Local validation of digital roadmaps

Abstract

The present invention relates to a method (100) for checking whether a digital road map (3) accurately reflects actual conditions visible from at least one set pose (1 a), the method (100) comprising: obtaining (110) observations (1 b) of a scene (1) from the at least one set pose (1 a); calculating (120) from the observations (1 b) an actual behavior (2a) of one or more other road users (2); calculating (130) a target behavior (2b) of the one or more other road users (2) based on one or more features (3 a) of the digital road map (3); and in response to the actual behavior (2a) matching (140) with the target behavior (2b), verifying (150) that the digital road map (3) accurately reflects the actual conditions visible from the set pose (1 a), at least with respect to the features (3 a) for which the target behavior (2b) was calculated.

Description

[Technical Field]

[0001] The present invention relates to the inspection of digital road maps, for example for use by vehicles or driver assistance systems with at least partially automated driving. [Background technology]

[0002] Prior art Driver assistance systems and systems for at least partially automated driving utilize digital road maps to plan vehicle actions, where information is retrieved from the road map and incorporated into the plan based on the vehicle's attitude determined, for example, based on sensor data and / or information from a fully or partially satellite-aided positioning system.

[0003] In order to adapt the vehicle actions performed by the vehicles to the respective traffic conditions, the digital roadmap must be kept up to date. However, even if all updates provided by the respective manufacturers are incorporated in a timely manner, there is still a possibility that traffic conditions may change for a short period of time and therefore may not be accurately reflected by the digital roadmap. For example, construction sites may be set up overnight, making lanes unusable.

[0004] WO 2019 / 038185 discloses using a mobile device to detect corrections for a digital map previously received from an external server and to return a high-precision map calculated for the corrections to the external server. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] International Publication No. 2019 / 038185 Summary of the Invention [Problem to be solved by the invention]

[0006] Disclosure of the Invention Within the framework of the present invention, a method has been developed for checking whether a digital roadmap accurately reflects the actual conditions visible from at least one set pose. In this context, pose includes the position and orientation relative to the coordinate system of the digital roadmap. For example, the pose of a vehicle includes both the position of the vehicle and the direction in which the vehicle is facing. Both the position and the orientation determine what is exactly visible from the vehicle. This pose can be determined, for example, by any localization device and represents an input for the method described herein. The localization device can, in particular, match features visible from the vehicle to features in the digital roadmap, for example. [Means for solving the problem]

[0007] Within the framework of the method, observations of the scene from at least one pose are obtained, so that in particular the pose of a vehicle to be controlled is located in the scene and observations here can be made from said vehicle, which in common parlance is also called ego-vehicle.

[0008] From the observations, the actual behavior of one or more other road users is calculated. This may include, in particular, identifying and tracking moving objects based on the observations. From the trajectories of other road users calculated in this way, various aspects of the actual behavior of these road users can be calculated. These aspects may include, in particular, aspects that cannot be inferred directly from the scene, as opposed to, for example, traffic signs, lane markings, or other directly assessable indications.

[0009] Based on one or more features of the digital road map, a target behavior of one or more other road users is calculated. Examples of features of the digital road map that affect the target behavior of a road user include: The layout of roads and / or lanes, including any connections that exist between roads or lanes; The boundaries between lanes, including the extent to which lane changes are permitted in each lane; - Preferred driving route for driving to a given destination; - prescribed direction of travel, and ·Priority traffic rules, Examples include:

[0010] Here, it is checked whether the actual behavior matches the target behavior. If this is true at least to a set extent or according to set criteria, it is confirmed that the digital roadmap accurately reflects the actual conditions visible from the set position, at least with respect to the features from which the target behavior was calculated. On the other hand, if the actual behavior does not match the target behavior, it can be determined that one or more features of the digital roadmap associated with the target behavior are incorrectly formed, i.e., do not match the actual conditions.

[0011] For such a check to proceed successfully, on the one hand, the actual behavior of other road users must be adequately estimated, which in turn requires that the observation of the scene be detected with a sufficiently good quality.

[0012] On the other hand, the actual behavior here must correspond to the target behavior. This also requires, first of all, that the features of the digital road map that are called up based on the set attitude are accurately related to the attitude at which the scene observation was also acquired. That is, if the vehicle is in a different attitude than the actual one, for example due to an incorrectly processed GPS module, the features of the digital road map for the incorrect attitude will be called up. Furthermore, a target behavior calculated based on a completely different attitude cannot correspond to the actual behavior for a different attitude and therefore for a different traffic situation. Furthermore, other road users must behave objectively to a degree that corresponds to the target behavior.

[0013] If all these conditions are met, it can be assumed with a high degree of probability that both the detection of observations and the calculation of the vehicle's eigenattitude are functioning properly, and that the digital roadmap is also mapped realistically with respect to this pose. That is, a single inspection tests many components of the entire system one by one. Therefore, in the normal case where the actual behavior of other road users matches the target behavior, it can be assumed that the vehicle is in the correct pose and is processing using the latest digital roadmap. Therefore, there is a high degree of probability that this pose and the vehicle behavior planned using this digital roadmap are adapted to the traffic situation.

[0014] On the other hand, if the actual behavior does not match the target behavior calculated based on the digital roadmap, the exact cause cannot yet be unambiguously estimated. However, in many cases, the possible contribution of various causes can be at least partially modeled probabilistically. For example, a driver ignoring a speed limit is more likely to occur than running a red light or ignoring a no-turn sign on a freeway, since the sanctions are relatively small.

[0015] The first important thing is to confirm that the expected functionality did not occur in the first place. Such information may already be sufficient to implement countermeasures. For example, It is possible to try to independently infer erroneous corrections of the digital road map from the actual behavior of other road users, - the driver of the vehicle can be asked to take full or partial control back; The driver assistance system or the system for at least partially automated driving can be put into an operating mode with reduced functionality that makes it more susceptible to errors, or This type of automation system can be completely shut down, which can lead to, for example, placing the vehicle on a pre-planned emergency stopping trajectory, It could be something like this.

[0016] One or more other road users and / or their actual behavior can be converted into a reference frame, in particular, a digital road map. In this reference frame, the other road users and / or their actual behavior can then be associated with features of the digital road map that are related to a target behavior. Not all features registered in the digital road map affect the target behavior of all road users. In this case, for example, entry onto a road may be prohibited only for vehicles above a certain size or weight.

[0017] In particular, for example, statements about multiple other road users associated with the same feature of the digital road map can be combined into a statement about how plausible that feature is. In this way, for example, the influence of individual road users who consciously do not follow the prescribed target behavior can be reduced. The basis for the validation test is that the majority of other road users follow the set target behavior. For example, the actual behavior of a large number of other road users with respect to the feature of the digital road map can be aggregated by majority voting or a similar mechanism.

[0018] In a particularly advantageous configuration, a conditional probability or conditional success rate is calculated for at least one feature of the digital road map, which indicates that the feature is valid under the conditions of the calculated actual behavior. This allows the probabilistic characteristics of the behavior of other road users to be mapped. In this case, the success rate represents the quotient of the conditional probability that the feature is valid under the conditions of the calculated actual behavior and the conditional probability that the feature is invalid under the same conditions of the calculated actual behavior.

[0019] In another advantageous configuration, the conditional probability or conditional probability is calculated under the additional assumption of the unconditional base probability or unconditional base probability that the feature is valid. In this way, prior knowledge regarding the confidence level in the correctness of the digital roadmap can be introduced. For example, the base probability or base probability can be monotonically decreased as the digital roadmap ages. This allows, for example, to take into account the empirical rule that the percentage proportion of certain information changes on average depending on the year in which a typical city plan was published.

[0020] In this case, the conditional probability or conditional outcome expressing the plausibility of the feature under the condition of the calculated actual behavior of a large number of other road users may in particular comprise a product of conditional probabilities or conditional outcomes each expressing the plausibility of the actual behavior of one of the other road users here under the condition of the feature of the digital road map, for example according to Bayes' theorem. Such probabilities or outcomes are relatively transparent and thus easily calculated.

[0021] b1,…,b N is the actual behavior of other road users, for example, 1,...,N, and p G Let (m=1) be the unconditional base probability that a given feature m of the digital roadmap is valid without any assumption of its actual behavior. In this case, feature m is assumed to have actual behaviors b1,...,b N The conditional probability p(m=1|b1,…,b N )teeth,

number

[0022] The normalization constant c is the normalization constant for the feature m when it is compared with the actual behavior b1,...,b N The conditional validity of o(m=1|b1,…,b N ) is calculated.

number

[0023] The calculation here is

number

number

[0024] Then, from here, the conditional probability p(m=1|b1,...,b N )of,

number

[0025] As mentioned above, the observations are preferably acquired by at least one sensor carried by the vehicle. The set attitude is then calculated based on a comparison of the observations with the digital roadmap. In this case, the check of the validity of the digital roadmap can be used, in particular, to continuously check, for example, while the vehicle is traveling, whether the vehicle's action plan is based on information that is valid in itself.

[0026] As mentioned above, the primary goal is to detect discrepancies in the first place and take action to address them. Optionally, there are various means to at least narrow down the source of the discrepancy.

[0027] In an advantageous configuration, in response to the digital roadmap not accurately reflecting the actual conditions visible from the set attitude, multiple proposed modifications of the set attitude are calculated. Starting from the modified attitude according to each proposed modification, the digital roadmap is re-examined in the manner described above to see how accurately it reflects the actual conditions visible from the modified attitude. In response to the achieved improvement in this case satisfying the set criteria, it is confirmed that the calculation of the set attitude from the observations acquired by the sensors was an error.

[0028] For example, if the compass that a vehicle uses to calculate its heading has a 10° deviation from its true heading, On the one hand, the observation of the scene and the actual behavior of other road users calculated therefrom; and On the other hand, the characteristics retrieved from the digital road map and the target behavior of other road users calculated therefrom relates to a traffic situation rotated by exactly 10° relative to one another. Such a rotation can be sufficient to significantly degrade the match between the actual behavior and the target behavior. However, if repeated testing is performed on various candidate poses, each rotated by a different angle relative to the original pose, the actual behavior will match the target behavior much better for a 10° rotation angle.

[0029] In a further advantageous configuration, in response to the digital roadmap not accurately reflecting the actual conditions visible from a set pose, multiple proposed modifications to the digital roadmap are calculated. Each of the modified digital roadmaps is then re-tested to see how accurately it reflects the actual conditions visible from the set pose. Then, in response to the results of the test meeting set criteria, the previous digital roadmap is replaced with the modified digital roadmap that best reflects the actual conditions. In this manner, certain errors or inaccuracies in the digital roadmap can be automatically "cured."

[0030] For example, the target behavior expected on a highway with two lanes for each direction of travel is: When the utilization rate is low to medium, vehicles are concentrated on the right lane of the direction of travel, and the left lane is used only for overtaking, while In the event of excessive utilization, two lanes will be filled with vehicles in the same way. That is the thing.

[0031] If the automated test described above indicates, for the moment, that the actual behavior of other road users does not match the target behavior, without any specific cause being identified, a candidate for a change to the digital roadmap could be, for example, that one lane is blocked. This is one of the frequent short-term changes caused, for example, by a newly installed construction site. If the automated test is then repeated using the modified digital roadmap, and the actual behavior matches the target behavior, the original cause of the discrepancy will become clear. This recognition can be fed back, for example, to the digital roadmap manufacturer, so that the digital roadmap can be updated immediately for all users.

[0032] As mentioned above, the automatic checking of the digital roadmap is primarily used for continuous monitoring of whether the entire system consisting of the digital roadmap, vehicle attitude determination, and vehicle environment detection is still functioning properly. Thus, advantageously, in response to the digital roadmap accurately reflecting the actual conditions visible from a set attitude, the digital roadmap is used for action planning of the at least partially automated driving vehicle and / or of a driver assistance system in the vehicle, in which case the vehicle is controlled based on said action plan.

[0033] The methods can in particular be fully or partly computer-implemented. The invention therefore also relates to a computer program comprising machine-readable instructions which, when executed on one or more computers, cause the one or more computers to carry out the above-described methods. In this sense, control devices for vehicles and embedded systems for technical devices can also be considered as computers, which are likewise capable of executing machine-readable instructions.

[0034] The invention likewise relates to a machine-readable data carrier and / or download product containing a computer program, a digital product that can be transmitted over a data network, i.e. downloaded by a user of the data network, which can be sold for instant download, for example in an online shop.

[0035] Furthermore, a computer containing the computer program or comprising a machine-readable data carrier or downloadable product may also be configured.

[0036] Further measures for improving the present invention will be described in detail below in conjunction with the description of preferred embodiments of the present invention in conjunction with the drawings. [Brief explanation of the drawings]

[0037] [Figure 1]FIG. 1 illustrates an embodiment of a method 100 for validating a digital roadmap. [Figure 2] 1 shows an exemplary traffic situation in which a mismatch occurs between the target behavior 2b and the actual behavior 2a of another road user 2. FIG. DETAILED DESCRIPTION OF THE INVENTION

[0038] Example FIG. 1 is a schematic flow chart illustrating one embodiment of a method 100 for checking whether a digital roadmap 3 accurately reflects the actual conditions visible from at least one set position 1a.

[0039] In step 110, observations 1b of a scene 1 from at least one pose 1a are obtained.

[0040] In block 111, observations 1b may be acquired by at least one sensor 51 carried by the vehicle 50. Then, in block 112, a set attitude 1a may be calculated based on a comparison of the observations 1b with the digital roadmap 3.

[0041] In step 120, from the observations 1b, the actual behaviour 2a of one or more other road users 2 is calculated.

[0042] In step 130, based on one or more features 3a of the digital roadmap 3, a target behavior 2b of one or more other road users 2 is calculated.

[0043] For this purpose, in block 131, one or more other road users 2 and / or their actual behavior 2a can be transformed into the frame of reference of the digital road map 3. Then, in block 132, in said frame of reference, one or more other road users 2 can be associated with features 3a of the digital road map 3 that are related to their target behavior 2b.

[0044] In step 140, it is checked whether the actual behavior 2a matches the target behavior 2b.

[0045] In block 141, statements about multiple other road users 2 associated with the same feature 3a of the digital road map 3 can be aggregated into a statement about how relevant that feature 3a is.

[0046] In block 142, for at least one feature 3a of the digital roadmap 3, a conditional probability or conditional success rate may be calculated that indicates the validity of the feature 3a under the conditions of the calculated actual behavior 2a.

[0047] In this case, a conditional probability or conditional probability can be calculated in block 142a under the additional assumption of the unconditional base probability or unconditional base probability that feature 3a is valid, which can be monotonically decreased with increasing age of digital roadmap 3 in block 142b.

[0048] If the actual behavior 2a matches the target behavior 2b (the truth value in step 140 is 1), then in step 150 it is confirmed that the digital roadmap 3 accurately reflects the actual conditions visible from the set posture 1a, at least with respect to the features 3a for which the target behavior 2b was calculated.

[0049] In this case, in step 180, the digital roadmap 3 can be used for a plan of action 180a of the vehicle 50 and / or a driving assistance system within the vehicle 50 for at least partially automated driving. The vehicle can then be controlled in step 190 based on the plan of action 180a.

[0050] On the other hand, if the actual behavior 2a does not match the target behavior 2b (the truth value in step 140 is 0), then multiple proposed modifications 1a' of the set attitude can be calculated in step 161. Then, in step 162, the digital roadmap 3 can be re-examined for each proposed modification 1a' to see how accurately it reflects the actual conditions visible starting from the modified attitude 1a.

[0051] It can now be checked in step 163 whether an improvement over the original pose 1a is achieved that meets the set criteria. If so (truth value is 1), it can be verified in step 164 that the calculation of the set pose 1a from the observations acquired by the sensor 51 was an error.

[0052] Alternatively or in combination, multiple proposed modifications 3' of the digital roadmap 3 can be calculated in step 171. Each of the modified digital roadmaps 3 can then be re-tested in step 172 to see how accurately it reflects the actual conditions visible from the set pose 1a.

[0053] In step 173, it can be checked whether the results of such a re-examination meet the set criteria, i.e., whether they particularly show improvements over the examination of the original digital roadmap 3. If so, in step 174 the previous digital roadmap 3 can be replaced by the modified digital roadmap 3 that best reflects the actual conditions.

[0054] FIG. 2 illustrates an exemplary scene 1 in which a mismatch exists between the desired behavior 2b and the actual behavior 2a of another road user 2. A highway 11 has a right lane 11a and a left lane 11b. Observation 1b of scene 1 indicates that vehicle 2 is currently traveling in the right lane 11a. Based on a digital map 3 including highway 11, vehicle 2 is expected to initially move forward in the right lane 11a as the desired behavior 2b. However, due to a short-term construction site 12 not shown on the digital road map 3, vehicle 2 will inevitably exhibit the actual behavior 2a of moving into left lane 11b. If this is the case, particularly for a large number of vehicles 2, it can be derived from this mismatch that right lane 11a is currently unavailable, according to method 100 described above.

Claims

1. A method (100) for checking whether a digital roadmap (3) accurately reflects real conditions visible from at least one set position (1 a), comprising: - obtaining (110) observations (1b) of a scene (1) from at least one pose (1a); - calculating (120) from said observations (1b) the actual behavior (2a) of one or more other road users (2); - calculating (130) a target behavior (2b) of said one or more other road users (2) based on one or more features (3a) of said digital road map (3); - in response to the actual behavior (2a) matching the target behavior (2b) (140), verifying (150) that the digital roadmap (3) accurately reflects the actual conditions visible from the set pose (1a), at least with respect to the characteristics (3a) for which the target behavior (2b) was calculated; Including, The method (100) calculates (142) a conditional probability or conditional probability of validity of at least one feature (3a) of the digital roadmap (3) given the calculated actual behavior (2a).

2. - the one or more other road users (2) and / or their actual behavior (2a) are transformed (131) into the reference frame of the digital road map (3), In the reference frame, the one or more other road users (2) are associated (132) with features (3a) of the digital road map (3) that are related to the target behavior (2b); The method (100) of claim 1.

3. The conditional probability or conditional probability is calculated (142a) under the additional assumption of the unconditional base probability or unconditional base probability that the characteristic (3a) is valid. The method (100) of claim 1.

4. The base probability or the base outcome is monotonically decreased with increasing age of the digital roadmap (3) (142b). The method (100) of claim 3.

5. the conditional probability or the conditional probability includes a product of conditional probabilities or conditional probability, respectively, representing the plausibility of an actual behavior (2a) of one road user (2) among a plurality of other road users (2) under the condition of the feature (3a) of the digital roadmap (3); The method (100) of claim 3 or 4.

6. The observations (1b) are acquired (111) by at least one sensor (51) carried by a vehicle (50); The set attitude (1 a) is calculated (112) based on a comparison of the observations (1 b) and the digital roadmap (3). The method (100) of claim 1.

7. In response to the digital roadmap (3) not accurately reflecting actual conditions visible from the established position (1 a), A plurality of proposed modifications (1a') of the set posture (1a) are calculated (161), The digital roadmap (3) is re-examined (162) for each proposed modification (1 a') to see how accurately it reflects the actual conditions visible starting from the modified pose (1 a); In response to the improvement achieved in this case meeting a set criterion (163), it is determined (164) that the calculation of the set attitude (1 a) from the observations acquired by the sensor (51) is an error. The method (100) of claim 6.

8. In response to the digital roadmap (3) not accurately reflecting actual conditions visible from the established position (1 a), A plurality of proposed modifications (3') of the digital roadmap (3) are calculated (171); The digital roadmap (3) thus modified is then re-examined (172) to see how accurately it reflects the actual conditions visible from the set position (1a), In response to the results of the test meeting established criteria (173), the previous digital roadmap (3) is replaced (174) by a modified digital roadmap (3) that best reflects the actual conditions; The method (100) of claim 1.

9. in response to the digital roadmap accurately reflecting actual conditions visible from the established position; The digital roadmap (3) is used (180) for action planning (180a) of a vehicle (50) performing at least partially automated driving and / or for action planning (180a) of a driving assistance system in the vehicle (50), The vehicle (50) is controlled (190) based on the action plan (180a); The method (100) of claim 1.

10. 10. A computer program comprising machine-readable instructions that, when executed on one or more computers, cause the one or more computers to perform the method (100) of claim 1.

11. A machine-readable data carrier containing a computer program according to claim 10.

12. One or more computers containing a computer program according to claim 10 or comprising a machine-readable data carrier according to claim 11.

Images