Method for determining a reliability value of a vehicle self-localization and vehicle

Abstract

A method for determining a reliability value of a self-position of a vehicle is described, in which method features of the surroundings of the vehicle are detected. According to the invention, the following method steps are carried out: - at a point in time (t1), a descriptor_environment_t1 is determined from the features of the surroundings, and at the determined self-position, a descriptor_map_t1 is determined from map data at the point in time (t1); - a similarity value (A1) between the descriptor_environment_t1 at the point in time (t1) and the descriptor_map_t1 is determined, and; - a similarity value (A2) between the descriptor_environment_t1 and a descriptor_environment_t0 determined from the features of the surroundings and stored at a point in time (t0) before the point in time (t1) is determined, wherein the reliability value of the determined self-position valid at the point in time (t0) is increased if the similarity value (A1) is above a predetermined threshold value (S1) and the similarity value (A2) is below a predetermined threshold value (S2).

Description

Technical Field

[0001] The present invention relates to a method for determining a reliability value for the determination of a vehicle's own position, and to a vehicle that executes the method. Background Art

[0002] It is disclosed in the publication "Proceedings of the 2021 IEEE International Conference on Robotics and Biomimetics, December 27-31, 2021, Sanya, China; Jingrui Yu and Jianbo Sui" that descriptors are compared for images captured at different locations and corresponding similarity values are determined.

[0003] A method is known from patent application DE 10 2018 210 765 A1, in which landmarks are detected based on collected ambient environment data, wherein the own position is determined based on landmarks recorded in map data, and the own position is used as a basis for determining starting points and end points and for planning driving trajectories for automated driving operations.

[0004] Here, high reliability of the determined own position is a prerequisite for safe automated driving operations. Summary of the Invention

[0005] Therefore, the object of the present invention is to provide a method and a device by which the reliability of geolocation can be improved.

[0006] This object is achieved by a method for determining a reliability value of a vehicle's own position having the features of claim 1, and a vehicle for executing the method according to claim 10. Dependent claims define preferred and advantageous embodiments of the present invention.

[0007] The method according to the present invention comprises the following steps:

[0008] - at time point t1, determining descriptor_environment_t1 based on features of the surrounding environment, and determining descriptor_map_t1 based on map data at the determined own position at time point t1,

[0009] - determining a similarity value A1 between descriptor_environment_t1 and descriptor_map_t1 at time point t1, and

[0010] - determining a similarity value A2 between descriptor_environment_t1 and descriptor_environment_t0 determined from features of the surrounding environment and stored at time point t0 which is before time point t1 (wherein t0<t1), wherein

[0011] If the similarity value A1 is higher than the predetermined threshold S1, and the similarity value A2 is lower than the predetermined threshold S2,

[0012] - This increases the reliability value of the determined self-position at time point t0.

[0013] Highly automated driving systems are typically permitted only for use within a specific Operational Design Domain (ODD), and the vehicle must be able to detect whether this condition is met; for example, highway driving assistance may only be activated on suitable highways, and parking assistance systems may only be activated in suitable parking lots. ODD boundaries are geographically well-defined, and therefore are usually recorded in high-precision maps already used by highly automated driving functions to support environmental detection. However, determining one's own position using satellite systems (GNSS), environmental feature detection, or other methods known in the art involves a degree of error, which should not be so large as to result in the actual vehicle position being outside the ODD boundary—for example, the vehicle might be on a parallel road where automated driving is not permitted. In other words, the error in determining the self-position must be small enough that the actual vehicle position cannot be outside the ODD boundary. To eliminate potential errors in position determination via GNSS or other methods, a descriptor—environment—determined from features of the surrounding environment determined using cameras, radar, or other sensors is compared with a descriptor—map—stored in a map at the position determined by GNSS or other methods. By using methods known in the prior art to determine the similarity between two descriptors, it is now possible to assess whether the current location is near that location on the map. However, if different locations appear similar, the descriptors may be similar to each other even if they are based on those different locations, because roads and their surrounding infrastructure often have similar structures. To eliminate this error, the method according to the invention, for a preferably moving vehicle, i.e., at different locations at time points t1 and t0, compares the current descriptor_environment_t1 at time point t1 with the descriptor_map_t1, and compares the current descriptor_environment_t1 at time point t1 with the previously stored descriptor_environment_t0 at time point t0. If a first condition is met, i.e., the current descriptor_environment_t1 and the descriptor_map_t1 are sufficiently similar, and simultaneously a second condition is met, i.e., the current descriptor_environment_t1 and the descriptor_environment_t0 are sufficiently dissimilar, it ensures that the surrounding environment is not determined to match the map data at different locations with similar appearances, thus preventing the vehicle from incorrectly determining its own location and mistakenly using that location as reliable for automated driving operations.

[0014] To this end, a similarity factor A1 between the descriptor_environment_t1 and the descriptor_map_t1, and a similarity factor A2 between the descriptor_environment_t1 and the descriptor_environment_t0 are determined. If A1 is greater than a predetermined threshold S1 (where A1 > S1), sufficient similarity as the first condition is satisfied, and if A2 is less than a threshold S2 (where A2 < S2), sufficient dissimilarity as the second condition is satisfied. Typically, S2 < S1 is selected. For the case where S2 ≤ S1, that is, S2 = S1 is also possible, the condition A1 > S1 = S2 and A2 ≤ S2 = S1 applies, so that A1 and A2 cannot be equal to satisfy both the first condition and the second condition. If these conditions are met, the reliability value valid at time t0 is increased, because errors in self-position determination caused by similar appearances of different locations can be excluded. The reliability value is increased or changed at a time point after t1, that is, after the comparison with the similarity values A1 and A2 is completed.

[0015] For the case of system startup, that is, when descriptor_environment_t1 is determined for the first time after the system or vehicle is started, there is no descriptor_environment_t0 that has been determined and stored before time t1. In this case, which is also referred to as cold start, since the value of descriptor_environment_t0 existing before time t1 is lacking, the determination of similarity value A2 is omitted, and the reliability value is changed only based on similarity value A1, that is, if the similarity value A1 is higher than the predetermined threshold S1, the initial predetermined reliability value is increased. Alternatively, the descriptor_environment_t0 can be initialized with such an artificially generated value each time the system is restarted, such that any theoretically determinable descriptor_environment_t1 is necessarily dissimilar, so that all method steps can be performed.

[0016] In another design of the method, the method steps are continuously performed at time points tn after t1. In other words, as long as the vehicle is moving and thus is located at different positions at different time points, the method is continuously executed. In this case, according to the foregoing method, the descriptor_environment_tn is compared with the descriptor_map_tn, and the descriptor_environment_tn is compared with the descriptor_environment_tn-1. Then, at each time point tn, the reliability value is incremented, decreased, or set to an initial value when appropriate. Advantageously, the content of the reliability value determined over a long period of time is significantly more reliable; in addition, in the case of a single non-satisfaction of the conditions, the reliability value is not immediately decreased, but is only decreased to the extent that automated driving operation has to be aborted when the conditions are not satisfied multiple times repeatedly.

[0017] In another design of this method, if A1 is not higher than the threshold S1, the reliability value is reduced. If the condition that the descriptor_environment and descriptor_map are sufficiently similar to each other at the same time point is not met, the previously determined reliability value will be reduced, i.e., either the predetermined value is reduced or the value is reset to the initial value. Advantageously, by reducing the reliability value, incorrect determination of its own position is avoided.

[0018] In an improved version of this method, if A2 is not less than the threshold S2, the reliability value is not changed. If the second condition is not met, i.e., the descriptor_environment at a certain acquisition time point is too similar to the descriptor_environment at the previous acquisition time point, then the surrounding environment has not changed enough to improve the reliability value.

[0019] In another design of this method, the control device for activating or deactivating automated driving is activated as long as the reliability value is higher than a predetermined minimum value. With the activation of the control device, the functionality of automated driving is activated; that is, after user authorization, the vehicle will automatically drive as long as all conditions are met (especially as long as the vehicle is on the pre-set route for automated driving, i.e., up to the ODD endpoint). Advantageously, when the reliability value is higher than the minimum value, the determined self-position can be considered correct, and automated driving can proceed.

[0020] In another preferred embodiment, machine learning is used to create descriptors_environments defined at different time points from data, i.e., from environmental features collected from radar, lidar, and / or cameras. Creating descriptors_environments based on multiple data sources of environmental data enables the large-scale collection of environmental features.

[0021] In an improved implementation, a descriptor_environment is determined at different points in time from semantic and geometric information of the environment. Semantic information such as road signs, billboards, or proper names can be combined with the geometric features of the surrounding environment to form a descriptor_environment that describes the environment with high precision.

[0022] In an embodiment of this method, for a given location on a map, machine learning is preferably used to determine a descriptor—a map—at different points in time from semantic or graphical information stored in a database (POI). This database is stored in the vehicle or on a server connected to the vehicle. Considering both semantic and graphical information makes it possible to provide a comprehensive description of a location based on map data.

[0023] In a favorable improvement to this method, the time intervals used to determine the descriptor_environment and descriptor_map are adjusted based on the vehicle's speed. The lower the speed, the larger the selected time interval should be to ensure that location changes are sufficiently significant relative to environmental changes, thereby allowing for adjustment of the reliability values.

[0024] The vehicle according to the invention is adapted for automated driving operations, the vehicle having a computing unit for performing the method according to any one of the preceding claims, wherein, for a preferably moving vehicle, the computing unit:

[0025] - At time point t1, the descriptor _environment_t1 is determined based on the characteristics of the surrounding environment, and at the determined self-location, the descriptor _map_t1 is determined based on map data at time point t1.

[0026] - Determine the similarity value A1 between descriptor_environment_t1 and descriptor_map_t1 at time point t1, and

[0027] - Determine the similarity value A2 between the descriptor_environment_t1 and the descriptor_environment_t0 determined and stored from the features of the surrounding environment at time point t0 before time point t1, where

[0028] If the similarity value A1 is higher than the predetermined threshold S1, and the similarity value A2 is lower than the predetermined threshold S2,

[0029] - This increases the reliability value of the determined self-position at time point t0. Attached Figure Description

[0030] Further advantages, features, and details are derived from the following description, in which at least one embodiment is described in detail—referring, if applicable, to the accompanying drawings. The described and / or illustrated features may constitute the subject matter of the invention individually or in any reasonable combination, and may also be independent of the claims, and in particular, may be the subject matter of one or more separate applications. Identical, similar, and / or functionally identical components are labeled with the same reference numerals.

[0031] in:

[0032] Figure 1 A flowchart of the method according to the present invention is shown; and

[0033] Figure 2 Showing the execution Figure 1 The method in the vehicle. Detailed Implementation

[0034] according to Figure 1 In step 100, a descriptor_environment_t1 is determined from features of the vehicle's surrounding environment, wherein these features are determined at time point t1 using sensors such as cameras, lidar, and / or radar. In the subsequent step 102, the vehicle's own position is determined using a suitable method (e.g., satellite positioning method such as GPS), and for that position, a descriptor_map_t1 is detected based on map data at time point t1.

[0035] In order to determine the similarity value A1, in step 104, the descriptor_environment_t1 is compared with the descriptor_map_t1. The similarity value between descriptors is determined by an algorithm known in the prior art, such as Levenshtein distance, histogram, Euclidean distance, etc., reference can also be made to the cited prior art.

[0036] The determined similarity value A1 is compared with a predetermined threshold S1. If the similarity value A1 does not exceed the threshold, that is, the first condition is not satisfied, it is considered that the descriptors are not sufficiently similar, the reliability value determined in the previous run of the method is decreased in step S110, and the method is restarted in step 100.

[0037] If the similarity value A1 is higher than the threshold S1 (A1>S1), it is confirmed that the descriptors are sufficiently similar, and the method proceeds to step A108.

[0038] In step S108, it is checked whether the descriptor_environment has been stored in a previous run at t1. If the descriptor_environment has not been stored, then in the branch of the method that is only applicable to the starting run, the current descriptor_environment is stored, the initial reliability value is increased in step S110, and the method is restarted in step S100.

[0039] If the descriptor_environment_t0 (where t0<t1) has been stored, then in step S112, the similarity value A2 between the descriptor_environment_t1 and the stored descriptor_environment_t0 determined at the time point t0 before the time point t1 is determined, and the second condition is checked, that is, whether the similarity value is less than or at least equal to the threshold S2. In other words, it is checked here whether the descriptor_environment_t1 and the descriptor_environment_t0 are sufficiently different.

[0040] If the two are not sufficiently different, a reliable self-position cannot be determined, because sufficient similarity between the descriptor_environment_t1 and the descriptor_map_t1 can be detected at multiple geographical positions, thus it is impossible to clearly determine which one is correct. For this case, the method is restarted in step S100.

[0041] If the descriptor_environment_t0 and the descriptor_environment_t1 are sufficiently different, the reliability value stored at the time point t0 is increased in step S110, and the method is restarted in step S100.

[0042] The method is continuously executed in a cyclic manner (that is, in continuous calculation cycles) for a moving vehicle, and the reliability value is updated regularly. The control unit reads the reliability value, and activates the automated driving operation only when the reliability value is higher than a predetermined minimum value.

[0043] Figure 2 Disclosed is a vehicle 1 adapted for automated driving operation according to the present invention. The vehicle comprises a sensing device 3 for collecting the surrounding environment of the vehicle, which is embodied as a camera for example, and a position detection device 5, which determines its own position based on geographic data via satellite signals. In order to control the automated driving operation, the control unit 7 receives geographic data and also receives signals related to surrounding environment data from the camera, as well as lidar and radar sensors not shown. At the determined own position at time point t1 determined by GPS, the computing unit 9 preferably arranged in the control unit 7 determines descriptor_environment_t1 from the features of the surrounding environment, and determines descriptor_map_t1 from the map data.

[0044] The computing unit 9 further determines a similarity value A2 between descriptor_environment_t1 and descriptor_environment_t0 determined and stored at time point t0 before time point t1 (where t0<t1). If, as a first condition, the similarity value A1 is higher than a predetermined first threshold S1 (where A1>S1), and as a second condition, the similarity value A2 is lower than or at least equal to a predetermined threshold S2 (where A2≤S2), the reliability value of the determined own position is increased. If the first condition is not satisfied, the current reliability value is decreased; if the first condition is satisfied but the second condition is not, the reliability value remains unchanged. The control unit 7 enables the automated driving operation of the vehicle 1 within the ODD boundary only when the reliability value is higher than a predetermined minimum value.

Claims

1. A method for determining a reliability value of a vehicle's determined self-position, the method comprising the following steps: - At time point t1, determine the descriptor _environment_t1 based on the characteristics of the surrounding environment (S100), and at the determined self-location, determine the descriptor _map_t1 based on the map data at time point t1 (S102). - Determine the similarity value A1 (104) between the descriptor_environment_t1 and the descriptor_map_t1 at time point t1, and - Determine the similarity value A2 between the descriptor_environment_t1 and the stored descriptor_environment_t0 determined at a time point t0 before time point t1 based on the features of the surrounding environment, where If the similarity value A1 is higher than a predetermined threshold S1, and the similarity value A2 is lower than a predetermined threshold S2, - This improves the reliability value of the determined self-position at time point t0.

2. The method according to claim 1, characterized in that, The method steps are executed continuously at time points after t1.

3. The method according to claim 1 or 2, characterized in that, If A1 is not higher than the threshold S1, then the reliability value is reduced or reinitialized.

4. The method according to any one of the preceding claims, characterized in that, If A2 is not less than the threshold S2, then the reliability value is not changed.

5. The method according to any one of the preceding claims, characterized in that, The control device for activating or deactivating automated driving operation is activated as long as the reliability value is higher than the predetermined minimum value.

6. The method according to any one of the preceding claims, characterized in that, The descriptor_environment is created using machine learning based on data from radar, lidar, and / or cameras.

7. The method according to any one of the preceding claims, characterized in that, The descriptor_environment is determined from the semantic and geometric information of the environment.

8. The method according to any one of the preceding claims, characterized in that, For a given location on the map, the descriptor_map is determined based on semantic or graphical information stored in the database.

9. The method according to any one of the preceding claims, characterized in that, The time interval used to determine the descriptor_environment and the descriptor_map is adjusted based on the vehicle's travel speed.

10. A vehicle adapted for automated driving operations, the vehicle having a camera for acquiring features of the surrounding environment and a computing unit (9) for performing the method according to any one of the preceding claims, wherein, for the vehicle, the computing unit (9): - At time point t1, a descriptor _environment_t1 is determined based on the characteristics of the surrounding environment, and at the determined self-location, a descriptor _map_t1 is determined based on the map data at time point t1. - Determine the similarity value A1 between the descriptor_environment_t1 and the descriptor_map_t1 at time point t1, and - Determine the similarity value A2 between the descriptor_environment_t1 and the stored descriptor_environment_t0 determined at time point t0 before time point t1 based on the features of the surrounding environment, where If the similarity value A1 is higher than a predetermined threshold S1, and the similarity value A2 is lower than a predetermined threshold S2, - This improves the reliability value of the determined self-position at time point t0.

Images