Method and apparatus for determining a position of a vehicle in a road network

Abstract

The invention relates to a method, wherein a position of a vehicle (102) in a road network is determined by means of a sensor set (104) of the vehicle (102), the road network comprising a plurality of edges (202a to 202f), which are respectively associated with a road section (106) or a lane on a road section, wherein first sensor data are detected by means of the sensor set (104), a first probability is respectively assigned to each edge of a plurality of first edges of the road network on the basis of the first sensor data, with which the vehicle (102) is in a road section (106) respectively assigned to the first edge or in a lane respectively assigned to the first edge, second sensor data are detected by means of the sensor set (104), a second probability is respectively assigned to each edge of a plurality of second edges of the road network on the basis of the second sensor data and the probability respectively assigned to the first edge, with which the vehicle (102) is in a road section (106) respectively assigned to the second edge or in a lane respectively assigned to the second edge, and the second edge with the highest assigned probability is determined as the position of the vehicle (102).

Description

TECHNICAL FIELD

[0001] 1. The invention relates to a method for determining a position of a vehicle in a road network by means of a sensor set of the vehicle. The invention also relates to a device for determining a position of a vehicle in a road network by means of a sensor set of the vehicle. BACKGROUND

[0002] 2. In the field of autonomous driving as well as in driving assistance systems or navigation systems the position of a vehicle is to be determined. Especially in autonomous driving systems as well as in driving assistance systems and navigation systems the position has to be determined with high precision in order to be able to ensure a high functional safety of the system. The precision of the position determination mainly depends on the quality of the sensor data on which the position determination is based. Inaccurate sensor data can lead to inaccurate or even incorrect position determination and catastrophic consequences. Therefore, it is desirable to be able to perform a position determination as precise as possible even with inaccurate sensor data.

[0003] 3. The document "LaneQuest. An Accurate and Energy-Efficient Lane Detection System" by Aly et al. (Preprint, arXiv: 1502.03038v1) discloses a method for determining a lane of a vehicle by means of a sensor set of the vehicle. With regard to the prior art, reference is also made to the document "Methods and Implementations of Road-Network Matching" by Zhang, M. (Thesis, Technical University of Munich, 2009), which discloses a method for comparing road maps by semantic methods, in particular. SUMMARY

[0004] 4. It is an object of the present invention to specify a method and a device for determining a position of a vehicle which allow a position determination even with inaccurate sensor data.

[0005] 5. This object is achieved by a method having the features of claim 1 and by a device having the features of the independent device claim. Advantageous improvements are specified in the dependent claims.

[0006] 6. In the method according to claim 1, the position of the vehicle in the road network comprising a plurality of edges, which are respectively associated with road sections or lanes on road sections, is determined by means of a sensor set of the vehicle. First sensor data are detected by the sensor set. On the basis of the first sensor data, a first probability is respectively assigned to each edge of a plurality of first edges of the road network, in which the vehicle is in a road section respectively assigned to the first edge or in a lane respectively assigned to the first edge. Second sensor data are detected by the sensor set. On the basis of the second sensor data and the probabilities respectively assigned to the first edges, a second probability is respectively assigned to each edge of a plurality of second edges of the road network, in which the vehicle is in a road section respectively assigned to the second edge or in a lane respectively assigned to the second edge. The second edge having the highest assigned probability is determined as the position of the vehicle.

[0007] 7. The road network is a graph, the edges of which respectively correspond to road sections or respectively to lanes on road sections. A road section is for example a section of a road between two intersections, i.e. between a road crossing at least one further road or opening into at least one further road, or between an intersection and the end of the road. However, shorter road sections are preferably chosen, for example shorter than sections having a fixed length. Different attributes can be assigned to a road section, for example lane numbers, driving directions, positioning points, landmarks, road names, etc.

[0008] 8. In an advantageous refinement, the edges are directly associated with lanes on road sections. This is for example possible if a more detailed map is present.

[0009] 9. The sensor set comprises at least one sensor of the vehicle. The detection of the first sensor data and the second sensor data can be carried out successively in time. If the sensor set comprises more than one sensor, the first sensor data and the second sensor data can also be detected simultaneously. The plurality of second edges is in particular a subset of the plurality of first edges.

[0010] 10. The plurality of second edges can also comprise edges which are not contained in the plurality of first edges.

[0011] 11. On the basis of the first sensor data, a coarse determination of the position of the vehicle within the road network is carried out by assigning a first probability to each first edge, in which the vehicle is in a road section respectively assigned to the first edge. The first probability is used together with the second probability as a basis for determining the second probability. Thereby, the determination of the position within the road network is improved and / or updated by means of the second sensor data.

[0012] 12. Since the determination of the position of the vehicle within the road network at a higher level of abstraction than for example the world coordinate system takes place, the method can better handle imprecise sensor data than other methods for example for determining the position of the vehicle within the world coordinate system. Since by the method a probability distribution of all possible positions of the vehicle within the road network is determined, it is also always possible to determine the degree of reliability of the currently determined position. The method also allows multiple position determinations, i.e. in which two or more edges are assigned the same or a similar size of the highest probability.

[0013] 13. It goes without saying that the detection of the second sensor data and the determination of the second probability can be repeated in temporally successive steps. Here, the second probability determined in the first step can in particular be used as a basis for determining a further second probability in a temporally subsequent second step as the first probability. In other words: In the second step, the second probability is determined on the basis of the second probability determined in the first step and the second sensor data detected in the second step. Thereby, a recursive improvement of the position determination within the road network is achieved.

[0014] 14. The method according to claim 1 can in particular be carried out in combination with a method for determining the position in a measuring manner, i.e. for determining the position of the vehicle within a coordinate system, for example the world coordinate system. The two methods are based on different functions and thus complement each other. Thereby, the functional safety of the system for determining the position of the vehicle can be significantly improved.

[0015] 15. In an advantageous refinement, the sensor coordinates of the vehicle within the world coordinate system are detected as at least a part of the first sensor data and / or as at least a part of the second sensor data by means of a receiver for signals of a satellite navigation system for the sensor group. The receiver is in particular a GPS antenna. By means of the satellite navigation system, the position of the vehicle can be determined with an accuracy of no more than a few meters. This is usually sufficiently accurate for limiting the position of the vehicle to a few edges of the road network. Furthermore, the driving direction of the vehicle can be determined by means of the satellite navigation system. Alternatively or additionally, the driving direction of the vehicle can be determined by means of the mileage data.

[0016] 16. Alternatively or additionally, the sensor coordinates of the vehicle within the world coordinate system can be determined by means of a receiver for mobile communication signals for the sensor group. Just in areas with a high density of mobile communication antennas, the determination of the position by means of mobile communication signals is usually more accurate than the determination of the position by means of a satellite navigation system.

[0017] 17. In a further advantageous refinement, each edge of the road network is assigned coordinates of the respective associated road section in the world coordinate system. The first probability is determined by comparing the sensor coordinates of the vehicle with at least a portion of the coordinates assigned to the first edge. For example, each first edge is assigned a first probability depending on the distance of the coordinates assigned to the first edge from the sensor coordinates. The further the coordinates assigned to the first edge are from the sensor coordinates, the smaller the first probability assigned to the first edge. It can be assumed here, inter alia, that the probability distribution of the actual coordinates of the vehicle in the world coordinate system is a normal distribution around the sensor coordinates. This makes it possible to quickly limit the position of the vehicle within the road network to a few edges and thus to reduce the outlay for determining the second probability.

[0018] 18. In a further advantageous refinement, image data are detected by means of a camera of the sensor group as at least a portion of the first sensor data and / or as at least a portion of the second sensor data. The image data correspond to an image of a detection region in front of or behind the vehicle. From the image data, a large amount of qualitative data can be obtained which allow an exact determination of the position of the vehicle. Since the method according to the application works with imprecise sensor data, it is not necessary, inter alia, to calibrate the camera.

[0019] 19. In a further advantageous refinement, at least a subset of the edges of the road network is assigned a localization point. It is determined on the basis of the image data and using an object recognition method whether a localization point is present in the detection region of the camera. If a localization point is present in the detection region of the camera, the first probability is also determined on the basis of the localization point. The presence (or absence) of a localization point is an example of qualitative data which allow an exact determination of the position of the vehicle. For example, if a bridge is seen in the detection region of the camera, a higher first probability is assigned to all first edges which are associated with the bridge as localization point than to first edges which are not associated with the bridge as localization point. It is not important here at first whether the bridge is located exactly where in the detection region of the camera. Alternatively or additionally, other known landmarks, in particular traffic lights, traffic signs or traffic signs indicating connection points / highway exits, can also be used as localization points.

[0020] 20. In a further advantageous refinement, at least a subset of the edges of the road network is assigned a road name. It is determined on the basis of the image data and using a text recognition method whether a road name or a portion of a road name is present in the image of the detection region. If a road name or a portion of a road name is present in the image of the detection region, the first probability is also determined on the basis of the road name or the portion of the road name. A road name is a further example of qualitative data which allow an exact determination of the position of the vehicle. If a road name or a portion of a road name is recognized, the position of the vehicle can generally be limited to a unique road.

[0021] 21. In a further advantageous refinement, self-motion data of the vehicle are detected by means of the image data as at least a portion of the second sensor data. In the present application, self-motion is understood to mean the motion of the vehicle in three-dimensional space. The self-motion data are correspondingly understood to mean motion data of the vehicle or of a camera fixedly connected to the vehicle, which are determined on the basis of the image data detected by means of the camera. Such motion data can be determined, for example, using optical flow-based image processing methods.

[0022] 22. In a further advantageous refinement, mileage data of the vehicle are detected by means of the sensor group as at least a portion of the first sensor data and / or as at least a portion of the second sensor data. By means of the mileage data and the displacement measurement data, the current position of the vehicle can be determined on the basis of the known past positions of the vehicle. The mileage data thus allow a very precise determination of the position of the vehicle, in particular when the second probability is determined repeatedly.

[0023] 23. In a further advantageous refinement, the road network comprises a plurality of nodes, which are respectively associated with a connection between at least two road segments or between at least two lanes. Nodes associated with a connection between three or more road segments or between three or more lanes are referred to as intersections. On the basis of the first sensor data or the second sensor data, it is determined whether the vehicle is located in a road segment or a lane adjacent to an intersection or whether the vehicle is located in a road segment not adjacent to an intersection. The first probability and / or the second probability are also determined on the basis of this information. This embodiment uses the information as to whether the vehicle is located in a road segment or a lane adjacent to an intersection or not in order to restrict the position of the vehicle. For example, if it is determined by means of the first sensor data that the vehicle is located in a road segment or a lane adjacent to an intersection, then all first edges adjacent to the intersection are assigned a higher first probability than all other first edges.

[0024] 24. In a further advantageous refinement, at least a subset of the edges of the road network is respectively assigned at least one lane, which corresponds to the lane of the road segment respectively assigned to the edge. The first probability is also determined on the basis of the lane respectively assigned to the first edge. In this refinement, the number of lanes of the road segment is used to restrict the position of the vehicle. For example, if it is determined by means of the first sensor data that the vehicle is on a road segment having three lanes, then all first edges associated with the three lanes are assigned a higher second probability than all other first edges.

[0025] 25. In a further advantageous refinement, a second probability is assigned to each of the second edges on the basis of the second sensor data, with which the vehicle is located in the respective lane. The lane having the highest assigned probability is determined as part of the position of the vehicle. For example, if it is determined by means of the second sensor data that a traffic sign is located to the right of the vehicle, then all second edges having a traffic sign of the same type as the detected traffic sign are assigned a second probability which is higher than that of all other second edges.

[0026] 26. For example, if it is determined by means of the second sensor data that there are left and right adjacent lanes next to the vehicle, respectively, then all second edges having a left and right adjacent lane are assigned a second probability which is higher than that of all other second edges.

[0027] 27. The probability with which the vehicle is in the respective lane is determined, in particular, in the case of the use of the probabilistic method. The lane can be determined, in particular, by means of image data of a camera of the sensor group of the vehicle. The determination of the lane of the road section in which the vehicle is located allows a very precise determination of the position of the vehicle. Such a determination of the position is important, for example, for an autonomous driving system or a navigation system.

[0028] 28. In a further advantageous refinement, a probability is assigned to each possible driving direction of the vehicle on the basis of the second sensor data, with which the vehicle is moving in the respective driving direction. The driving direction having the highest assigned probability is determined as part of the position of the vehicle. The probability with which the vehicle is moving in the respective driving direction is determined, in particular, in the case of the use of the probabilistic method. The driving direction can be determined, in particular, by means of mileage data, self-motion data and / or from the time course of the sensor coordinates of the vehicle. The determination of the driving direction allows, in particular, to restrict the position of the vehicle to specific edges. For example, only those edges can be selected which are located in the driving direction of the vehicle starting from a known position of the vehicle.

[0029] 29. In a further advantageous refinement, at least one possible driving direction is assigned to at least one subset of the edges of the road network. The second probability is likewise determined on the basis of the possible driving direction assigned to the respective second edge. For example, if the vehicle has to move against the possible driving direction in order to reach an edge, then this edge is assigned a low probability. Thereby, in particular, edges which are located behind the vehicle in the driving direction are discarded.

[0030] 30. In a further advantageous refinement, the probabilities assigned to the first edges are compared to a minimum probability. The first edges to which the probabilities assigned thereto are greater than or equal to the minimum probability are determined as second edges. In this embodiment, a plurality of possible positions of the vehicle within the road network is determined. The minimum probability can either be predetermined or determined, for example, on the basis of the first probabilities. By selecting the second edges, i.e. restricting to the most probable positions of the vehicle, the expenditure for determining the second probabilities can be significantly reduced. Alternatively or additionally, the second edges can be selected on the basis of the first sensor data or the second sensor data. For example, the edges of the road network which lie in the driving direction of the vehicle can be determined as second edges.

[0031] 31. In a further advantageous refinement, the position of the vehicle is used as an input for an autonomous driving system and / or a driving assistance system of the vehicle. Alternatively or additionally, the position is output to a driver of the vehicle. For example, the position is displayed on a map.

[0032] 32. Preferably, the second probabilities are determined in the case of a probabilistic method, in particular a Bayesian filter. In this preferred embodiment, the first probabilities correspond to the probabilities p(z t | x) of detecting the first sensor data z t detected at the point in time t in the case of the assumption that the vehicle is located in the position x. The second probabilities correspond to the probabilities p(y) of the vehicle being located in the position y at the point in time t.

[0033] 33. The second probabilities are determined with the aid of a transition matrix P t+1,y,t,x and the first probabilities. The transition matrix is determined, in particular, on the basis of the ego-motion data and / or the image data. For example, a transition to a position opposite the driving direction of the vehicle is very unlikely. Such a transition is also referred to as a forbidden transition. Other examples of forbidden transitions are a jump from one road to another road, a driving onto a road section against the permitted driving direction, and a driving onto a multi-lane motorway via a lane which is not connected to a connection point.

[0034] 34. Another aspect of the application relates to an apparatus for determining a position of a vehicle in a road network. The road network comprises a plurality of edges, which are respectively assigned to a road section or a lane. The apparatus comprises a sensor group, which is configured to detect first sensor data and second sensor data. The apparatus further comprises a processor, which is configured to respectively assign, based on the first sensor data, a first probability for each edge of a plurality of first edges of the road network, that the vehicle is in a road section respectively assigned to the first edge or in a lane respectively assigned to the first edge with the first probability; respectively assign, based on the second sensor data and the probabilities respectively assigned to the first edges, a second probability for each edge of a plurality of second edges of the road network, that the vehicle is in a road section respectively assigned to the second edge or in a lane respectively assigned to the second edge with the second probability. The plurality of second edges is at least a subset of the plurality of first edges. The processor is further configured to determine the second edge with the highest assigned probability as the position of the vehicle.

[0035] 35. The apparatus has the same advantages as the claimed method and can be improved in the same way, especially with the features of the attached patent claims. BRIEF DESCRIPTION OF DRAWINGS

[0036] 36. Further features and advantages will become apparent from the following description, taken in connection with the accompanying drawings, which illustrate embodiments. In the drawings:

[0037] 37. Figure 1 A schematic illustration of an apparatus for determining a position of a vehicle in a road network is shown;

[0038] 38. Figure 2 A fragment of a road network and positions of sensor coordinates relative to the road network are shown;

[0039] 39. Figure 3 A fragment of a road network is shown to illustrate a first step for determining a position;

[0040] 40. Figure 4 A fragment of a road network is shown to illustrate a second step for determining a position;

[0041] 41. Figure 5 A fragment of a road network is shown to illustrate a repetition of a second step for determining a position; and

[0042] 42. Figure 6 A fragment of a road network is shown to illustrate a repetition of a second step for determining a position;

[0043] 43. Figure 7 A preferred embodiment is shown, in which edges are associated with different lanes on a road section. DETAILED DESCRIPTION

[0044] 44. Figure 1 A schematic illustration of determining a position of a vehicle 102 in a road network by means of a sensor set 104 of the vehicle 102 is shown.

[0045] 45. The vehicle 102 is in a road section 106, which is assigned one of a plurality of edges of the road network, and which road section has exemplarily a unique lane. Each of the plurality of edges of the road network is assigned a respective coordinate of the respective road section 106 in the world coordinate system.

[0046] 46. The device 100 comprises a sensor set 104, which in the shown embodiment comprises a video camera, a receiver for signals of a satellite navigation system and an interface for receiving mileage data of the vehicle 102. The video camera is oriented in the driving direction of the vehicle 102 and is configured for detecting image data, which corresponds to an image of a detection area 108 in front of the vehicle 102. The receiver for signals of a satellite navigation system is configured for determining sensor coordinates 206 (cf. Fig. 2) of the vehicle 102 in the world coordinate system. Figures 2 to 6 The interface for receiving mileage data of the vehicle 102 is configured for receiving mileage data, i.e. displacement measurement data, of the vehicle 102. The displacement measurement data comprises inter alia a distance covered within a certain period of time.

[0047] 47. The device 100 further comprises a processor 110, which is exemplarily connected with the sensor set 104 by means of a cable 112. The processor 110 is configured for determining, based on the image data and / or in case of using a text recognition method and an object recognition method, whether a road name or a part of a road name or other positioning points can be seen in the image of the detection area 108 of the video camera and for further processing this information as positioning point information. Furthermore, the processor 110 is configured for determining self-motion data based on the image data and in case of using a light flow based image processing method and for further processing it. The processor 110 is further configured for determining a position of the vehicle 102 in the road network based on the positioning point information, the self-motion data and the sensor coordinates 206.

[0048] 48. To determine the position of the vehicle 102, in a first step for determining the position, the processor 110 compares the sensor coordinates 206 with coordinates of respective ones of a plurality of first edges of the road network that are assigned to the respective edges. Here, the first edges can be a subset of all edges of the road network or all edges. In case the probability distribution of the actual position of the vehicle 102 is assumed to be a normal distribution, the processor 110 assigns a first probability to each edge, that the vehicle 102 is located in the road segment 106 associated with the respective edge with the first probability. Then, the processor 110 modifies the first probabilities based on the localization point information. For example, if a bridge is identified in the image of the detection area 108, the processor 110 adjusts the first probabilities such that the first probabilities of all first edges that are not associated with the bridge as a localization point are reduced.

[0049] 49. In a second step for determining the position, which immediately follows the first step in time, the processor 110 assigns a second probability to each of a plurality of second edges of the road network that the vehicle 102 is located in the road segment 106 assigned to the respective second edge with the second probability based on the ego-motion data, the image data, the odometry data and the probabilities associated with the first edges, respectively, and in case a Bayesian filter is used. Then, the processor 110 determines the second edge for which the assigned probability is highest as the position of the vehicle 102.

[0050] 50. The following flow for determining the position of the vehicle 102 in the road network is further illustrated by means of Figures 2 to 6 Fig. 3, which shows a fragment 200 of the road network, respectively. The fragment 200 comprises six edges 202a to 202f, which correspond to road segments 106, respectively, and three junctions 204a to 204c, which connect the edges 202a to 202f. Two of the three junctions 204a to 204c, 204a, 204c, connect more than two of the edges 202a to 202f. The junctions 204a, 204c are also referred to as intersections in the following. The fragment 200 corresponds to a one-way road from left to right in Fig. 3, which intersects another road 202f, 202f, respectively, at two intersections 204a, 204c. Figures 2 to 6 Fig. 4, which shows a fragment 200 of the road network, respectively. The fragment 200 comprises six edges 202a to 202f, which correspond to road segments 106, respectively, and three junctions 204a to 204c, which connect the edges 202a to 202f. Two of the three junctions 204a to 204c, 204a, 204c, connect more than two of the edges 202a to 202f. The junctions 204a, 204c are also referred to as intersections in the following. The fragment 200 corresponds to a one-way road from left to right in Fig. 4, which intersects another road 202f, 202f, respectively, at two intersections 204a, 204c.

[0051] 51. Figure 2 Fig. 5 shows the position of the sensor coordinates 206 as a point with respect to the road network. In the illustrated embodiment, the sensor coordinates 206 correspond to first sensor data and have a measurement error, which is assumed to be a normal distribution. Exemplarily, the contour lines 208 of a two-dimensional normal distribution of the measurement error are shown as dashed circles in Fig. 5. The dashed circles at least partially overlap with Figure 2 Fig. 6, which shows the position of the sensor coordinates 206 as a point with respect to the road network. In the illustrated embodiment, the sensor coordinates 206 correspond to second sensor data and have a measurement error, which is assumed to be a normal distribution. Exemplarily, the contour lines 208 of a two-dimensional normal distribution of the measurement error are shown as dashed circles in Fig. 6. The dashed circles at least partially overlap with Figure 2The three edges 202a, 202b, 202f shown in the middle left. This means that the three edges 202a, 202b, 202f within the measurement error of the sensor coordinates 206 are possible locations of the vehicle 102.

[0052] 52. Figure 3 A segment 200 of a road network according to Figure 2 is shown. Figure 3 A first step for determining a position according to Figure 1 is shown. All edges 202a to 202f of the segment 200 shown in Figure 3 are selected as first edges. The three edges 202a, 202b, 202f within the measurement error of the sensor coordinates 206 are shown in Figure 3 with dashed lines. In the first step for determining a position, the three edges 202a, 202b, 202f are assigned a high first probability. In other words: In the first step, it is assumed that the vehicle 102 is most likely located in one of the three first edges 202a, 202b, 202f.

[0053] 53. Figure 4 A segment 200 of a road network according to Figure 2 is shown, and a second step for determining a position according to Figure 1 is shown. In the second step, a second probability is determined using a Bayesian filter using the ego-motion data, the odometry data and the first probability assigned to the first edges. The ego-motion data states that the vehicle 102 moved in a certain direction forward, but there was no big change in the driving direction. The latter excludes that the vehicle 102 is or has been located in the edge 202e leading into the one-way street, because in this case the vehicle 102 must have turned. In view of the odometry data and the fact that the vehicle 102 is located on a one-way street, the two edges 202b, 202c between the two intersections 204a, 204c can be assigned the highest second probability. These two second edges 202b, 202c are shown in Figure 4 with dashed lines.

[0054] 54. Figure 5 A segment 200 of a road network according to Figure 2 is shown, and a third step for determining a position according to Figure 1The first repetition of the second step for determining the position. In the first repetition of the second step, a new second probability is determined using a Bayesian filter, using new self-motion data, new mileage data, and the second probability correspondingly assigned to the second edges 202b and 202c in the previous second step. Similarly, the new self-motion data indicates that vehicle 102 is moving forward but not turning. Therefore, vehicle 102 cannot be located on edge 202f leading into the one-way road, which is shown... Figure 5 On the right side. Taking mileage data into account, compared to... Figure 4 The two adjacent edges 202c and 202d at the intersection shown on the right can be assigned the highest second probability. These two second edges... Figure 5 It is shown in dashed lines.

[0055] 55. Figure 6 It shows according to Figure 2 The road network segment 200 is shown, and it illustrates how improvements can be made based on image data. Figure 5 The location is determined. Traffic signs can be seen in the image of the camera located in the detection area 108 before vehicle 102, corresponding to the signs on the one-way street. Figure 5 The edge 202f shown on the right is associated. Traffic signs are part of the location point information. Because the traffic signs are visible in front of vehicle 102, it is possible to exclude vehicle 102 from being located on a one-way road, taking into account the direction of travel of vehicle 102 and the fact that vehicle 102 is on a one-way road. Figure 6 This is shown in the rightmost edge 202d. Therefore, the highest second probability can be assigned to this unique edge. The second edge is in Figure 6 It is shown in dashed lines.

[0056] 56. Figure 7 A preferred embodiment is shown, in which edges 702a to 702d are associated with different lanes on a road segment. The lanes are located within road segments 701a to 701e. Lane 702a is not connected to lanes 702b, 702c, and 702d. Node 703 connects lane 702b to lanes 702c and 702d.

[0057] 57. The road network in Figure 7 All edges 702a to 202d of segment 700 shown are selected as first edges. In the first step for determining position, a high first probability is assigned to the three edges 702b, 702c, and 702d within the measurement error of sensor coordinates 206. In other words, in the first step, it is assumed that vehicle 102 is most likely located at one of the three first edges 702b, 202c, and 702d.

[0058] 58. In a second step, a second probability is determined using a Bayesian filter using the ego-motion data, the odometry data and the first probability assigned to the first edge. The ego-motion data states that the vehicle 102 moves in a certain direction along the road, but there is no big change in the driving direction. The latter excludes that the vehicle 102 is or has been in the edge 702c leading in, because in this case the vehicle 102 must have turned. In consideration of the odometry data and the fact that the vehicle 102 is on a single-lane road, the highest second probability can be assigned to both edges 702b, 702d.

[0059] 59. In a first repetition of the second step, a new second probability is determined using a Bayesian filter using the new ego-motion data, the new odometry data, the new image data and the second probability assigned to the second edge 702b, 702d in the previous second step.

[0060] 60. In the image of the detection area 108 in front of the vehicle 102, a traffic light associated with the edge 702d can be seen. Because the traffic light in front of the vehicle 102 can be seen, it can be excluded that the vehicle 102 is in the edge 702b in consideration of the driving direction of the vehicle 102 and the fact that the vehicle 102 is on a single-lane road. Therefore, the highest second probability can be assigned to the only edge (702d).

[0061] 61. The method according to the application and the device 100 according to the application are described by way of example with reference to the embodiments. Figures 1 to 7 The sensor group 104 can comprise further or additional sensors not present in the shown embodiments, among others. Examples of further sensors include a receiver for mobile communication signals, a further camera for detecting image data corresponding to an image of a detection area 108 behind and / or to the side of the vehicle 102, an accelerometer or gyroscope for detecting odometry data and a radar or lidar sensor.

[0062] List of reference signs:

[0063] 100 device

[0064] 102 vehicle

[0065] 104 sensor group

[0066] 106 road section

[0067] 108 detection area

[0068] 110 processor

[0069] 112 cable

[0070] 200 segment

[0071] 202a-f edge

[0072] 204a-c junction

[0073] 206 sensor coordinate

[0074] 208 contour

[0075] 700 segment of lane

[0076] 701a-e edge

[0077] 702a-d lane

[0078] 703 junction

Claims

1. Method for determining a position of a vehicle (102) in a road network by means of a sensor set (104) of the vehicle (102), the road network comprising a plurality of edges (202a to 202f), which are respectively associated with a road section (106) or a lane on a road section, wherein detecting first sensor data by means of the sensor set (104), respectively assigning a first probability to each edge of a plurality of first edges of the road network on the basis of the first sensor data, the first probability being a probability of the vehicle (102) being in a road section (106) respectively assigned to the first edge or in a lane respectively assigned to the first edge, detecting second sensor data by means of the sensor set (104), respectively assigning a second probability to each edge of a plurality of second edges of the road network on the basis of the second sensor data and the probability respectively assigned to the first edge, the second probability being a probability of the vehicle (102) being in a road section (106) respectively assigned to the second edge or in a lane respectively assigned to the second edge, determining the second edge with the highest assigned probability as the position of the vehicle (102), the plurality of second edges being a subset of the plurality of first edges, The second probability is determined with the aid of a transition matrix P t+1,y,t,x and a first probability determination, the first probability corresponding to a probability p(z t | x) of the first sensor data z t detected at a point in time t, assuming that the vehicle is located in a position x, the second probability corresponding to a probability p(y) of the vehicle being located in a position y at a point in time t+1, the transition matrix being determined on the basis of ego-motion data and / or image data, a transition to a position opposite the driving direction of the vehicle being assessed as impossible, image data being detected with the aid of a camera of the sensor group (104) as at least a portion of the first sensor data and / or at least a portion of the second sensor data, wherein the image data corresponds to an image of a detection region (108) in front of or behind the vehicle (102), the ego-motion data of the vehicle (102) being detected with the aid of the image data as at least a portion of the second sensor data.

2. The method of claim 1, wherein, detecting sensor coordinates (206) of the vehicle (102) in a world coordinate system by means of a receiver for signals of a satellite navigation system for the sensor set (104) as at least a part of the first sensor data and / or at least a part of the second sensor data.

3. The method of claim 2, wherein, assigning to each edge of the road network a respectively associated road section (106) coordinates in the world coordinate system, wherein the first probability is determined by comparing the sensor coordinates (206) of the vehicle (102) with at least a part of the coordinates respectively assigned to the first edge.

4. The method of claim 1, wherein, respectively assigning to at least one subset of edges (202a to 202f) of the road network a localization point, wherein it is determined on the basis of the image data and in the case of use of an object recognition method whether a localization point is present in the detection area of the camera, and if a localization point is present in the detection area of the camera, the first probability is also determined on the basis of the localization point.

5. The method of claim 1, wherein, respectively assigning to at least one subset of edges (202a to 202f) of the road network a landmark, the landmark being a traffic light or a traffic sign, wherein it is determined on the basis of the image data whether a landmark is present in the image of the detection area (108), and if a landmark is present in the image of the detection area (108), the second probability is also determined on the basis of the landmark.

6. The method of any one of claims 1-5, wherein, The road network comprises a plurality of nodes (204a to 204c) which are associated with a connection between at least two road sections (106) or between at least two lanes, wherein a node (204a to 204c) which is associated with a connection between three or more road sections (106) or between three or more lanes is referred to as an intersection, wherein an information is determined on the basis of the first sensor data or the second sensor data, which information relates to whether the vehicle (102) is located in a road section (106) or a lane which is located close to an intersection or whether the vehicle (102) is located in a road section (106) or a lane which is not adjacent to an intersection, wherein the first probability and / or the second probability are / is determined on the basis of the information as well.

7. The method of any one of claims 1-5, wherein, At least one subset of edges (202a to 202f) of the road network is respectively assigned at least one lane which corresponds to a lane of a road section (106) which is respectively assigned to the edge, wherein the first probability and / or the second probability are / is determined on the basis of the lane which is respectively assigned to the second edge as well.

8. The method of claim 7, wherein, For each edge of the second edges, a probability is respectively assigned that the vehicle (102) is located in the respective lane on the basis of the second sensor data of the respective lane, wherein the lane which is assigned the highest probability is determined as part of the position of the vehicle (102).

9. The method of any one of claims 1-5, wherein, For each possible driving direction of the vehicle (102), a probability is assigned that the vehicle (102) is moving in the respective driving direction on the basis of the second sensor data, the driving direction which is assigned the highest probability is determined as part of the position of the vehicle (102).

10. The method of any one of claims 1-5, wherein, At least one subset of edges (202a to 202f) of the road network is respectively assigned at least one possible driving direction, wherein the first probability is determined on the basis of the possible driving direction which is respectively assigned to the first edge as well.

11. The method of any one of claims 1-5, wherein, The probabilities which are respectively assigned to the first edges are compared to a minimum probability, wherein the first edges whose respectively assigned probabilities are greater than or equal to the minimum probability are determined as the second edges.

12. An apparatus (100) for determining a position of a vehicle (102) in a road network, the road network comprising a plurality of edges (202a to 202f), the edges being associated with road segments (106) or lanes, respectively, wherein, The apparatus (100) comprises: a sensor group (104) which is configured to detect first sensor data and second sensor data; a processor (110) which is configured to assign, on the basis of the first sensor data, a first probability to each edge of a plurality of first edges of the road network, in which the vehicle (102) is located with the first probability in a road section (106) which is respectively assigned to the first edge or in a lane which is respectively assigned to the first edge, assign, on the basis of the second sensor data and the probabilities which are respectively assigned to the first edges, a second probability to each edge of a plurality of second edges of the road network, in which the vehicle (102) is located with the second probability in a road section (106) which is respectively assigned to the second edge or in a lane which is respectively assigned to the second edge, wherein the plurality of second edges is at least one subset of the plurality of first edges, and determine the second edge which is assigned the highest probability as the position of the vehicle (102), The apparatus is configured to perform the method according to any one of claims 1-11. The apparatus is configured to perform the method according to any one of claims 1-11.

Images