METHOD AND SYSTEM FOR DETERMINING AN ENVIRONMENTAL MODEL FOR A VEHICLE

Description

[0001] The present invention relates to a method for determining an environment model for a vehicle, in which an initial position estimate for the vehicle is acquired and map data is acquired. The map data includes information about the spatial arrangement of geographical areas, and the geographical areas are assigned to different area categories. Environmental data is acquired in a detection space, and objects are detected based on the environmental data. Each detected object is assigned an object position and an object category. The environment model is determined based on the detected objects.The invention further relates to a system for determining an environment model for a vehicle, comprising a position detection unit for detecting an initial position estimate for the vehicle and a map data detection unit for detecting the map data based on the initial position estimate, wherein the map data comprises information about the spatial arrangement of geographical areas and the geographical areas are assigned to different area categories. The system further comprises an environment data detection unit for detecting environmental data in a detection space and a detection unit for detecting objects based on the environment data and for assigning an object position and an object category to each detected object. It further comprises an environment model determination unit for determining the environment model based on the detected objects. The invention also relates to a computer program product.

[0002] When vehicle sensors and systems detect objects, misdetections are a recurring problem. For example, data can be misinterpreted if the side of a truck is mistaken for a house wall, or an incorrect classification can occur if, for example, a bike path marking is mistaken for a lane marking on the road. If the detected objects are used to perform navigation or determine a vehicle's position, this can lead to incorrect results.

[0003] It is known to use multiple sensors and perform a fusion of the sensor data in order to achieve more reliable object detections.

[0004] US 2017 / 0337435 A1 proposes a method for detecting objects in which images are captured with a camera. First, a coarse detection is performed, in which objects are detected in non-consecutive images. These objects are then analyzed in a fine detection.

[0005] In the method described in US 2017 / 0008521 A1 for calibrating a speed indicator for an autonomous vehicle, landmarks are detected in a large number of images and the speed of the vehicle is calculated based on their movement relative to the vehicle.

[0006] US 2017 / 0350712 A1 describes a method for determining the position of a vehicle. Features in the vehicle's surroundings are extracted and the probability that the objects are stationary is determined. The vehicle's own position is corrected based on the stationary objects.

[0007] JP 2014 145666 A describes a navigation system that issues driving instructions relative to a specific object, such as a shop sign. This takes into account how well the object can be detected by the vehicle.

[0008] JP 2008 051612 A describes a landmark recognition system in which data is collected by a vehicle's camera and landmarks are recognized. A check is then carried out to determine whether the conditions are sufficient for reliable recognition. If the detected landmarks deviate from the information on a centrally provided map, this is only reported to the control center if the conditions were sufficiently good.

[0009] US 2017 184 407 A1 describes a device that receives position information transmitted from a nearby mobile object. A position correction process is performed to correct the position information of the nearby mobile object based on map data so that a corrected position of the object lies within a travel area of ​​a road.

[0010] In the autonomous vehicle described in US 2014 379 247 A1, environmental information indicating the lane is compared with image data representative of the environment to determine a portion of the image data that corresponds to the lane of the environment. Based on this portion of the image data and by ignoring a remaining portion of the image data, the vehicle can determine whether an object is present in the lane and provide instructions for controlling the vehicle in autonomous mode within the environment.

[0011] EP 3 078 937 A1 describes a system for estimating the position of a vehicle, comprising a device that estimates the position of the vehicle using a characteristic object contained in an image captured by an image recording device on the vehicle, and a server connected to this device via a network. The characteristic object is used, in particular, to further improve the position accuracy of an initial position determination based on satellite navigation.

[0012] DE 10 2016 203086 A1 describes a method for driver assistance in which objects in the vicinity of the vehicle are detected while a vehicle is executing a driving maneuver. Stationary objects forming bottlenecks and oncoming vehicles are classified as objects. Subsequently, the movement of the vehicle and the movements of oncoming vehicles are predicted, open spaces in the vicinity of the stationary objects forming bottlenecks are determined, and blockage situations are identified in which the vehicle can no longer continue driving at a bottleneck due to an oncoming vehicle.

[0013] The present invention is based on the object of providing a method, a system and a computer program product that allow particularly reliable detection of objects in the environment of the vehicle.

[0014] According to the invention, this object is achieved by a method having the features of claim 1, a system having the features of claim 12, and a computer program product having the features of claim 14. Advantageous embodiments and further developments emerge from the dependent claims.

[0015] In the method according to the invention, assignment rules are provided which define an assignment of the object categories to the area categories, wherein the detection of the objects takes place depending on the assignment rules.

[0016] This advantageously allows detections of objects to be identified as faulty and filtered out if the detected objects are not allowed to occur in certain areas.

[0017] An "environment model" within the meaning of the invention comprises information about the spatial surroundings of the vehicle. It particularly includes information about the arrangement of objects relative to the vehicle and relative to each other. It can also include information about parameters of the objects, such as orientation, length or height, width, angle, or radius.

[0018] The environment model specifically depicts the vehicle's surroundings at a specific position or pose. The term "position" is used below to include a "pose," i.e., a combination of position and orientation in space.

[0019] In one embodiment of the method according to the invention, the area categories comprise at least one passable area, one non-passable area, one roadway area, one built-up area, and / or one vegetation area. This advantageously allows for distinguishing between areas that are of particular relevance to a model of the vehicle's surroundings. Other area categories can be provided alternatively or additionally.

[0020] For example, the map data may include information that a specific area is a road, whereby the area category "road" may also be assigned to the category "driveable area." Within the area categorized as a road, further areas may be located, such as several different lanes, a hard shoulder, or a peripheral area. Furthermore, a specific area may be categorized as a "development area," which may contain buildings. Another area may be categorized as a "footpath" and "not driveable," for example.

[0021] Based on the initial position estimate, the arrangement of the areas relative to the vehicle can be determined. For example, the vehicle may be in a certain position on a road, and the map data can be used to determine the distance and direction relative to the vehicle to the other areas.

[0022] In a further development of the method, the accuracy of the initial position estimate is determined, and based on this accuracy, the arrangement of the geographical areas relative to the vehicle is determined. This advantageously ensures optimal use of the available data, even with varying degrees of position estimate accuracy.

[0023] The method particularly determines how the geographical areas defined by the map data are arranged relative to the vehicle. The initial position estimate is used to determine the position or pose of the vehicle within the reference system of the map data. The accuracy of the initial position estimate can be determined in a conventional manner. It can also be predetermined, for example, for certain positioning methods, or the accuracy can be output by a device for determining the position, for example, as an error range or confidence interval.

[0024] This determines how the various areas are arranged relative to the vehicle. The method can take into account that, for example, if the vehicle's own position estimate is less accurate, the arrangement of the areas can be determined with correspondingly lower accuracy. In particular, areas and their arrangement relative to the vehicle are determined for which their affiliation to a specific area defined in the map data cannot be precisely determined. It can be provided that the assignment rules are modified in such areas to handle ambiguous assignments.

[0025] The environmental data is acquired in a manner known per se, in particular by means of sensors. For example, lidar, radar, or ultrasonic sensors can be used; a camera can also be used. Other sensors can also be used, in particular combinations of different sensors. The acquisition space in which environmental data is acquired is defined in particular as the spatial area in which the sensors used can acquire data, for example a range or a field of view of the sensors. It can be provided that the acquisition space is adaptable, for example by restricting the acquisition of environmental data to a specific spatial area.

[0026] The detection of objects in the environmental data can also be performed in a conventional manner, for example, using pattern recognition. This method particularly detects semantically defined objects, i.e., objects that are recognizable as distinct entities and assigned to specific object categories.

[0027] The assignment to object categories also occurs in a conventional manner, with each detected object being assigned at least one object category. In particular, a distinction is made between static and dynamic objects. Static objects are those that exist for a longer period of time in a fixed arrangement within the coordinate system of the map data, for example, traffic facilities, buildings, road markings, or vegetation elements. In particular, these can be so-called "landmarks." Dynamic objects can be, for example, other road users or objects temporarily located at a specific position. When an object is detected, the position of the object relative to the vehicle is also determined, such as a distance and an angle at which the object was detected.

[0028] The mapping rules can be provided in various ways known per se. For example, they can be specified by means of a storage unit or received from an external unit. In particular, the map data can include mapping rules, or the mapping rules can be captured together with the map data. Alternatively or additionally, the mapping rules can be captured through user input. Machine learning methods can also be used to determine the mapping rules.

[0029] According to the invention, a negative assignment for a specific area category and a specific object category is determined based on the assignment rules. Objects assigned to the specific object category are detected based on those environmental data that do not relate to geographical areas assigned to the negatively assigned specific area category. This advantageously ensures that no objects based on false detections are taken into account in the environmental model, and detection can be performed more efficiently.

[0030] The negative assignment of specific object categories and area categories determined based on the assignment rules excludes certain object categories from certain area categories. Therefore, to detect objects of the specific object category, less extensive environmental data can be used, making the detection step faster and more efficient. For example, the environmental data can be filtered to exclude areas of the negatively assigned area category. Furthermore, it can be specified that no environmental data is collected in areas assigned to a negatively assigned area category.

[0031] For example, objects in the "road markings" category can be negatively linked to areas in the "house wall" category. To detect road markings, the environmental data can then be filtered so that no environmental data collected in the area of ​​house walls is taken into account. It can also be planned to control a sensor so that it does not collect environmental data in the area of ​​house walls. Alternatively or additionally, detected objects can be discarded if they are assigned to a specific object category and are detected in an area for which this object category is excluded.

[0032] Conversely, it can also be provided that specific object categories are positively linked to specific area categories. Objects of these object categories are therefore only detected and accepted for further processing if they are located in areas of the positively linked area categories. The positive assignment can be used to specifically use environmental data from areas of certain area categories to detect objects of a specific object category and to exclude other areas. This also leads to a reduction in the data to be processed. Analogous to the above example, objects of the category "road markings" can be positively linked to

[0033] Areas in the "road surface" category can be linked. To detect road markings, the environmental data can then be filtered so that only environmental data collected in the area of ​​the road surface is taken into account. It can also be configured to control a sensor so that it preferentially or exclusively collects environmental data in the area of ​​the road surface.

[0034] In a further development, preliminary detections are first determined during object detection, and these preliminary detections are then filtered based on the matching rules. This advantageously allows the detections to be verified based on the matching rules.

[0035] During filtering, each object is checked to determine which object category it belongs to and which area category the position of the detected object is assigned to. If the object and area categories are mutually exclusive, the detected object is discarded and not considered when determining the environment model. If, however, the object and area categories are positively assigned to each other, the detected object is considered.

[0036] In a further development, a subset of the recorded environmental data is determined based on the assignment rules when detecting objects, and the objects are detected based on this subset of environmental data. This advantageously allows for particularly efficient processing of the environmental data, as unnecessary environmental data is not used for detection, and real-time processing or processing at a speed sufficient for use during vehicle travel can be ensured.

[0037] For example, it can be specified that objects of a certain object category are to be detected specifically. Based on the assignment data, it can be determined which area categories these objects are positively linked to, i.e. in which areas the desired objects can be detected. Based on the map data, it is determined how the areas of the corresponding area categories are arranged relative to the vehicle. The recorded environmental data is processed in such a way that a subset is formed that is essentially limited to these areas. This can be done, for example, using filtering. The environmental data is reduced in the process. For example, it can be specified that road markings are only detected in the area of ​​the road surface and other areas, such as built-up areas, are excluded from detection.

[0038] In a further embodiment, the acquisition space in which the environmental data is acquired is defined based on the mapping rules. This advantageously limits the acquired data to the necessary extent. Unlike in the case explained above, in which the acquired environmental data is reduced to a suitable subset, here the reduction can already occur during acquisition. For example, a solid angle within which data is acquired can be limited.

[0039] Here, too, it can be taken into account that the initial position estimation is carried out with a certain inaccuracy and that the relevant areas must therefore be selected as large as possible in order to be able to capture all relevant objects.

[0040] In a further development, the detected objects are static objects. These are particularly well-suited for use in position determination. Static objects are assigned to specific object categories, which are referred to as static. They exhibit essentially constant properties, such as a fixed position and orientation, size, height, width, or an arrangement relative to one another.

[0041] For example, the object categories can include posts, poles, traffic structures, peripheral development elements, road markings, signs, and traffic lights. Such elements are advantageously widespread in road environments and therefore universally applicable.

[0042] The initial position estimate is performed in a known manner, for example, using a global navigation satellite system, such as GPS. The position or pose of the vehicle can be determined in various coordinate systems, such as a global coordinate system or a coordinate system relative to a specific reference point. In addition, known methods can be used to optimize the position determination, for example, using landmarks. The initial position estimate can also be generated by an external unit and received by the vehicle, for example, provided by other road users or a traffic monitoring device.

[0043] In a further development, an optimized position estimate for the vehicle is determined based on the initial position estimate and the environment model. The optimized position estimate can be used as a new initial position estimate in an iterative implementation of the method, for example, to update and redetermine a further optimized position estimate or the environment model. The environment model can thus be advantageously used to determine the position or pose of the vehicle with particular precision.

[0044] The optimized position estimate is specified, in particular, in a global coordinate system or in a coordinate system of the map data. In particular, a map comparison is performed in which the information about landmarks contained in the map data, in particular detectable objects, is compared with the actually detected objects. In particular, an optimization problem is solved in which the positions of the detected objects are approximated as closely as possible to the landmarks contained in the map data. The position and orientation of the vehicle in the map coordinate system can thus be determined based on the relationship between the position and orientation of the vehicle and the arrangement of the detected objects.

[0045] In a further development, the detected objects include dynamic objects. This advantageously allows for the plausibility of detections of dynamic objects.

[0046] For example, the object categories to which dynamic objects are assigned may include other road users. The method according to the invention can be used in this case to ensure that such objects are only detected in permissible areas, avoiding false detections and inappropriate reactions to false detections. For example, known methods have the difficulty that the irregular shape of vegetation at the edge of a roadway can lead to the erroneous detection of other road users in these areas. For example, a vehicle located within the vegetation may be erroneously detected.

[0047] Dynamic objects can be included in the environment model or processed in addition to it, for example to transfer them to a driver assistance system.

[0048] The map data can be acquired in a manner known per se, for example by retrieving it from a storage unit of the vehicle, for example as part of a navigation system. In one embodiment of the method according to the invention, the map data is at least partially received from an external unit. This can be, for example, an external server or a backend device that is at least temporarily connected to the vehicle via data technology. For this purpose, a data connection is established between the vehicle and the external unit, for example via a data network such as the Internet, and the map data can be requested and received by the vehicle. This advantageously makes it possible to always provide up-to-date map data.

[0049] The map data is collected, in particular, based on the initial position estimate of the vehicle, for example, so that the map data relates to a geographical area in the vicinity of the vehicle, such as within a certain minimum radius around the initially estimated position of the vehicle. The map data can also relate to a geographical area in which the vehicle is expected to be located in the future, taking into account, for example, a planned route. This allows map data to be collected well in advance, before it becomes relevant at a later point in time.

[0050] In the invention, the map data are designed to include information about at least two geographical areas and their spatial arrangement. For example, the arrangement of the geographical areas relative to one another and / or in a specific global or relative coordinate system can be included. The map data particularly includes information about the boundaries of specific areas, which define the shape and location of the geographical areas. The geographical areas can be defined as planar structures or as points or lines, whereby points and lines can in particular be assigned a width, a radius, and / or other information about their areal extent.

[0051] The map data also includes information about the area categories to which a geographical area is assigned. The area categories can be formed such that an area category is provided for each of the areas of the map data. A geographical area can also be assigned to multiple area categories.

[0052] The map data can also be updated as needed by providing update data from the external unit. Furthermore, map data can be requested when the vehicle enters a specific geographical area, which can be determined, for example, based on a planned route from a navigation system.

[0053] The data received from the external unit may also include supplementary information, such as information on the arrangement of geographical areas, their assignment to area categories or assignment rules.

[0054] In a further development, transmission data is generated based on the specific environment model and transmitted to an external unit, such as a server or a backend device. The environment model generated for a vehicle can thus be advantageously used for other purposes.

[0055] For example, the environment model or part of the information it contains can be made available to another road user. Furthermore, it can be checked whether existing map data matches the environment model or requires updating. If, for example, several vehicles detect deviations between the actually detected objects and the map data, such as if a building or road layout has changed, these deviations can be identified. The map data can subsequently be corrected or updated, for example, by commissioning a new acquisition or by evaluating the information from the vehicle's environment model.

[0056] In the system according to the invention, the detection unit is designed to detect the objects depending on assignment rules, wherein the assignment rules define an assignment of the object categories to the area categories. The system according to the invention is characterized in that, based on the assignment rules, a negative assignment can be determined for a specific area category and a specific object category; and objects assigned to the specific object category can be detected based on those environmental data that do not relate to geographical areas assigned to the negatively assigned specific area category.

[0057] The system according to the invention is particularly designed to implement the above-described method according to the invention. The system thus has the same advantages as the method according to the invention.

[0058] In one embodiment of the system according to the invention, the environmental data acquisition unit comprises a lidar, radar, or ultrasonic sensor, or a camera for visible or infrared light. Furthermore, other sensor types can be included, or combinations of several sensors and sensor types can be provided. In this way, a multitude of possibly already existing vehicle sensors can be used to acquire the environmental data.

[0059] The computer program product according to the invention comprises instructions which, when executed by a computer, cause the computer to carry out the method described above.

[0060] The invention will now be explained using embodiments with reference to the drawings. Figure 1 shows a vehicle with an embodiment of the system according to the invention, Figure 2 shows an embodiment of the method according to the invention and Figures 3A to 3D show an embodiment of a traffic situation in which the method according to the invention can be carried out.

[0061] With reference to Figure 1 A vehicle with an embodiment of the system according to the invention is explained.

[0062] The vehicle 1 comprises a position detection unit 2, a map data acquisition unit 3, and an environmental data acquisition unit 4. In this exemplary embodiment, the position detection unit 2 comprises a GPS module. The map data acquisition unit 3 comprises an interface to a storage unit (not shown) that provides map data for retrieval. In other exemplary embodiments, the map data acquisition unit 3 can be designed as an interface to an external unit from which map data or supplementary map data can be retrieved. In this exemplary embodiment, the environmental data acquisition unit 4 comprises a camera that is arranged in the front area of ​​the vehicle 1 and captures image data about the area in front of the vehicle 1. In further exemplary embodiments, other sensors can be provided alternatively or additionally, such as an infrared camera, a 3D camera, or an ultrasound, radar, or lidar sensor.

[0063] The vehicle 1 further comprises a computing unit 7, which in turn comprises a detection unit 5 and an environment model determination unit 6. The computing unit 7 is coupled to the position detection unit 2, the map data acquisition unit 3, and the environment data acquisition unit 4. The computing unit 7 is further coupled to a driver assistance system 8.

[0064] In the exemplary embodiment, the driver assistance system 8 comprises a module for at least partially autonomous control of the vehicle 1. Various driver assistance systems 8 known per se can be provided.

[0065] With reference to Figure 2 An embodiment of the method according to the invention is explained. This is based on the above-explained embodiment of the system according to the invention, which is further specified by the description of the method.

[0066] In step 21, an initial position estimate is acquired. In the exemplary embodiment, this is done using the position detection unit 2 with its GPS module. Various known methods for determining the position can be used, with particular provision being made for providing an indication of the accuracy of this estimate along with the position estimate. In the exemplary embodiment, the position determination also includes determining the orientation of the vehicle, i.e., the pose of the vehicle is determined.

[0067] In a further step 22, map data is acquired. In the exemplary embodiment, map data about a geographical area within a specific radius around the vehicle 1 are acquired based on the initial position estimate. In further exemplary embodiments, map data about a geographical area that the vehicle 1 is expected to travel through in the future can be acquired alternatively or additionally, for example, because a route to this area is planned.

[0068] The map data includes information about the location and arrangement of geographical areas as well as information about the area categories to which the individual areas are assigned. For example, information about the course of streets and roads is included, as well as their width or other spatial extent. Furthermore, the exemplary embodiment includes information about where buildings are located next to a road. Examples of area categories include passable and non-passable areas, a roadway area, a built-up area, or a vegetation area. The arrangement of the areas is described below with reference to the Figures 3A to 3D explained in more detail.

[0069] In a step 23, environmental data are acquired, whereby in the exemplary embodiment the camera of the environmental data acquisition unit 4 is used for this purpose.

[0070] In a further step 24, objects are detected based on the environmental data, in particular using known pattern recognition methods, such as for detecting landmarks. Machine learning or artificial intelligence methods can also be used in particular. If an object is detected in the environmental data, it is assigned a position and an object category. Furthermore, it can be provided that further characteristics are determined, such as the orientation, size, length, width or color design of an object. The assignment to an object category is carried out in particular based on the characteristics detected for an object. Examples of object categories include posts, piles, traffic structures, elements of peripheral development, road markings, signs or traffic lights.

[0071] In this embodiment, static objects are detected, which are assigned to corresponding static object categories. In further embodiments, dynamic objects can be detected alternatively or additionally. Examples of dynamic object categories include other road users, cars, trucks, pedestrians, or bicycles.

[0072] In step 25, the detected objects are filtered. For this purpose, assignment rules are recorded, which in the exemplary embodiment are provided by the map data acquisition unit 3. The assignment rules comprise a positive and / or negative assignment of object categories and area categories. In particular, it is defined that objects of a certain object category should not be detected in areas of a certain area category. For example, it can be assumed that there are no building edges on a roadway. In the present exemplary embodiment, these objects are therefore discarded in step 25.

[0073] In step 26, an environment model for vehicle 1 is determined based on the detected objects. This model includes the positions and, in particular, the orientations of the detected objects relative to vehicle 1. It also includes information about the objects, in particular the object categories assigned to them. The environment model may include further information.

[0074] The initial position estimate is optimized, in particular by means of map matching. This solves an optimization problem in which the detected objects, whose positions relative to vehicle 1 and relative to each other have been determined, are overlaid with the positions contained in the map data for the objects. This is done in a conventional manner, in particular using methods for determining positions using landmarks. In this way, an optimized position and, in particular, the orientation of vehicle 1 in the coordinate system of the map data is determined.

[0075] In a further step 27, the optimized position is output to the driver assistance system 8, for example in order to carry out a partially or fully autonomous control of the vehicle 1.

[0076] In further embodiments, it is provided that the accuracy of the initial position estimate is taken into account when filtering the detected objects. For example, it cannot be precisely determined for a position relative to vehicle 1 whether it belongs to an area of ​​a first area category or to a second area category. The assignment rules can be applied here such that, for example, objects of a certain object class are accepted at a position even though they are located with a certain probability in an area in which they should not be detected. Conversely, it can be provided that detected objects are discarded in such cases of doubt.In addition, a method can be provided in which, upon detection of a specific object, especially at a position not clearly assigned to an area category, a plausibility check is performed, for example by searching for corresponding objects in the map data. In this way, it can be determined whether a specific detection of an object is plausible and whether the object should therefore be considered in the environment model or rejected.

[0077] When determining the environment model, one difficulty can arise from the fact that a position is to be determined based on the environment model, while simultaneously requiring an initial position estimate. It can be planned that the optimized position is used in an iterative process to improve the initial position estimate in a subsequent step. Furthermore, the inaccuracy of the initial position estimate can be taken into account in various ways, for example, by checking individual detected objects in parallel with their assignment to specific area categories.

[0078] In a further embodiment, it is provided that the assignment rules comprise a positive assignment between object categories and area categories. In this case, the assignment rules define that objects of a certain object category can only be found in areas that belong to a certain area category. It can also be provided that, when detecting the objects, it is already taken into account in which areas objects of a certain object category should or should not be detected. In this case, the acquired environmental data is filtered such that only data that was acquired in suitable areas is taken into account. This means that the environmental data is restricted to a subset such that only environmental data from areas that are positively or at least not negatively linked to a certain object category are analyzed.This reduces the amount of data within which object detection is to be performed, which significantly shortens the computing time, especially for complex pattern recognition methods. This allows the method to be performed more effectively in real time, or at runtime, i.e., during operation of vehicle 1.

[0079] In further embodiments, the environmental data acquisition unit 4 is configured to acquire environmental data only in specific spatial areas. To acquire objects of a specific object category, environmental data is then acquired only in areas that are positively or at least not negatively linked to the object category. In this way, the amount of data to be processed can be reduced.

[0080] With reference to the Figures 3A to 3DThe exemplary embodiment of the method according to the invention is explained in more detail in an exemplary traffic situation. This is based on the exemplary embodiments of the system and method according to the invention explained above.

[0081] Figure 3A shows a representation of the ego vehicle 30 on a roadway 31. An arrow symbol symbolizes the forward direction of the vehicle 30. Furthermore, a detection area 36 of the environmental data acquisition unit 4 is indicated, which extends symmetrically forward from the vehicle 30 at a specific angle. The detection area 36 can be configured differently, particularly depending on the sensors used in the environmental data acquisition unit 4, in particular with regard to the aperture angle, the direction, the range, and the dependence on environmental conditions, such as visibility.

[0082] A lane marking 35, running centrally in the exemplary embodiment, is applied to the surface of the roadway 31, and the roadway 31 is laterally bordered by a curb 34. In the exemplary embodiment, the roadway 31 curves to the right. Posts 33 are arranged next to the roadway 31, which can be identified as delineators, particularly based on their thickness and height. Furthermore, buildings 32 are arranged laterally next to the roadway 31.

[0083] In the exemplary embodiment, the map data that are acquired by the map data acquisition unit 3 for carrying out the method according to the invention include information about the course and spatial extent of the roadway 31, the lane markings 35, the presence and certain features of the curb 34, the posts 33 and the buildings 32. In particular, the map data include information about the arrangement of certain areas and their assignment to certain area categories. This will be explained below with reference to the Figures 3B to 3D explained in more detail.

[0084] In Figure 3BA drivable area 37 is highlighted, or rather, the intersection of such a drivable area 37 with the detection area 36. The drivable area 37 extends across the surface of the roadway 31. Positions within this area are assigned to the area category "drivable area." In the exemplary embodiment, it is provided that certain object categories, for example, objects in the category "posts" or "building edges," are not to be detected here; that is, such detections are discarded as false detections. On the other hand, this area category is positively linked to other object categories, for example, the category "road markings." In the exemplary embodiment, it is therefore provided that, to detect the road marking 35, only a subset of the recorded environmental data that includes the drivable area 37 is analyzed.Another exemplary object category can be a "preceding vehicle", the detection of which is only expected in the drivable area 37 and not, for example, in the area of ​​the buildings 32.

[0085] In Figure 3Ca building front area 38 is highlighted, which is arranged at a certain distance from the roadway 31 and along the buildings 32. The arrangement of the building front area 38 can, for example, be included in the map data, or a unit of the vehicle 1 can determine where the building fronts 38 are located based on the position of the buildings 32 included in the map data. Here, for example, edges, facade structures, or other features of buildings 32 can be detected as static objects. The building front area 38 can be positively linked to object categories in which features of the buildings 32 are detected; it can be negatively linked to object categories that, for example, affect other road users.Analogous to the procedure described above, it can be provided that only a subset of the recorded environmental data is analyzed for the detection of building edges and corners, which includes the building front area 38.

[0086] In 3D figureA road edge area 39 is highlighted, which is adjacent to, at a short distance from, or with a slight overlap with the roadway 31. For example, the posts 33 are located in this road edge area 39. Furthermore, the curb 34 can be found in this area. Accordingly, it can be provided that these objects are only detected in a corresponding subset of the environmental data with the road edge area 39. Other example objects can be pedestrians who are on a footpath in the road edge area 39; analogously, cyclists can be detected on a cycle path.By appropriately defining the areas, it can be ensured, for example, that markings on a roadway located in the road edge area 39 are not interpreted as relevant for the travel of the vehicle 1, while at the same time the lane marking 35 in the drivable area 37 is recognized as relevant.

[0087] In a further embodiment, vehicle 1 includes an interface to an external unit. Map data or supplementary data, as well as mapping rules, are acquired by the external unit. This can be configured as an external server or as backend devices. Furthermore, data can be transmitted from vehicle 1 to the external unit, for example, to enable an evaluation of map data when, for example, checking the plausibility of the map data based on the acquired environmental data or the environmental model generated by vehicle 1. List of reference symbols

[0088] 1Vehicle 2Position acquisition unit 3Map data acquisition unit 4Environmental data acquisition unit 5Detection unit 6Environment model determination unit 7Calculation unit 8Driver assistance system 21Acquisition of an initial position estimate 22Acquisition of map data 23Acquisition of environmental data 24Detection of objects 25Filtering detected objects 26Determination of an environment model, position optimization 27Output of the optimized position 30Ego-vehicle (symbol) 31Roadway 32Static object, building 33Static object, post 34Static object, curb 35Static object, lane marking 36Detection area 37Drivable area, roadway area 38Development area, building fronts 39Roadside area

Claims

1. Method for determining an environment model of a vehicle (1), in which method an initial position estimate for the vehicle (1) is acquired; map data are acquired, the map data comprising information about the spatial arrangement of geographical regions (37, 38, 39) and the geographical regions (37, 38, 39) being assigned to different region categories; environment data are acquired in an acquisition space (36); objects (32, 33, 34, 35) are detected based on the environment data, each detected object (32, 33, 34, 35) being assigned an object position and an object category; and the environment model is determined based on the detected objects (32, 33, 34, 35); assignment rules are provided which define an assignment of the object categories to the region categories; the detection of the objects (32, 33, 34, 35) taking place depending on the assignment rules, characterized in that a negative assignment is determined for a specific region category and a specific object category based on the assignment rules; and objects (32, 33, 34, 35) assigned to the specific object category are detected based on those environment data that do not relate to geographical regions (37, 38, 39) assigned to the negatively assigned specific region category.

2. Method according to claim 1, characterized in that the region categories comprise at least one drivable region (37), one non-drivable region, one roadway region (37), one developed region (38) and / or one vegetation region.

3. Method according to either claim 1 or claim 2, characterized in that a precision of the initial position estimate is determined and, depending on the precision, an arrangement of the geographical regions relative to the vehicle (1) is determined.

4. Method according to any of the preceding claims, characterized in that when detecting the objects (32, 33, 34, 35), preliminary detections are first determined and the preliminary detections are filtered based on the assignment rules.

5. Method according to any of the preceding claims, characterized in that when detecting the objects (32, 33, 34, 35), a subset of the acquired environment data are determined based on the assignment rules and the objects (32, 33, 34, 35) are detected based on the subset of the environment data.

6. Method according to any of the preceding claims, characterized in that the acquisition space (36) in which the environment data are detected is formed based on the assignment rules.

7. Method according to any of the preceding claims, characterized in that the detected objects (32, 33, 34, 35) are static objects (32, 33, 34, 35).

8. Method according to claim 7, characterized in that the object categories comprise posts (33), poles, traffic structures, elements of a peripheral development (32), road markings (35), signs or traffic signals.

9. Method according to any of the preceding claims, characterized in that an optimized position estimate for the vehicle (1) is determined based on the initial position estimate and the environment model.

10. Method according to any of the preceding claims, characterized in that the detected objects (32, 33, 34, 35) comprise dynamic objects.

11. Method according to any of the preceding claims, characterized in that the map data are at least partially received from an external unit.

12. System for determining an environment model of a vehicle (1), the system comprising a position acquisition unit (2) for acquiring an initial position estimate for the vehicle (1); a map data acquisition unit (3) for acquiring map data, the map data comprising information about the spatial arrangement of geographical regions (37, 38, 39) and the geographical regions (37, 38, 39) being assigned to different region categories; an environment data acquisition unit (4) for acquiring environment data in an acquisition space (36); a detection unit (5) for detecting objects (32, 33, 34, 35) based on the environment data and for assigning an object position and an object category to each detected object (32, 33, 34, 35); and an environment model determination unit (6) for determining the environment model based on the detected objects (32, 33, 34, 35); the detection unit (5) being designed to detect the objects (32, 33, 34, 35) depending on assignment rules; the assignment rules defining an assignment of the object categories to the region categories, characterized in that a negative assignment for a specific region category and a specific object category can be determined based on the assignment rules; and objects (32, 33, 34, 35) assigned to the specific object category can be detected based on those environment data that do not relate to geographical regions (37, 38, 39) assigned to the negatively assigned specific region category.

13. System according to claim 12, characterized in that the environment data acquisition unit (4) comprises a lidar, radar or ultrasonic sensor or a camera for visible or infrared light.

14. Computer program product having instructions which, when executed by means of a computer, cause the computer to carry out the method according to any of claims 1 to 11.