is a street scenario 200 shown schematically. This street scenario 200 is determined by several parameterized environmental characteristics: Hierarchy class equals highway 220 ' , right-sided green strip 211 , on the left side solid road markings 210-2a ' , right-sided road markings 210-2b , central dashed road marking 210-1 ,

It would be possible that the street scenario 200 only by a lower number of environmental features 210-2a . 210-2b . 210-1 ' . 220 ' . 211 is characterized.

The level of detail with which the parameterized environmental characteristics 210-2a . 210-2b . 210-1 ' . 220 . 211 ' are variable, is variable. For example, in addition, a certain thickness of the road markings 210-2a . 210-1 ' . 210-2b ' in the catalog in the database 113 deposited his usf.

In 4 is the catalog 300 of road scenarios 200 for the hierarchy class highway. The catalog 300 , in the embodiment of 4 , is in the hierarchical structure of a search tree 310 shown. Various nodes 311-1 - 311-7 correspond to different road scenarios 200 , The knots 311-1 - 311-7 or the road scenarios 200 are interconnected with each other 321-2-1 . 312-3-2 . 312-2-4 . 312-3-6 . 312-6-5 . 312-6-7 associated.

With increasing hierarchical level (in 4 shown above), the number of environmental features provided in each road scenario increases. This includes a higher hierarchical node 311-1 - 311-7 all environment features of the low hierarchy nodes to which it is linked. This is in 4 through the illustrated road scenarios 200 in conjunction with the 3 seen.

In addition to this hierarchy, which takes into account the number of environmental features, as the hierarchy increases, so too could a level of detail of the various environmental features increase or decrease (in 4 Not shown).

Basically, it is possible to use the search tree structure 310 starting from any node 311-1 - 311-7 to go through. In the following, an example is discussed in a purely illustrative and not restrictive way, in which the search tree structure 310 starting from the node 311-1 (in 4 bottom left).

For example, the sensor interface 120-1 Provide sensor data from an optical camera describing the environment feature "solid road marking". This corresponds to the street scenario 200 of the node 311-1 , By comparing these sensor data with the parameterized environment feature, a probability value may be that 311-1 In the present case, this probability value would be comparatively high, since there is a good match between the corresponding road scenario 200 that's the corresponding one environment feature 210-2a (Cf. 3 ) and the measured environmental feature.

Then you can use the shortcut 312-2-1 be hypothesized that the street scenario 200 of the node 311-2 is present; So in addition to the solid road marking even - right there is - a dashed road mark is present. The road scenarios 200 the knot 311-1 . 311-2 So describe congruent road situations, because the road situation of the street scenario 200 of the node 311-1 under the street situation of the street scenario 200 of the node 311-2 can be subsumed.

This hypothesis can be verified by evaluating the node 311-4 , the street scenario 200 concerning only the dashed line. This rating can be determined by comparing the corresponding parameterized environmental feature 210-1 ' , here dashed lane marking, done with measured environmental features. The measured environmental characteristics for the node 311-4 in turn can be used as sensor data via the sensor interface 120-1 obtained from the optical camera.

Assign both the knots 311-1 . 311-4 comparatively large probability values, it can be assumed that the associated higher hierarchical node 311-2 has a comparatively high probability. This corresponds to the verification of the previously established hypothesis (hypothesis verification).

This probability for the street scenario 200 of the node 311-2 can be obtained in more detail, in which also a geometric relationship between the measured and predefined, parameterized environmental features 210-2a . 210-1 ' , here solid and dashed road marking, is performed. For example, it can be checked by comparison whether the dashed lane marking is at a certain distance to the right of the solid lane marking. Such geometric relationships do not necessarily have to be fixed values, but may be over a tolerance range, e.g. B. be defined by a statistical normal distribution. If the measured environmental features deviate from an average of the normal distribution, then obtaining the probability value may determine this deviation by the properties of the normal distribution, eg. B. half-width, take into account exactly.

The method may, for. For example, continue through hypothesis generation and hypothesis verification until all nodes 311-1 - 311-7 have assigned a probability value. It would also be possible to abort the process if a probability value falls below a predefined threshold.

The street scenario of the knot 311-3 further includes a right-sided turf (cf. 3 ). For example, it may be possible the presence of the sod, for instance in relation to the road scenario 200 of the node 311-7 to verify by sensor data received from a radar sensor via the sensor interface 120-2 to be obtained.

From the above, it can be seen that fusion of multiple sensor data becomes possible according to such techniques. The various sensor data are taken into account in stages, namely when passing through the search tree structure 310 ,

4 is a non-limiting, purely illustrative example. In general, it may be desirable to have the catalog 300 with a larger or smaller number of nodes 311-1 - 311-7 provide. For example, the catalog 300 also such street scenarios 200 include, which do not affect highway, z. B. highways or city streets.

In this regard, first of all, namely when selecting the first node 311-1 To evaluate, already a comparison with sensor data obtained from a navigation system via the sensor interface 120-3 be obtained. This sensor data can specify that the car 100 in a street scenario 200 which describes a road situation of the hierarchy class highway. All other nodes (in 4 not shown), which relate to a road situation of another hierarchy class, can then be assigned a comparatively low probability value.

In 5 FIG. 3 illustrates a flowchart of a method for providing driver assistance functionality according to various embodiments. The process starts in step s1.

First, the catalog is loaded 300 from the database 113 (Step S2). About the sensor interfaces 112-1 . 112-2 . 112-3 then sensor data is obtained (step S3). The sensor data describe the environment of the car 100 by means of environmental features, eg. As road markings, crash barriers, road signs, etc. The Senordaten can continue a hierarchy class of the road on which the car 100 currently located, denote.

Then, in step S4, a road scenario 200 from the catalogue 300 selected. For example, for the given hierarchy class of the street, a low-hierarchical street scenario 200 which contains only one or a few environmental features in parameterized form. This environmental feature can be compared with measured environmental features and a probability value obtained therefrom (step S5). If there is at least one measured environment feature that matches well with the parameterized environment feature, then the probability value may indicate a good match.

In step S6, it is then checked whether there is another relevant road scenario 200 is present in the catalog. For example, only those road scenarios can be considered, which have the appropriate hierarchy class.

If there are other road scenarios, steps S4 and S5 are performed again.

In step S7 becomes a street scenario 200 selected based on the obtained probability values, e.g. B. the street scenario 200 with the highest probability value or a street scenario 200 while taking into account the level of detail and / or the number of parameterized environmental features, as well as the probability value.

In step S8, the driver assistance functionality is provided based on the selected road scenario and based on the measured environmental characteristics. For example, a predictive lane for the car 100 can be calculated and it can be implemented partially or fully autonomous driving.

The process ends in step S9.

As can be seen from the above description, it allows to use the catalog 300 of road scenarios controlling several sensor data in a targeted and flexible way. It can also be a street scenario 200 be selected, which is the actual environment of the car 100 describes very well. If, in step S8, the driver assistance functionality is based both on the measured sensor data and on the selected road scenario describing the environment parameterized 200 is provided, this can be provided with a high accuracy, low error rate and especially far-sighted.

Various embodiments of the invention are also based on the knowledge that by providing the catalog 300 in the form of a search tree 310 a particularly efficient determination of the selected street scenario 200 is possible. Such techniques, as part of step S5 of the 5 can be performed in the flowchart of 6 shown.

In step T1, a significant environmental feature is determined from measured sensor data. For example, this can be an environmental feature with a high confidence level. Typically, this may be a current-street hierarchy class or a succinct, safely-recognized lane marker.

Then a corresponding node 311-1 - 311-7 from the search tree 310 selected (step T2). Basically, this node can 311-1 - 311-7 at each hierarchical level of the search tree 310 are located. However, it may be desirable to have a low-hierarchy node 311-1 - 311-7 select. Which node 311-1 - 311-7 is selected depends in particular on the sensor data.

In step T3, the probability value for this node is calculated by comparing the measured environmental characteristics with the parameterizing environmental features of the selected road scenario 200 ,

If not a higher hierarchical node 311-1 - 311-7 is present (step T4), z. B. because the node selected in step T2 was already located at the highest hierarchical level, the method ends 6 and it can with step S6 of 5 be continued.

Otherwise, in step T5, selecting a higher hierarchical node is done 311-1 - 311-7 , This node typically includes all environment features that the node 311-1 - 311-7 from step T2 also includes, and additionally one or more other environmental features. In addition, the higher node can parameterize any or all environment features with a different level of detail.

In step T6, the probability value for the higher hierarchical node from step T5 is again calculated based on measured environmental characteristics from the sensor data.

Steps T5 and T6 may also be referred to as hypothesis generation because, starting from the low hierarchy node 311-1 - 311-7 , the assumption is made that a higher hierarchical node 311-1 - 311-7 is present.

Accordingly, the steps T7-T10 may be referred to as hypothesis verification: In step T7, it is checked whether a hierarchy Node 311-1 - 311-7 is available. If so, this node becomes 311-1 - 311-7 is selected (step T8) whose probability value is calculated (step T9) and, if necessary, in step T10 the probability value of the higher hierarchical node 311-1 - 311-7 from step T5 based on the probability value of the low-hierarchy node 311-1 - 311-7 ,

The above description illustrates the simplified z. B. partially recursive traversal of the search tree 310 , In general, it may be desirable to consider additional termination criteria, e.g. B. Falling below predefined threshold values of the calculated probability values. It may also be desirable to calculate probability values for all nodes. In addition, it may be possible that there is no appropriate sensor data for directly calculating a probability value for a node 311-1 - 311-7 z. B. higher hierarchical level are available; then it may be necessary to use the hypothesis values of lower hierarchical nodes to calculate the probability values of higher hierarchical nodes by means of hypothesis verification.

For example, it would be generally possible for low-hierarchy nodes 311-1 - 311-7 their probability values for all higher hierarchical nodes 311-1 - 311-7 with which they are linked (see step T10 of the 6 ). This indexing step makes it through the search tree 310 very efficient, since already calculated probability values for low-hierarchy nodes 311-1 - 311-7 between several higher hierarchical nodes 311-1 - 311-7 be shared and therefore redundant calculations are avoided.

Thus, techniques have been described above that allow efficient fusion of multiple sensor data. While environmental features such as lane markings or column lanes often allow an estimate of the entire lane course (high-level information), z. B. image-based object detectors only in a near and local environment of the vehicle information about traffic signs, delineators or other road users ready (low-level information). Various reference implementations therefore estimate the course mainly based on the meaningful high-level environmental features, in particular the lane markings. However, once these are missing (this is often the case in inner-city road scenarios), no robust course can be estimated. Low-level environmental features play a crucial role here. The techniques described above have been used to merge low-level and high-level environmental features based on a catalog of different road scenarios.

The data processing of the merger is hierarchical, z. In the search tree. The hierarchy describes, for example, the decomposition of a road scene (represented by the root node of the hierarchy) into individual lanes, the breakdown of the lanes into features such as lane markings, raised obstacles, guide posts, or arrows, and these in turn in low-level environmental features such as lines in the camera image ( represented by the leaves in the hierarchy). The merger contains prior knowledge of the different types of roads in the form of the catalog. This describes the different road types, eg. Highway, highways, city streets, junction and intersection types. Each street type is defined by its own node in the hierarchy. The nodes therefore describe semantically different scene types, the edges describing spatial relations or geometric relationships between the individual nodes. The geometric relationships between a roadway and its lane markings is z. B. given by the street width. But other orientation aids such. B. Guide posts have a fixed geometric relationship to the road and can therefore be integrated into the hierarchy.

The data processing or the search tree structure distinguishes between two process directions: bottom-up (from the leaves in the direction of the root) and top-down (from the root toward the leaves). Since the sensors typically provide sensor information at lower hierarchical levels (low-level information), the data processing starts with the bottom-up process in which hypotheses for the nodes are generated at higher levels (hypothesis generation). A recognized marker generated, such. Example, a hypothesis that a lane is right, and a hypothesis that a lane is on the left. Each hypothesis is then verified in a top-down step (hypothesis verification). This z. For example, if there is a recognized left mark, it checks to see if there is also a right mark supporting the lane hypothesis.

Abstract and generalized can be obtained by propagating probability values from lower hierarchical nodes to higher hierarchical nodes (see step T10 of the 6 ) may be able to break down a complex road scenario of the higher hierarchical node into several simpler road scenarios of low hierarchical nodes. Such techniques can be used in the context of "parts and Herrsche "(English:" divide and conquer ") can be regarded as a reductionist approach.

The hypothesis generation process can basically be started on the different hierarchy levels of the search tree. Thus, low-level distance information from the radar sensor can be used to generate high-level hypotheses. These hypotheses can then z. B. be verified using high-level information such as column traces. But it is also the reverse direction possible, d. H. the column traces are fed in at higher hierarchical levels and used for the direct generation of lane hypotheses. In the top-down verification process you can then u. a. be verified by the low-level distance information from the radar sensor.

Of course, the features of the previously described embodiments and aspects of the invention may be combined. In particular, the features may be used not only in the described combinations but also in other combinations or per se, without departing from the scope of the invention.

LIST OF REFERENCE NUMBERS

100

vehicle

110

Driver assistance system

111

computer unit

112

User interface

113

Database

120-1

Sensor interface

120-2

120-3

200

road scenario

210-1a

Dashed road marking

210-1b

210-1c

210-2a

Solid road marking

210-2b

211

sward

212

hard shoulder

213

barrier

214

acceleration lane

220

Highway

250-1

lane

250-2

250-3

250-4

300

Catalog

310

Search tree

311-1-311-7

node

312-2-1

edge

312-3-2

312-2-4

312-3-6

312-6-5

312-6-7

QUOTES INCLUDE IN THE DESCRIPTION

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

Cited patent literature

• DE 10254806 A1 [0003]

Claims (13)

Method for providing driver assistance functionality in a vehicle ( 100 ), comprising: - Loading a predefined catalog ( 300 ) of road scenarios ( 200 ) from a database ( 113 ), whereby the catalog ( 300 ) several road scenarios ( 200 ), each of which represents a road situation by at least one parametric environmental feature ( 210-1 ' . 210-2a ' . 210-2b ' . 211 ' . 220 ' ), obtaining sensor data from at least one sensor of the vehicle ( 100 ), each having at least one measured environmental feature ( 210-1a . 210-1b . 210-1c . 211-2a . 211-2b . 211 . 212 . 213 . 214 ) in an environment of the vehicle ( 100 ), for at least some of the road scenarios ( 200 ) of the predefined catalog ( 300 ): Assess the respective street scenario ( 200 ) by assigning a probability value to the respective road scenario ( 200 ), which is at least obtained from: a comparison of the at least one measured environmental feature ( 210-1a . 210-1b . 210-1c . 211-2a . 211-2b . 211 . 212 . 213 . 214 ) with the at least one parameterized environment feature ( 210-1 ' . 210-2a ' . 210-2b ' . 211 ' . 220 ' ) of the respective street scenario ( 200 ), - choosing a street scenario ( 200 ) as a function of the assigned probability values of the at least some road scenarios ( 200 ), - providing the driver assistance functionality depending on the selected road scenario ( 200 ) and / or depending on the measured environmental characteristics ( 210-1a . 210-1b . 210-1c . 211-2a . 211-2b . 211 . 212 . 213 . 214 ). Method according to claim 1, characterized in that the catalog ( 300 ) a search tree ( 310 ) for the several road scenarios ( 200 ), the links between the several road scenarios ( 200 ) and that the probability value of the respective road scenario ( 200 ) is further obtained from: traversing the search tree ( 310 ) up to the respective street scenario ( 200 ). Method according to one of claims 1 or 2, characterized in that the different road scenarios ( 200 ) of the catalog ( 300 ) are classified with regard to a degree of detail and / or a number of the associated at least one environmental feature ( 210-1 ' . 210-2a ' . 210-2b ' . 211 ' . 220 ' ), whereby the catalog ( 300 ) of road scenarios ( 200 ) Links between the different road scenarios ( 200 ) containing such two road scenarios ( 200 ), the congruent road situations with a different degree of detail and / or a different number of associated environmental features ( 210-1 ' . 210-2a ' . 210-2b ' . 211 ' . 220 ' ). Method according to claim 3, characterized in that the probability value with which the respective road scenario ( 200 ) is further obtained from: probability values of road scenarios ( 200 ), which correspond to the respective street scenario ( 200 ) and which preferably have a lower degree of detail and / or a small number of associated environmental features ( 210-1 ' . 210-2a ' . 210-2b ' . 211 ' . 220 ' ) exhibit. Method according to one of claims 3 or 4, characterized in that the selection of the road scenario ( 200 ), the level of detail and / or the number of associated at least one environment feature ( 210-1 ' . 210-2a ' . 210-2b ' . 211 ' . 220 ' ) for the different road scenarios ( 200 ) considered. Method according to one of claims 3-5, characterized in that the different road scenarios ( 200 ) of the catalog ( 300 ) are classified in terms of a number of the associated at least one environmental feature, the links between the different road scenarios ( 200 ) such two road scenarios ( 200 ), of which a higher hierarchical of the two road scenarios ( 200 ) all environmental features of a low-hierarchy of the two road scenarios ( 200 ) includes. Method according to one of the preceding claims, characterized in that the rating of the respective road scenario ( 200 ) is performed stepwise for the obtained sensor data, wherein preferably the sensor data of at least two sensors are obtained. Method according to one of the preceding claims, characterized in that the parameterized environmental characteristics ( 210-1 ' . 210-2a ' . 210-2b ' . 211 ' . 220 ' ) are selected from a group comprising the following elements: - number of lanes; - width of a lane; - Hierarchy class of the road, z. B. highway, highway, city streets; - green area ( 211 ) right and / or green stripes on the left; - guardrail ( 213 ) right and / or guardrail left; - hard shoulder ( 212 ) right and / or hard shoulder left; - acceleration track ( 214 ) right and / or acceleration lane on the left; - road markings ( 210-1a . 210-1b . 210-1c . 210-2a . 210-2b ); - guide posts, - street signs; - other road users; - column track; - edge development; - grid map; - Map of a navigation system. Method according to one of the preceding claims, characterized in that in the catalog ( 300 ) included several road scenarios ( 200 ) the characteristic road situation is further defined by at least one geometric relationship between at least some of the parameterized environmental characteristics ( 210-1 ' . 210-2a ' . 210-2b ' . 211 ' . 220 ' ) and that the probability value with which the respective street scenario ( 200 ) is further obtained from: a comparison of at least one geometric relationship between the plurality of measured environmental features ( 210-1a . 210-1b . 210-1c . 211-2a . 211-2b . 211 . 212 . 213 . 214 ) with the at least one geometric relationship between the respective parameterized environment features ( 210-1 ' . 210-2a ' . 210-2b ' . 211 ' . 220 ' ). A method according to claim 9, characterized in that the at least one geometric relationship between the parameterized environmental features ( 210-1 ' . 210-2a ' . 210-2b ' . 211 ' . 220 ' ) of the several road scenarios ( 200 ) is further characterized by a predefined statistical distribution that describes an expected actual variation of the at least one geometric relationship, and that this statistical distribution is taken into account in obtaining the probability value from the comparison between the at least one geometric relationship. Method according to one of the preceding claims, characterized in that the method further comprises: - predictive determination of a lane ( 250-1 . 250-2 . 250-3 . 250-4 ) for the vehicle ( 100 ) depending on the selected street scenario ( 200 ) and depending on the measured environmental characteristics ( 210-1a . 210-1b . 210-1c . 211-2a . 211-2b . 211 . 212 . 213 . 214 ), wherein providing the driver assistance functionality on the anticipated lane ( 250-1 . 250-2 . 250-3 . 250-4 ). Driver assistance system ( 110 ) in a vehicle ( 100 ), comprising: - a sensor interface ( 120-1 . 120-2 . 120-3 ) arranged to receive sensor data from at least one sensor of the vehicle ( 100 ), each having at least one measured environmental feature ( 210-1a . 210-1b . 210-1c . 211-2a . 211-2b . 211 . 212 . 213 . 214 ) in an environment of the vehicle ( 100 ), to obtain - a computer unit ( 111 ), which is set up to carry out the following steps: Loading a predefined catalog ( 300 ) of road scenarios ( 200 ) from a database ( 113 ), whereby the catalog ( 300 ) several road scenarios ( 200 ), each of which represents a road situation by at least one parametric environmental feature ( 210-1 ' . 210-2a ' . 210-2b ' . 211 ' . 220 ' ), for at least some of the road scenarios ( 200 ) of the predefined catalog ( 300 ): Assess the respective street scenario ( 200 ) by assigning a probability value to the respective road scenario ( 200 ), at least from a comparison of the at least one measured environmental feature ( 210-1a . 210-1b . 210-1c . 211-2a . 211-2b . 211 . 212 . 213 . 214 ) with the at least one parameterized environment feature ( 210-1 ' . 210-2a ' . 210-2b ' . 211 ' . 220 ' ) of the respective street scenario ( 200 ), - selecting a street scenario ( 200 ) as a function of the assigned probability values of the at least some road scenarios ( 200 ), - providing the driver assistance functionality depending on the selected road scenario ( 200 ) and / or depending on the measured environmental characteristics ( 210-1a . 210-1b . 210-1c . 211-2a . 211-2b . 211 . 212 . 213 . 214 ). Driver assistance system ( 110 ) according to claim 12, characterized in that it is further adapted to perform a method according to any one of claims 1-11.

Images