DE102016214030A1 - Method and system for detecting a traffic environment of a mobile unit - Google Patents

Abstract

The invention relates to a method for detecting a traffic environment of a mobile unit (1). In this case, data records with data points are recorded by means of a detection unit (3), wherein a data point is assigned to each data record. Landmark observations are respectively determined for the recorded data records, with landmarks parameters and object classes being assigned to the landmark observations, and environment model data are generated based on the determined landmark observations, the environmental model data comprising a description of the traffic environment. On the basis of the environment model data, distance calculations for landmark observations are carried out on the basis of their object class and their landmark parameters, whereby distance values for the landmark observations are determined. In the process, landmarks are assigned to certain landmarks on the basis of the distance values. The environment model data are assigned probabilities and the assigned probabilities are changed based on the landmark observations. The environmental model data is updated based on the probabilities associated with the landmark observations. Finally, the environment model data is output. The invention further relates to a system for detecting a traffic environment of a mobile unit (1).

Description

The present invention relates to a method and a system for detecting a traffic environment of a mobile unit, in particular of a vehicle.

For the increasing integration of automatic driving functions in modern vehicles, the exact location of the vehicle is an important prerequisite. It is not sufficient to know the position of the vehicle to a few meters exactly, but it may even minor deviations, such as along the width of a lane, have a significant impact on the ride.

In order to enable a sufficiently accurate and reliable localization of the vehicle, approaches for landmark-based positioning are pursued.

In the in the DE 10 2004 003 850 A1 described method for the detection of markings on a roadway are detected by a laser scanner in temporal sequence distance images of a detection area and the distance of a detected mark is estimated. By recognizing straight lines on the road markings are identified.

The DE 10 2007 020 791 A1 proposes a lane marking recognizing apparatus in which a center line and a lane width are extracted from data of a lidar instrument. A parameter tracking is used.

Of particular importance in such methods is the management of landmarks detected by the sensors of the mobile unit. In this case, several different sensors can be used to provide on the one hand by redundancy for a reliable detection and on the other hand to ensure as complete a detection by complementary sensor principles.

Particularly in such cases, there is a need for methods of processing the detected landmarks to reliably generate a comprehensive representation of the mobile unit's environment. It is the object of the present invention to provide such a method and system.

According to the invention, this object is achieved by a method having the features of claim 1 and a system having the features of claim 9. Advantageous embodiments and further developments emerge from the dependent claims.

In the method according to the invention, data records with data points are recorded by means of a detection unit, wherein a data point is assigned to each data record. In each case, landmark observations are determined for the recorded data records, the landmark observations being assigned landmark parameters and object classes. Based on the particular landmark observations, environmental model data is generated, where the environmental model data includes a description of the traffic environment, and distance modeling for landmark observations is performed based on the environmental model data based on its object class and landmark parameters, with distance values determined for the landmark observations. In the process, the landmark observations are assigned to specific landmarks on the basis of the distance values. The environment model data are assigned probabilities and the assigned probabilities are changed based on the landmark observations. The environment model data is updated based on the probabilities associated with the landmark observations and the environmental model data is output.

Thus, advantageously, a comprehensive representation of the traffic environment can be detected and provided for further processing. In particular, the method may be applied in the context of a localization of a mobile unit, such as a vehicle.

The inventive method operates in particular vector-based, that is, there is no comprehensive spatially resolved mapping of a space made, as is the case with point or grid-based method is provided, but it individual objects are parameterized detected and stored on the basis of this parameterization. As a result, the method can advantageously be operated quickly and with a comparatively low requirement for memory and computing power, which is advantageous, in particular, in mobile units, for example in vehicles.

The term "position" in the narrower sense comprises the coordinates of a point in a coordinate system, wherein a coordinate system relative to the mobile unit or a global coordinate system can be considered. In a broader sense, however, the term "position" should also include features that typically fall under the term "pose", in particular an orientation of the mobile unit in a particular direction in the coordinate system. In the following, to simplify the presentation of the "position" the speech, the other features of a "pose" should not be excluded.

Landmarks within the meaning of the invention are, inter alia, orientation points and topographical objects which are recognizable and classifiable and suitable for structuring a spatial area or relative to which a position in the sense of the invention can be defined. In particular, objects that can be mapped and parameterized as geometric shapes, for example based on their shape, position, area, length and width, can be classified. In particular, the landmarks include elements of the traffic infrastructure, such as lane markings, traffic signs, lighting systems, bridges or other traffic structures. Further, this includes other structures, such as houses, and plants, such as trees, and scenic features, such as soil surveys. Landmarks should moreover be recognizable, that is, their appearance and their further properties should remain substantially constant over a certain period of time, in particular so that recognizability based on the landmarks over a sufficiently long period is possible, if necessary, to take account of changes by means of an updating functionality can.

In particular, according to the invention, semantically interpretable elements of the road infrastructure are recorded as landmarks. These can also be grouped according to their geometry as well as the possibilities of processing and acquisition. In particular, a distinction can be made between point-based landmarks, such as piles, posts and pillars, which can be detected in particular by means of lidar, line-based landmarks, such as dashed lines, arrows and stop lines, which can be detected in particular by lidar or a camera, as well as polylinia-based landmarks, such solid lines that can be represented as polylines and for the detection of which in particular a camera can be used.

und absoluten Positionen (p wgs / 1, ..., p wgs / n) sowie modifizierte Parameter wie die Breite und Länge der (Poly-)Linie.To characterize a landmark observation, a state vector is generated, wherein the data stored in this state vector differ according to the type of landmark: The state vector s ^{PO of} a point-based landmark observation comprises a position p ^vrf in the vehicle ^{coordinate system} , an uncertainty Σ _p and an absolute position p ^wgs in a global coordinate system. The points are given as triples p = (x, y, z). Further, a diameter and a height are defined and a time stamp for the observation as well as an identification of the observing sensor are stored. Furthermore, fields for a probability value and an identification of a particular landmark are provided, which are used in the further course of the method for further processing. For the state vector s ^{DL of} a line-based landmark observation or the state vector s ^{PL of} a polyline-based landmark observation are two or more positions (p vrf / 1, ..., p vrf / n) provided as well as correspondingly many uncertainties

and absolute positions (p wgs / 1, ..., p wgs / n) and modified parameters such as the width and length of the (poly) line.

Alternatively or additionally, further parameterizations of a landmark observation, in particular for other types of landmarks, may be provided.

The "traffic environment" of the mobile unit in the sense of the invention comprises a spatial environment of a position of the mobile unit, in particular with regard to the properties relevant for traffic in this environment. In particular, according to the invention, the surroundings of a road user, for example on a roadway, can be considered. The spatial area is essentially determined by the range with which data in the environment can be recorded and the environment in which landmarks can be used to locate the mobile unit.

A "description of the traffic environment" includes in particular information about landmarks in the traffic environment. It may include positions and characteristics of the landmarks.

The data sets are recorded in a first step of the method in a manner known per se. In particular, sensors with different principles of action can be used, for example lidar, radar or ultrasonic sensors, cameras in the field of visible light or infrared cameras and time-of-flight cameras. Furthermore, data may be collected from external devices, in particular from other vehicles or other off-board devices, in particular methods from the field of "Car 2 X" communication can be used.

A recorded data record is detected and / or output by a sensor, in particular at a specific time, such as a single scan of a laser scanner or an image captured by a camera. In this case, a data record may include the data points of a detection at a particular time or data points of several measurements may be accumulated.

Depending on the type of sensor used, various features of the traffic environment can be detected, in particular, a spatially resolved detection takes place. For example, properties of reflection, scattering, transmission or absorption of electromagnetic radiation as well as distances can be detected. The data points correspond to individual measurements, in particular at different locations of the detection range of a sensor.

In a special case, the data sets can be recorded by a laser scanner. A recorded data set then comprises in particular a single scan of the laser scanner, wherein parameters such as distances and reflection properties at points in the traffic environment can be detected by pivoting a light beam and detecting the reflected light within a certain detection range. The data points correspond to individual measurements, especially at different angles relative to the laser scanner.

In this case, the acquisition of the data sets by the laser scanner can be performed periodically. Furthermore, the method can also be carried out periodically. As a result, current data can advantageously be provided continuously in order to be able to record the landmarks in the traffic environment as accurately as possible at any desired time. Further processing of the acquired landmarks allows for ongoing refinement and tracking of landmark-based location based on current data.

In other embodiments, the detection may also be done in other ways, such as asynchronously.

Furthermore, it can be provided that data is acquired by further sensors. In particular, data relating to movement and guidance of the vehicle can be detected in a vehicle. For example, a speed and acceleration can be detected as well as a direction of movement and a steering angle.

In addition, provision can be made for data to be recorded by further devices, wherein in particular a data-technical connection can be used. By way of example, a data connection of the mobile unit to other units can exist and data acquired by these can be used. In this case, the data connection can be provided in a different, known manner and data can be transmitted in a likewise known manner, whereby, for example, a method for retrieving specific data from a remote unit can be used. For example, data collected by other vehicles may be transmitted to a vehicle, or other devices may be used in an analogous manner.

In particular, the captured raw data can be filtered, for example to limit the detected landmarks to features of the roadway, such as lane markings, which are particularly important landmarks for the localization of vehicles.

The collected data can be combined to form output data for further processing. In particular, a single data record is used here in a special case. Furthermore, an accumulation can be performed. The suitability of a specific number of data points and / or data sets for further processing as well as the requirement for accumulation also results from the properties of the recorded data records, for example the number of data points of a single data record. In this case, the output data can be stored in a ring memory in a manner known per se, with the respectively last recorded data being stored while the oldest existing data are deleted. In this way, the data of a specific, immediately past period can be stored together.

In a further step, a segmentation based on the output data can now take place. In the process, data points belonging to a landmark are combined into "segments". For each potential landmark, therefore, a subset of the output data is defined for the associated data points.

In a further development, the segmentation is carried out by means of an Euclidean Cluster Extraction Algorithm, wherein data points whose distance from each other is less than a predetermined threshold value are each assigned to a segment. This advantageously allows a simple and fast determination of the segments.

Alternatively, or in addition, other algorithms may be used, such as a Random Sample Consensus (RANSAC) algorithm, where various segmentation techniques may be implemented.

A principal component analysis (PCA) is then followed by the analysis of the segments. In this case, parameters are determined for each segment and stored as "landmark parameters". The landmark parameters are particularly suitable for defining the segment as comprehensively as possible. For example, a geometric shape describing the distribution of the data points of a segment can be determined.

The principal component analysis may include determining a proximity ellipse for the segment to determine the landmark parameters for each segment. The geometric shape of the approximate ellipse advantageously allows a variety of other geometric shapes to be approximated easily and with high accuracy.

The parameters of the approximation ellipse allow conclusions to be drawn about the actual dimensions and parameters of the geometric shape. In particular, points and lines can be approximated particularly well by the approximation ellipse. Furthermore, the PCA allows a particularly fast and robust analysis of the segments. In particular, the relevant parameters of the approximate ellipse are determined jointly in one step, so that the number of calculation steps and the computational effort for detecting the landmarks are substantially reduced compared to other known methods. In particular, two- or multi-dimensional data can be easily processed and a suitable parameterization can be determined. For example, it can be determined particularly quickly and easily what length and width an elongate object, such as a line, has.

In an embodiment, the landmark parameters comprise at least one relative position relative to the detection unit. This advantageously makes a position determination for the detected object relative to a detection coordinate system, in particular relative to a vehicle coordinate system.

In particular, this position can be determined together with the other landmark parameters in the context of the main component analysis. For example, Cartesian coordinates can be used or a position can be determined by distance and angle.

The origin of the coordinate system can be assumed in particular in the case of the mobile unit, so that first of all a position determination takes place relative thereto. In a further step, a transformation into another coordinate system, in particular a global coordinate system, can take place.

In a further embodiment, filtering is carried out on the basis of the landmark parameters of the segments. As a result, the segments considered as real landmarks are advantageously selected and the further calculations can be limited to these.

For example, limits can be used for filtering to verify the plausibility of detected objects to correspond to actual landmarks and to avoid artifacts of detection. In the case of point-based landmarks, an upper and / or lower limit for the detected radius can be checked. Furthermore, upper and / or lower limits for the detected length and / or width may be used for line-based and polyline-based landmarks.

In addition, legal regulations may be used that provide specific dimensions for certain traffic infrastructure landmarks, in particular lane markings. For example, a minimum and / or maximum width and / or length of lines may be predetermined, which are accepted as potential lane markings. Furthermore, certain shapes can be recognized and their plausibility can be assessed for recognition as landmarks.

The information on individual landmarks collected and determined on the basis of a data record is combined into a "landmark survey". In this way the data can be structured. According to the invention, a landmark observation comprises all information about a particular sighting of a particular landmark within a set of initial data. Thus, if the method is used repeatedly, a series of landmark observations for the same real landmark can be determined if the landmark is viewed several times, in particular also by a plurality of sensors and detection units. In particular, an assignment of the landmark observation to the corresponding landmark can take place, for example by means of an identification number, for example in order to be able to compare several landmark observations of the same landmark with one another.

In a further embodiment, the landmark observations are associated with elements of a road infrastructure. As a result, the data can advantageously be processed so that an assignment of the landmark observations to certain landmarks can be made.

For example, a match with landmarks that have been detected by means of further data records can also be determined in this way, and further processing is made possible. In particular, a position determination can subsequently be carried out, wherein different methods for landmark-based localization can be used.

In a further step, object classes are determined on the basis of the landmark parameters. The classification is based on geometrical basic forms that allow a basic assignment of the segments to different landmarks.

In a further development, the landmark observations are classified as point-based, line-based or polyline-based objects on the basis of the object classes assigned to them. As a result, advantageously different landmarks, in particular lane markings and / or other traffic signs such as piles, can be well recognized and parameterized.

Punk, line or polyline-based landmarks allow a particularly convenient classification of landmarks, especially road markings. The classification is carried out depending on the required for the complete characterization of a landmark landmark parameter, in particular the number of positions to be determined in order to detect the course of a lane marking.

The classification step may already be performed for the particular segments. As already described above, different parameters can be determined and stored depending on the object class. Thus, a particularly efficient data management is possible.

According to one embodiment, the particular landmark parameters for segments of the line-based or polyline-based feature class include a length parameter and a width parameter, and a start position and an end position.

This advantageously provides the most relevant parameters for describing these classes of objects as well as their variances for further analysis.

Optionally, a simplification of polyline-based objects may be made by the Ramer-Douglas-Peucker algorithm or a similar suitable method.

In the method according to the invention, the landmark observations are processed further. For this purpose, the data can be transmitted, for example, to a unit for landmark-based localization, so there is not necessarily a perceptible for a user output, such as in the form of a display.

In the case of the method according to the invention, surrounding model data are determined on the basis of the determined landmark observations, the environmental model data comprising a description of the traffic environment of the mobile unit. By means of the environment model data, in this way data is stored that is particularly relevant for a landmark-based localization of the mobile unit.

Specifically, the particular environmental model data includes data about landmarks and positions associated with the landmarks. Thereby, the environment model data may be determined to represent a comprehensive description of the landmarks in the mobile unit traffic environment and to enable landmark based localization on its basis.

In particular, the environment model data includes information about the landmarks in the mobile unit's traffic environment. In this case, object classes assigned to the landmarks can be determined and stored, and the landmark parameters with data about the positions of the landmarks can be included. The requirement that the generated environmental model should be consistent concerns primarily those landmarks for which there are several landmark observations from several sightings: these are fused and tracked over a certain period of time, whereby so-called "tracks" are generated. The tracks represent observations associated with a landmark and their tracing based on the positions determined during the observations. The tracks of the environmental model data are used to manage the data collected about a landmark in the traffic environment in such a way that the landmarks are finally represented as precisely as possible by the environmental model data.

In particular, since the data sets can be captured by different sensors, they may be collected at different times. For example, by a sensor, the data acquisition can be done periodically in a certain clock, with another sensor in another clock or offset in time can capture data. Accordingly, both the data sets and the data of the landmark observations, which are determined on the basis of the data sets, are assigned to different times. In this case, the different points in time, and in particular a movement in the meantime, should be taken into account. For example, a shift of a coordinate system origin of the mobile unit between the times of the acquisition of different data sets can be taken into account.

In order to be able to process the landmark observations on the basis of different data sets, possibly of different sensors, a tracking is carried out in one step, whereby a landmark from different data sets can be recognized. In this way, landmark observations can be confirmed as belonging to a real landmark if they are recognized in multiple datasets.

Therefore, based on the environmental model data, distance calculations for landmark observations are carried out on the basis of their object class and their landmark parameters. The landmark observations are assigned to specific landmarks based on the distance values. As a result, advantageously corresponding landmark observations can be determined very simply and quickly.

In particular, this step serves the fusion of newly determined landmark observations with the environmental model data. For example, a landmark in the traffic environment has been observed earlier and is encompassed by the environmental model data. In this case, the environment model data already includes a "track" for this landmark. If a further landmark observation is added, it must be determined whether the new landmark observation corresponds to an already existing track of the surrounding model data or whether a new landmark has been detected here for the first time. The assignment is made on the basis of the determined distance values, it is thus assumed that landmark observations, which were acquired at substantially the same location or sufficiently close to each other, are to be assigned to each other and belong to the same actual landmark.

The positions of landmarks represented by the tracks of the environmental model data are compared based on the landmark parameters of various landmark observations, and it is determined by distance calculations which detected landmarks belong together. In particular, the distance calculations are performed on the basis of the landmarks and landmark observations associated object classes. This takes particular account of the fact that different shapes of landmarks require different methods of distance calculation. For example, point-based landmarks a single position can be assigned, the geometric distance of such landmarks can be clearly determined by the respective positions. The computation is complicated for line- and polyline-based landmarks by the fact that they can not be assigned a single, unique position value that could be compared to another position value. Instead, several positions for such landmarks are characteristic, such as start and end positions and intermediate points.

In a further embodiment, a Mahalanobis distance is determined in the distance calculations for point-based objects, for line-based objects a modified Mahalanobis distance based on the start and end positions and / or for polyline-based objects an area enclosed by the polyline-based objects. As a result, suitable distance metrics for different object classes are advantageously provided.

The method for determining the Mahalanobis distance for point-based landmarks is known per se. This equation is modified for line-based landmarks, each having a start and an end position, so that the four possible combinations for the distance measurement are considered together (start position 1 - start position 2, end position 1 - start position 2, start position 1 - end position 2, end position 1 - end position 2). As a measure of the distance between polyline-based landmarks (polylines), an area enclosed by them is used. Alternatively or additionally, other suitable methods may be used by which distance metrics may be determined.

In particular, the data for landmark observations are evaluated for each real landmark, that is, for each detected in the traffic environment landmark the data of the associated landmark observations are stored and the data of various landmark observations can be processed collectively.

In order to further determine whether a landmark observation corresponds to a real landmark, the registered landmarks of the surrounding model can be assigned probabilities indicating with which security a real landmark was detected. By comparing the data of different landmark observations associated with the same landmark, the value of the probability can be changed.

In a further development, the probabilities assigned to the environment model data are modified by means of a sequential probability test, wherein the probabilities are used to check whether a landmark observation corresponds to an actual detection of a landmark of the traffic environment. The probabilities assigned to the landmarks of the environment model, ie the tracks, are thus calculated by means of a Sequential Probability Ratio Test. Thus, advantageously, a statistical test can be applied, which allows a particularly reliable determination of the probabilities in a variable database.

In one embodiment, a probability of landmark observation is increased if it is assigned another landmark observation, and the probability of landmark observation is lowered if it is not confirmed by another landmark observation. Thus, the probability can advantageously be determined more and more accurately in an iterative process.

In a further refinement, the landmark observations of the surrounding model are accepted or rejected by means of a comparison of their associated probabilities with a threshold value. As a result, it can advantageously be tested in an iterative procedure whether a decision for or against the acceptance of a detected landmark can be assumed to be actually present. This allows a particularly simple and reliable identification and deletion of misdetections.

In a further development, the environment model data are assigned landmarks and include positions of the landmarks. The environment model data with the confirmed landmarks of the tracks can thus be advantageously used to represent a description of the traffic environment.

In this case, the environment model data can correspond approximately to a map of the traffic environment of the mobile unit. In particular, a consistent environment model is generated, that is to say that the environment model data are generated on the basis of the data acquired by a plurality of data records in such a way that the recorded data records are synchronized and spatially coincident relative to each other.

The environment model data are subsequently output according to the method according to the invention, wherein in particular a further processing of the data for a landmark-based localization can take place. For this purpose, the environment model data can be transmitted, for example, to a further unit for evaluation or they can be made available to a program module for processing, for example.

In the case of a landmark-based localization of the mobile unit by means of the environmental model data determined according to the invention, an estimated position of the mobile unit can also be detected and based on the acquired estimation position, reference map data can be retrieved, the reference map data including reference positions of landmarks. On the basis of the environment model data and the reference map data, the position of the mobile unit can now be determined by means of a particle filter-based method. Furthermore, alternatively or additionally, the position determination can be carried out using other methods, for example by means of a Kalman filter. This can advantageously be localized using reference card data.

The reference map data is retrieved in response to an estimated position of the mobile unit. The estimation position is detected by methods known per se, such as by means of a satellite positioning system such as the Global Positioning System (GPS). The retrieval itself likewise takes place in a manner known per se, in particular either by a storage device of the mobile unit or by means of a data-technical connection to an external unit.

The acquisition of the estimation position and the retrieval of the reference map data can be done periodically. As a result, it can advantageously be ensured that constantly matching reference card data is available and can be used for the localization. For example, polling may occur every 10 seconds or at a different specified time interval. Further, similarly, the fetch may be regular based on other criteria, for example, a distance from the position of the last fetch, so that the fiduciary card data is fetched whenever the position of the mobile unit has changed by a certain amount.

In the particulate filter based method, a plurality of statistically distributed hypotheses about the actual position of the mobile unit are generated and these hypotheses can be represented as individual particles. The particles are assigned probabilities corresponding to the true position of the mobile unit, and the probabilities are used to determine the hypothesis that can best be assumed to be the true position of the mobile unit. This position can then be output, for example by transmitting to another unit, such as a driver assistance system or a navigation system.

The system according to the invention of the type mentioned above comprises a detection unit, can be detected by the data sets with data points, each record is a time can be assigned. It also comprises a landmark detection unit, by the respective landmark records observations can be determined for the collected records, the landmark observations and landmarks parameters and object classes are assigned. It also comprises an arithmetic unit by means of which environmental model data can be generated on the basis of the determined landmark observations, the environmental model data comprising a description of the traffic environment, and a tracking unit by means of which distance calculations for landmark observations can be carried out on the basis of the environmental model data based on their object class and landmark parameters Landmark observations are determinable. In this case, the environment model data probabilities are attributable to the arithmetic unit and the associated probabilities are changeable based on the landmark observations; wherein further the environmental model data can be updated on the basis of the probabilities associated with the landmark observations. The environment model data can finally be output.

The system according to the invention is in particular designed to implement the method according to the invention described above. The device thus has the same advantages as the method according to the invention.

Furthermore, it can be provided that map data can be generated and output on the basis of the environment model data. In this case, the map data include information about the landmarks, in particular their positions. It can also be provided that the map data can be transmitted to, for example, an external server and / or further devices.

In one embodiment, the detection unit comprises at least two different sensor types. As a result, data is advantageously acquired in complementary ways and redundancy is created by multiple detection, making the process more robust. Different types of sensors in particular use different detection principles. For example, a camera, a laser scanner, a RADAR detector and / or another sensor may be provided.

Further, the system may include other sensors. In particular, data relating to a movement of the mobile unit, in particular of a vehicle, can be detected. For example, a Speed and acceleration of a vehicle are recorded as well as a direction of movement and a steering angle.

In a development of the system, the environment model data can be transferred to a position determination unit, in particular to a unit for landmark-based localization. This advantageously allows the captured and determined data to be used for localization. In particular, a landmark-based localization of a vehicle can be performed.

Alternatively or additionally, the environment model data can also be transmitted to other devices, such as driver assistance systems, which can support or independently execute functions of the control of the vehicle by means of the detected landmarks, or remote devices, such as other vehicles or vehicle-external devices.

The invention will now be explained by means of embodiments with reference to the drawings.

1A and 1B show an embodiment of the system according to the invention,

2 shows an embodiment of a landmark-based localization,

3 shows an embodiment of a detection of landmarks,

4A and 4B show a generation of a consistent environment model,

5 shows performing a particle filter based localization and

6 shows an embodiment of an update functionality.

Related to the 1A and 1B an embodiment of the system according to the invention will be described.

In the embodiment is as a mobile unit 1 a vehicle 1 intended.

The vehicle 1 comprises a detection unit 3 , in turn, a laser scanner 3a , a camera 3b and a GPS sensor 3c includes. The registration unit 3 is with a control unit 4 coupled. Furthermore, with the control unit 3 coupled: One arithmetic unit 6 , a driver assistance system 12 and an interface 5 , The arithmetic unit 6 includes an accumulation unit 7 , a segmentation unit 8th , a landmark recognition unit 9 and a tracking unit 10 , The control unit further comprises an output unit 11 ,

the interface 5 is in the embodiment data-technically connected to an external server 2 , The data connection is through the interface 5 produced in a known per se and takes place wirelessly in the embodiment via the Internet. The connection is made via a telephone network. In further embodiments, the connection is made in other ways, such as indirectly via a mobile telephone connected to the Internet, which itself connects to the external server 2 can produce.

The registration unit 3 may alternatively or additionally include further sensors in other embodiments, in particular use other detection principles. For example, lidar, radar or ultrasonic sensors, cameras in the visible light or infrared cameras and time-of-flight or stereo cameras may be provided.

The provided in the embodiment GPS sensor 3c is formed in a manner known per se and enables the detection of a position by means of a navigation satellite system such as the Global Positioning System (GPS). Alternatively or additionally, other positioning systems may be used to determine the position of the vehicle 1 make.

Furthermore, it is provided in the exemplary embodiment that data from further sensors of the vehicle 1 be recorded. It collects data showing the movement and guidance of the vehicle 1 affect. In particular, the speed and acceleration are detected and the direction of movement and the steering angle.

The driver assistance system 12 is a device of the vehicle 1 which the driver while driving the vehicle 1 supported. The driver assistance system 12 in the embodiment can act as a pure information system which supports the driver, but it can also fully or partially automatically the locomotion of the vehicle 1 influence, in which control functions are intervened.

Without the driver assistance system 12 the driver directly influences the movement of the vehicle 1 , At most, signals or movements of driver-operated controls, such as the pedals, the gear lever or the steering wheel, to corresponding devices of the vehicle 1 transmit the movement of the vehicle 1 influence. Such locomotion of the vehicle 1 corresponds to the lowest level of automation. With a higher degree of automation is partly intervened automatically in facilities, which is the movement of the vehicle 1 serve. For example, in the steering of the vehicle 1 or the acceleration intervened in a positive or negative direction. At a still higher degree of automation will be so far in vehicle facilities 1 intervened that certain types of movement of the vehicle 1 , for example, a straight ahead, can be performed automatically. At the highest level of automation, a given route can be driven essentially automatically.

In this case, the operation of the proposed in the embodiment driver assistance system 12 the most accurate knowledge of the position of the vehicle 1 provided. In particular, the knowledge of a precise position on the roadway is needed, for example, a guide of the vehicle 1 to allow in the transverse direction and / or longitudinal direction and to be able to drive about a certain lane or trajectory automatically.

In 1B are some elements of a typical traffic environment of the vehicle 1 shown. Of course, the drawings are to be understood as schematic sketches in which the scales and size ratios used are of no significance.

The traffic environment is the spatial environment of the position of the vehicle 1 especially the characteristics relevant to traffic in this environment. The traffic environment can therefore, for example, other road users in the vicinity of the vehicle 1 include (in 1B these have not been shown for the sake of simplicity), and lane markings and traffic signs, infrastructural facilities and other features may be included, such as structures, vegetation or topographical features in the environment. The spatial area is essentially determined by the range with which data in the environment can be recorded and in which environment landmarks can be used for localization.

The vehicle 1 is on a lane 18a a roadway 18 , which also has a long lane 18b includes. Between the lanes 18a . 18b runs a guideline 16 which includes several sub-lines and to which there is a lane boundary 15 in the form of a solid line connects. The roadway 18 is also at the outer edges by lane boundaries 14 limited. At the edge of the road 18 are also guide posts 13 arranged.

Further, there is a detection area 17 the registration unit 3 of the vehicle 1 indicated. In the embodiment, in particular, the laser scanner 3a and the camera 3b the registration unit 3 of the vehicle 1 set up so that they are in the detection area 17 Data about the traffic environment of the vehicle 1 to capture. In this case, different sensors of the detection unit 3 have different detection ranges.

In reference to 2 For example, one embodiment of landmark based localization will be described as a method for determining a position of a vehicle 1 explained. It is of the reference to the 1A and 1B explained system gone out.

In a first step S1a to S1n, data records are read by a number of n sensors of the detection unit 3 of the vehicle 1 recorded, so for example by the laser scanner 3a and the camera 3b , It is in 2 in each case a separate step S1a to S1n provided for a number of n sensors.

In further embodiments, alternatively or additionally, data records can be detected by sensors of other vehicles and / or devices. It can then data by means of a data connection to the vehicle 1 be transferred, in particular directly from the vehicles and / or facilities that have collected the data, or indirectly via the server 2 , The processing can be done in this case both outside the vehicle 1 done by the server 2 or the other vehicles and / or facilities, or the records may be attached to the vehicle 1 dominate and continue to be processed there.

The acquired data are evaluated in a further step S2a to S2n, where landmarks are detected and landmark observations are generated. Again this is in 2 in each case a separate step S2a to S2n provided for all n sensors.

The "landmark observations" include all information about a particular sighting of a particular landmark within the data acquired in steps S1a-S1n. The landmark observations are each assigned time points, in particular this is the time of recording the corresponding data records. Further, the landmark observations include relative positions relative to the vehicle 1 as well as parameters which characterize the state of the landmark detected during a sighting, such as the extent of the detected landmark.

The part of the process performed in steps S1a to S1n and S2a to S2n is described below with reference to FIG 3 explained in more detail.

In a further step S3, a tracking and a fusion of the previously acquired data is performed. This data is generated from a consistent environment model, that is, the information about the traffic environment of the vehicle 1 are merged so that the data relevant to landmark based localization is available. In particular, the environment model includes data from the landmark observations, such as data about landmarks in the traffic environment and the positions of the landmarks.

The generation of a consistent environment model executed in step S3 is described below with reference to FIG 4 explained in more detail.

In a step S4, an estimated position of the vehicle 1 detected. This is in the embodiment by means of the GPS sensor 3c the registration unit 3 executed, in other embodiments, alternatively or additionally, other methods for determining the estimated position can be used. It is assumed that the accuracy with which the estimated position of the vehicle 1 can be determined for the operation of the driver assistance system 12 not enough. For example, the estimated position can be determined to within 1 m, whereas a guidance of the vehicle 1 along a trajectory requires a much higher accuracy.

The estimation position is in the embodiment in a step S6 by means of the interface 5 to the external server 2 and this provides in the sense of an "electronic horizon" reference card data and transmits it to the vehicle 1 , The step S6 is carried out in particular in the context of a retrieval of the reference card data from the external server 3 , The reference map data includes landmark data in the traffic environment of the vehicle 1 , The estimation position can be used to determine which for the traffic environment of the vehicle 1 relevant data should include the reference card data.

In particular, the reference map data and the environment model data are structured such that they can be compared with one another, for example, in that positions of landmarks in the respective data can be compared with one another.

In further embodiments, it is provided that the reference map data alternatively or additionally from a memory unit of the vehicle 1 be retrieved. In this case, there is a data connection to the external server 2 not required and the method may be operated, for example, under the conditions of incomplete network connection.

In a further step S5 a self-motion estimation is performed. In particular, data from vehicle sensors are used, such as the speed, acceleration, inclination, direction, steering angle and more for the movement of the vehicle 1 relevant data. Based on this data can be determined how the vehicle 1 moves, in particular, is determined with higher accuracy, at which position the vehicle 1 located at the time in which a record using the detection unit 3 is detected. In particular, this can change the position of the vehicle 1 since the last detection of an estimation position.

The estimation position and self-motion estimation determined in steps S4 and S5 are considered in addition to the data of the landmark observations for the creation of the consistent environment image by tracking and fusion in step S3.

Based on the environment image and the reference map data, a particle filter-based localization is performed in a step S7. Statistical hypotheses for the position are determined on the basis of the data of the environment model and the reference map data, whereby the hypotheses are represented as "particles".

By means of a weighting method, probabilities of the hypotheses are determined on the basis of the reference map data and the most probable position is determined. This is the position of the vehicle determined by the method 1 ,

Further details of the method used in step S7 are given below with reference to 5 explained in more detail.

In a further step S8, an output of the position, wherein in the embodiment, the specific position by means of the output unit 11 to the driver assistance system 12 is transmitted. Alternatively or additionally, the position in this step can be transmitted to different units of the vehicle or vehicle-external units.

In reference to 3 an embodiment of a detection of landmarks will be explained. It is of the reference to the 1A and 1B explained system gone out. In particular, the steps described below may be considered as part of the above with reference to 2 explained landmark-based localization. The method explained here for a sensor can be applied in particular to the data detected by the sensors of the system.

In a first step S1a, one of the sensors of the detection unit 3 of the vehicle 1 Recorded records. In particular, the special case is considered that data through the laser scanner 3a but the method can be applied to other sensors with appropriate modifications.

The acquired data sets comprise data points, wherein the individual data sets each comprise the data points that occur during a period of pivoting of the laser beam of the laser scanner 3a be recorded. In particular, a distance value is acquired for each data point, as well as data for reflection of the surface from which the laser beam is reflected. The acquisition of the data takes place periodically by regularly pivoting the laser beam over a certain angular range.

Optionally, filtering may be performed in this first step S1a. In particular, a suitable laser scanner 3a Already provide an integrated classification of the data points that includes information about whether a surface point on the roadway is at a data point 18 was detected.

For example, all other data points other than those on the road can be neglected if lane markers are to be detected as landmarks.

In a further step S21a an accumulation is carried out. In a special case, the data points of a single data set, in this case a scan of the laser scanner, can be used 3a , already form a sufficient database, so that an accumulation of the data points of a data set is made. In other cases, the data points of multiple collected records may be accumulated. The accumulated data form the output data, on the basis of which further processing is then carried out.

In particular, the data points can be stored in a ring memory, in which the newly acquired data points are always stored, while at the same time the older data points are deleted. This means that the most up-to-date data for a particular period or a specific volume of data is always made available.

In a further step S22a, a segmentation is carried out on the basis of the output data. In this case, such data points are each assigned to a segment that appears to belong together. In particular, an Euclidean Cluster Extraction Algorithm can be used for such points be assigned to a common segment, which have less than a certain maximum distance from each other. Alternatively or additionally, other methods of segmentation may be used.

In a subsequent step S23a, the segments are analyzed by means of a principal component analysis (PCA) and an approximate ellipse is determined for each of the segments. Landmark parameters are determined for each of the segments, which uniquely define the particular approximate ellipse in particular. In particular, the landmark parameters include a direction parameter, a length parameter, a width parameter, and a start and an end position. Furthermore, the variances for the respective values can be determined. In this case, the landmark parameters can be determined on the basis of the PCA, in particular in one step.

In a step S24a, a classification of the segments is made on the basis of the determined landmark parameters. In particular, three classes are provided, namely point-based, line-based and polyline-based landmarks. Point-based landmarks are characterized by a certain extent and position.

Line-based landmarks have a starting and ending position, while polylinear-based landmarks have additional intermediate positions.

For example, solid lines can be recognized as polyline-based landmarks because neither a clear start point nor a clear end point is detected for them. In particular, such lines can be characterized as solid lines extending across the detection area 17 the registration unit 2 out and are not completely covered by the data points.

In particular, positions are defined for the landmark parameters. That is, for an approximate ellipse determined by PCA, at least one position is determined that is initially relative to the detection unit 3 or to the vehicle 1 is specified.

On the basis of the specific landmark parameters, a filtering can be carried out, which in particular includes a plausibility check. For example, it checks whether the length and width of detected line-based landmarks complies with legal requirements for the dimensions of road markings so formed. In particular, the landmark observations are each assigned elements of a road infrastructure, that is, parts or whole elements of the in 1B presented guideline 16 , the road limit 14 , the lane boundary or the delineator 13 ,

Furthermore, it can be determined on the basis of certain variances whether a landmark, for example a lane marking, has been detected with sufficient certainty. Additional filtering steps may be used to eliminate badly or misidentified landmarks.

Finally, in a step S25a, landmark observations are generated, that is, the information determined for the acquired data points is summarized in such a way that they include all data relevant for further processing for a landmark-based localization. The individual landmark observations are each assigned to a landmark, that is, a landmark observation comprises data that were determined during a sighting of the landmark and therefore can be assigned to the detected state of a landmark at the time of detection.

der Zustandsvektor s^DL einer linienbasierten Landmarkenbeobachtung als

der Zustandsvektor s^PL einer polylinienbasierten Landmarkenbeobachtung als

The generated landmark observations indicate, in particular, a state vector of a landmark, the data for which differ depending on the type of landmark: The state vector s ^{PO of} a landmark based landmark observation is specified as

the state vector s ^{DL of} a line-based landmark observation as

the state vector s ^{PL of} a polyline-based landmark observation as

angegeben. Der Zeitpunkt der Wahrnehmung ist im Feld „Zeitstempel” angegeben, ebenso die Identifikation des Sensors als „Sensor^ID” und die Identifikation der Landmarkenbeobachtung als „ID”. Das Feld „Belief” wird bei der weiteren Verarbeitung der Landmarkenbeobachtungen für statistische Einträge einer Wahrscheinlichkeit benötigt. Die Charakterisierung der erfassten Landmarken wird durch die Felder „Durchmesser”, „Höhe”, „Länge” und „Breite” angegeben.Designate p vrf / n Coordinates in the vehicle coordinate system and p wgs / n Coordinates in the global coordinate system, where each point is given as a triple p = (x, y, z). Furthermore, uncertainties

specified. The time of the perception is indicated in the field "Timestamp", as well as the identification of the sensor as "Sensor ^ID " and the identification of the landmark observation as "ID". The field "Belief" is needed in the further processing of the landmark observations for statistical entries of a probability. The characterization of the detected landmarks is indicated by the fields "diameter", "height", "length" and "width".

Alternatively or additionally, the state vectors may comprise other elements, for example an area or a reflectance.

The landmark observations generated in step S25a may be output and transmitted to a module or unit for further processing. In particular, so that with respect to 2 described step S3 are performed. In addition, landmark observations of the n sensors can be combined, for example in a specific data structure, wherein in particular the state vectors described above can be used.

Related to the 4A and 4B a generation of a consistent environment model is explained. It is of the reference to the 1A and 1B explained system gone out. Especially can follow the steps described below as part of the above with reference to 2 explained landmark-based localization. In the landmark-based localization, a seonsor- and landmark-specific preprocessing of the raw data is carried out in particular in a first layer, and in a second layer a consistent environmental model is built up by means of a tracking and fusion method. Finally, in a third layer, a map is constructed based on the environment model.

The described method steps can be summarized in particular in a tracking and fusion module, this being done in particular by the tracking unit 10 is performed. The recorded landmark observations are merged and stabilized into a consistent landmark environment model, referred to below as the "environmental model". The problems to be solved relate in particular to the fact that the data sets at different times and therefore during a movement of the vehicle 1 can also be recorded at different positions. In addition, misdetections should be detected and removed.

First, landmark observations are received, each at specific times and relative positions relative to the vehicle 1 assigned. This is done in particular with reference to the above 3 described method. This preparatory step describes a sensor-based layer of the process.

Based on the estimation position (see 2 , Step S4) and the self-motion estimation (see 2 , Step S5) are assigned to the landmark observations estimated absolute positions. So it can be during the continuous movement of the vehicle 1 a coordinate system tracked and the relative coordinate system of the vehicle 1 can be transformed into a global coordinate system by transformation.

In a first step S31, a temporal synchronization is performed. This is in 4B shown in detail. Parallel to a time axis (t) are axes for each sensor 3a . 3b . 3n the registration unit 3 shown. Furthermore, an axis (T) for the so-called tracks is shown, to which the landmark observations are merged at specific times. A track represents the data collected, merged and synchronized to a real landmark.

At certain times u _t , which are shown as dashed lines, data about the self-motion of the vehicle 1 receive. In each case amounts of landmark observations B n / t in particular, which are output as a data object from one of the n sensors. The landmark observations B n / t are shown here as bold vertical lines. The landmark observations B n / t are not detected synchronously with each other and with the vehicle's intrinsic motion data, so the movement must be balanced during the time difference between two times to compare the data. This is indicated by horizontal dashed arrows. The landmark observations B n / t are buffered until data is available for a new move step. Then the landmarks observations B n / t for every sensor 3a . 3b . 3n synchronized for the times u _t . Curved arrows pointing downwards make it clear that these synchronized data are merged into tracks T _t . Tracks created in this way are tracked with each movement step. So that is the landmark observations of Bn / t-1 and T _t-1 in the coordinate system at time t of the moving state u _t .

In a further step S32, distance calculations for landmark observations and landmarks of the tracks are made in order to obtain corresponding landmark observations of the landmarks B n / t to assign to the tracks T _t . At first, point, line and polyline-based landmarks can be grouped, since only they can correspond with each other.

By means of known methods, a gating is performed, with only landmark observations B n / t in a given environment of the respective tracks T _t are eligible for association.

einer Landmarkenbeobachtung

und einem Punkt

eines Tracks

berechnet werden. Dabei ist ein Vektor zwischen

und

gegeben als

. Es wird eine Berechnung der Mahalanobisdistanz durchgeführt anhand der Gleichung

wobei S^–1 die inverse Kovarianzmatrix der Track-Landmarke

bezeichnet.It is now determined whether a landmark observation B n / t Assign a specific track T _t is. This is determined by the spatial proximity between the two points. In order to calculate the distances, the following methods are used in the exemplary embodiment:
For point-based landmarks, the state vector s ^{PO of} a point-based landmark is considered, in which the position at time t is given as point p = (x, y, z) in the vehicle coordinate system vrf. We now want a distance function d ^PO between a point

a landmark observation

and one point

of a track

be calculated. There is a vector between

and

given as

, A Mahalanobis distance calculation is performed using the equation

where S ^{-1 is} the inverse covariance matrix of the track landmark

designated.

For line-based landmarks, the landmark observations include start and end points. The determination of a geometric distance is not so clearly possible with typical methods, since clearly different results can be expected, for example, by interchanging the start and end positions. For this reason, the distance metric is modified according to the Malahanobis equation.

mit Startpunkt

und Endpunkt

sowie eine Landmarke eines Tracks

mit Startpunkt

und Endpunkt

. Es können nun Distanzen zwischen Start- und Endpunkten von

und

berechnet werden, wobei sich vier Kombinationsmöglichkeiten (Abstandsmessungen Startposition 1 – Startposition 2; Endposition 1 – Startposition 2; Startposition 1 – Endposition 2; Endposition 1 – Endposition 2) ergeben:

wobei mit S^–1 wieder die inverse Kovarianzmatrix der Track-Landmarke

bezeichnet wird.Considered are a landmark observation

with starting point

and endpoint

as well as a landmark of a track

with starting point

and endpoint

, Now distances between start and end points of

and

four possible combinations (distance measurements start position 1 - start position 2; end position 1 - start position 2; start position 1 - end position 2; end position 1 - end position 2) result:

where S ^-1 again the inverse covariance matrix of the track landmark

referred to as.

As a result, values for a quadratic mean (QM) can be calculated according to the formula:

Finally, the distance function d ^DL can be calculated according to:

Alternatively or additionally, further methods for determining the distance metric may be used.

Another distance metric is defined for polyline-based landmarks, that is, polylines. In addition to the start and end positions, further intermediate points are taken into account in order to determine a distance measure. In particular, an area between the lines is determined.

betrachtet, deren Verlauf als zusammenhängende Punkteliste (p₁, p₂, ..., p_i) modelliert wird, und eine Landmarkenbeobachtung

, mit der zusammenhängenden Punkteliste (p₁, p₂, ..., p_q). Um die zwischen den Polylinien eingeschlossene Fläche zu bestimmen, wird zunächst eine Projektion der Linienzüge aufeinander durchgeführt und die Linienzüge werden so abgeschnitten, dass nur diejenigen Abschnitte der Linienzüge weiter verarbeitet werden, die jeweils einem Abschnitt eines anderen Linienzugs gegenüberliegen.It becomes a landmark observation

whose history is modeled as a contiguous point list (p ₁ , p ₂ , ..., p _i ) and a landmark observation

, with the contiguous point list (p ₁ , p ₂ , ..., p _q ). In order to determine the area enclosed between the polylines, first a projection of the lines is performed on each other and the lines are cut so that only those sections of the lines that are opposite to a section of another line are further processed.

For this purpose, the points along the two lines can be determined, in which a perpendicular connection to an end point of the other line can be determined. In these places the truncation is performed. This avoids, for example, that differences in length of the lines result in an overestimation or underestimation of the distance measure.

The area enclosed by the lines is calculated, for example, by means of the Gaussian trapezoidal formula. For this purpose, the polygon enclosed by the polylines is first divided into subpolygons by means of intersections of the polylines, if present. Subsequently, the areas of the subpolygons are summed and normalized by the length of the sections of the lines.

The criterion for the assignment of the corresponding landmark observations to the landmarks of the tracks is the smallest distance is used as a criterion, the Greedy Nearest Neighbor method can be applied. Alternatively or additionally, other methods, such as the Hungarian method, may be used for the mapping.

In a further step S33, a track management is performed wherein probabilities are assigned and changed to the landmarks using a Sequential Probability Ratio Test (SPRT) to check whether a detected landmark of an actual landmark is in the traffic environment of the vehicle 1 equivalent.

If, for example, a landmark is observed at one time and not at another, it can be checked whether a real landmark has been detected or whether a landmark has been detected. If, on the other hand, a landmark is detected in several successive landmark observations, this is evaluated as confirmation and the value of the assigned probability increases.

In the embodiment, the detected landmarks are first assigned the same probability, the probability is increased when the landmark is confirmed, and lowered when the landmark is not confirmed. If the value of the probability exceeds or falls below a certain threshold, the hypothesis that it is a detection of an actual landmark is accepted or rejected. As a result of the SPRT it is determined as a statistical test which landmarks should continue to be considered.

wobei P_T und P_F die zu den Hypothesen (H₀/H₁ beziehungsweise true/false) gehörigen Wahrscheinlichkeiten und D die Daten bezeichnen.A null hypothesis H ₀ ("All observations are from a false detection") and an alternative hypothesis H ₁ ("All observations come from a real detection") are checked. It is therefore checked with what security based on a landmark observation a real landmark was detected. This test is based on the likelihood quotient LR:

where P _T and P _F denote the probabilities associated with the hypotheses (H ₀ / H ₁ and true / false, respectively) and D denotes the data.

To explain the method, individual landmarks observed at time t are referred to as z _t . Furthermore, it is assumed that, in the method carried out up to this point, all landmarks of the measurements B _t for which no corresponding landmark was found at the tracks T _t with the distance functions are already recorded as potentially new tracks in T _t . The set of observations at time t is called Z _t . The observation of the ith landmark z at time t is therefore called zl / t ∈ Z _t designated.

The observations are assigned an initialization probability of L ₀ at the time of data input. In doing so, the same value for all new observed landmarks can be selected for the initialization or a sensor-specific value can be assigned. To avoid numerical instabilities, the log-likelihood quotient LLR ^{i can be used} for the i-th landmark. Now, the probability that the i-th landmark from a true detection originates (alternative hypothesis H _1), developed over the detection points in time on the basis of mutually corresponding observed landmarks:

On the basis of this value, it is checked whether LLR ⁱ exceeds (A) or falls below a specific threshold value (B): B = ln (β / 1-α) ≤ LLR ⁱ ≤ ln (1-β / α) = A

LLR ⁱ corresponds to a probability for each i-th landmark of the environment model; where this probability is understood as the confidence value for the track assigned to the respective landmark. In particular, the field "Belief" of the above-explained state vector can be used for this purpose. The system design parameters α and β are specified and describe the probability of confirming a wrong track or deleting a real track. Further, the value of the probability of a landmark is thereafter increased or decreased whether it is confirmed by further landmark observations or not. In case of exceeding or falling below the iterative procedure is aborted and the landmark is confirmed or rejected:

In a step S34, environmental model data are finally generated, in particular this takes place in a "Track Innovation" step. The amount of tracked landmarks in the environment model is updated. This includes data on the traffic environment of the vehicle 1 registered landmarks, in particular their positions. In particular, based on the data determined in the previous method steps, positions, uncertainties and confidence values of the landmark observations are updated.

Based on the confirmed tracks, the respective assigned landmarks can be included in a map. Such a map includes the determined and optimized locations for the landmarks.

Various methods known per se can be used to update the positions, for example. For example, at several specific positions of a landmark, the last observed position may be adopted, exponential smoothing may be performed, a moving or arithmetic mean may be calculated, or, if the landmark has a length, a length-based weight may be used for location determination become.

The environment model data are structured so that they can be compared with the reference card data. In particular, for example, the positions of certain landmarks can be compared.

The environment model data are then output, that is transmitted to another module or unit for further processing.

In reference to 5 a particle filter based localization is explained. It is of the reference to the 1A and 1B explained system gone out. In particular, the steps described below may be considered as part of the above with reference to 2 explained landmark-based localization.

In the particle-based method, a particle cloud is created, each particle having a position hypothesis, that is, a hypothetical position of the vehicle 1 stands. The "particle cloud" thus denotes a quantity of particles. Thus, each particle represents a particle state that hypothesizes the state of the vehicle 1 is understood. It is now statistically tested in which hypothesis, that is for which particle, the landmark observations best match the reference map data. An embodiment of the method will be described in detail below.

In a first step S40 it is checked whether an initialization has already been carried out. If this is denied, the initialization is carried out in a further step S41. At initialization, as described above for step S4, the estimated position of the vehicle 1 for example by means of the GPS sensor 3c detected at a certain time, in particular, is based on a greater inaccuracy. Furthermore, data about the proper motion of the vehicle 1 be taken into account.

By applying a movement model in a step S42, the position is tracked with the self-movement of the vehicle 1 determined, that is, the change in the position of the vehicle 1 since the last acquisition of the estimation position is considered. For an initialization can also be a stored estimation position, such as the last drive of the vehicle 1 , or a position entered by the user.

There is a corresponding transformation of the particle cloud made to the displacement of the position of the vehicle 1 to balance by his own movement. Here, too, an inaccuracy of the corresponding sensors and thus also the resulting results can be assumed. In particular, uncertainties in sensors for detecting the intrinsic position (for example by means of GPS) or the intrinsic movement (for example by means of odometry) can be taken into account.

In a further step S43, a sensor model is applied. A sensor-related measurement noise is taken into account and the probability is determined with which the landmark observations were recorded assuming the respective hypothetical particle state.

From this probability, a weighting results for the individual hypotheses and this weighting is assigned to the hypotheses in a step S44. In particular, a normalization of the probabilities can be carried out.

In a further step S45, a smoothing is also performed to stabilize the results, smoothing out strong fluctuations of the probabilities for individual particles. Such fluctuations can occur, in particular, when major differences occur in the detection of a landmark at successive times, for example due to measurement errors, hidden landmarks or the disappearance of the landmarks from the detection area 17 the registration unit 3 ,

For example, the smoothed weighting ω _{f, t of} the particle f in the time step t can be calculated after: ω _{f, t} = αω _{f, t} + (1-α) ω _{f, t-1}

In this case, an attenuation factor α is selected empirically and the intensity of the attenuation can also be determined by a suitable choice of the time window between time steps t and (t-1).

In a step S46, the position corrected according to the preceding steps is output. In particular, the state of the particle with the maximum associated probability is taken into account.

In particular, during the output, the specific position is transmitted to a module or a unit for further processing, in particular the output can be sent to the driver assistance system 12 respectively.

Furthermore, resampling can be performed in a further step S47. This may become necessary if, by iterating the method described above, the number of particles converges to a single particle. This is due in particular to the fact that the scattering of the particle cloud by accumulated measurement errors increases in each step, while the proper motion of the vehicle 1 is tracked repeatedly. More and more particles move farther and farther from the most probable position and are assigned a correspondingly low probability. This means that fewer and fewer particles are taken into account with a particularly high probability. After initial scattering of the particles has been generated during sampling, resampling generates so-called "recovery particles" in the further iterations of the method, by which the particle cloud is supplemented by new hypotheses in order to statistically select the most probable hypothesis based on a sufficiently large database to meet.

In reference to 6 an embodiment of an update functionality is explained. It is of the reference to the 1A and 1B explained system and with reference to 2 explained procedures.

In a step S50, the estimated position of the vehicle is first 1 detected. This is done for example by the GPS sensor 3c , as described above. The estimation position becomes when retrieving reference map data to the external server 2 transfer.

Based on the estimation position, in a step S51, the external server 3 Reference card data generated and sent to the vehicle 1 transfer. The server attacks 2 to a reference map database and generates the reference map data so that they are all for landmark-based localization of the vehicle 1 relevant information, in particular landmarks and their positions in the traffic environment of the vehicle 1 ,

By the vehicle 1 Now a landmark-based localization takes place based on the reference map data. There may be differences from the vehicle 1 detected landmark observations and the expected according to the reference map data landmark observations. For example, landmarks may have disappeared or changed, or new landmarks may be detected.

A method may be used to ensure that these are actual deviations of the real landmarks from the reference map data. In particular, thereby artifacts of the detection of landmarks by the vehicle 1 be excluded. For this purpose, in particular known methods can be used.

In a step S52 are by the vehicle 1 generates update data based on the detected discrepancies between the reference map data and the actual acquired landmark observations. These include information that is suitable for being sent by the server in a further step S53 2 an update of the reference map database is made. In this case, when the procedure is repeated, such as the repeated retrieval of reference map data at a particular estimation position, new reference map data can now be provided which better reflect the actual features of the landmarks.

That way, the vehicle can 1 be used for a new survey of landmarks in the traffic environment. Furthermore, the server can 2 Receive update data from multiple sources, in particular multiple vehicles, in the same way and, for example by means of statistical methods, decide on an update of the reference map database. Furthermore, it can be determined on the basis of the update data whether a new measurement of the landmarks is to be undertaken in a specific traffic environment, for example by a special surveying vehicle which enables a particularly precise acquisition of the data for the reference map database.

In a further embodiment it is provided that the reference map data alternatively or additionally from a device other than the external server 2 to be provided. For example, this may be a device of the vehicle 1 be, even without a data connection to the server 2 allows retrieval of reference map data for landmark-based localization. The method can be carried out here in an analogous manner, whereby the device of the vehicle 1 accesses a reference map database and updates it using the update data. In another embodiment, both the external server 2 as well as an on-vehicle device for retrieving reference card data and updating the respective reference map database of the device and the server 2 can be done by a vote between the two.

LIST OF REFERENCE NUMBERS

1

Mobile unit; vehicle

2

External server

3

acquisition unit

3a

laser scanner

3b

camera

3c

GPS sensor

3n

Another sensor

4

control unit

5

interface

6

computer unit

7

accumulation unit

8th

segmentation unit

9

Landmark recognition unit

10

tracking unit

11

output unit

12

Driver assistance system

13

Landmark; delineators

14

Landmark; road boundary

15

Landmark; lane boundary

16

Landmark; Guideline

17

detection range

18

roadway

18a, 18b

lane

S1a, ..., S1n

Capture of records

S2a, ..., S2n

Landmark detection

S3

Tracking and fusion

S4

Acquisition of the estimated position

S5

Self-motion estimation

S6

Retrieval of the reference card data

S7

Particle filter based localization

S8

Output of the position

S21a

accumulation

S22a

segmentation

S23a

Principal Component Analysis

S24a

classification

S25a

Generating landmark surveys

S31

Timing synchronization

S32

distance calculations

S33

Assign and change probabilities

S34

Generation of environment model data

S40

Check the initialization

S41

Make the initialization

S42

Apply the movement model

S43

Apply the sensor model

S44

Assign the weight

S45

smoothing

S46

Output of the corrected pose

S47

Perform resampling

S50

Capture the estimated position (vehicle)

S51

Generating the reference card data (server)

S52

Generating update data (vehicle)

S53

Updating the reference map database (server)

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 102004003850 A1 [0004]
• DE 102007020791 A1 [0005]

Claims (10)

Method for detecting a traffic environment of a mobile unit ( 1 ), in which by means of a detection unit ( 3 ) Records are collected with data points, each time being assigned to each record; Landmark observations are respectively determined for the recorded data records, the landmark observations being assigned landmark parameters and object classes; environment model data are generated based on the particular landmark observations, the environmental model data including a description of the traffic environment; Distance model calculations for landmark observations are performed on the basis of the environment model data using their object class and their landmark parameters, distance values being determined for the landmark observations, the landmark observations being assigned to specific landmarks on the basis of the distance values; the environmental model data are assigned probabilities and the assigned probabilities are changed on the basis of the landmark observations, the environmental model data are updated on the basis of the probabilities associated with the landmark observations; and the environmental model data is output. Method according to one of the preceding claims, characterized in that the landmark observations are classified on the basis of their associated object classes as point-based, line-based or polyline-based objects. A method according to claim 2, characterized in that the determined landmark parameters for segments of the line-based or polyline-based feature class include a length parameter and a width parameter, and a start position and an end position. Method according to one of the claims 2 or 3, characterized in that in the distance calculations for point-based objects a Mahalanobis distance is determined, for line-based objects a modified Mahalanobis distance is determined on the basis of the start and end positions and / or for polyline-based objects one of the polyline-based objects enclosed area is determined. Method according to one of the preceding claims, characterized in that the probabilities associated with the environment model data are changed by means of a sequential probability test, wherein the probabilities are used to check whether a landmark observation of an actual detection of a landmark ( 13 . 14 . 15 . 16 ) corresponds to the traffic environment. A method according to claim 5, characterized in that a probability for a landmark observation is increased if it is assigned to a further landmark observation, and the probability for a landmark observation is lowered, if it is not confirmed by another landmark observation. Method according to one of the preceding claims, characterized in that the landmark observations of the environment model are accepted or rejected on the basis of a comparison of their associated probabilities with a threshold value. Method according to one of the preceding claims, characterized in that the environmental model data are assigned landmarks and include positions of the landmarks. System for detecting a traffic environment of a mobile unit ( 1 ) with a registration unit ( 3 ), can be detected by the records with data points, each record is a time assignable; a landmark recognition unit ( 9 ), by the landmark records observations can be determined by the recorded data records, the landmark observations and landmark classes parameters and object classes are assigned; an arithmetic unit ( 6 ) by which environment model data can be generated based on the determined landmark observations, the environment model data comprising a description of the traffic environment; a tracking unit ( 10 ), on the basis of the environmental model data distance calculations for landmark observations on the basis of their object class and their landmark parameters are feasible, with distance values for the landmark observations can be determined; wherein by the arithmetic unit ( 6 ) the environment model data probabilities are assignable and the associated probabilities are changeable based on the landmark observations; further wherein the environmental model data is updatable based on the probabilities associated with the landmark observations; and the environment model data can be output. System according to claim 9, characterized in that the detection unit ( 3 ) at least two different sensor types ( 3a . 3b ).

Images