DE102019119585A1 - Determination of the installation location and the alignment of ultrasonic sensors by means of neural networks - Google Patents

Abstract

The present invention relates to a method for monitoring the installation position and / or the alignment of at least one ultrasonic sensor of a vehicle, the method comprising the following method steps: a) sending an ultrasonic pulse by an ultrasonic sensor arranged on the vehicle and detecting an ultrasonic signal reflected from the surface of a surface covering ; b) outputting one or more measured values corresponding to the recorded ultrasonic signal; c) entering the measured value or values into an artificial neural network; d) determining the position and alignment of the ultrasonic sensor by the artificial neural network, e) outputting one or more values characterizing the ultrasonic sensor position and alignment from the artificial neural network. The present invention also relates to the use of the method according to the invention in a motor vehicle, a non-volatile, computer-readable storage medium and a driving support system which is set up to carry out the method according to the invention.

Description

The present invention relates to a method for monitoring the installation position and / or the alignment of at least one ultrasonic sensor of a vehicle, the method comprising the following method steps: a) sending an ultrasonic pulse by an ultrasonic sensor arranged on the vehicle and detecting an ultrasonic signal reflected from the surface of a surface covering ; b) outputting one or more measured values corresponding to the recorded ultrasonic signal; c) entering the measured value or values into an artificial neural network; d) determining the position and alignment of the ultrasonic sensor by the artificial neural network, e) outputting one or more values characterizing the ultrasonic sensor position and alignment from the artificial neural network. The present invention also relates to the use of the method according to the invention in a motor vehicle, a non-volatile, computer-readable storage medium and a driving support system which is set up to carry out the method according to the invention.

The volume of traffic as such and its complexity has increased significantly in recent years. Not only the fact that more road users can be found on the streets, but also the participation of new traffic groups, such as pedelecs, with new movement profiles and properties, has led to increasingly confusing traffic situations. One way of reducing the stress on road users is the use of technical aids, such as electronic driving support systems, which relieve vehicle drivers of some of the information gathering and processing and are intended to contribute to safer and more relaxed travel. The basic requirement for the systems to function properly is, of course, that the sensors used to obtain information function properly.

Using the example of ultrasonic sensors, this means that in today's driver assistance systems, installation in a new vehicle model can only take place after a complex manual calibration. This calibration is usually carried out by specially trained people who examine a wide variety of scenarios and adjust the sensors accordingly. This process is not only time-consuming, it also requires special environmental conditions. A flexible adjustment of the settings while driving, e.g. as a function of the roadway, is not provided in most cases. Another important point are the tolerances when installing the sensors. These tolerances can lead to the fact that the function of the ultrasonic sensors does not correspond to the calibrated one even in a new vehicle and thus the system does not function optimally. In addition, the alignment of the ultrasonic sensors can change in everyday operation, e.g. due to damage to the bumper. With this type of damage, the ultrasonic sensors can usually no longer be used nowadays, although the sensors themselves have not been damaged.

Various possibilities have been suggested in the patent literature to improve the functionality and to expand the possible uses.

For example, the DE 10 2017 208 268 B3 a method for determining the load condition of a vehicle and / or for determining the installation height (h) of at least one ultrasonic sensor of the vehicle, wherein an ultrasonic sensor of the vehicle is designed to emit acoustic signals and to receive echo signals from the surroundings of the vehicle, and wherein the ultrasonic sensor for Detection of the surroundings of the vehicle is operated in a first operating mode and the ultrasonic sensor for determining the load condition of the vehicle and / or for determining the installation height is operated in a second operating mode, characterized in that the ultrasonic sensor is operated in the second operating mode in such a way that the Ultrasonic sensor has increased sound radiation in one direction of a roadway compared to the first operating mode, echo signals reflected from the roadway starting to determine the load condition of the vehicle and / or to determine the installation height of the ultrasonic sensor be evaluated.

Another function test procedure is described in the DE 10 2012 216 290 A1 suggested. The document discloses a method for monitoring the function of ultrasonic sensors of a vehicle, the vehicle comprising at least one ultrasonic sensor that is oriented parallel to the direction of movement and at least one ultrasonic sensor that is oriented perpendicular to the direction of movement, the ultrasonic sensors emitting ultrasonic signals with a frequency fo and wherein the ultrasonic sensors receive floor echoes reflected from the floor while driving, characterized in that a floor echo is recorded as a reference signal with at least one ultrasonic sensor aligned perpendicular to the direction of movement and a Doppler shifted floor echo is recorded as a measurement signal with at least one ultrasonic sensor aligned parallel to the direction of movement, depending on the Doppler frequency f1 des Doppler-shifted ground echoes, a ratio (A) is formed from the measurement signal and reference signal and the frequency response curve of the at least one ultrasonic sensor aligned parallel to the direction of movement is determined from at least two ratios (A) determined at different Doppler frequencies f1, and a malfunction of an ultrasonic sensor due to deviations of the determined frequency response curve from a reference frequency response curve is detected.

Another functional check procedure is in the DE 10 2005 057 973 A1 shown. The method relates to the functional test of an ultrasonic sensor on a motor vehicle, wherein the ultrasonic sensor emits an ultrasonic signal and the ultrasonic signal is reflected from a floor surface in front of the vehicle and received again by the ultrasonic sensor or another ultrasonic sensor is characterized in that the duration and / or the amplitude of the emitted ultrasonic signal is selected to be large enough to enable reception of the signal reflected from the floor surface under conventional conditions, and that a function of the ultrasonic sensor is determined when the reflected signal is received.

Such solutions known from the prior art can offer further potential for improvement, in particular with regard to the simplicity of the structure used and the accuracy of the results obtained therefrom.

It is the object of the present invention to at least partially overcome the disadvantages known from the prior art. In particular, it is the object of the present invention to provide a flexibly adaptable solution by means of which reliable monitoring of the alignment of ultrasonic sensors is provided for a wide variety of sensor and vehicle structures and environmental conditions.

The object is achieved according to the invention by a method with the features of claim 1. The object is also achieved according to the invention by using the method according to claim 10 and a driving assistance system with the features of claim 12. Preferred embodiments of the invention are set out in the subclaims, in the description or the figures, with further features described or shown in the dependent claims or in the description or the figures individually or in any combination being able to represent an object of the invention if the context does not clearly indicate the opposite.

1. a) Aussenden eines Ultraschallimpulses durch einen am Fahrzeug angeordneten Ultraschallsensor und Erfassen eines von der Oberfläche eines Oberflächenbelags reflektierten Ultraschallsignals;
2. b) Ausgeben eines oder mehrerer dem erfassten Ultraschallsignal entsprechender Messwerte;
3. c) Eingeben des oder der Messwerte in ein künstliches neuronales Netzwerk;
4. d) Ermitteln der Position und Ausrichtung des Ultraschallsensors durch das künstliche neuronale Netzwerk,
5. e) Ausgeben eines oder mehrerer für die Ultraschallsensorposition und -Ausrichtung charakterisierenden Werte aus dem künstlichen neuronalen Netzwerk.

According to the invention, a method for monitoring the installation position and / or the alignment of at least one ultrasonic sensor of a vehicle is proposed, the method comprising the following method steps:

1. a) sending an ultrasonic pulse by an ultrasonic sensor arranged on the vehicle and detecting an ultrasonic signal reflected from the surface of a surface covering;
2. b) outputting one or more measured values corresponding to the recorded ultrasonic signal;
3. c) entering the measured value or values into an artificial neural network;
4. d) Determination of the position and alignment of the ultrasonic sensor by the artificial neural network,
5. e) outputting one or more values characterizing the ultrasonic sensor position and alignment from the artificial neural network.

Surprisingly, it has been found that the above method is able to monitor the location-specific functioning of ultrasonic sensors very precisely without further technical effort and, if necessary, to display this reliably in the event of deviations between the target and the actual behavior. These advantages are achieved without additional hardware outlay on the sensors, which generally reduces the process costs. Current situations can be mapped easily and flexibly and errors in alignment can be detected, either due to the installation as such or, for example after an accident, incorrectly positioned ultrasonic sensors. This makes it possible to adjust the calibration of the sensor accordingly in the event of deviations or to manually reposition only incorrectly positioned sensors. This can reduce the general reject rate in production and keep repair costs low. Thanks to the use of the neural network, the method is also so flexible that it is generally possible to include other influencing factors in the method, so that optimal functionality can be guaranteed. This result is also surprising, since it is generally assumed in the prior art that the intensity of reflected ultrasonic waves in normal operation, without further measures to reliably determine the ultrasonic sensor position, is too low or generally unsuitable. It is thus a further advantage of the method that this method can be operated in a “usual” operating mode of an ultrasonic sensor. It is therefore a neural network-based calibration method with which an online calibration can be carried out, the calibration being underground is independent and can also be carried out with other objects in the immediate vicinity of the vehicle. Furthermore, only the recent improvements made in the context of ultrasonic sensors make it possible to create a reliable and precise model using an artificial neural network, since "old" sensors only pass on specific data extracts or integral values instead of the entire, completely time-resolved data due to a lack of computing and processing capacity could. However, the latter largely determines the accuracy of the available models. In this respect, the increased sensitivity of the sensors and the increased data rates and volumes can be used for automated self-monitoring by an artificial neural network. A customary calibration by manual setting of threshold values as such, as previously known from the prior art, is thus superfluous.

The method according to the invention is a method for monitoring the installation position and / or the alignment of at least one ultrasonic sensor of a vehicle. The method can thus be used to determine both the height of the ultrasonic sensor above the floor and the angular alignment of the ultrasonic sensor to the floor. The orientation describes the angular component and the installation position describes the spatial component in relation to the floor or another fixed point, for example a vehicle fixed point. The individual parameters of the installation position are shown and defined further below in the description of the figures. By means of suitable coordinate transformations, however, further possibilities for the installation position can optionally also be specified, for example in the form of Cartesian coordinates in relation to a vehicle reference point. The angle can also be specified with reference to a suitable vehicle axis.

Vehicles within the meaning of the invention can be land, water or air vehicles. It is therefore possible to use the method according to the invention both for the production of aircraft, rail vehicles, motor vehicles or ships, as long as the vehicles are at least temporarily in the vicinity of structures connected to the earth. A vehicle is preferably a motor-driven land vehicle, in particular a motor vehicle.

In method step a), an ultrasonic pulse is transmitted by an ultrasonic sensor arranged on the vehicle and an ultrasonic signal reflected from the surface of a surface covering is recorded. The data of the method according to the invention are supplied via an ultrasonic sensor which has a permanent or only a temporary connection to the vehicle. The ultrasonic sensor can be permanently mounted on the vehicle or it can only be in brief contact with the vehicle. The latter can also include, for example, “stopping” the ultrasonic sensor on the vehicle via a robot arm. It is essential that both the position and the alignment of the ultrasonic sensor are known during the training and are precisely adhered to. It is of course also possible for the entire method to be carried out with one or more ultrasonic sensors.

From its position on the vehicle, the ultrasonic sensor emits sound waves which at least partially also hit the ground on which the vehicle is standing. In this case, the ultrasonic sensor emits an ultrasonic signal of this type which is at least partially aligned with the surface on which the vehicle is moving. As a rule, no additional measures are required for this, since the profile of the ultrasonic signals, which are sent out by conventional ultrasonic sensors for distance measurement, at least partially hit the surface on which the motor vehicle is located. In this context, it is very preferred that the ultrasonic signal hits the surface with such an angle of incidence that is greater than 30 °, preferably greater than 40 ° and very particularly preferably greater than 50 °. As the angle of incidence increases, the reflection rate decreases, but it has been shown that the ultrasonic sensor can still detect such reflections that enable the determination of the sensor position relative to the surface of the surface covering on which the motor vehicle is traveling with the aid of the artificial neural network. Typical values for the transmission frequency of conventional ultrasonic sensors are 40 to 60 kHz.

In method step b), one or more measured values corresponding to the recorded ultrasonic signal are output. The ultrasonic sensor therefore forwards the recorded sound waves as a function of the time interval and in the form of their amplitude as electrical signals or scanning points. In addition, it is possible that the sensor works continuously and forwards all signals as continuous measured values of one or more measurements. However, it is also possible that only some of the recorded sound waves are converted into measured values, for example as a function of the time of impact. Furthermore, both the raw data and mathematically transformed, i.e. processed data, for example in the form of a mathematical description of the envelope curve of the signals, can be passed on. In this way, for example, the amount of data or also the number of interfering signals that cannot originate from the surface can be reduced.

In method step c) the input of the measured value or values takes place in an artificial neural network. The measured values of the sound waves reflected from the surface are used as input values of an artificial neural network. To this end, the measured values can be input as such, for example in the form of a frequency spectrum, a time envelope or in a mathematically transformed form. The transformation can be selected on the basis of the expected data and as the form of the artificial neural network used. As already stated above, the entirety of the measured values or only an appropriately selected part of the data can be used as input values. This allows the amount of data to be reduced and the processing of the data to be accelerated. Different structures for artificial neural networks can be used for processing, for example artificial neural networks with FFNN or CNN structures.

In method step d), the position and alignment of the ultrasonic sensor is determined by the artificial neural network. To determine the position, the artificial neural network can, for example, have already been trained on the basis of different data sets from different application situations with different substrates, different ultrasonic sensor positions and orientations. In such a training phase, the artificial neural network learns to differentiate between different positions based on different measured values output by the ultrasonic sensor. It applies here that a corresponding training is basically only required once for each arrangement of an ultrasonic sensor or a plurality of ultrasonic sensors and for each type of vehicle. Once the system has been trained, the determination of the ultrasonic sensor position and orientation can be applied directly to the currently measured data from the vehicle currently in use.

In method step e), one or more values characterizing the ultrasonic sensor position and alignment are output from the artificial neural network. By entering the reflections on the ground, fundamental data such as the height above the ground, the alignment of the ultrasonic sensor in relation to a vehicle axis or to the ground can be determined for the alignment of ultrasonic sensors. One possibility for defining the ultrasonic sensor position is given further back in the figures, for example. However, it is also possible to specify other coordinate systems for outputting the ultrasonic sensor position and alignment, for example in relation to a vehicle reference point and a reference axis. These data result as output values of the artificial neural network.

In a preferred embodiment of the method, the artificial neural network can be trained using ultrasonic sensor echo data that was obtained with a reference object different from the surface covering at one or more predefined distances. It has proven to be advantageous to train the artificial neural network using ultrasound sensor data in which the ultrasound echo has both components from reflections on the surface and additional echo components from another object which is positioned at a defined distance from the ultrasound sensor . The calibration of the artificial neural network is therefore carried out with a reference object at predefined distances. This method is particularly suitable for motor vehicles and here for new vehicles, as the precisely defined structure of the additional object together with the precisely known distance between the sensor, the ground and the object can achieve a high level of accuracy in determining the position and orientation of the artificial neural network . Accuracy can also be increased by including different objects, surfaces, distances and ultrasonic sensor positions and orientations in the training. It applies here that a corresponding training is basically only required once for each arrangement of an ultrasonic sensor or a plurality of ultrasonic sensors and for each type of vehicle. Once the system has been trained, the determination of the position and the orientation can be applied directly to the measured data of the vehicle currently in use.

In a further preferred embodiment of the method, the artificial neural network can be trained using ultrasonic sensor echo data that originate from the surface of one or more defined surface coverings. Another preferred option for training is to use a specific floor covering, or the floor echoes on this covering, for calibration or for teaching the artificial neural network. Asphalt, for example, is a possible floor covering. In this case, a new calibration can advantageously also be carried out by the end user or automatically by the system itself.

In a further alternative of the method, the artificial neural network can be trained using ultrasonic sensor echo data, the echo data coming from an ultrasonic sensor in different installation positions and / or orientations. It has been found to be particularly cheap pointed out that regardless of the choice of the structure selected for training the artificial neural network, for example including a special surface, vehicle, other reference objects, various installation angles and installation heights, the response signals are recorded for training purposes. This means that these variances in the position and alignment of the sensor can be combined with calibration using known floor coverings or with calibration using an additional reference object at a known distance. These combinations make it possible to obtain particularly robust and precise models. It has also been shown that the training of the artificial neural network becomes better the higher the number of different ultrasonic sensor positions and orientations. In order to generate the large number of necessary training data, it is possible to automate the measurements, for example with the help of a robot arm, which attaches the sensor to the vehicle only temporarily but with a fixed position and orientation.

In a preferred variant of the method, the ultrasonic signal echo data can be divided into different time ranges before being input into the artificial neural network. To reduce the amount of data and to maintain the possibility of processing the data immediately, i.e. without a significant time delay, it has proven advantageous not to use all the data supplied by the ultrasonic sensor, but only certain data windows from it. For training, the amount of data transferred to the network can be reduced by using certain distance ranges for the echoes. By suitable preprocessing, the size of the input data of the network can typically be reduced to such an extent that particularly efficient artificial neural network structures can be used, for example multilayer perceptrons. Due to the small network depth, architectures can be created that are suitable for online calibration on a vehicle ECU. So it is no longer necessary to switch to larger computer capacities.

In a further preferred embodiment of the method, the measured value can be entered into the artificial neural network in vector or matrix form. With the method according to the invention, the main part of the signal processing can typically take place in the sensor itself. The data that are available on the ECU can, however, already be filtered, sampled and quantized. Only a relevant section of this data is required for the network. The data are then advantageously available as a vector that can be transferred to the network.

In a preferred embodiment of the method, the artificial neural network can be a forward neural network, FFNN. It has turned out to be particularly suitable to use an artificial neural network with the structure of a “feed forward neural network” FFNN in the form of a multilayer perceptron. To implement the process according to the invention, up to 5 hidden layers are conceivable and sufficient for the artificial neural network. The exact number of layers and neurons can be determined during the training process. The output layer can consist of two neurons, for example, which in particular determine a height and an angle. In an extension of the method to a second angle and a relative distance to the next sensor, four neurons would then be required. The number of neurons in the input layer depends on the selected section and the sampling rate of the data. Typically, however, fewer than 100 neurons might be suitable. Heuristic methods such as SGD (“Stochastic Gradient Descent”), for example, can be used as an optimization function, which optimize the parameters in the direction of a local minimum in the deviation. ReLU can typically be used for the hidden neurons. Due to the selected regression, a sigmoid activation can expediently be used in the output layer, for example in cases in which the input data is normalized or no activation function is used.

In a further preferred embodiment of the method, the artificial neural network can include as input values, in addition to the ultrasonic signal echo data, the constructively provided values of the installation angle and height of the ultrasonic sensor used. For training the artificial neural network, it has been found to be particularly effective that, in addition to the ultrasonic sensor echo data, the constructional installation positions of the sensors currently in use are also fed into the network as input values. In other words, data is also fed into the network, which includes, for example, the actual target height and the target alignment of the ultrasonic sensor. The target height and orientation are those values that should be optimally adhered to according to the vehicle planning. These installation data can of course differ from the data of the current situation. This can contribute to a faster learning phase and improved results. These parameters are only mentioned here as preferred, since the signal properties that the neural network actually uses for the regression of the angle and height are determined by the network itself during training.

In a further alternative of the method, in a further method step f), the position and orientation determined via the artificial neural network can be compared with a target position and orientation stored for the ultrasonic sensor. In a further step, the setpoint and actual values for one or all of the ultrasonic sensors present on the vehicle can be compared within the method according to the invention. The comparison can be used to determine whether the positions match or whether there are significant deviations in position and / or alignment. In the latter case, the calibration of the affected sensor can be adjusted or a warning can be issued to the vehicle driver.

The use of the method according to the invention in a motor vehicle is also according to the invention. The method according to the invention is particularly suitable for motor vehicles, since no further investment in sensor technology has to be made via the ultrasonic sensors already present on the vehicle. In addition, it has been shown that the floor coverings usually found in road traffic allow particularly precise calibration or efficient training of the artificial neural network. This use enables the position of the ultrasonic sensors present to be monitored automatically and immediately over time.

Furthermore, according to the invention is a non-volatile, computer-readable storage medium with commands stored thereon which, when executed on a processor, bring about the method according to the invention.

Furthermore, according to the invention, a driving support system comprising at least one ultrasonic sensor which is designed to transmit acoustic signals and to receive echo signals from the vehicle environment, a control unit for controlling the at least one ultrasonic sensor and an evaluation unit for evaluating the received echo signals in the form of an artificial neural network, wherein the A driver assistance system is designed to carry out a comparison of a target position and orientation stored for the ultrasonic sensor with a position and orientation determined via the artificial neural network according to the method according to the invention. With regard to the advantages of the driving assistance system that can be used according to the invention, reference is made explicitly to the advantages of the method according to the invention. It should also be noted that, based on the selected evaluation via an artificial neural network, the computing power of commonly used vehicle central units (ECU) is sufficient, so that no further investments are required here to carry out the method according to the invention.

• 1 schematisch ein Kraftfahrzeug mit Ultraschallsensoren, die zum Senden und Empfangen von Ultraschallsignalen eingerichtet sind;
• 2 ein mögliches Teilkoordinatensystem zur Kennzeichnung der Position und Ausrichtung eines Ultraschallsensors; und
• 3 eine mögliche erfindungsgemäße Ausgestaltung einer Sensoranordnung im Rahmen eines erfindungsgemäßen Fahrunterstützungssystems.

Further advantages and advantageous configurations of the subject matter according to the invention are illustrated by the drawings and explained in the following description. It should be noted that the drawings are only of a descriptive nature and are not intended to restrict the invention in any way. In the drawings show:

• 1 schematically, a motor vehicle with ultrasonic sensors which are set up for sending and receiving ultrasonic signals;
• 2 a possible partial coordinate system to identify the position and orientation of an ultrasonic sensor; and
• 3 a possible embodiment according to the invention of a sensor arrangement within the scope of a driving support system according to the invention.

From the 1 is schematically a motor vehicle 1 with ultrasonic sensors 2 that are set up to send and receive ultrasonic signals. The ultrasonic sensors 2 are on the one hand in the front area of the motor vehicle 1 and on the other hand in the rear area of the motor vehicle 1 arranged and can, as is generally known from the prior art, serve to measure the distance for parking processes.

The 2 shows schematically a possibility for defining the installation location of an ultrasonic sensor 2 . The position of the ultrasonic sensor 2 can be relative to the ground 3 or to the reference background 3 Specify the location of the ultrasonic sensor 2 through the height 4th from the ground 3 and the angle 5 to the ground 3 is marked. However, other coordinate systems can also be used, which determine the relative position between the ground 3 and ultrasonic sensor 2 mark.

In 3 is shown schematically that the ultrasonic sensors 2 According to the preferred exemplary embodiments of the invention described here, they are part of a sensor arrangement which, via the ultrasonic sensors 2 also a central unit 6th (ECU) with which the ultrasonic sensors 2 are connected. The central unit 6th controls the ultrasonic sensors 2 and can from the ultrasonic sensors 2 process recorded ultrasonic signals. The central unit 6th including an artificial neural network 7th , in the present case a forward-looking artificial neural network 7th (FFNN), on.

List of reference symbols

1

vehicle

2

Ultrasonic sensor

3

ground

4th

height

5

angle

6th

Central unit

7th

artificial neural network

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

• DE 102017208268 B3 [0005]
• DE 102012216290 A1 [0006]
• DE 102005057973 A1 [0007]

Claims (12)

Method for monitoring the installation position and / or the alignment of at least one ultrasonic sensor (2) of a vehicle (1), characterized in that the method comprises the following method steps: a) Emission of an ultrasonic pulse by an ultrasonic sensor (2) arranged on the vehicle (1) and detecting an ultrasonic signal reflected from the surface of a surface covering; b) outputting one or more measured values corresponding to the recorded ultrasonic signal; c) input of the measured value or values into an artificial neural network (7); d) determining the position and alignment of the ultrasonic sensor (2) by the artificial neural network (7), e) outputting one or more values characterizing the ultrasonic sensor position and alignment from the artificial neural network (7). Procedure according to Claim 1 , wherein the artificial neural network (7) is trained via ultrasonic sensor echo data which was obtained with a reference object different from the surface covering (3) at one or more predefined distances. Method according to one of the preceding claims, wherein the artificial neural network (7) is trained using ultrasonic sensor echo data which originate from the surface of one or more defined surface coverings (3). Method according to one of the preceding claims, wherein the artificial neural network (7) is trained via ultrasonic sensor echo data, the echo data originating from an ultrasonic sensor (2) in different installation positions and / or orientations. Method according to one of the preceding claims, wherein the ultrasonic signal echo data are divided into different time ranges before being input into the artificial neural network (7). Method according to one of the preceding claims, wherein the measured value is entered into the artificial neural network (7) in vector or matrix form. Method according to one of the preceding claims, wherein the artificial neural network (7) is a forward neural network (7), FFNN. Method according to one of the Claims 2 - 7th , the artificial neural network (7) including as input values, in addition to the ultrasonic signal echo data, the constructively intended values of the installation angle and height of the ultrasonic sensor used. Method according to one of the preceding claims, wherein in a further method step f) the position and alignment determined via the artificial neural network (7) is compared with a target position and alignment stored for the ultrasonic sensor (2). Use of a method according to one of the preceding claims in a motor vehicle (1). Non-volatile, computer-readable storage medium with instructions stored on it, which when executed on a processor, a method according to one of the Claims 1 - 9 cause. Driving support system comprising at least one ultrasonic sensor (2) which is designed to send out acoustic signals and to receive echo signals from the vehicle environment, a control unit (6) for controlling the at least one ultrasonic sensor (2) and an evaluation unit for evaluating the received echo signals in the form of an artificial neural one Network (7), characterized in that the driver assistance system is designed to compare a target position and orientation stored for the ultrasonic sensor (2) with that of the artificial neural network (7) according to a method of Claims 1 - 9 determined position and alignment.

Images