METHOD FOR DETERMINING A RELIABILITY VALUE FOR A VEHICLE'S SELF-POSITION DETERMINATION AND VEHICLE

Description

[0001] The invention relates to a method for determining a reliability value for determining the self-position of a vehicle and to a vehicle for carrying out the method.

[0002] From the publication "Visual Place Recognition via Semantic and Geometric Descriptor for Automated Valet Parking", Proceedings of the 2021 IEEE International Conference on Robotics and Biomimetics December 27-31, 2021, Sanya, China, pages 1142-1147; Jingrui Yu and Jianbo Su, a descriptor comparison of images taken at different locations and a determination of associated similarity values ​​is known.

[0003] From the patent application DE 10 2018 210 765 A1, a method is known in which landmarks are detected using recorded environmental data, whereby a self-position is determined based on the landmarks recorded in map data, which is used as a basis for determining a start and end point as well as a driving trajectory planning for automated driving operation.

[0004] A method for vehicle localization using road markings is known from US Patent 2014 / 343842 A1. A database of templates for classifying road markings contains templates for training images from various navigation environments. Corners of the road marking can be calculated based on a rectified or enhanced image. Positions can be defined for these corners and stored as part of a template. Similarly, a runtime image can also be rectified and enhanced. Corners of the runtime image can be calculated and matched with corners of templates from the template database; a detected match can be used to determine the location or pose of a vehicle.

[0005] High reliability of the determined self-position is a prerequisite for safe automated driving operation.

[0006] The object of the invention is therefore to provide a method and a device by which the reliability of geolocation is increased.

[0007] The problem is solved by a method for determining a reliability value for a self-position of a vehicle having the features of claim 1 and a vehicle for carrying out the method according to claim 10. The dependent claims define preferred and advantageous embodiments of the present invention.

[0008] The method according to the invention comprises the following steps: Determination of a Descriptor_Environment_t1 from features of the environment at time t1 and determination of a Descriptor_Map_t1 from map data at time t1 at the determined self-position, determination of a similarity value A1 between the Descriptor_Environment_t1 and the Descriptor_Map_t1 at time t1 and determination of a similarity value A2 between the Descriptor_Environment_t1 and a Descriptor_Environment_t0 determined and stored from features of the environment before time t1 at time t0 (with t0 <t1), wobei sofern der Ähnlichkeitswert A1 über einem vorgegebenen Schwellenwert S1 liegt und der Ähnlichkeitswert A2 unterhalb einem vorgegebenen Schwellenwert S2 liegt, Erhöhung eines zum Zeitpunkt t0 gültigen Zuverlässigkeitswertes für die ermittelte Eigenposition, where a control system is activated to enable automated driving operation as long as the reliability value is above a predetermined minimum value.

[0009] Highly automated driving systems are typically approved for a specific operational design domain (ODD), and the vehicle must be able to recognize whether this condition is met; for example, a highway chauffeur may only be offered for activation on a suitable highway, and a parking pilot only in a suitable parking lot. ODD boundaries can be readily defined geographically and are therefore often included in the high-precision maps that highly automated driving functions already use to support environmental perception. However, determining a vehicle's own position using a satellite system (GNSS), by capturing environmental features, or by any other method known from the prior art is subject to a certain degree of inaccuracy. Any potential error must never be so large that the actual vehicle position lies outside the ODD boundaries due to the possible error in determining the vehicle's own position.This means, for example, that the vehicle could be located on a parallel road not authorized for automated driving. In other words, the error in the determined self-position must be so small that the actual vehicle position cannot lie outside the ODD boundaries. To eliminate any errors in the position determined by GNSS or other methods, a descriptor (environment) determined from features of the surroundings using cameras, radar, or other sensors is compared with a descriptor map stored on a map at the position determined by GNSS or other methods. Based on the similarity of the two descriptors, determined using methods known from the prior art, it can then be assessed whether the current position is close to the position on the map or not.Nevertheless, individual descriptors can be similar even though they are based on different locations if these locations look similar, because roads and their surrounding infrastructure are often similarly constructed. To eliminate this error, the method according to the invention, preferably in a moving vehicle (i.e., at different locations at times t1 and t0), compares the current Descriptor_Environment_t1 with the Descriptor_Map_t1 at time t1 and the current Descriptor_Environment_t1 at time t1 with a Descriptor_Environment_t0 previously stored at time t0.If the first condition is met, that the current descriptor_environment_t1 is sufficiently similar to the descriptor_map_t1, and at the same time the second condition is met, that the current descriptor_environment_t1 is sufficiently dissimilar to the descriptor_environment_t0, then it is ensured that a match between the environment and map data is not determined at different locations of similar appearance, and thus a self-position is not incorrectly determined and used as supposedly reliable for automated driving operation.

[0010] For this purpose, the similarity factor A1 between Descriptor_Environment_t1 and Descriptor_Map_t1 and a similarity factor A2 between Descriptor_Environment_t1 and Descriptor_Environment_t0 are determined. Sufficient similarity, as the first condition, is given if A1 is greater than a predefined threshold S1, with A1 > S1; sufficient dissimilarity, as the second condition, is given if A2 is less than the threshold S2, with A2 <S2 ist. Üblicher Weise wird S2 <s1 gewählt. für den fall, dass s2≤s1 ist, d.h. es kann auch s2="S1" sein, gilta1>S1=S2 and A2≤S2=S1, so that A1 and A2 cannot be equal in order to satisfy the first and second conditions. If the conditions are met, then a reliability value valid at time t0 is increased, since an incorrectly determined self-position due to the similar appearance of different locations can be ruled out. The reliability value is increased or changed at a time following t1, i.e., after the comparison with the similarity values ​​A1 and A2 has been carried out.

[0011] In the case of a system start, i.e., the initial determination of the descriptor_environment_t1 after a system or vehicle start, no descriptor_environment_t0, determined and stored before time t1, exists. In this case, also known as a cold start, the determination of the similarity value A2 is omitted due to the lack of a value for the descriptor_environment_t0 existing before time t1. Instead, the reliability value is modified solely based on the similarity value A1; that is, if the similarity value A1 exceeds a predefined threshold S1, an initially specified reliability value is increased. Alternatively, a descriptor_environment_t0 can be initialized with artificial values ​​at each restart such that every theoretically determinable descriptor_environment_t1 is guaranteed to be dissimilar, thus enabling all process steps to be carried out.

[0012] In a further refinement of the procedure, the process steps are carried out continuously at time points tn following t1. In other words, the procedure is performed continuously as long as the vehicle is moving and thus is in different locations at different times. Here, according to the previously described procedure, a Descriptor_Environment_tn is compared with a Descriptor_Map_tn, and the Descriptor_Environment_tn is compared with a Descriptor_Environment_tn-1. At each time point tn, the reliability value is then either incremented, decremented, or, if necessary, set to an initial value.Advantageously, the reliability value determined over a longer period is significantly more reliable; moreover, in the event of a single failure to meet the conditions, the reliability value is not immediately reduced, but only after several repetitions to such an extent that automated driving operation has to be terminated.

[0013] In a further refinement of the procedure, the reliability value is reduced if A1 does not exceed the threshold S1. If the condition that the descriptor_environment and the descriptor_map are sufficiently similar at the same time is not met, then the previously determined reliability value is reduced, i.e., either by a predefined value or reset to an initial value. Advantageously, reducing the reliability value prevents an incorrect determination of the self-position.

[0014] In a further development of the procedure, the reliability value remains unchanged as long as A2 is not less than the threshold S2. If the second condition is not met, i.e., the descriptor_environment at one recording time is too similar to that of the previous recording time, then the environment has not changed sufficiently to increase the reliability value. In a further refinement of the procedure, a control system is activated to enable automated driving mode as long as the reliability value is above a predefined minimum value. With the activation of the control system, the functionality of automated driving is enabled; that is, the vehicle drives after authorization by a user as long as all prerequisites are met, in particular as long as the vehicle is on a route designated for automated driving, i.e., until the end of the ODD (On-Demand Data).Advantageously, if the reliability value is above the minimum value, it can be assumed that the determined own position is correct and can be driven automatically.

[0015] In another preferred embodiment, the environmental descriptor, determined at various times, is generated using machine learning from data, i.e., environmental features from radar, lidar, and / or camera. Generating the environmental descriptor based on multiple environmental data sources enables the wide-area detection of environmental features.

[0016] In a further developed embodiment, the environment descriptor, determined at various times, is derived from semantic and geometric information about the environment. Semantic information such as street signs, billboards, or proper names can be combined with geometric features of the environment to create a highly accurate environment descriptor.

[0017] In this process, a descriptor map, determined at various times for a given position on a map, is preferably generated using machine learning from semantic or graphical information stored in a database (POI). The database is located in the vehicle or on a server connected to the vehicle. Considering both semantic and graphical information enables a comprehensive description of a position based on the map data.

[0018] In a further advantageous development of the method, time intervals for determining the descriptors_environment and descriptors_map are adjusted depending on the vehicle's speed. The lower the speed, the longer the time intervals for determining the descriptors must be, so that a change in location is sufficiently large to allow for an adjustment of the reliability values.

[0019] The vehicle according to the invention is equipped for automated driving operation with a computing device for carrying out the method according to one of the preceding claims, wherein the computing unit is preferably for the driving vehicle. A descriptor_environment_t1 is determined from features of the environment at time t1, and a descriptor_map_t1 is determined from map data at time t1 at a determined self-position. A similarity value A1 is determined between the descriptor_environment_t1 and the descriptor_map_t1 at time t1, and a similarity value A2 is determined between the descriptor_environment_t1 and a descriptor_environment_t0 determined and stored from the features of the environment before time t1 at time t0. If the similarity value A1 is above a predefined threshold S1 and the similarity value A2 is below a predefined threshold S2, a reliability value valid at time t0 for the determined self-position is increased.

[0020] Further advantages, features, and details will become apparent from the following description, in which—possibly with reference to the drawing—at least one exemplary embodiment is described in detail. Features described and / or illustrated can, individually or in any meaningful combination, constitute the subject matter of the invention, possibly also independently of the claims, and can, in particular, also be the subject of one or more separate applications. Identical, similar, and / or functionally equivalent parts are designated with the same reference numerals. This shows:

[0021] Fig. 1 Flowchart of the method according to the invention and Fig. 2 Vehicle for carrying out the method Fig. 1 .

[0022] According to Fig.1 In step 100, a descriptor_environment_t1 is determined from features of a vehicle's environment, whereby the features are determined at time t1 using sensors such as a camera, lidar, and / or radar. In the following step 102, a self-position is determined using a suitable method, for example, a satellite positioning method such as GPS, and a descriptor_map_t1 is recorded for this self-position from map data at time t1.

[0023] To determine a similarity value A1, a comparison is made in step 104 between the descriptor_environment_t1 and the descriptor_map_t1. The determination of similarity values ​​between descriptors is carried out using algorithms known from the prior art, such as Levenshtein distance, histograms, Euclidean distance, etc., see also the cited prior art.

[0024] The determined similarity value A1 is compared with a predetermined threshold value S1. If the similarity value A1 does not exceed the threshold value, i.e., this first condition is not met, it is assumed that the descriptors are not sufficiently similar and the reliability value determined in a previous run of the procedure is reduced in step S110 and the procedure is restarted in step 100.

[0025] If the similarity value A1 is above the threshold value S1 A1>S1, then sufficient similarity of the descriptors is confirmed and the procedure continues in step S108.

[0026] In step S108, it is checked whether a descriptor_environment has already been stored in a previous iteration at t1. If no descriptor_environment has yet been stored, then in this branch of the procedure (relevant only for a single start iteration), the current descriptor_environment is stored, an initial reliability value is increased in step S110, and the procedure is restarted in step S100.

[0027] Is there already a Descriptor_Environment_t0 with t0 <t1 gespeichert, dann wird im Schritt S112 ein Ähnlichkeitswert A2 zwischen dem Deskriptor_Umfeld_t1 und einem vor dem Zeitpunkt t1 zum Zeitpunkt t0 bestimmten und gespeicherten Deskriptor_Umfeld _t0 bestimmt und eine zweite Bedingung geprüft, nämlich ob der Ähnlichkeitswert kleiner oder zumindest gleich einem Schwellenwert S2 ist. Mit anderen Worten wird an dieser Stelle geprüft, ob der Deskriptor_Umfeld_t1 und der Deskriptor_Umfeld _t0 ausreichend ungleich sind. Sind diese nicht ausreichend ungleich, dann kann keine zuverlässige Eigenposition bestimmt werden, da an mehreren Geo-Positionen die ausreichende Ähnlichkeit zwischen den Deskriptor_Umfeld _t1 und Deskriptor_Karte_t1 feststellbar ist und damit offen bleibt, welche die richtige ist. Für diesen Fall wird das Verfahren im Schritt S100 neu gestartet.If Descriptor_Environment_t0 and Descriptor_Environment_t1 are sufficiently unequal, then in step S110 the reliability value stored at time t0 is increased and the procedure is restarted in step S100.

[0028] The process is continuously executed in loops, i.e., in successive calculation cycles, while the vehicle is in motion, and the reliability value is updated regularly. A control unit reads the reliability value and activates automated driving mode only if the reliability value is above a predefined minimum value.

[0029] In Fig. 2 The vehicle 1 according to the invention, equipped for automated driving operation, is shown. The vehicle comprises a sensor device 3 for detecting the vehicle's surroundings, which is implemented, for example, as a camera, and a location detection device 5, which determines its own position via satellite signals and geodata. For controlling automated driving operation, a control unit 7 receives signals relating to environmental data from the camera and / or lidar and radar sensors (not shown) in addition to the geodata. At the determined own position at time t1, as determined by GPS, a processing unit 9, preferably arranged in the control unit 7, calculates a descriptor_environment_t1 from the features of the environment and a descriptor_map_t1 from map data.

[0030] The processing unit 9 further determines a similarity value A2 between the descriptor_environment_t1 and a descriptor_environment_t0 determined and stored before time t1 at time t0 with t0 <t1. Sofern als erste Bedingung der Ähnlichkeitswert A1 über einem vorgegebenen ersten Schwellenwert S1 mit A1> If the first condition is met and the similarity value A2 is below or at least equal to a predefined threshold S2 with A2 ≤ S2, the reliability value for the determined self-position is increased. If the first condition is not met, the current reliability value is decreased; if the first but the second condition is not met, the reliability value remains unchanged. Control unit 7 only enables automated driving of vehicle 1 within the ODD limits as long as the reliability value is above a predefined minimum value.

Claims

1. Method for ascertaining a reliability value for an ascertained ego position of a vehicle, which method comprises the steps of: - ascertaining a descriptor Descriptor_Environment_t1 (S100) from features of the environment, which features are ascertained by means of sensors, at a time t1 and ascertaining a descriptor Descriptor_Map_t1 (S102) from map data at a time t1 in the ascertained ego position, - ascertaining a similarity value A1 (104) between the Descriptor_Environment_t1 and the Descriptor_Map_t1 at the time t1 and - ascertaining a similarity value A2 between the Descriptor_Environment_t1 and a Descriptor_Environment_t0 determined and stored from the features of the environment before the time t1 at a time t0, wherein if the similarity value A1 is above a specified threshold value S1 and the similarity value A2 is below a specified threshold value S2, - increasing a reliability value valid at the time t0 for the ascertained ego position, wherein control for activating automated driving is switched to active if the reliability value is above a specified minimum value.

2. Method according to claim 1, characterized in that the method steps are carried out continuously at times following t1.

3. Method according to claim 1 or claim 2, characterized in that if A1 is not above the threshold value S1, the reliability value is reduced or reinitialized.

4. Method according to any of the preceding claims, characterized in that if A2 is not below the threshold value S2, the reliability value is not changed.

5. Method according to any of the preceding claims, characterized in that the Descriptor_Environment is generated by means of machine learning from data from a radar, lidar and / or camera.

6. Method according to any of the preceding claims, characterized in that the Descriptor_Environment is ascertained from semantic and geometric information from the environment.

7. Method according to any of the preceding claims, characterized in that the Descriptor_Map is ascertained from semantic or graphical information, stored in a database, with respect to a position on a map.

8. Method according to any of the preceding claims, characterized in that time intervals for determining the Descriptor_Environment and Descriptor_Map are adjusted depending on a speed of the vehicle.

9. Vehicle designed for automated driving, having a camera for capturing features of the environment ascertained by means of sensors and having a computing unit (9) for carrying out the method according to any of the preceding claims, wherein the computing unit (9) for the vehicle ascertains - a descriptor Descriptor_Enviroment_t1 from features of the environment at the time t1 and a descriptor Descriptor_Map_t1 from map data at the time t1 in an ascertained ego position, - a similarity value A1 between the Descriptor_Environment_t1 and the Descriptor_Map_t1 at the time t1 and - a similarity value A2 between the Descriptor_Environment_t1 and a Descriptor_Environment_t0 determined and stored from the features of the environment before the time t1 at the time t0, wherein if the similarity value A1 is above a specified threshold value S1 and the similarity value A2 is below a specified threshold value S2, - a reliability value valid at the time t0 for the ascertained ego position increases, wherein control for activating or deactivating automated driving is switched to active if the reliability value is above a predetermined minimum value.