DE102019213533A1 - Device for adapting a driving strategy of a vehicle - Google Patents

Abstract

The present invention relates to a device (30) for adapting a driving strategy of a vehicle (12), having an input interface (34) for receiving an image signal from an optical sensor (22) containing information about the surroundings of the vehicle (12), the image signal at least includes two images generated in two different locations; a processing unit (36) with a pre-trained artificial neural network (52) for determining an angle of incline (16) of an incline (14) in the vicinity of the vehicle (12) and a distance (20) of the vehicle (12) to the incline (14) ) based on the at least two images of the image signal; an adaptation unit (38) for adapting the driving strategy based on the ascertained gradient angle (16) and the ascertained distance (20); and an output interface (40) for outputting the driving strategy to a vehicle control unit (32). The present invention also relates to a method and a system (10) for adapting the driving strategy of a vehicle (12).

Description

The present invention relates to a device as well as a system and a method for adapting a driving strategy of a vehicle.

Modern vehicles (cars, vans, trucks, motorcycles, commercial vehicles, etc.) include a large number of systems that provide the driver with information and control individual functions of the vehicle in a partially or fully automated manner. The surroundings of the vehicle and other road users are recorded via sensors. Based on the recorded data, a model of the vehicle environment can be generated and changes in this vehicle environment can be reacted to. Due to the progressive development in the field of autonomous and semi-autonomous vehicles, the influence and scope of such driver assistance systems (Advanced Driver Assistance Systems, ADAS) are increasing.

In the case of agricultural utility vehicles and agricultural machinery (tillage machines, tractors, etc.), autonomous or semi-autonomous operation can be implemented comparatively easily, since the vehicles are mainly used outside of high-traffic areas, for example in fields and fields.

For vehicles in general and in particular for commercial vehicles, but also for agricultural machines, it is important to recognize changes in the surroundings of the vehicle at an early stage. A lot of energy is saved, especially when transporting heavy loads, if the vehicles are moved as evenly as possible. If the vehicle speed is reduced by braking, the original speed can only be regained by subsequent acceleration, which leads to higher energy consumption and a loss of time. The same applies to an incline in the course of the route. If a vehicle climbs an incline in a high gear of the transmission, the speed is reduced and the energy consumption increases significantly compared to a lower gear. This applies in particular to vehicles that transport large loads or that have little power and can lead to the vehicle coming to a standstill if there is no timely response. However, a stationary vehicle on the lane represents a source of danger for other road users. Acceleration on an incline following a standstill additionally increases the energy consumption compared to an adapted driving with a smaller gear ratio. If changes in the route are recognized in good time, the vehicle's driving strategy can be adjusted accordingly.

A method for recognizing the route is proposed, for example, in “Road Segmentation Supervised by an Extended V-Disparity Algorithm for Autonomous Navigation”, Nicolas Soquet, Didier Aubert, Nicolas Hautiere, 2007, iEEE Intelligent Vehicles Symposium, iSSN 1931-0587. In the method, a disparity image is generated based on the data from a stereo camera and the lane in the front area of a vehicle is identified by colored segments.

From the US 8,867,790 B2 a further method for detecting objects in a frontal area in front of a vehicle is known which is based on disparity maps based on images from a stereo camera. This method is suitable for detecting objects above the road surface.

The known methods require high computing power and complex rule-based programming.

Based on this, the task of the present invention is to adapt the driving strategy of a vehicle, be it autonomous, partially autonomous or driving manually, to the surroundings of the vehicle, in particular to terrain information, in order to avoid dangerous situations and accidents. In addition, the driving strategy should be able to be designed in such a way that the vehicle can be operated in a time- and energy-efficient manner.

• einer Eingangsschnittstelle zum Empfangen eines Bildsignals eines optischen Sensors mit Informationen zu einer Umgebung des Fahrzeugs, wobei das Bildsignal wenigstens zwei Bilder umfasst, die an zwei unterschiedlichen Orten erzeugt wurden;
• einer Verarbeitungseinheit mit einem vortrainierten künstlichen neuronalen Netz (KNN) zum Ermitteln eines Steigungswinkels einer Steigung in der Umgebung des Fahrzeugs und eines Abstands des Fahrzeugs zu der Steigung basierend auf den wenigstens zwei Bildern des Bildsignals;
• einer Anpassungseinheit zum Anpassen der Fahrstrategie basierend auf dem ermittelten Steigungswinkel und dem ermittelten Abstand; und
• einer Ausgangsschnittstelle zum Ausgeben der Fahrstrategie an ein Fahrzeugsteuergerät.

To achieve this object, the present invention relates in a first aspect to a device for adapting a driving strategy of a vehicle, in particular a commercial vehicle, with:

• an input interface for receiving an image signal from an optical sensor with information about the surroundings of the vehicle, the image signal comprising at least two images that were generated at two different locations;
• a processing unit with a pre-trained artificial neural network (KNN) for determining an angle of incline of an incline in the vicinity of the vehicle and a distance of the vehicle to the incline based on the at least two images of the image signal;
• an adaptation unit for adapting the driving strategy based on the ascertained gradient angle and the ascertained distance; and
• an output interface for outputting the driving strategy to a vehicle control unit.

• einer wie zuvor beschriebenen Vorrichtung;
• einem optischen Sensor zum Übermitteln von Bildsignalen mit Informationen zu der Umgebung des Fahrzeugs; und
• einem Fahrzeugsteuergerät zum Empfangen und Verarbeiten der Fahrstrategie.

In a further aspect, the present invention relates to a system for adapting a driving strategy of a vehicle, with:

• a device as previously described;
• an optical sensor for transmitting image signals with information about the surroundings of the vehicle; and
• a vehicle control unit for receiving and processing the driving strategy.

Further aspects of the invention relate to a vehicle, a method embodied in accordance with the device and a computer program product with program code for performing the steps of the method when the program code is executed on a computer, as well as a storage medium on which a computer program is stored is executed on a computer, causes an execution of the method described herein.

Further refinements of the invention are described in the dependent claims. It goes without saying that the features mentioned above and those yet to be explained below can be used not only in the respectively specified combination, but also in other combinations or on their own, without departing from the scope of the present invention. In particular, the system, the vehicle, the method and the computer program product can be designed in accordance with the configurations described for the device and the system in the dependent claims.

A computer program can be stored / distributed on a non-volatile data carrier, for example on an optical memory or on a semiconductor drive (SSD). A computer program can be distributed together with hardware and / or as part of hardware, for example by means of the Internet or by means of wired or wireless communication systems.

According to the invention, it is provided that, from the received image signals with two images that were generated at different locations, an angle of incline of an incline in the vicinity of the vehicle and a distance between the vehicle and the incline are determined. The angle of inclination and the distance are determined based on the at least two images of the image signal, a pre-trained artificial neural network (ANN) being used. In the context of the invention, it has been shown that the detection of a slope in the frontal area in front of the vehicle from the received image signal is possible in principle. However, this requires the establishment and development of a large number of rules that make processing and recognition slow and computationally intensive. It has been shown that an artificial neural network can be pre-trained with data sets and that processing based on the artificial neural network is much more reliable and faster. Computing power is saved. Rule-based programming is not required.

Based on the gradient angle and the distance of a recognized gradient in front of the vehicle, the driving strategy is adapted in an adaptation unit of the device according to the invention. This adapted or changed driving strategy is transmitted to a vehicle control unit in the vehicle. For example, depending on the gradient and the distance, it is possible to prevent the transmission from shifting to a higher gear, i.e. a higher gear ratio. This can be done automatically in an automatic transmission. In the case of a manual gearbox of a vehicle, either a warning signal can be output to the user of the vehicle, for example as an optical or acoustic signal, or a higher gear with a larger gear ratio can be temporarily blocked. In this case, the device according to the invention would inform the vehicle control unit that switching is to be prevented and / or output a warning signal.

Alternatively, a lower gear with a lower gear ratio could be selected as soon as the vehicle has reached a predefined critical distance from the incline. The distance can be dependent on the slope angle of the slope. Furthermore, the engine electronics can be prepared so that the engine can provide more power as soon as the incline is reached. This applies in particular to vehicles with low power, when transporting large loads or when the vehicle speed is to be kept constant. As an alternative or in addition, the driver can receive a message that an incline is to be expected. For example, the angle of incline could be given to the driver so that he can adjust to the upcoming incline. It is also possible, in particular in the case of commercial vehicles and vehicles that transport heavy loads, to output a warning signal to other road users. This can be done, for example, by controlling the brake lights. Alternatively or additionally, other vehicles located in the vicinity can be informed via communication units when there is communication between the vehicles. For example, a radio signal or other signals could be emitted that vehicles in the vicinity of the vehicle that is about to slow down could receive.

In a further development, the processing unit of the device is designed in such a way that the artificial neural network is pre-trained with data records that include two images, a slope angle of a slope and a distance to the slope. The pitch angle and distance are known. The two images are deposited. All data serve as training data for the artificial neural network. The two images are fed to the input layer of the neural network, which uses this to generate the output layer (using several hidden layers) with information on the angle of incline and the distance between the incline.

In a further development, the artificial neural network of the processing unit is pre-trained by using pre-processed images. The images can be in the form of filtered data, compressed data or reduced data, for example. For example, it is possible not to use all the image data of an image, but only selected ones. For example, an area could be selected which is described by a square or a trapezoid in which the area of interest is depicted in front of the vehicle. It would also be possible for data to be pre-scaled.

In a further-developing embodiment, the images are first converted into a disparity map and processed in which the differences between the two images or the image data representing the images are used. The disparity map serves as part of the data set with which the artificial neural network is trained. The disparity map is preferably processed further for the training, for example it can be compressed or filtered or smoothed. Correspondingly, a slope is also recognized by means of a disparity map. The images of the optical sensor are converted into a disparity map, on the basis of which the slope is recognized.

The artificial neural network of the processing unit is preferably pre-trained with data sets which are based on measured data. For this purpose, a further processing of generated images is carried out with a vehicle at a known distance and a known angle of incline of an incline. The more images that can be generated and processed, the better the artificial neural network can be trained. Distance and angle of incline can be measured with additional measuring devices. Since training with measurement data is quite time-consuming, the artificial neural network can alternatively or additionally be pre-trained with data sets based on simulated data. The simulated data are preferably generated by means of a simulation program. Simulation data and measurement data can also be used in combination.

As soon as a training phase of the artificial neural network of the processing unit has been completed, the data can be verified in a further step. For this purpose, image data are made available to the device, from which a (previously known) gradient is to be determined or forecast. Since the slope with the slope angle and distance is known, the quality of the device can be checked. If necessary, more data sets can be used in order to make the artificial neural network more reliable.

Overall, the artificial neural network has the advantage that a significantly faster and more reliable forecasting and detection of an incline in front of a vehicle with incline angle and distance is possible. There is no need for traditional rule-based programming.

In a further-developing embodiment, the processing unit has an artificial neural network that is pre-trained with data records that include a time specification and / or additional vehicle information. For example, the speed of the vehicle and a steering angle of the vehicle can be recorded and fed to the data set for training the artificial neural network. In this way, when using the device, it can be recognized when an incline is not located directly in front of the vehicle, but rather the vehicle is moving at a different angle, that is, at an incline, towards the incline. On the basis of this information, the reliability of the device can be increased further. When the device is in operation, the time information and / or the vehicle information are then also determined and fed to the artificial neural network (ANN) of the processing unit.

If the additional vehicle information and a time specification are available, the accuracy of the prediction or the accuracy of the training data can be increased, in particular when creating a disparity map. In addition, filtering over time can optionally take place in order to take into account inaccuracies in the forecast, deviations in the data and information. For example, variances and covariances can be determined in order to improve the detection of the slope. For example, a Kalman filter could be used. Other filter techniques known to those skilled in the art can also be used. These filters are used in particular when the image signal includes images that differ from one another in time and are created at different locations, for example when using a Mono camera with two images at different times of a moving vehicle.

In a further development, the input interface is designed to receive an image signal from a stereo camera. The image signal preferably includes a first image from a first camera sensor and a second image from a second camera sensor. The stereo camera with the two camera sensors is preferably arranged at the front of the vehicle in order to detect the frontal area in front of the vehicle. The two camera sensors are preferably arranged in a common housing. It is of course possible to arrange one camera sensor in the right area of the vehicle and the other camera sensor in the left area of the vehicle. The stereo camera then generates an image signal with two images, synchronized in time, which are generated at the same distance from the slope.

In an alternative refinement, the input interface is designed to receive an image signal from a mono camera. The image signal comprises a first image from the mono camera and a second image from the mono camera. The two images are taken one after the other. Since the vehicle moves between the two recordings of the mono camera, the first image and the second image of the mono camera deviate from one another. For example, a disparity map can be created from the deviation, which enables statements to be made about the distance to the gradient and the gradient angle. In the case of a stereo camera, the image of the first camera sensor and the image of the second camera sensor differ from one another, so that a difference can also be formed and, for example, a disparity map can be formed in order to determine the distance to the slope and its slope angle.

In a further-developing embodiment, the system according to the invention comprises, as an optical sensor, a stereo camera with two camera sensors or a mono camera, as already described above.

The vehicle control unit of the system according to the invention preferably generates control commands based on the driving strategy that was transmitted by the device. The control commands are transmitted to components of the vehicle in order to implement the specified driving strategy. For example, engine control components can be prepared so that more power will soon be available. It is also possible to prevent the transmission from shifting up to a higher gear ratio or to initiate a change in gear selection to a lower gear ratio.

The surroundings of a vehicle include, in particular, an area of the surroundings of the vehicle detected by the optical sensor or surroundings sensor, the optical sensor being attached to the vehicle. An optical sensor can comprise several individual sensors. In particular, it is arranged in the front area of the vehicle in order to detect a frontal area in front of the vehicle. The optical sensor preferably comprises a stereo camera or a mono camera. The sensor creates two images that were created in different locations. In the case of a stereo camera, these are two images from different camera sensors; in the case of a mono camera, images recorded at different times when the camera is moving. The artificial neural network can be implemented in an electronic control unit of the processing unit. For example, the artificial neural network can be integrated in a chip or processor. It can be in the form of software or, for example, as program code or implemented in a chip close to the hardware. The filter used for training or applying the artificial neural network can include customary filters in order to take into account the deviation, dispersion, variance or covariance. Different filter techniques are known to those skilled in the art, for example a Kalman filter could be used. The filters can be initialized at the beginning of the execution of a discovery or training.

• 1 eine schematische Ansicht des erfindungsgemäßen Systems zum Anpassen einer Fahrstrategie eines Fahrzeugs;
• 2 eine schematische Darstellung des erfindungsgemäßen Systems;
• 3 eine schematische Darstellung eines Fahrzeugs mit dem erfindungsgemäßen System;
• 4 eine schematische Darstellung eines Datensatzes zum Trainieren eines künstlichen neuronalen Netzes (KNN) der erfindungsgemäßen Vorrichtung;
• 5 eine Prinzipskizze des Trainierens eines künstlichen neuronalen Netzes;
• 6 eine schematische Darstellung des erfindungsgemäßen Verfahrens; und
• 7 eine alternative Ausführungsform des erfindungsgemäßen Verfahrens.

The invention is described and explained in more detail below using a few selected exemplary embodiments in connection with the accompanying drawings. Show it:

• 1 a schematic view of the system according to the invention for adapting a driving strategy of a vehicle;
• 2 a schematic representation of the system according to the invention;
• 3 a schematic representation of a vehicle with the system according to the invention;
• 4th a schematic representation of a data set for training an artificial neural network (ANN) of the device according to the invention;
• 5 a schematic diagram of training an artificial neural network;
• 6th a schematic representation of the method according to the invention; and
• 7th an alternative embodiment of the method according to the invention.

The 1 shows schematically a system according to the invention 10 for adapting a driving strategy of a vehicle 12th , here schematically as Tractor is shown. The illustration shows a side view of a traffic situation with an incline 14th with a slope angle 16 that are in a frontal area 18th in front of the vehicle 12th lies. A distance 20th reflects the distance between the vehicle 12th and a base point of the slope 14th contrary. The frontal area 18th in front of the vehicle 12th is made by means of an optical sensor 22nd captured as a stereo camera 24 in a front area 26th at the vehicle 12th is arranged.

Based on an image signal from the optical sensor 22nd , which includes two images, each from one of the two cameras of the stereo camera 24 are recorded, a disparity map is determined. On this basis, by means of the system according to the invention 10 an estimate of the distance 20th and the pitch angle 16 the slope 14th respectively.

2 shows the system according to the invention 10 with a device 30th , the optical sensor 22nd and a vehicle control unit 32 . The device 30th includes an input interface 34 to get an image signal from the optical sensor 22nd to receive the image signal information about the frontal area 18th in the vicinity of the vehicle 12th includes. The image signal comprises two images that were generated at two different locations, namely by the two cameras of the stereo camera 24 .

The device 30th further has a processing unit 36 , an adjustment unit 38 for adapting the driving strategy and an output interface 40 to output the driving strategy to the vehicle control unit 32 .

The processing unit 36 comprises a pre-trained artificial neural network (ANN) in order to determine the angle of inclination based on the two images of the image signal 16 the slope 14th and the distance between the vehicle 12th and the slope 14th to determine. Based on pitch angle 16 and distance 20th becomes the driving strategy in the adaptation unit 38 customized.

The image signals sent by the optical sensor are preferred 22nd supplies, in the processing unit 36 or in another unit of the device 30th preprocessed. For example, a so-called disparity map can be created, which can be reduced (sparse), compressed (dense) or filtered. The image data of the images can be scaled, complete and / or reduced to areas of interest. Downscaling can also be done optionally. The preferably preprocessed image data are then made available to the artificial neural network, which calculates the angle of inclination from these data 16 and the distance 20th determined.

3 shows a vehicle 12th that is a commercial vehicle 42 is and about the system according to the invention 10 disposes. A stereo camera 24 is in the front area 26th of the commercial vehicle 42 arranged. The one with the stereo camera 24 detected image signals are in the device 30th processed by means of the artificial neural network and a driving strategy to the vehicle control unit 32 transmitted. The vehicle control unit 32 In the present case, for example, prevents the automatic upshifting into a gear with a larger gear ratio and therefore generates a control signal for an automatic transmission 44 . The commercial vehicle shown here 42 thus includes a vehicle control unit 32 , which makes it possible to adapt the driving strategy and thus the vehicle behavior according to the determined driving strategy. For example, via the vehicle control unit 32 a downshift of the automatic transmission 44 be initiated so that just before reaching the slope 14th a lower gear ratio is selected.

4th shows a number of typical data sets 46 used for artificial neural network training. The records 46 each include image data 48 a first and second image from a stereo camera 24 , optional vehicle information 50 (e.g. speed, steering angle and / or time information) and a known angle of incline 16 and a known distance 20th . These records 46 are used to train the artificial neural network. It can be seen that the first three data sets shown 46 each have a pitch angle 16 of 10 degrees and the distance 20th varies from 5 meters to 0.5 meters. In the second group of data sets, the slope angle is 16 always 5 degrees, while the distance assumes different values.

When using the artificial neural network in the processing unit 36 , if an unknown slope is to be recognized, the artificial neural network receives the image data 48 made available, alternatively further processed image data, such as a disparity map 56 . The output data sets that are generated by the artificial neural network then give the slope angle sought 16 and the distance 20th at.

5 shows schematically using the example of a disparity map 56 an artificial neural network as the input variable 52 with an input layer 70 , several hidden layers 72 and an output layer 74 . The number of neurons used here depends on the rasterization of the disparity map 56 or the pixels of the disparity image that was obtained from the two images of the image signal. When determining the angle of incline 16 and the distance 20th can in addition to the disparity map 56 also other vehicle information 50 include, for example, the vehicle speed and the Can include steering angle. Filtering is also possible in order, for example, to take into account variances and covariances of the data and information used, as well as image signals.

Of course it is possible, in addition to using a disparity map 56 also the images and the corresponding image data 48 to compare directly with each other and from this distance 20th and pitch angle 16 to determine.

In order to save computing time and data volumes, the images or the disparity map 56 only an interesting and relevant area can be selected, for example in the form of a rectangle or a column or alternatively in the form of a trapezoid. A region of interest (ROI) in the images is reflected in the disparity map 56 also in the form of a rectangle, a column or a trapezoid.

6th shows the method according to the invention for adapting a driving strategy. In a first step S10 of receiving, an image signal from an optical sensor 22nd with information about an environment, in particular a frontal area 18th of the vehicle 12th received. The image signal comprises two images that were generated at different locations, for example by means of a stereo camera 24 .

In an optional step S12, a disparity map is created from the image data of the images 54 56 that can be filtered, compressed or reduced. The disparity map 56 can be complete, scaled (e.g. scaled down) and / or reduced to areas of interest. The same applies to the images or the image data themselves.

The received images of the image signal or those for the disparity map 56 Further processed data are used in a further step S14 of the determination to determine an unknown pitch angle 16 an incline 14th and an unknown distance 20th between the vehicle 12th and the slope 14th to determine. Another step of adapting S16 the driving strategy based on the ascertained gradient angle 16 and the determined distance 20th follows. In a further step S18, the driving strategy is sent to a vehicle control unit 32 issued.

7th shows a further developing embodiment of the method according to the invention, in which optional, additional steps are carried out. In a step S20, an existing filter for filtering the ascertained incline angle and distance is initialized. Steps S10 and optionally S12 follow as in the method according to FIG 6th . In a step S22, the vehicle speed and steering angle are determined by the vehicle 12th determined. Step S24 includes calculating the from the vehicle 12th Distance covered between two recordings of the stereo camera 24 based on the vehicle speed. In the subsequent step S14, the slope angle is determined by means of the artificial neural network 16 and the distance 20th determined, which are filtered in step S26. The filtered values are used to adapt the driving strategy. Variances or covariances can be taken into account. In step S18, the driving strategy is sent to the vehicle control unit 32 transmitted.

The disparity map is not used for each measurement 56 calculated and trained, but directly the images 54 from the stereo camera 24 is used, the disparity map is not calculated 56 . The artificial neural network learns the slope angle 16 and the distance 20th to the slope 14th directly from the images of the camera.

In an alternative training for the artificial neural network, a mono camera is used. A so-called motion stereo is generated by recording the mono camera one after the other. Two mono camera images recorded one after the other are combined to form a motion stereo pair. These images can be whole, scaled and / or for an area of interest. The time between the two paired motion stereo images is determined on the basis of the speed of the vehicle, so that a fixed value of the distance traveled by the vehicle in vehicle and / or camera coordinates is maintained. For example, the vehicle speed can be selected in such a way that a predetermined distance traveled, for example one meter, is achieved between two motion stereo images. This requires a vehicle speed of 1 meter per second and a time difference between the recordings of the mono images of 1 second. In this case, there is no need to calculate a disparity map or use a stereo camera. For this, however, the information on the vehicle speed must be very accurate. The artificial neural network learns the angle of incline 16 and the distance 20th to the slope 14th directly from a series of paired images.

List of reference symbols

10

system

12th

vehicle

14th

pitch

16

Pitch angle

18th

Frontal area

20th

distance

22nd

optical sensor

24

Stereo camera

26th

Front area

30th

contraption

32

Vehicle control unit

34

Input interface

36

Processing unit

38

Adjustment unit

40

Output interface

42

Commercial vehicle

44

automatic transmission

46

record

48

Image data

50

Vehicle information

52

Artificial Neural Network (ANN)

56

Disparity Map

70

Input layer

72

Hidden layer

74

Output layer

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

• US 8867790 B2 [0006]

Claims (15)

Device (30) for adapting a driving strategy of a vehicle (12), in particular a commercial vehicle, with: an input interface (34) for receiving an image signal from an optical sensor (22) with information on the surroundings of the vehicle (12), the image signal comprising at least two images which were generated at two different locations; a processing unit (36) with a pre-trained artificial neural network (52) for determining an angle of incline (16) of an incline (14) in the vicinity of the vehicle (12) and a distance (20) of the vehicle (12) to the incline (14) ) based on the at least two images of the image signal; an adaptation unit (38) for adapting the driving strategy based on the ascertained gradient angle (16) and the ascertained distance (20); and an output interface (40) for outputting the driving strategy to a vehicle control unit (32). Device (30) after Claim 1 , wherein the processing unit (36) is designed such that the artificial neural network (52) is pre-trained with data sets which include two images, a slope angle (16) and a distance (20). Device (30) according to one of the preceding claims, wherein the processing unit (36) is designed such that the artificial neural network (52) is pre-trained with data sets in which the images are pre-processed. Device (30) according to one of the preceding claims, wherein the processing unit (36) is designed such that the artificial neural network (52) is pre-trained with data sets in which the images are in the form of a disparity map (56), the disparity Map (56) is further processed. Device (30) according to one of the preceding claims, wherein the processing unit (36) is designed such that the artificial neural network (52) is pre-trained with data sets which are based on measured data and / or which are based on simulated data. Device (30) according to one of the preceding claims, wherein the processing unit (36) is designed in such a way that the artificial neural network (52) is pre-trained with data records which include a time specification and / or additional vehicle information. Device (30) according to one of the preceding claims, wherein the input interface (34) is designed to receive an image signal from a stereo camera (24); and the image signal comprises a first image from a first camera sensor and a second image from a second camera sensor. Device (30) according to one of the preceding claims, wherein the input interface (34) is designed to receive an image signal from a mono camera; and the image signal comprises a first image from the mono camera and a second image from the mono camera, which were recorded one after the other. System (10) for adapting a driving strategy of a vehicle (12), with: a device (30) according to any one of the preceding claims; an optical sensor (22) for transmitting image signals with information about the surroundings of the vehicle (12); and a vehicle control unit (32) for receiving and processing the driving strategy. System (10) according to Claim 9 , wherein the optical sensor (22) is a stereo camera (24) with two camera sensors or a mono camera. System (10) according to Claim 9 or 10 wherein the vehicle control unit (32) generates control commands based on the driving strategy, which are transmitted to components of the vehicle (12) in order to implement the driving strategy. Vehicle (12) with a system according to one of the Claims 9 to 11 wherein the vehicle (12) is a utility vehicle (42). Method for adapting a driving strategy of a vehicle (12), with the steps: Receiving an image signal from an optical sensor (22) with information about the surroundings of the vehicle (12), the image signal comprising at least two images that were generated at two different locations; Determining an angle of incline (16) of an incline (14) in the vicinity of the vehicle (12) and a distance (20) of the vehicle (12) to the incline (14) based on the at least two images of the image signal; Adapting the driving strategy based on the ascertained gradient angle (16) and the ascertained distance (20); and outputting the driving strategy to a vehicle control unit (32). Method according to the preceding claim, with the further step of: using a pre-trained artificial neural network (52) to determine the angle of incline (16) the incline (14) in the vicinity of the vehicle (12) and the distance (20) of the vehicle (12) to the incline (14) based on the at least two images of the image signal. Computer program product with program code for performing the steps of the method according to Claim 13 or 14th when the program code is executed on a computer.

Images