DE102020117638A1 - Computer-implemented method and system for validating a sensor-based vehicle function - Google Patents

Abstract

The invention relates to a computer-implemented method for validating an output value, in particular a numerical, alphanumeric, vector-based or structured output value, of a machine learning algorithm which approximates a first sensor-based vehicle function. The invention further relates to a system for validating an output value, in particular a numerical, alphanumeric, vector-based or structured output value, of a machine learning algorithm, which approximates a first sensor-based vehicle function. The invention also relates to a computer program and a computer-readable data carrier.

Description

The invention relates to a computer-implemented method for validating an output value, in particular a numeric, alphanumeric, vector-based or structured output value, of a machine learning algorithm which approximates a first sensor-based vehicle function. The invention further relates to a system for validating an output value, in particular a numeric, alphanumeric, vector-based or structured output value, of a machine learning algorithm which approximates a first sensor-based vehicle function.

Large amounts of data are required to design algorithms for approximating sensor-based vehicle functions reliably and efficiently. The necessary data must be transferred from the vehicle and, depending on the area of application, show vehicle occupants and can therefore also affect their privacy. Data protection guidelines, e.g. in accordance with the GDPR, regulate the purpose and use of this data. So-called federated learning or federated learning can be used as a possible solution to this problem.

In this model, the processes for training the algorithm take place on the individual devices, i.e. an in-vehicle data processing device. Therefore only the result, i.e. the corresponding parameters, is sent to a backend server. Sensitive data does not leave the device. However, this method is not easy to implement, especially in the automotive industry.

EP 349 45 22 A1 discloses a communication efficient federated learning method. This includes obtaining global values for a set of parameters of a machine-learned model by a client computing device, training the machine-learned model based on a local data set to obtain an update matrix which contains updated values for the set of parameters of the machine-learned model describes, wherein the update matrix is restricted to have a predefined structure, and wherein the local data set is stored locally by the client computing device. Furthermore, the method comprises a transmission of information describing the update matrix by the client computer device to a server device.

The disadvantage of this procedure is that the data in the vehicle can only be processed unlabeled or unmarked. Thus it cannot be verified how well the corresponding algorithms have been trained. A cross-vehicle improvement of the algorithms in question in federated learning is therefore also more difficult without labeling or marking the data.

There is therefore a need to provide an improved method and system which enables the algorithms used to be improved across vehicles as part of federated learning.

The object is achieved with a computer-implemented method for validating an output value, in particular a numeric, alphanumeric, vector-based or structured output value, a machine learning algorithm which approximates a first sensor-based vehicle function, with the features of claim 1.

Furthermore, the object is achieved with a system for validating an output value, in particular a numeric, alphanumeric, vector-based or structured output value, a machine learning algorithm which approximates a first sensor-based vehicle function, with the features of claim 13.

The present invention creates a computer-implemented method for validating an output value, in particular a numeric, alphanumeric, vector-based or structured output value, of a machine learning algorithm which approximates a first sensor-based vehicle function. The method comprises comparing a value of an objective function for sensor data of a first vehicle sensor output by a first machine learning algorithm of an in-vehicle data processing device with a value of an objective function for sensor data of a second vehicle sensor output by a second machine learning algorithm of the in-vehicle data processing device.

The value of the objective function for the sensor data of the first vehicle sensor is validated in-vehicle by the value of the objective function for the sensor data of the second vehicle sensor, if the value of the objective function for the sensor data of the second vehicle sensor is in a predetermined numerical range of values of the objective function for the sensor data of the second Vehicle sensor lies. The predefined range is determined as a function of the value of the objective function for the sensor data of the first vehicle sensor.

The present invention further provides a system for validating an output value, in particular a numeric, alphanumeric, vector-based or structured output value, of a machine learning algorithm which approximates a first sensor-based vehicle function.

The system comprises an in-vehicle data processing device which is configured to compare a value of a target function for sensor data of a first vehicle sensor output by a first machine learning algorithm with a value of a target function for sensor data of a second vehicle sensor output by a second machine learning algorithm of the in-vehicle data processing device .

The value of the objective function for the sensor data of the first vehicle sensor is validated within the vehicle by the value of the objective function for the sensor data of the second vehicle sensor or can be validated within the vehicle if the value of the objective function for the sensor data of the second vehicle sensor is in a predetermined numerical range of values of the objective function for the sensor data of the second vehicle sensor. The predefined range can be determined as a function of the value of the objective function for the sensor data of the first vehicle sensor.

The present invention also provides a computer program with program code in order to carry out the method according to the invention when the computer program is executed on a computer.

The present invention also provides a computer-readable data carrier with program code of a computer program in order to carry out the method according to the invention when the computer program is executed on a computer.

One idea of the present invention is to enable the accuracy and plausibility of the output data of the first machine learning algorithm to be improved by validating the value of the objective function for the sensor data of the first vehicle sensor by the value of the objective function for the sensor data of the second vehicle sensor.

The data recorded by the first vehicle sensor, which are processed unlabeled or unlabelled by the first machine learning algorithm, can thus advantageously be indirectly labeled or labeled by the second machine learning algorithm based on the validation of the output data or parameters of the first machine learning algorithm. be marked.

The present invention thus enables the advantages of federated learning to be used, i.e. the training and implementation of algorithms on a local data processing device in the vehicle, advantageously eliminating the need to transmit large amounts of data from the vehicle to the backend server.

Due to the transmission of only parameters or output data of the algorithms from the vehicle to the backend server, the data protection guidelines, e.g. according to the GDPR, are maintained. Nevertheless, the solution according to the invention enables a significant improvement in a detection accuracy or an improved approximation result of the respective sensor-based vehicle function through the corresponding machine learning algorithm.

Advantageous embodiments and developments emerge from the subclaims and from the description with reference to the figures.

According to a preferred development, it is provided that only when the value of the objective function for the sensor data of the second vehicle sensor is outside the predetermined numerical range, predetermined data sequences of the sensor data of the first vehicle sensor are transmitted to the server external to the vehicle.

There is thus advantageously no continuous data exchange of the respective sensor data between the data processing device of the vehicle and the backend server. Data is only transmitted from the data processing device of the vehicle to the server in exceptional cases, provided that the specified conditions exist, according to which the value of the objective function for the sensor data of the second vehicle sensor is outside the specified numerical range. Likewise, only predetermined data sequences of the sensor data are transmitted from the vehicle-internal data processing device to the vehicle-external server within a limited period of time, for example within the last 10 to 30 seconds before a defined event.

According to a further preferred development, it is provided that the predetermined range is dependent on the value of the objective function for the sensor data of the first vehicle sensor from the first machine learning algorithm, the second machine learning algorithm, a vehicle-external server, or a combination thereof is determined. The specified range can be delimited, for example, by an upper and / or lower threshold value, an error message being generated if the threshold value is exceeded or not reached and, in such a case, the specified data sequence of the sensor data from the first vehicle sensor is transmitted to the server external to the vehicle.

According to a further preferred development, it is provided that the validation of the value of the target function for the sensor data of the first vehicle sensor includes assigning a designation relating to the sensor-based vehicle function. The designation relating to the vehicle function is also known to the person skilled in the art as a label.

The designation represents an object to be classified by the machine learning algorithm or a result to be achieved by a regression. An example of such a target function is given in a behavior analysis of a driver of the vehicle. Exact or correct detection of the behavior of the user can then, for example, activate or deactivate certain functions or generate certain messages.

According to a further preferred development, it is provided that the comparison of the value of the target function for the sensor data of the first vehicle sensor output by the first machine learning algorithm with the value of the target function for sensor data of the second vehicle sensor output by the second machine learning algorithm is carried out by an observation algorithm . Thus, in an advantageous manner, a comparison is only carried out between the output data or parameters of the first algorithm and the second algorithm, so that no data that affect the privacy of the user have to be used for this.

According to a further preferred development, it is provided that the value of the target function for the sensor data of the first vehicle sensor is transmitted from the vehicle-internal data processing device to the vehicle-external server at predetermined intervals. The intervals can advantageously be selected according to certain systemic and / or data transmission-relevant aspects such as, for example, a bandwidth of a data transmission.

According to a further preferred development, it is provided that a third machine learning algorithm of the vehicle-external server is trained using values received from vehicle-internal data processing devices from a large number of vehicles, and the trained third machine learning algorithm from the vehicle-external server to the vehicle-internal data processing device of the large number of Vehicles is transmitted. Thus, within the scope of the present invention, the improved detection accuracy of the vehicle-specific algorithms can be used in an advantageous manner with a federated learning method which enables only small amounts of data to be transmitted and compliance with corresponding data protection guidelines.

According to a further preferred development, it is provided that the trained first machine learning algorithm of the vehicle-internal data processing device is replaced by the trained third machine learning algorithm of the vehicle-external server. The third machine learning algorithm of the server external to the vehicle can thus advantageously be trained on the basis of the data received from the vehicle-specific algorithms and then delivered again to the vehicle fleet.

According to a further preferred development, it is provided that the trained third machine learning algorithm transmitted to the vehicle-internal data processing device and the first machine learning algorithm of the vehicle-internal data processing device are executed in parallel. In this constellation, the vehicle user thus benefits in an advantageous manner both from the specific adaptation of the first machine learning algorithm to the individual usage behavior of the driver and from generic improvements to the third machine learning algorithm, which is based on feedback from all fleet participants.

According to a further preferred development, it is provided that the first machine learning algorithm of the vehicle-internal data processing device is integrated with the trained third machine learning algorithm. This advantageously creates a machine learning algorithm which is composed of elements of the first machine learning algorithm and the third machine learning algorithm.

According to a further preferred development, it is provided that the first vehicle sensor and / or the second vehicle sensor is formed by a vehicle interior camera, a vehicle exterior camera, a steering wheel sensor, a seat sensor and / or an audio sensor arranged in a vehicle interior. The before The sensors mentioned above can thus cover a large number of vehicle functions such as face recognition of a driver and / or a passenger, emotion recognition of vehicle occupants, recording of vital data of vehicle occupants, speech analysis of vehicle occupants and recognition of the outside of the vehicle such as the recognition of lane markings, Signs, etc.

According to a further preferred development, it is provided that if negative usage behavior of the sensor-based vehicle function is detected by a user of the vehicle, in particular ignoring a behavior recommendation generated by the sensor-based vehicle function, predetermined data sequences including sensor data from the first vehicle sensor are transmitted to the server external to the vehicle. A negative usage behavior of a vehicle user or driver can indicate a malfunction or incorrect identification of a desired object by the algorithm. In order to improve the algorithm performance, relevant data sequences are transmitted to the server external to the vehicle.

The refinements and developments described can be combined with one another as required.

Further possible configurations, developments and implementations of the invention also include combinations, which are not explicitly mentioned, of features of the invention described above or below with regard to the exemplary embodiments.

Figure list

The accompanying drawings are intended to provide a further understanding of the embodiments of the invention. They illustrate embodiments and, in conjunction with the description, serve to explain principles and concepts of the invention.

Other embodiments and many of the advantages mentioned emerge with a view to the drawings. The elements shown in the drawings are not necessarily shown true to scale with one another.

• 1 eine schematische Darstellung eines Verfahrens und Systems zum Validieren eines Ausgabewerts, insbesondere eines numerischen, alphanumerischen, vektorbasierten oder strukturierten Ausgabewerts, eines Algorithmus maschinellen Lernens, welcher eine erste sensorbasierte Fahrzeugfunktion approximiert gemäß einer bevorzugten Ausführungsform der Erfindung.

It shows:

• 1 a schematic representation of a method and system for validating an output value, in particular a numeric, alphanumeric, vector-based or structured output value, a machine learning algorithm which approximates a first sensor-based vehicle function according to a preferred embodiment of the invention.

In the present illustration, a first vehicle 1 and a second vehicle 2 are shown schematically. An in-vehicle data processing device 10 of the first vehicle 1 and a data processing device 20 of the second vehicle 2 are connected to a server 16 external to the vehicle, preferably by a wireless connection.

The in-vehicle data processing device 10 of the first vehicle 1 has a first machine learning algorithm A1 and a second machine learning algorithm A2. The first machine learning algorithm A1 is set up to receive sensor data D1 from a first vehicle sensor 12. The second machine learning algorithm A2 is set up to receive sensor data D2 from a second vehicle sensor 14.

In the present embodiment, only two algorithms are shown in a simplified manner. Of course, more than two algorithms can be used, for example. The first machine learning algorithm A1 approximates at least one value of an objective function for the sensor data of the first vehicle sensor 12, which is applied to the sensor data D1 of the first vehicle sensor 12. The objective function here represents a first sensor-based vehicle function. Such a vehicle function can, for example, be a behavioral recognition of a vehicle user, depending on the sensor. For example, image data, audio data or other data generated by corresponding sensors can be used for this purpose.

The second machine learning algorithm A2 also approximates at least one value of an objective function for the sensor data of the second vehicle sensor 14, the machine learning algorithm being applied to the sensor data of the second vehicle sensor 14. The objective function here represents a second sensor-based vehicle function.

If the first vehicle sensor is formed, for example, by an interior camera, the second vehicle sensor can, for example, be formed by an audio sensor or other sensor described in this application.

The in-vehicle data processing device 10 is set up to perform machine learning A1 using the first algorithm To compare the given value W1 of the target function for sensor data D1 of the first vehicle sensor 12 with the value W2 of the target function for sensor data D2 of the second vehicle sensor 14 output by the second machine learning algorithm A2 of the in-vehicle data processing device 10.

The values W1, W2 are preferably numerical values. Alternatively, the values W1, W2 can be, for example, alphanumeric, vector-based or structured values.

The value W1 of the target function for the sensor data D1 of the first vehicle sensor 12 is validated inside the vehicle by the value W2 of the target function for the sensor data D2 of the second vehicle sensor 14.

If the value W2 of the objective function for the sensor data D2 of the second vehicle sensor 14 lies in a predetermined numerical range of values of the objective function for the sensor data D2 of the second vehicle sensor, the output of the first machine learning algorithm A1 is validated by the second machine learning algorithm A2 instead of. The predefined range can be determined as a function of the value W1 of the objective function for the sensor data D1 of the first vehicle sensor 12.

Only when the value W2 of the objective function for sensor data D2 of the second vehicle sensor 14 lies outside the specified numerical range are the specified data sequences of the sensor data D1 of the first vehicle sensor 12 transmitted to the server 16 external to the vehicle. If the value W2 of the objective function for the sensor data D2 of the second vehicle sensor 14 is within the predetermined numerical range, only output or parameter data of the first machine learning algorithm A1 and / or the second machine learning algorithm A2 are received from the in-vehicle data processing device 10 the vehicle-external server 16 transmitted.

The predefined range is determined by the first machine learning algorithm A1 as a function of the value W1 of the objective function for the sensor data D1 of the first vehicle sensor 12. Alternatively, the determination can be made, for example, on the basis of data from the second machine learning algorithm A2, the server 16 external to the vehicle, or a combination thereof.

The validation of the value W1 of the target function for the sensor data D1 of the first vehicle sensor 12 includes assigning a designation or a label relating to the sensor-based vehicle function. Thus, through the feedback of the second machine learning algorithm A2, indirect or implicit labeling of the objective function of the first machine learning algorithm A1 can be carried out or achieved.

The comparison of the value W1 of the target function for the sensor data D1 of the first vehicle sensor 12 output by the first machine learning algorithm A1 with the value W2 of the target function for the sensor data D2 of the second vehicle sensor 14 output by the second machine learning algorithm A2 is carried out by an observation algorithm BA1 accomplished.

The value W1 of the objective function for the sensor data D1 of the first vehicle sensor 12 is transmitted from the vehicle-internal data processing device 10 to the vehicle-external server 16 at predetermined intervals. The same applies to the value W2 of the target function for the sensor data D2 of the second vehicle sensor 12.

A third machine learning algorithm A3 of the vehicle-external server 16 is trained using vehicle-internal data processing devices 10 of a multiplicity of values received from vehicles 1, 2. The trained third machine learning algorithm A3 of the vehicle-external server 16 is then transmitted to the respective vehicle-internal data processing devices 10 of the plurality of vehicles 1, 2. In this context, the first machine learning algorithm A1 of the in-vehicle data processing device 10 is preferably replaced by the trained third machine learning algorithm A3 of the server 16 external to the vehicle.

Alternatively, it is also possible that the first machine learning algorithm A1 remains unchanged and, in addition, the third machine learning algorithm A3 is transmitted from the vehicle-external server 16 to the vehicle-internal data processing device 10 and the two algorithms are then executed in parallel.

Furthermore, as an alternative, it is possible, for example, for the first machine learning algorithm A1 and the third machine learning algorithm A3 to be integrated with one another.

In this case, the third machine learning algorithm A3 can have the function of an observation algorithm, for example, which compares the output parameters of the algorithms A1, A2.

When the vehicle-internal data processing device 10 is updated by the vehicle-external server 16, there is an update UD1 relating to the first vehicle 1 and an update UD2 relating to the second vehicle 2.

The second vehicle 2 is shown only by way of example in the context of the present embodiment and has a structure or configuration that is analogous to the first vehicle 1. In particular, the second vehicle 2 has the vehicle-internal data processing device 20, which in turn is assigned a first machine learning algorithm A11 and a second machine learning algorithm A12. The vehicle 2 also has an observation algorithm BA2, which compares the output parameters of the algorithms A11, A12. A first vehicle sensor 22 transmits sensor data D11 to the first machine learning algorithm A11. A second vehicle sensor 24 transmits sensor data D12 to the second machine learning algorithm A12 of the second vehicle 2. The vehicle-internal data processing device 20 also transmits the output data or parameters W11, W12 of the algorithms A11, A12 to the vehicle-external server 16.

The first vehicle sensor 12 and / or the second vehicle sensor 14 can be embodied, for example, by a vehicle interior camera, a vehicle exterior camera, a steering wheel sensor, a seat sensor and / or an audio sensor arranged in a vehicle interior.

The first vehicle sensor 12 and / or the second vehicle sensor 14 can furthermore be formed by any sensors present in the vehicle.

A large number of other vehicle sensors can also be used. If negative usage behavior of the sensor-based vehicle function is detected by a user of the vehicle 1, 2, in particular ignoring a behavior recommendation generated by the sensor-based vehicle function, predefined data sequences including sensor data D1 from the first vehicle sensor 12 are transmitted to the server 16 external to the vehicle.

The term vehicle includes cars, trucks, buses, mobile homes, motorcycles, etc. that are used to transport people, goods, etc.

In particular, the term includes motor vehicles for passenger transport. Additionally or alternatively, a hybrid or electric vehicle according to embodiments can be a purely electric vehicle (BEV) or a plug-in hybrid vehicle (PHEV). However, other forms of drive can also be used, for example in the form of a diesel or gasoline-powered vehicle. The vehicle can also be in the form of a rail vehicle.

Although the invention has been illustrated and explained in more detail by preferred exemplary embodiments, the invention is not restricted by the disclosed examples and other variations can be derived therefrom by the person skilled in the art without departing from the scope of protection of the invention.

It is therefore clear that there is a multitude of possible variations. Embodiments cited by way of example only represent examples that are not to be interpreted in any way as a limitation, for example, of the scope of protection, the possible applications or the configuration of the invention.

Rather, the preceding description and the description of the figures enable the person skilled in the art to specifically implement the exemplary embodiments, whereby the person skilled in the art, with knowledge of the disclosed inventive concept, can make various changes, for example with regard to the function or the arrangement of individual elements mentioned in an exemplary embodiment, without the To leave the scope of protection, which is defined by the claims and their legal equivalents, such as further explanations in the description.

The validation of the value W1 of the target function for the sensor data D1 of the first vehicle sensor 12 takes place in the present embodiment in the vehicle by the value W2 of the target function for the sensor data (D2) of the second vehicle sensor 14.

In addition, however, it is possible for the value W2 of the objective function for the sensor data D2 of the second vehicle sensor 14 to be validated by the value W1 of the objective function for the sensor data D1 of the first vehicle sensor 12.

Alternatively, the value W1 of the target function for the sensor data D1 of the first vehicle sensor 12 can be validated, for example, by further values of target functions for the sensor data of further vehicle sensors.

QUOTES INCLUDED IN THE DESCRIPTION

This list of the documents listed by the applicant was generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Patent literature cited

• EP 3494522 A1 [0004]

Claims (15)

Computer-implemented method for validating an output value, in particular a numeric, alphanumeric, vector-based or structured output value, of a machine learning algorithm which approximates a first sensor-based vehicle function, with the following steps: comparing a machine learning (A1) of a vehicle-internal data processing device (10 ) the output value (W1) of an objective function for sensor data (D1) of a first vehicle sensor (12) with a value (W2) outputted by a second machine learning algorithm (A2) of the vehicle-internal data processing device (10) of an objective function for sensor data (D2) of a second Vehicle sensor (14), the value (W1) of the objective function for the sensor data (D1) of the first vehicle sensor (12) being validated internally in the vehicle by the value (W2) of the objective function for the sensor data (D2) of the second vehicle sensor (14), if the value (W2) of the target radio tion for the sensor data (D2) of the second vehicle sensor (14) lies in a predetermined numerical range of values of the target function for the sensor data (D2) of the second vehicle sensor (14), the predetermined range being dependent on the value (W1) of the target function for the sensor data (D1) of the first vehicle sensor (12) is determined. Computer-implemented method according to Claim 1 , only if the value (W2) of the objective function for the sensor data (D2) of the second vehicle sensor (14) is outside the predetermined numerical range, predetermined data sequences of the sensor data (D1) of the first vehicle sensor (12) to the server external to the vehicle ( 16) can be transferred. Computer-implemented method according to Claim 1 or 2 , the predetermined range depending on the value (W1) of the objective function for the sensor data (D1) of the first vehicle sensor (12) from the first machine learning algorithm (A1), the second machine learning algorithm (A2), a server (16 ), or a combination thereof. Computer-implemented method according to one of the preceding claims, wherein the validation of the value (W1) of the objective function for the sensor data (D1) of the first vehicle sensor (12) comprises assigning a designation relating to the sensor-based vehicle function. Computer-implemented method according to one of the preceding claims, wherein the comparison of the value (W1) output by the first machine learning algorithm (A1) of the objective function for the sensor data (D1) of the first vehicle sensor (12) with the value (W1) output by the second machine learning algorithm (A2 ) The output value (W2) of the objective function for sensor data (D2) of the second vehicle sensor (14) is carried out by an observation algorithm (BA1). Computer-implemented method according to one of the Claims 2 until 5 , the value (W1) of the objective function for the sensor data (D1) of the first vehicle sensor (12) being transmitted from the vehicle-internal data processing device (10) to the vehicle-external server (16) at predetermined intervals. Computer-implemented method according to Claim 6 , wherein a third machine learning algorithm (A3) of the server external to the vehicle (16) is trained using values received from in-vehicle data processing devices (10) from a plurality of vehicles, and the trained third machine learning algorithm (A3) of the server external to the vehicle (16) is transmitted to the in-vehicle data processing device (10) of the plurality of vehicles. Computer-implemented method according to Claim 7 , wherein the trained first machine learning algorithm (A1) of the in-vehicle data processing device (10) is replaced by the trained third machine learning algorithm (A3) of the server (16) external to the vehicle. Computer-implemented method according to Claim 7 wherein the trained third machine learning algorithm (A3) transmitted to the in-vehicle data processing device (10) and the first machine learning algorithm (A1) of the in-vehicle data processing device (10) are executed in parallel. Computer-implemented method according to one of the Claims 7 until 9 , the first machine learning algorithm (A1) of the in-vehicle data processing device (10) being integrated with the trained third machine learning algorithm (A3). Computer-implemented method according to one of the preceding claims, wherein the first vehicle sensor (12) and / or the second vehicle sensor (14) is formed by a vehicle interior camera, a vehicle exterior camera, a steering wheel sensor, a seat sensor and / or an audio sensor arranged in a vehicle interior. Computer-implemented method according to one of the preceding claims, wherein if a negative usage behavior of the sensor-based vehicle function by a user of the vehicle (1), in particular ignoring a behavior recommendation generated by the sensor-based vehicle function, is recorded, predetermined data sequences comprising sensor data (D1) of the first vehicle sensor ( 12) to the server (16) external to the vehicle. System for validating an output value, in particular a numeric, alphanumeric, vector-based or structured output value, of a machine learning algorithm which approximates a first sensor-based vehicle function, comprising a vehicle-internal data processing device (10) which is set up to perform a machine learning ( A1) an output value (W1) of an objective function for sensor data (D1) of a first vehicle sensor (12) with a value (W2) of an objective function for sensor data (D2) output by a second machine learning algorithm (A2) of the in-vehicle data processing device (10) to compare the second vehicle sensor (14), the value (W1) of the target function for the sensor data (D1) of the first vehicle sensor (12) being validated inside the vehicle by the value (W2) of the target function for the sensor data (D2) of the second vehicle sensor (14) is if the value (W2) of the objective function for the sensor data (D2) of the second vehicle sensor (14) lies in a predetermined numerical range of values of the target function for the sensor data (D2) of the second vehicle sensor (14), the predetermined range being dependent on the value (W1) of the target function for the Sensor data (D1) of the first vehicle sensor (12) can be determined. Computer program with program code for the method according to one of the Claims 1 until 12th when the computer program is executed on a computer. Computer-readable data carrier with the program code of a computer program to perform the method according to one of the Claims 1 until 12th when the computer program is executed on a computer.

Images