EP4003802A1

EP4003802A1 - Method and system for providing a context-dependent knowledge base for plausibilizing at least one detection function

Info

Publication number: EP4003802A1
Application number: EP20746157.5A
Authority: EP
Inventors: Fabian HÜGER; Peter Schlicht; John Serin VARGHESE; Nikhil KAPOOR
Original assignee: Volkswagen AG
Current assignee: Volkswagen AG
Priority date: 2019-07-31
Filing date: 2020-07-22
Publication date: 2022-06-01
Also published as: WO2021018704A1; DE102019211457A1; CN114144795A

Abstract

The invention relates to a method for providing a context-dependent knowledge base (20) for plausibilizing at least one detection function (30), wherein subscribers (40) of a vehicle fleet (51) locally each perform at least one detection function (30) and generate at least one context-dependent knowledge datum (10) on the basis of at least one output from the at least one detection function (30), wherein the generated context-dependent knowledge data (10) are transmitted from the subscribers (40) of the vehicle fleet (51) to a backend server (3), wherein the transmitted context-dependent knowledge data (10) are collected on the backend server (3) and have their contents compared against one another, wherein a comparison result is taken as a basis for generating or updating a context-dependent knowledge base (20), wherein the generated or updated context-dependent knowledge base (20) is transmitted to the subscribers (40) of the vehicle fleet (51), and wherein the transmitted context-dependent knowledge base (40) is used by the subscribers (40) locally to plausibilize and/or support the at least one detection function (30). The invention also relates to a system (1) for providing a context-dependent knowledge base (20) for plausibilizing at least one detection function (30).

Description

description

Method and system for providing a context-dependent knowledge base for

Checking the plausibility of at least one perception function

The invention relates to a method and a system for providing a context-dependent knowledge base for plausibility checking at least one perception function. The invention also relates to a vehicle and a backend server.

It is known to generate a knowledge base on the basis of manual knowledge modeling from existing knowledge in the form of, for example, graphs and, on the basis of this, to form hypotheses about an environment or objects. This can be used, for example, for classification and object recognition. However, the problem here is that all relevant semantic relationships have to be modeled manually. It is also known to generate knowledge in the form of class-specific occupancy maps (also referred to as heat maps) from data records and to weight a decision based on this, for example in the case of semantic segmentation. However, the creation and use of occupancy cards is data record-specific, so that the

Occupancy cards can vary greatly depending on the context. It is also known to use methods of machine learning in order to extract patterns from a data set and then to use the extracted patterns to increase the performance of a function, e.g. when extracting shapes or functions in

two-dimensional or three-dimensional mapping of objects. Here, too, the extracted patterns are very much dependent on the data set used during training.

The invention is based on the object of a method and a system for providing a context-dependent knowledge base for checking the plausibility of at least one

To improve perceptual function.

The object is achieved according to the invention by a method with the features of

Claim 1 and a system with the features of claim 13 solved. Furthermore, the object is achieved by a vehicle with the features of claim 14 and a backend server with the features of claim 19. Beneficial

Refinements of the invention emerge from the subclaims. In particular, a method for providing a context-dependent knowledge base for plausibility checking of at least one perception function is provided, whereby participants of a vehicle fleet each locally execute at least one perception function and generate at least one context-dependent knowledge datum based on at least one output of the at least one perception function, the context-dependent knowledge data generated by the participants of the vehicle fleet to one

Backend server are transmitted, the transmitted context-dependent knowledge data are collected on the backend server and their content is compared with each other, with a context-dependent knowledge base being generated or updated based on a comparison result, the generated or updated context-dependent knowledge base being transmitted to the participants in the vehicle fleet, and the The context-dependent knowledge base transmitted is used locally by the participants to check plausibility and / or to support the at least one perception function.

Furthermore, a system for providing a context-dependent knowledge base for checking plausibility of at least one perception function is created, comprising

Control devices of participants in a vehicle fleet, and a backend server, the control devices each being designed to generate at least one context-dependent knowledge data based on at least one output of at least one perception function performed locally by the participant and to transmit the generated at least one context-dependent knowledge data to the backend server , wherein the backend server is designed to collect the knowledge data transmitted by the participants and to compare their content with one another and based on a

To generate or update a context-dependent knowledge base based on the comparison results, and to transmit the generated or updated context-dependent knowledge base to the participants in the vehicle fleet, and wherein the control devices are also each designed in such a way that the transmitted context-dependent knowledge base is locally checked for plausibility and / or support of the at least one perception function to use.

The method and the system enable an improved context-sensitive

Provide knowledge base with which the at least one perception function can be checked for plausibility or checked. This is done in that the water for the water base is generated on a fleet basis. Participants in a vehicle fleet each run at least one locally

Perception function and generate at least one context-dependent knowledge datum based on at least one output of the at least one perception function. The The context-dependent knowledge data generated are transmitted to a backend server by the participants in the vehicle fleet. There the transmitted context-dependent

Knowledge data collected. The content of the context-dependent knowledge data collected is compared with one another in the back-end server, that is, in particular, merged. On the basis of a comparison result, a context-dependent knowledge base is generated if no knowledge base is available yet, or an already existing context-dependent knowledge base is updated on the basis of the currently transmitted context-dependent knowledge data and their comparison. The updating here comprises in particular an expansion of the context-dependent knowledge base. The contextual knowledge base includes

in particular a set of merged context-dependent knowledge data. The generated or updated context-dependent knowledge base is transmitted from the backend server to the participants in the vehicle fleet. The transmitted context-dependent knowledge base is used locally by the participants to check plausibility and / or to support the at least one perception function.

One advantage of the invention is that a context-dependent knowledge base can be generated and updated based on the fleet. As a result, the knowledge included in the knowledge base is not limited in scope and can also be kept up to date and expanded continuously and depending on the context. Updating and expanding can be done automatically. Furthermore, the creation of the knowledge base by the fleet ensures that only knowledge that can be experienced by fleet participants is stored in the knowledge base.

A context or the term context-dependent should in particular mean that the

Knowledge date is generated or valid for a specific context. The context can be local (e.g. in the form of coordinates, or through an infrastructural or geographical reference), temporal (e.g. time, day, day of the week, week, season), situation-related (e.g. village, city, country, parking lot), weather-related ( e.g.

Sunshine, rain, snow, fog, wind) and / or traffic-related (e.g. little traffic, dense traffic, traffic jam, many different road users, traffic light phase). In principle, it is also possible to take into account other context dimensions, in particular also dimensions to which humans have no intuitive access, but which can only be determined using machine learning methods. A context for a knowledge datum can be, for example, from sensor data of a current environment of the participant in the vehicle fleet, recorded by means of at least one sensor and / or can be derived from other sources (weather service, etc.) taking into account data, for example by means of a context recognition device.

A perception function particularly refers to a function that is carried out for the purpose of detecting and / or interpreting the surroundings. The perception function can serve, for example, semantic segmentation, object recognition and / or a prediction of the behavior of other road users. The perception function can for example be based on a trained deep neural network and / or on

Based on a Bayesian network. In principle, a

Perception function can also be realized differently. The perception function is performed, for example, by means of a control device of a participant in the vehicle fleet

provided. For this purpose, the perception function, in particular the control device, is supplied in particular with sensor data that was recorded by means of at least one sensor of the participant in the vehicle fleet. A sensor can be, for example, a camera, a lidar sensor, a radar sensor or an ultrasonic sensor. The perception function is carried out on the recorded sensor data and, based on the sensor data, provides a result or output (e.g. semantic segmentation, object recognition, etc.). The result or the output of the at least one perception function is then fed in particular to an automated driving function which, based on the result or output, controls the participant in the vehicle fleet, in particular an at least partially automated vehicle.

The at least one perception function used to generate the at least one knowledge datum can in principle be a different perception function than the at least one that is plausible or checked by means of the context-dependent knowledge base

Perceptual function. Ideally, however, the functions of perception are the same.

A context-dependent knowledge datum denotes, in particular, a context-dependent relationship, that is to say a relationship between objects and / or environment properties in a specific context. The context-dependent knowledge data or context-dependent knowledge data can be generated in particular in the form of occupancy maps, object occurrences, object relationships, object properties and / or behavior of the objects. In simple terms, the context-dependent knowledge date or the

give or enable context-dependent knowledge data about which context, which environment with which objects and their object properties, etc.

is typically expected. An occupancy card indicates, for example, how likely it is that certain objects (eg pedestrians or traffic signs) will be present in certain areas of the environment. For example, in the area of a street crossing or a zebra crossing, it is more likely that pedestrians will be found on the road than on the lane of a motorway.

Object occurrences generally indicate which objects or which types or classes of objects can occur in a context. In the previous example, for example, in the context of a motorway, the pedestrian object would not occur or would be linked with an extremely low probability. The motor vehicle object, on the other hand, would be considered extremely likely.

Object relationships indicate relationships between different or similar objects or relationships within or between object classes, in particular a common occurrence of objects in a specific context. Examples of this are the following statements: a person has legs, people can be found at a bus stop, or people can be found at a zebra crossing.

Object properties denote, for example, the colors, shapes, reflection properties or relationships between classes (person on a bicycle, trucks are larger than pedestrians, children move faster than trees) of objects.

Behaviors describe the typical behavior of objects in a specific context. For example, a pedestrian at a zebra crossing with high

It is likely that the pedestrian wants to cross the street or will cross it. Another example is the assumption of a continuous / changing speed of a motor vehicle or a pedestrian etc. For example, another motor vehicle in a right-turn lane is very likely to turn right, etc.

A participant in the vehicle fleet is in particular a vehicle, in particular a

Motor vehicle. The vehicle is in particular an at least partially automated driving vehicle, that is to say at least one (partially) automated driving function is carried out in the vehicle. In principle, however, a participant can be another land, water, air or space vehicle that forms a vehicle fleet with other participants. The control devices of the participants and the backend server communicate in particular via communication interfaces designed for this purpose and a communication link. In particular, this is a wireless communication connection (WLAN, 4G, 5G, ...). The context-dependent knowledge data and the context-dependent knowledge base are

for example in the form of data packets.

Parts of the control devices can be designed individually or collectively as a combination of hardware and software, for example as program code that is executed on a microcontroller or microprocessor. However, it can also be provided that parts are designed individually or combined as an application-specific integrated circuit (ASIC).

The transmission of the at least one context-dependent knowledge data from the

Participants in the backend server take place in particular at time and / or local intervals and / or as a function of progress in generating the context-dependent knowledge data, in particular as a function of a number and / or a data volume of the context-dependent knowledge data generated.

The comparison of the context-dependent knowledge data transmitted to the back-end server takes place in particular by consolidating the context-dependent knowledge data. For this purpose, the individual knowledge data can be offset against one another,

for example by averaging and / or mapping the relationships using statistical distributions and / or expected values, etc. In particular, this can

Matching or consolidating the knowledge data comprise correlation analyzes, on the basis of which the knowledge data is generalized. The correlation analysis answers the question of which constellations of objects and / or

Environmental properties etc. occur simultaneously (positive correlation) or do not occur simultaneously (negative correlation). If the corresponding correlations are found in the knowledge data of several participants, the underlying relationships are generalized. In simple terms, after performing the comparison, you want to save as

Matching result averaged or generalized relationships between objects and / or environmental properties etc. are available for the contexts under consideration via the participants in the vehicle fleet. The context-dependent knowledge base is then generated from the comparison result, i.e. from the averaged relationships

context-dependent knowledge base is updated. When updating, in particular be provided that an already existing context-dependent knowledge base when

Matching is taken into account.

When comparing on the backend server, it is also possible, in particular, for a knowledge datum transmitted by a participant to be discarded, that is to say is not taken into account when the context-dependent knowledge base is generated and / or updated. This is particularly the case when the knowledge date is far apart from other averaged

Knowledge data is and therefore there is a high probability that a sensor error and / or measurement error and / or a transmission error was present. Whether a knowledge datum should be discarded can be determined statistically, for example. For a simple example in which context-dependent knowledge data include values that statistically follow a normal distribution, a knowledge data can be discarded, for example, if it lies outside a predetermined range (e.g. twice or three times the standard deviation) of this normal distribution.

The generated or updated context-dependent knowledge base is used on the part of the participants in the vehicle fleet to check plausibility and / or to support the at least one

Perception function used. This means in particular that a result of the at least one perception function is checked for plausibility on the basis of the context-dependent knowledge base. On the other hand, for example, before an application of the

Perception function parameters of the perception function can already be selected and / or set on the basis of the context-dependent knowledge base, so that the at least one perception function is supported in a current context. If the perception function is, for example, a function for recognizing pedestrians in camera images captured from the surroundings, support can be provided in that an area is identified on the basis of the context-dependent knowledge base in which

usually pedestrians can be found. The perception function can then evaluate this area with a particularly high resolution. It is also possible to search for properties of the pedestrians that have changed depending on the context in order to identify pedestrians in the vicinity. If the current context includes rainy weather, for example, it can be extracted for pedestrians from the context-dependent knowledge base that, depending on the context, pedestrians are around half a meter taller than in dry weather due to open umbrellas. For example, search requests for pedestrians in certain areas of the environment can also be specified on the basis of the context-dependent knowledge base (eg "find pedestrian / object y in area x"). The perception function is then adjusted accordingly or a result of the perception function is checked for plausibility accordingly.

It is provided in particular that for a result of the at least one

Perception function based on the context-dependent knowledge base

Plausibility value is calculated. This plausibility check value can then be used to evaluate the result of the at least one perception function. A plausibility value specifies in particular how plausible a result or an output of the at least one perception function is in the given context, i.e. in particular how high a probability is for the correctness of the result or the output (statistical or derived from a history) the given context. On the basis of the plausibility check value, it can be decided, for example, that a result of the at least one perception function can be trusted or that the result should be discarded. This is done, for example, by comparing the determined, in particular calculated, plausibility check value with a (minimum plausibility) threshold value. If the threshold value is not reached, the result of the at least one perception function can be discarded or, for example, taken into account with a lower weighting in subsequent further processing, in particular when performing an automated driving function of a vehicle. In particular, when using the result of the at least one perception function to implement an automated driving function, this can increase safety.

It is therefore provided in one embodiment that the context-dependent knowledge base is used during the plausibility check to determine a plausibility check value for a result provided by the at least one perception function.

In one embodiment it is provided that a plausibility check result of

Plausibility check, in particular in the form of the specific plausibility check value, at

Establishing and / or setting at least one parameter of a driving strategy and / or maneuver planning of a participant in the vehicle fleet is taken into account. For example, it can be provided that the at least one parameter relates to a type of driving strategy and / or maneuver planning. For example, the parameter can determine whether the driving strategy is more defensive or more offensive. For this purpose, for example, a corresponding control signal or a corresponding control command is generated and provided which defines and / or defines the parameter for the driving strategy as a function of the plausibility check result adjusts. A maneuver planning can, for example, be changed in terms of comfort or “sportiness”.

In one embodiment it is provided that in order to support the at least one perception function, parameters of the at least one

Perception function can be selected and / or adjusted on the basis of the context-dependent knowledge base. Examples of this embodiment have already been described above.

In one embodiment it is provided that a relevance of the context-dependent knowledge data transmitted to the back-end server is assessed by means of the back-end server. This can improve the comparison, since, for example, only knowledge data are taken into account that achieve a certain evaluation. In the context of the evaluation, for example, compliance with the law or conventional rules can be assessed. An ethical assessment can also be carried out.

The basis for such an assessment can in particular be:

Regulatory requirements (the minimum quality to be achieved is stipulated and any knowledge that leads to an expected increase in functional quality is assessed as relevant);

Latency requirements (maximum knowledge representation is allowed, which allows plausibility checking in a given maximum latency in every context);

Usability requirements (only such knowledge data are used that occur in a minimum amount of plausibility checks);

Generality requirements (only knowledge that is relevant to a minimum amount of different situations is considered relevant);

Ethical requirements (certain knowledge, even if it is potentially factually correct, can bring negative bias into the behavior of systems - such knowledge should not have any plausibility-checking influence).

In one embodiment it is provided that the correctness of the context-dependent knowledge data transmitted to the backend server is assessed by means of the backend server. The assessment of the correctness should in particular include checking a logical consistency of the context-dependent knowledge data. The logical consistency can be based, for example, on physical laws and / or traffic rules and / or other behavioral conventions and social rule systems are checked.

The basis for such an assessment can in particular be:

- World knowledge: an assessment based on existing rule-based knowledge can be used for certain knowledge data; this applies in particular to physical, geometric or traffic regulation-related knowledge;

Expert assessments: certain, interpretable knowledge data may not be able to be covered by existing rule-based knowledge, but can definitely be assessed by experts; such experts can be: lawyers, government authorities, scientists, technicians (engineers, computer vision experts, etc.), ethics committees etc .;

Evaluations on a large basis of training data or fleet experience;

Logic (applications of logical operations such as links, logical

Inferring, generalizing or specializing).

In one embodiment it is provided that the at least one context-dependent knowledge datum generated locally by the participants is locally for plausibility checking and / or

Support at least one perceptual function is used. As a result, context-dependent knowledge data collected locally by the individual participants can already be used to check plausibility and / or to support the at least one perception function. This already improves the performance of the at least one without the context-dependent knowledge base that is compared or consolidated by the backend server

Perceptual function.

In particular, it can be provided that the participants in the vehicle fleet locally generate or update a context-dependent knowledge base on the basis of the at least one generated knowledge data and use this to check and / or support the at least one perception function.

In one embodiment it is provided that context-dependent knowledge data are annotated by the participants in the vehicle fleet as a function of the degree to which they have contributed to an increase in the performance of the at least one perception function and / or to the plausibility check of the at least one perception function, with at least annotations that contain a predetermined Exceeding the threshold value will be transmitted to the backend server. In this context, annotating is intended to mean in particular that a context-dependent knowledge data item is linked to a value that is representative of the Degree to which this increases the performance of at least one

Perception function and / or to check the plausibility of the at least one

Perceptual function has contributed. This allows an assessment of the

context-dependent knowledge base stored context-dependent knowledge data according to their relevance or a level of support for which at least one perception function are performed. The context-based knowledge base can be improved step by step as a result, since non-relevant context-dependent knowledge data that allow no or only a slight increase in the performance of the at least one perception function can be removed in favor of context-dependent knowledge data that have a

Enable performance improvement. In the simplest case, it is only annotated whether a

context-dependent knowledge date for an increase in performance of at least one

Perception function and / or to check the plausibility of the at least one

Perceptual function contributed (positive annotation) or not (negative annotation). However, it can also be provided that further intermediate stages are provided for the evaluation.

It can also be provided that only context-dependent knowledge data is sent to the

Backend servers are transferred for which an associated annotation exceeds the specified threshold. In this way, a preselection can already be made, so that the number of context-dependent knowledge data to be transferred and compared can be reduced. This can save storage space and data volume in the

Communication between the participant and the backend server can be saved.

In a further-developing embodiment it is provided that the context-dependent

Knowledge base is updated taking into account the transmitted annotations. In particular, the context-dependent knowledge data, which enable a high level of support, are given preference.

In one embodiment it is provided that, in order to update the context-dependent knowledge base, a probability-based assessment of the context-dependent knowledge base

Knowledge data is carried out. In particular, this can include all known methods of probabilistic, that is to say probability-based, inference.

It can also be provided that, in order to update the context-dependent knowledge base, the context-dependent knowledge data are evaluated or linked to one another by means of fuzzy logic. In one embodiment it is provided that a contextual refinement of the context-dependent knowledge base takes place in order to update the context-dependent knowledge base. As a result, the context-dependent knowledge data can be made available even more specifically for a specific, namely more finely delimited, context. The contextual refinement includes in particular the dimensions of place, time, situation, weather and / or traffic. In principle, it is also possible to take into account other context dimensions, in particular also dimensions to which humans have no intuitive access, but which can only be determined using machine learning methods.

In one embodiment it is provided that only a selected part of the participants in the vehicle fleet generates context-dependent knowledge data and transmits it to the back-end server. As a result, these selected participants can be equipped, for example, with improved (and therefore usually more expensive) environmental sensors and increased data storage capacity. In other words, a distinction is made between one

Extraction fleet, the participants of which generate the context-dependent knowledge data, and a total fleet, which in addition to the extraction fleet also includes all the other participants in the vehicle fleet, the other participants only being provided with the context-dependent knowledge base from the backend server, but not generating or providing context-dependent knowledge data themselves. Overall, this can save resources and costs.

It can be provided that the backend server transmits at least one specification to the participants in the vehicle fleet for which context context-dependent knowledge data are to be generated, the participants generating the at least one context-dependent knowledge data based on the transmitted at least one specification. For example, context-dependent knowledge data can only be generated in those contexts that correspond to the specifications of the backend server. In this way, context-dependent knowledge data can be generated and collected in a targeted manner for certain contexts for which, for example, little or no knowledge is available in the context-dependent knowledge base.

Further features for the configuration of the system emerge from the description of configurations of the method. The advantages of the system are the same as in the case of the refinements of the method. In addition, a vehicle is also created, the vehicle being a

Participant of a vehicle fleet, comprising a control device, wherein the

Control device is designed such that a transmitted from a backend server

to receive context-dependent knowledge base and to use this locally for plausibility checking and / or support of at least one perception function.

In one embodiment of the vehicle it is provided that the control device is also designed in such a way that, depending on the context, at least one context-dependent knowledge data item based on at least one output of at least one output locally in the vehicle

To generate perception function and to transmit the generated at least one context-dependent knowledge date to the backend server.

In one embodiment of the vehicle it is provided that the control device is further designed in such a way that it uses the context-dependent knowledge base during plausibility checks to establish a plausibility check value for one of the at least one

Perceptual function to determine the result provided.

In one embodiment it is provided that the control device is also designed to take into account a plausibility check result of the plausibility check, in particular in the form of the determined plausibility check value, when defining and / or setting at least one parameter of a driving strategy and / or maneuver planning of the vehicle.

In a further embodiment of the vehicle it is provided that the control device is further designed to provide support before the at least one is used

Perception function to select and / or set parameters of the at least one perception function on the basis of the context-dependent knowledge base.

The advantages of the vehicle are in each case the same as in the configurations of the

Procedure. Further features for the configuration of the vehicle emerge from the description of configurations of the method.

A back-end server is also created, in particular, the back-end server being designed to collect knowledge data transmitted by participants in a vehicle fleet and to compare their content with one another and to generate or update a context-dependent knowledge base based on a comparison result, and the generated or to transmit updated contextual knowledge base to the participants of the vehicle fleet.

Further features for the configuration of the backend server emerge from the description of configurations of the method. The advantages of the backend server are the same as in the case of the refinements of the method.

The invention is explained in more detail below on the basis of preferred exemplary embodiments with reference to the figures. Here show:

1 shows a schematic representation of an embodiment of the system for providing a context-dependent knowledge base for checking the plausibility of at least one

Perceptual function;

2 shows a schematic flow diagram of an embodiment of the method for

Clarification of the procedural steps when generating or updating the context-dependent knowledge base;

3 shows a schematic flow diagram of an embodiment of the method for

Clarify the procedural steps on the part of the participants.

In Fig. 1 is a schematic representation of an embodiment of the system 1 for

Providing a context-dependent knowledge base 20 for plausibility checking at least one perception function is shown. The system 1 comprises control devices 2 in

Motor vehicles 50, which form participants 40 of a vehicle fleet 51, and one

Backend server 3.

The control devices 2 are for example as a combination of hardware and

Software designed, for example as program code that is executed on a microcontroller or microprocessor.

The backend server 3 comprises a computing device 4, which can access a storage device 5.

At least one perception function, for example for object recognition in the respective local surroundings of the motor vehicle 50, is carried out in each of the motor vehicles 50. The A perception function can be carried out on the control devices 2, for example. The control devices 2 receive sensor data recorded by an environment sensor system (not shown), for example a camera, on which the perception function is then carried out. The perception function can, for example, pedestrians in the vicinity of the

Detect motor vehicles and determine their position.

The control devices 2, depending on the context, generate at least one context-dependent knowledge datum 10 on the basis of at least one output of the at least one locally executed perception function. The generated at least one context-dependent knowledge data item 10 is then transmitted from the control devices 2 to the backend server 3. This takes place in particular at certain time and / or local intervals or as a function of certain criteria, such as a number of collected knowledge data or a route traveled since the last transmission, etc.

The backend server 3 receives and collects the knowledge data 10 transmitted by the control devices 2 and compares the content of the knowledge data 10 transmitted by the participants 40 with one another. On the basis of a comparison result, the back-end server 3 generates or updates a context-dependent knowledge base 20. The back-end server 3 then transmits the generated or updated context-dependent knowledge base 20 back to the participants 40 of the vehicle fleet 51.

The control devices 2 of the participants 40 or the motor vehicles 50 receive the transmitted context-dependent knowledge base 20 and use it locally for plausibility checking and / or support of the at least one perception function.

The steps described are then repeated, that is to say the participants 40 collect further context-dependent knowledge data 10 and transmit them to the

Backend server 3. There the context-dependent knowledge base 20 is updated by comparison and the updated knowledge base 20 is again transmitted to the control devices 2 of the subscribers 40.

It can be provided that a relevance of the context-dependent knowledge data 10 transmitted to the back-end server 3 is assessed by means of the back-end server 2. The

Matching then takes place taking into account the relevance of the context-dependent

Knowledge data 10. Provision can be made for the correctness of the context-dependent knowledge data 10 transmitted to the backend server 3 to be assessed by means of the backend server 2. The

Matching is then carried out taking into account the correctness of the context-dependent

Knowledge data 10.

Provision can be made for the at least one context-dependent knowledge datum 10 generated locally by the participants 40 to be used locally for plausibility checking and / or support of at least one perception function. This takes place in particular by means of the control devices 2, but can in principle also by means of other devices of the

Motor vehicle 50 take place.

It can be provided that context-dependent knowledge data 10 are annotated by the participants 40 of the vehicle fleet 51 as a function of a degree with which they contribute to an increase in the performance of the at least one perception function and / or to

Have contributed to plausibility checking of the at least one perception function, at least annotations 11 that exceed a predetermined threshold value being transmitted to the backend server 3.

In a further development, it can be provided that the backend server 3 is the context-dependent

Knowledge base 10 updated taking into account the transmitted annotations 11.

Provision can be made for the backend server 3 to carry out a probability-based assessment of the context-dependent knowledge base 20 in order to update the context-dependent knowledge base

Knowledge data 10 performs.

It can be provided that the back-end server 3 carries out a contextual refinement of the context-dependent knowledge base 20 in order to update the context-dependent knowledge base 20.

It can be provided that only a selected part of the participants 40 of the

Vehicle fleet 51 generates context-dependent knowledge data 10 and transmits it to backend server 3. The motor vehicles 50 of the selected part in particular have an improved environment sensor system and a larger memory. It can be provided that the context-dependent knowledge base 20 is used during the plausibility check to determine a plausibility check value for one of the at least one

To determine perception function 30 provided result.

Provision can be made for a plausibility check result of the plausibility check, in particular in the form of the determined plausibility check value, to be taken into account when defining and / or setting at least one parameter of a driving strategy and / or maneuver planning of a participant 40 of the vehicle fleet 51. For example, it can be provided that, as a function of the determined plausibility check value, the control device 2 generates a control signal or a control command which defines a more offensive or defensive driving style for an automated driving function of the vehicle 50.

Provision can be made for parameters of the at least one perception function 30 to be selected and / or set on the basis of the context-dependent knowledge base 20 for support before the at least one perception function 30 is applied.

FIG. 2 shows a schematic flow diagram of an embodiment of the method for clarifying the method steps when generating or updating the context-dependent knowledge base 10. Some of the method steps shown are each carried out by the participants 40 of the vehicle fleet 51, another part by the backend server 3.

After starting 100 the method, in a method step 101 the

Perception functions and the local knowledge bases 12 of the participants 40 (or in the motor vehicles 50) of the vehicle fleet 51 are initialized.

In a method step 101, the participants 40 of the vehicle fleet 51 each locally execute the at least one perception function and generate at least one context-dependent knowledge datum 10 based on at least one output of the at least one

Perceptual function. The context-dependent knowledge data 10 or context-dependent knowledge data 10 can, in particular, be in the form of occupancy cards 13,

Object occurrences 14, object relationships 15, object properties 16 and / or

Behaviors 17 of the objects are generated. In a method step 103, the context-dependent knowledge data 10 generated are transmitted from the participants 40 of the vehicle fleet 51 to the backend server 3. This is done using a communication link. The transmission can take place at predetermined time and / or local intervals and / or as a function of a number of context-dependent knowledge data 10 generated.

In a method step 104, the backend server 3 collects the transmitted

context-dependent knowledge data 10 of the participants 40 of the vehicle fleet 51.

In a method step 105, the backend server 3 carries out an assessment of the relevance and the correctness of the context-dependent knowledge data 10 that has been collected.

If a context-dependent knowledge datum 10 is not relevant and incorrect, it is discarded in a method step 106. If, on the other hand, the context-dependent knowledge data 10 is judged to be relevant and correct, the context-dependent knowledge base 10 is updated in a method step 107.

The updated context-dependent knowledge base 10 is then transmitted to the participants 40 of the vehicle fleet 51. The participants 40 can then, for example, each replace the local knowledge base 12 with the updated context-dependent knowledge base 10 transmitted by the backend server 3.

The transmitted updated context-dependent knowledge base 10 is then used locally by the participants 40 to check plausibility and / or to support the at least one

Perception function used.

FIG. 3 shows a schematic flow chart of an embodiment of the method for clarifying the method steps on the part of the participants.

After method step 101 (cf. FIG. 2), a method step 200 checks whether the method is being carried out for the first time. If this is the case, one

Method step 201 a local knowledge base 12 and the at least one

Perception function 30 initialized.

In a method step 202, the at least one perception function 30 is carried out, that is to say a result is inferred. For example, by means of a trained Neural network object recognition can be carried out in camera data captured from the environment.

In a method step 203 it is checked whether a result of the at least one

Perception function 30 is to be checked for plausibility or whether a context-dependent

Knowledge date is to be generated.

If a context-dependent knowledge datum is to be generated, it is stored in a

Method step 204 generated or collected on the basis of the inferred result and a current context.

In a method step 205 it is checked whether a time interval has expired or whether a certain number of context-dependent knowledge data has already been collected. If this is not the case, a return is made to method step 202.

If this is the case, however, the context-dependent knowledge data are in a

Method step 206 transmitted to the backend server.

If, on the other hand, method step 203 shows that a plausibility check of the inferred result of the at least one perception function is to be carried out, then, following method step 204, a method step 207 checks whether the result inferred by the at least one perception function is with the local knowledge base 12

matches. In this context, agreement should denote the existence of a plausibility between the inferred result and the context-dependent knowledge base and should not be understood in the sense of a strict identity. If this is not the case, a

Step 208 generates a warning and reverts to a fallback state. The fallback state can include a more defensive driving behavior, for example. If, however, the inferred result and the local knowledge base 12 match, then one

Method step 209 continued with the method. In particular is a

The plausibility check result is positive in this case. As part of the check in method step 207, provision can be made for a plausibility check value to be generated or determined and checked against a threshold value.

Furthermore, every time the backend server sends an updated

context-dependent knowledge base 20 is transmitted by the local knowledge base 12 in one

Method step 210 updated. In method step 208 it can be provided in particular that the in

Method step 207, a certain plausibility check value is taken into account when defining and / or setting at least one parameter of a driving strategy and / or maneuver planning of a participant in the vehicle fleet. For example, it can be provided that a control signal or a control command changes the parameters of the driving strategy that relates to a variant of a driving style (offensive or defensive), in particular in the direction of a more defensive driving style.

List of reference symbols

system

Control device

Backend server

Computing device

Storage facility

contextual knowledge date

Annotation

local knowledge base

Occupancy card

Object occurrence

Object relationship

Object property

Behavior

context-dependent knowledge base

Perceptual function

Attendees

Motor vehicle

Vehicle fleet

-107 procedural steps

-210 procedural steps

Claims

1. A method for providing a context-dependent knowledge base (20) for

Checking the plausibility of at least one perception function (30),

whereby participants (40) of a vehicle fleet (51) each have at least one locally

Carry out perception function (30) and at least one context-dependent

Generate knowledge datum (10) based on at least one output of the at least one perception function (30),

the context-dependent knowledge data (10) generated by the participants (40) of the vehicle fleet (51) being transmitted to a backend server (3),

wherein the transmitted context-dependent knowledge data (10) on the

Backend servers (3) are collected and their content is compared with one another, with a context-dependent knowledge base (20) being generated or updated on the basis of a comparison result,

wherein the generated or updated contextual knowledge base (20) to the

Subscriber (40) of the vehicle fleet (51) is transmitted, and

the context-dependent knowledge base (20) transmitted by the participants (40) locally for plausibility checking and / or support of the at least one

Perception function (30) is used.

2. The method according to claim 1, characterized in that a relevance of the context-dependent knowledge data (10) transmitted to the backend server (3) is assessed by means of the backend server (3).

3. The method according to claim 1 or 2, characterized in that the correctness of the context-dependent knowledge data (10) transmitted to the backend server (3) is evaluated by means of the backend server (3).

4. The method according to any one of the preceding claims, characterized in that the locally generated by the participants (40) at least one context-dependent

Knowledge date (10) is used locally to check plausibility and / or support at least one perception function (30).

5. The method according to any one of the preceding claims, characterized in that context-dependent knowledge data (10) are annotated by the participants (40) of the vehicle fleet (51) depending on a degree with which they lead to an increase in the performance of the at least one perception function (30) and / or have contributed to the plausibility check of the at least one perception function (30), at least annotations (11) that exceed a predetermined threshold value being attached to the

Backend server (3) are transmitted.

6. The method according to claim 5, characterized in that the context-dependent

Knowledge base (20) is updated taking into account the transmitted annotations (11).

7. The method according to any one of the preceding claims, characterized in that for updating the context-dependent knowledge base (20) a

probability-based evaluation of the context-dependent knowledge data (10) is carried out.

8. The method according to any one of the preceding claims, characterized in that for updating the context-dependent knowledge base (20) a contextual

Refinement of the context-dependent knowledge base (20) takes place.

9. The method according to any one of the preceding claims, characterized in that only a selected part of the participants (40) of the vehicle fleet (51) generates context-dependent knowledge data and transmits it to the backend server (3).

10. The method according to any one of the preceding claims, characterized in that the context-dependent knowledge base (20) is used during plausibility checking to determine a plausibility check value for a result provided by the at least one perception function (30).

11. The method according to any one of the preceding claims, characterized in that a plausibility check result of the plausibility check is taken into account when defining and / or setting at least one parameter of a driving strategy and / or maneuver planning of a participant in the vehicle fleet.

12. The method according to any one of the preceding claims, characterized in that for support before applying the at least one perception function (30) parameters of the at least one perception function (30) are selected and / or set on the basis of the context-dependent knowledge base (20).

13. System (1) for providing a context-dependent knowledge base (20) for

Checking the plausibility of at least one perception function (30), comprising:

Control devices (2) of participants (40) in a vehicle fleet (51), and

a backend server (3),

wherein the control devices (2) are each designed in such a way that, depending on the context, at least one context-dependent knowledge datum (10) based on at least one output of at least one locally executed by the participant (40)

To generate perception function (30) and the generated at least one

To transmit context-dependent knowledge data (10) to the back-end server (3), the back-end server (3) being designed to collect the knowledge data (10) transmitted by the participants (40) and to compare their content with one another and to create a context-dependent knowledge base based on a comparison result (20) to generate or update, and to transmit the generated or updated context-dependent knowledge base (20) to the subscribers (40) of the vehicle fleet (51), and wherein the control devices (2) are further designed in such a way that the transmitted context-dependent knowledge base (20) to be used locally for plausibility checking and / or support of the at least one perception function (30).

14. Vehicle (50), wherein the vehicle (50) is a participant (40) of a vehicle fleet (51), comprising:

a control device (2),

wherein the control device (2) is designed to receive a context-dependent knowledge base (20) transmitted by a backend server (3) and to use this locally for plausibility checking and / or support of at least one perception function (30).

15. Vehicle (50) according to claim 14, characterized in that the

The control device (2) is also designed to assign at least one context-dependent knowledge datum (10) based on at least one output of at least one perception function (30) performed locally in the vehicle (50) generate and transmit the generated at least one context-dependent knowledge datum (10) to the backend server (3).

16. Vehicle (50) according to one of claims 14 or 15, characterized in that the control device (2) is further designed in such a way that the context-dependent

To use knowledge base (20) during plausibility checking to determine a plausibility check value for a result provided by the at least one perception function (30).

17. Vehicle (50) according to claim 16, characterized in that the

Control device (2) is further designed to take into account a plausibility check result of the plausibility check when defining and / or setting at least one parameter of a driving strategy and / or maneuver planning of the vehicle (50).

18. Vehicle (50) according to one of claims 14 to 17, characterized in that the control device (2) is further designed to support parameters of the at least one before applying the at least one perception function (30)

Select and / or set perception function (30) based on the context-dependent knowledge base (40).

19. Backend server (3), wherein the backend server (3) is designed in such a way from

To collect knowledge data (10) transmitted to participants (40) of a vehicle fleet (51) and to compare their content with one another and to generate or update a context-dependent knowledge base (20) on the basis of a comparison result, and to send the generated or updated context-dependent knowledge base (20) to the participants (40) to transmit the vehicle fleet (51).