CN115963499A - Method and control unit for ensuring measurement data of at least one sensor of a vehicle - Google Patents

Abstract

The invention relates to a method for securing measurement data (108) of at least one sensor (104) of a vehicle (100), wherein the measurement of the measurement data (108) takes place at a measurement location (112), a stored location-dependent error probability (120) of the sensor (104) at the measurement location (112) is read out from a location-dependent database (122) using the measurement location, and the measurement (110) is processed taking into account the error probability (120).

Description

Method and control unit for ensuring measurement data of at least one sensor of a vehicle

Technical Field

The present invention relates to a method for ensuring measurement data of at least one sensor of a vehicle, a corresponding controller and a corresponding computer program product.

Background

The sensor is capable of detecting a detection area and providing measurements regarding a detection object within the detection area as measurement data. The measurement data is composed of the individual measured values measured. The measured value may, for example, reflect the distance between the sensor and the detected point on the object. The measurement may also reflect the direction from the sensor to the point detected.

However, the measurement may also be erroneous. In this case, the erroneous measurement may be false positive or false negative. In the measurement of false positives, a point on the object is reflected in the measurement data, although no object is present at that point. In the measurement of false negatives, the presence of an object is not reflected in the measurement data.

Erroneous measurements may affect the behavior of the auxiliary systems of the vehicle. For example, a detected object which is detected as false positive may lead to an avoidance response and/or a braking response of the assistance system, which is unpredictable, in particular, for other traffic participants.

An attempt is therefore made to distinguish between correct and incorrect measurements before the measurement data is further processed.

For example, DE 10 2017 209 667 A1 describes a storage of speed information for predicting future speed trajectories.

Disclosure of Invention

Against this background, a method for ensuring measurement data of at least one sensor of a vehicle, a corresponding controller and a corresponding computer program product according to the invention are proposed in the solution proposed here. Advantageous embodiments and improvements of the solution proposed here can be achieved by the measures listed in the preferred embodiments.

In the solution proposed here, the location-related empirical values are used for plausibility checking of the measurements. The empirical values are reflected here: whether an erroneous measurement has been detected in the past at a location. The empirical values are stored here in a database with geographical reference. The database may for example represent a map with location-related empirical values.

The empirical value is stored for a specific sensor or a specific sensor class and reflects the probability that the sensor or a sensor of the same class measures an incorrect measurement at the location.

In a further process, the measurements of the locations with an increased probability of erroneous measurements are ranked as less reliable or weighted less heavily.

A method for securing measurement data of at least one sensor of a vehicle is proposed, wherein a measurement of the measurement data takes place at a measurement location, a stored location-dependent error probability of the sensor at the measurement location is read out from a location-dependent database using the measurement location, and the measurement is processed taking into account the error probability.

The idea behind embodiments of the invention may be primarily seen as based on the concepts and knowledge explained below.

The sensor is capable of detecting a detection area and reflecting an object within the detection area in the measurement data. The sensor may be an active sensor, such as a radar sensor or a lidar sensor. The sensor may likewise be a passive sensor, for example a camera. The sensor may be an integral part of a sensor system of the vehicle.

The measurement data consists of individual measurements. A measurement represents a detected point on the object in the detection area. In the case of an active sensor, the measurement represents, for example, a reflection on the object. In the case of passive sensors, for example, image points representing an image of the object are measured. The object can be reflected by multiple measurements.

The measurements may include measurement values, such as distance values and/or direction values from the measurement location to the measurement.

The measurement location may represent the position of the sensor during the measurement. I.e. the measurement is here performed at the measurement site. The measurement location may also represent the position of the vehicle during the measurement.

A false measurement may represent a point on a non-existent object that is mistaken for detection. An object that does not exist may be referred to as a ghost object. The erroneous measurements may not reflect the actual presence of the object.

Ghost objects or erroneous measurements may for example be caused by specular reflections on flat objects. Here, the mirror reflection reflects a ghost object at a position where the mirror reflection is not arranged. The position of the ghost object is here related to the angle of the flat object with respect to the measurement site. Not only active sensors but also passive sensors may be susceptible to specular reflections.

In the case of passive sensors, such as video cameras, objects that are mirrored, such as glass plates, for example, can lead to ghost objects. If the object is strongly specular, the angle of the object with respect to the measurement site is hardly important. The ghost object can then for example also be its own vehicle.

In the case of active sensors, ghost objects may appear in particular on such objects: the surface of which is irradiated at a flat incident angle. Ghost objects may occur, for example, on tunnel walls, guardrails, or other surfaces that are approximately parallel to the roadway.

The error probability may be a numerical value. The error probability may reflect: what the probability of a faulty measurement of a sensor at a location is. The error probability may be different for different sensor classes.

The location-dependent database may contain, inter alia, data of road maps. Different road sections can be assigned different error probabilities. If the measurement location is located in the region of a road section, the error probability stored for the road section can be read out. Different error probabilities may be stored in the database for different sensor classes.

If the error probability is greater than a threshold, the measurement may be removed when fusion with other measurement data is performed. The threshold may be related to the use of the measurement data. For example, for data fusion, the threshold for error probability may be between 10% and 50%. In particular, measurement data with an error probability of more than 50% can be removed. The more the purpose of the measurement data is safety-relevant the lower the threshold value may be.

The location-related database can be read in from a superordinate data processing system. The database may be centrally stored and accessible to a plurality of vehicles. The database can thus be maintained centrally. For example, changes in the course of a worksite/road may be provided centrally, and multiple participants may benefit therefrom.

Using the measurement, a location-dependent error probability of the sensor at the measurement location can be determined and stored in a location-dependent database. The saved error probability may be verified. If the determined error probability is different from the stored error probability, the stored error probability may be changed. The stored error probabilities may be changed stepwise. The change in the error probability can be reflected quickly and easily. The determined error probability may also be saved on the upper level data processing system. For example, a trained machine learning method can be used to determine the error probability.

The error probability can be determined temporally offset using additional measurements detected before and/or after the measurement location. The error probability for a measurement location may be determined after the vehicle has traveled past the measurement location. The error probability for future travel through the measurement location can be learned by retrospectively determining the error probability. The error probability can be determined, for example, by carrying out a plausibility check retrospectively on the object. In this case, longer test periods can be tested than in the case of a time-synchronized determination. For example, the object can be identified retrospectively as being not authentic and an increased error probability can be determined from such a measurement location: the object suddenly appears at the measurement site, although the object should have been detected for a longer time. Likewise, if an object suddenly disappears, although it should be able to continue to be detected, it is possible to retroactively identify the object as untrustworthy and to determine an increased error probability for the measurement location that has already traveled. Furthermore, if an object moves in a physically impossible manner, for example moves too fast, moves with too great an acceleration and/or moves in a jerky manner, the object can be identified retrospectively as being untrustworthy and an increased error probability can be determined. Likewise, if an object appears to be moving through its own vehicle, the object may be retroactively identified as being untrustworthy and an increased probability of error determined.

The error probability can be determined using at least one further measurement of at least one further sensor measured at the measurement location. The further sensor may have a detection area at least partially overlapping the aforementioned sensor. The further sensor may be a further sensor of the vehicle. The sensor may also be an external sensor that at least partially detects the detection zone. The further sensor may have a different measurement principle. The further sensors may be arranged in the region around the measurement location, i.e. with different sensor perspectives, or at adjacent measurement locations. Cross-comparison of measurements may be performed by taking multiple measurements from the same or similar measurement locations.

A further error probability of the further sensor at the measurement location can be determined and stored in a database. Due to the different error probabilities of the different sensors, measurements of sensors with low error probabilities can be used.

The error probability can be determined using the environmental information of the measurement location. For example, the environmental information may be read from a stored map. Such environmental information may also be provided by a sensor system of the vehicle. The disturbing object and/or the cause may be reflected in the environmental information. Such as tunnels and guardrails, may be reflected in the environmental information. In addition, the possible traffic region may be reflected in the environment information. Also, the location reflecting the ghost object may be reflected as impossible in the environmental information. For example, ghost objects may be arranged outside the traffic area and thus identified as impossible ghost objects. Ghost objects may also float in the air and thus be identified as impossible ghost objects.

The error probabilities determined at the measurement location at different points in time can be aggregated in a database. For example, a counter can be assigned to a measurement location in the database. The counter may be incremented at each misidentification. Another counter may be incremented each time it travels past the measurement location. The ratio of these two counters directly represents the error probability. By means of a plurality of drives, the error probability can be determined more and more accurately. In particular on commuting sections traveled almost daily, the error probability can thus be determined for a plurality of measurement points along the commuting section.

The error probabilities determined by the different vehicles at the measurement locations may be aggregated in a database. These different vehicles may have different sensors. However the working principle of these sensors may be the same. The centrally collected information enables a plurality of error probabilities to be determined for a measurement location. When the error probability at a measurement location changes, vehicles that travel past the measurement location earlier in time may retain an expected error probability for vehicles that travel past the measurement location later in time.

The method may be implemented in software or hardware or in a mixture of software and hardware, for example, in a controller.

The solution proposed here also proposes a control unit which is designed to carry out, control or carry out the steps of the variants of the method proposed here in a corresponding device.

The controller can be an electrical device having at least one computing unit for processing signals or data, at least one memory unit for storing signals or data, and at least one interface for reading in or outputting data embedded in a communication protocol and/or a communication interface. The computation unit may be, for example, a signal processor, a so-called system ASIC or a microcontroller, for processing the sensor signals and outputting data signals as a function of these sensor signals. The memory unit may be, for example, a flash memory, an EPROM or a magnetic memory unit. The interface can be designed for reading in sensor signals from the sensors and/or as an actuator interface for outputting data signals and/or control signals to the actuator. The communication interface can be designed for wireless and/or wired reading in or outputting of data. These interfaces may be, for example, software modules that are present on the microcontroller together with other software modules.

A computer program product or a computer program with a program code may be stored on a machine-readable carrier or storage medium, such as a semiconductor memory, a hard disk memory or an optical memory, and in particular when the program product or the program is executed on a computer or a device, is used to carry out, implement and/or manipulate the steps of the method according to one of the embodiments described above.

It should be noted that some of the possible features and advantages of the present invention are described herein with reference to different embodiments. Those skilled in the art will recognize that the features of the controller and method may be combined, adjusted or interchanged as appropriate to arrive at additional embodiments of the present invention.

Drawings

Embodiments of the present invention are described below with reference to the drawings. Wherein neither the drawings nor the description are to be construed as limiting the invention.

FIG. 1 shows a diagram of a vehicle having a controller according to one embodiment; and

FIG. 2 shows a diagram of a database, according to one embodiment.

The figures are purely diagrammatic and not to scale. The same reference numerals in the figures denote features which are the same or have the same effect.

Detailed Description

Fig. 1 shows a diagram of a vehicle 100 on a road. The vehicle enters a left-hand bend of the road. On the outside of the curve of the left-hand curve a guard rail 102 is arranged. The vehicle 100 has at least one forward facing sensor 104. Here, the sensor 104 is a radar sensor and emits radar waves into a detection area in front of the vehicle 100. Therefore, the radar waves are irradiated onto the road ahead of the vehicle 100, the guardrail 102, and other vehicles 106 traveling on the road ahead of the vehicle 100. The sensor 104 provides measurement data 108 reflecting measurements 110 in the detection area. The measurement 110 represents the radar waves that return to the measurement site 112 of the sensor 104. Thus, the measurement location 112 substantially corresponds to the position of the vehicle 100 at the time of the measurement 110. The measurement 110 contains at least one piece of directional information about the emission direction of the radar wave incident at the measurement location 112 and distance information about the propagation time of the radar wave between the emission and reception points in time.

The other vehicle 106 has traveled farther through the left hand curve than the vehicle 100. Radar waves from the radar sensor are reflected off the other vehicle 106 and back to the sensor 104. By means of at least one direct measurement 110, the further vehicle 106 is reflected in the measurement data 108 as a moving object at an actual distance from the vehicle 100 and in an actual direction relative to the measurement location 112 or the vehicle 100. Thus, the other vehicle 106 is reflected in its actual position in front of the left of the vehicle 100 by a direct measurement 110.

However, the further vehicle 106 is also reflected in the measurement data 108 by at least one error measurement 116 as a ghost object 118 moving at a seemingly distance and in a seemingly direction relative to the measurement location 112 or the vehicle 100. Thus, in the measurement data 108, the ghost object 118 is substantially reflected in direct front of the vehicle 100. In the case of a faulty measurement 116, this other vehicle 106 is thus reflected in the measurement data 108 at a location where it is not actually located.

The erroneous measurement 116 is here caused by the reflection of the radar wave on the guard rail 102. The radar waves are incident on the guard rail 102 at a flat angle and are deflected in the direction of the other vehicle 106, i.e. incident on the other vehicle 106 from the direction of the guard rail 102. Radar waves reflected at other vehicles 106 are likewise incident at a straight angle on the guard rail 102 and are deflected by reflection in the direction of the sensor 104. These radar waves are therefore incident on the sensor 104 from the direction of the guard rail 102. The moving ghost object 118 is therefore reflected behind the guardrail 102 in the measurement data 108.

The fixed position guard rail 102 is also reflected in the measurement data 108. But the guardrail 102 may also be filtered out of the measurement data 108 as an infrastructure object.

In the solution proposed here, the error probability 120 of the sensor 104 at the measurement location 112 is stored in a location-dependent database 122. The error probability 120 is illustrated here, for example, for the measurement location 112, that a moving object reflected in a particular direction is a ghost object 118 with a high probability, while an object reflected in another direction is a ghost object 118 with a low probability. The vehicle controller 124 reads the error probabilities 120 from the database 122 and assigns the error probabilities 120 of the respective measurement 110 or the error measurement 116 at the measurement location 112. The direct measurement 110 of the vehicle 106 is therefore assigned a low probability of error 120. A high error probability 120 is therefore associated with the erroneous measurement 116 of the vehicle 106 with respect to the guard rail 102. Due to the different error probabilities, the measurements 110 and the error measurements 116 are weighted differently in the further processing 126.

In one embodiment, if the error probability 120 is greater than the threshold 128, the measurement 110 or the erroneous measurement 116 is deleted from the sensor data 108 before further processing 126. The deletion of the measurement 110 or of the error measurement 116 can be used in particular when safety-relevant functions are controlled in a further process 126 on the basis of the sensor data 108.

In one embodiment, database 122 is stored on upper level data processing system 130 and is read in wirelessly by controller 124. The upper level data processing system 130 may be, for example, a cloud server. Thus, multiple vehicles can access the same database 122.

In one embodiment, the process measurement data 108 is analyzed in the controller 124 to determine a current error probability 120 of the measurement 110 and the erroneous measurement 116. Changes in the environment of the measurement site 112 can be detected and reflected in the database 122.

In one embodiment, the current error probability 120 is not determined time synchronously. Instead, the error probability is determined temporally offset and retrospectively. The course of the measured data 108 before and/or after the measurement location 112 can be recorded and examined later. The measurement 110 can therefore be checked for plausibility and better distinguished from a faulty measurement 116. The error probability 120 can thus be determined with high accuracy and stored in the database 122 for the next run. The determination of the error probability 120 may be performed, for example, when computing power is available in the controller 124, i.e., when no other vehicles 106 are detected and tracked, for example.

For example, in the example shown, the ghost object 118 suddenly appears at the beginning of the guardrail 102 and also suddenly disappears at the end of the guardrail 102. These sudden appearances and disappearing are very well recognized from the course of the change in the measurement data 108. Objects that appear and/or disappear suddenly are very unreliable.

Furthermore, due to reflections on the guard rail 102, the course of motion of the ghost object 118 is much more dynamic than the course of motion of the vehicle 106 measured directly. Here, the ghost object 118 can move very rapidly sideways, for example, during its movement in front of the vehicle 100, which is not usually possible with real vehicles 106 due to their inertia. Highly dynamic objects are also highly untrusted.

Since the probability of other vehicles traveling behind the guardrail 102 is very small, erroneous measurements 116 can likewise be distinguished from correct measurements 110, since there are no vehicles there with a near positive probability. Measurements 110 at locations where they are not possible are also highly unreliable. The impossible locations may be extracted from a map, for example. It is also highly unreliable if the ghost object 118 is traveling outside the road space reflected on the map at the speed of the other vehicle 106.

In one embodiment, the measurement data of at least one further sensor is evaluated in order to check the plausibility of the measurement data 108 of the sensor 104. The further sensor at least partially detects the detection area of the sensor 104. In particular, measurement data of sensors with different measurement principles can be used for plausibility checking. For example, the image of the camera can be evaluated. In these camera images it can be clearly recognized that no vehicle is driving behind the guardrail 102.

The further sensor may be a component of a sensor system of the vehicle 100. The further sensor may also be an infrastructure sensor. In particular if the further sensor is part of a sensor system, the error probability 120 at the measuring location 112 can likewise be determined for the further sensor and stored in the database 122.

In one embodiment, the vehicle 100 travels the same road segment through a left-hand turn multiple times. Here, a new error probability 120 is determined each time and combined with the previously determined error probability 120 in the database 124. Due to the slight differences between each run, a slightly different error probability 120 is obtained each time.

In one embodiment, a plurality of identified ghost objects 116 are accumulated for the measurement location 112 by traveling a plurality of times along the road segment. The ratio of the number of ghost objects 116 to the number of travels yields an error probability 120 for the measurement location 112.

In one embodiment, the error probabilities 120 determined by different vehicles are combined in the upper data processing system 130. Thus, the controller 124 can already access the error probabilities 120 stored in the database 122 when driving along the first route of the route and thus ensure the measurement data 108.

FIG. 2 shows a diagram of database 122, according to one embodiment. The database is shown here in the form of a map. The illustrated map partially reflects a left-hand curve in fig. 1. The region 200 with a high probability of error is marked along the left hand bend. Each region 200 is assigned to a measurement point 112 and is located on the outer side of the curve in the direction of the guard rail, i.e. in the direction in which more ghost objects are detected.

In one embodiment, each region 200 is assigned a numerical error probability. The measurements from these regions 200 are assigned a stored error probability before further processing.

In other words, the identification of erroneous radar measurement data on a route section which is repeatedly traveled is proposed.

Driver assistance functions and automated driving require detailed information about objects in the own vehicle environment. These objects are, for example, other traffic participants, obstacles or the course of the driving route.

Environmental sensors such as radar, video or lidar scan the surroundings and provide the necessary measurement data about objects in the vehicle environment. The measurement data is summed or filtered over time by the tracking system and the fusion system and supplemented with additional attributes, such as velocity or acceleration. This results in a consistent environment model on the basis of which driver assistance or automated driving functions can be implemented.

Unfortunately, the measurement data of the environmental sensors is often not perfect. On the one hand, these measurement data have measurement inaccuracies, which can be partially corrected by filtering over time. But the object is not measured even if it is actually present (false negative) and is provided to the measurement data even if it is not actually present (false positive).

These sensor-type defects can ultimately lead to incorrect system responses which, in the worst case, can lead to the occurrence of a traffic accident. One example is unreasonable emergency braking of an object that does not actually exist, which may lead to a rear-end collision following the vehicle.

The solution proposed here solves the problem of false positive measurement data, in particular from radar sensors. In particular for the in-home application of automated driving zones on repeatedly traveled road sections (commuter sections).

Currently available automotive radar sensors may have a relatively large amount of false positive measurement data. These false positive measurements may be as high as single digit percentage. Thus, for example, 1% of all the measured values provided may be erroneous. The reasons for these erroneous measurement data are manifold.

Modern tracking methods can in particular perform a plausibility check over time in order to filter out those measurement data which are not correlated over time.

Methods are also known from the literature which attempt to distinguish and filter out erroneous measurement data from the correct measurement data by means of classifiers or machine learning.

If further environmental sensors with overlapping fields of view are available in the vehicle, it is possible to distinguish correct measurement data from incorrect measurement data by means of a cross-comparison. The precondition for this is that different sensors do not simultaneously produce the same measurement error, which unfortunately is not always the case.

In particular, time-dependent measurement errors that occur repeatedly over a plurality of measurement cycles are problematic. One example for this is an ambiguity in the angle measurement of the radar sensor, which may be caused by the measurement principle.

In this case, for example, a real object at an angle of 45 ° may be incorrectly measured at an angle of 0 °. If the error is repeated a number of times in succession, the error can no longer be identified simply by checking the plausibility in time. The system takes as its starting point an object that does not actually exist. These objects are also called ghost objects and are a known problem for driver assistance functions or automated driving.

The solution proposed here can be used for driver assistance functions or automated driving functions in which one or more vehicles repeatedly travel on the same route. In the case of only one vehicle, this route section may be, for example, a commuting section of repeated travel.

The basic idea is to reduce the technical challenges of automated driving by accurate knowledge of the application area and its challenges. Knowledge of the distance section is used for the targeted compensation of the radar sensor defects.

The known methods for detecting faulty measured values in radar sensors are effective only to a limited extent. There are always still false measurements which may lead to undesired system reactions.

Using the solution proposed here, it is possible to mark areas in the map in which faulty measured values (see below for the solution for identification) frequently occur when driving through the same route section multiple times. This knowledge can be used in subsequent driving operations in order to assign an increased error probability to the measured values occurring there. This can take the form of an additional attribute attached to the measured value, for example. Erroneous measurement values can therefore not be taken into account in the system or at least to a lesser extent. It is thus possible to eliminate false identifications in advance and thus improve the system behavior.

In a common method, for example, a potentially erroneous measurement value of a radar sensor is marked each time a journey is made on a commute section, and the location of its occurrence is marked on a map. One example of a false measurement may be, for example, a guardrail mirror reflection that is detected by a radar sensor and is nearly indistinguishable from direct reflection.

The vehicle, the vehicle travelling in front and the guard rail at the edge of the travelling road can be seen in this figure. Radar sensors mounted on the vehicle detect a vehicle traveling ahead on a direct path (left star) and on an indirect path by mirroring on the guard rails (right star). Measurements from guardrail mirror reflections potentially lead to undesirable ghost objects because the mirror reflections are stable in time and are not filtered out by conventional tracking algorithms. Depending on the curve curvature and the speed ratio, the ghost object may become relevant to the driving function.

A common approach is to attempt to identify erroneous measurements. This can be achieved, for example, by using information from a map which indicates that there is no ground surface on the other side of the railing which can be traveled over and therefore that there is no possibility of real, moving traffic participants. Erroneous measurement values can likewise be identified by cross-comparison with other sensors (e.g. video) which may not have identified an object there. Furthermore, the measurement data can be classified by means of a classifier which has been trained by machine learning for distinguishing correct measurement values from incorrect measurement values. A false measured value can also be found by plausibility checking if the object moves in a physically impossible manner (too quickly, too quickly accelerating, too hard of a blow, through the vehicle, through other objects which are particularly highly reliable, colliding with other objects without a recognizable movement change).

If an incorrect measurement value, such as a guardrail mirror reflection, has been identified, the location is marked on the map (circle). The marking can also obtain a counter which is incremented each time a commuter segment is traveled, if a faulty measurement value again occurs there.

In addition to the location, further object properties can be used for distinguishing between repeated faulty measurements and real objects. For example, velocity vectors, acceleration vectors and/or object classes may be used.

If, in addition, the number of drives is also recorded, the counter indicates the following: how high the probability of an erroneous measurement at that location is.

In the following travel, all radar measurements are compared with the map and the error probability estimated there is obtained as an additional property.

If the error probability is above the threshold, the measurements can be removed directly.

The error probability can also be used to weight the measured values weakly during object formation (object tracking) in order to assign a low probability of existence to the tracked object or to avoid new object tracking being established at this location.

In an exemplary map, areas where erroneous radar data was identified in past driving are marked.

In the proposed solution, the tracking of objects that have been incorrectly tracked over a repeatedly traveled route section is evaluated. This enables fewer and fewer false object tracks with an estimated high probability of being present during repeated driving.

The proposed method can be used for driver assistance functions or automated driving if they are based on measurement data of environmental sensors, in particular radar sensors, and repeatedly travel over the same route.

Finally, it is pointed out that terms such as "having," "including," and the like do not exclude additional elements or steps, and that terms such as "a" or "an" do not exclude a plurality. Reference signs in the claims shall not be construed as limiting.

Claims (13)

1. A method for ensuring measurement data (108) of at least one sensor (104) of a vehicle (100), wherein a measurement (110) of the measurement data (108) takes place at a measurement location (112), a stored location-dependent error probability (120) of the sensor (104) at the measurement location (112) is read out from a location-dependent database (122) using the measurement location, and the measurement (110) is processed taking into account the error probability (120).

2. The method of claim 1, wherein the measurement (110) is removed when performing data fusion with other measurement data (108) if the error probability (120) is greater than a threshold (128).

3. The method according to any one of the preceding claims, wherein the location-related database (122) is read in from a superordinate data processing system (130).

4. The method according to any one of the preceding claims, wherein, using the measurements (110), a location-dependent error probability (120) of the sensor (104) at the measurement location (112) is determined and saved in the location-dependent database (122).

5. The method according to claim 4, wherein the error probability (120) is determined temporally offset using further measurements (110) detected before and/or after the measurement location (112).

6. The method according to any one of claims 4 to 5, wherein the error probability (120) is determined using at least one further measurement (110) of at least one further sensor (104) measured at the measurement location (112).

7. The method according to claim 6, wherein a further error probability (120) of the further sensor (104) at the measurement site (112) is also determined and saved in the database (122).

8. The method according to any one of claims 4 to 7, wherein the error probability (120) is determined using environmental information at the measurement site (112).

9. The method according to any one of claims 4 to 8, wherein the error probabilities (120) determined at the measurement locations (112) at different points in time are aggregated in the database (122).

10. The method according to any one of claims 4 to 9, wherein the error probabilities (120) determined by different vehicles (100, 106) at the measurement site (112) are aggregated in the database (122).

11. A controller (124), wherein the controller (124) is configured for implementing, implementing and/or handling the method according to any of the preceding claims in a respective device.

12. A computer program product arranged for, when executed, directing a processor for implementing, implementing and/or handling a method according to any one of claims 1 to 10.

13. A machine-readable storage medium on which the computer program product of claim 12 is stored.

Images