DE102021003597A1 - Method for influencing the behavior of at least one occupant in a vehicle - Google Patents

Abstract

The invention relates to a method for influencing the behavior of at least one occupant in a vehicle. According to the invention, a behavior-triggering context for a behavior is recognized and, on the basis of the existing behavior-triggering context recognized, it is predicted that undesired behavior of the occupant will occur without influencing the occupant, and an intervention is carried out to influence the behavior of the occupant in order to suppress or change the undesired behavior of the occupant.

Description

The invention relates to a method for influencing the behavior of at least one occupant in a vehicle according to the features of the preamble of claim 1.

From the prior art, as in DE 11 2019 001 601 T5 described, a method and a device for safety-enhanced computer-assisted driving known. The device comprises a neural network in order to determine a classification for a behavior of a driver of a vehicle having the device, based at least in part on data collected in real time about the vehicle and a current level of stress or fatigue of the driver determined in real time . The device further comprises a safety action machine, which is coupled to the neural network, in order to determine a safety-related action based at least in part on the determined driver behavioral classification and data relating to the current traffic or road condition of a route on which the vehicles are traveling just move. The safety-related action can be performed by an infotainment system or a navigation system of the vehicle to help the driver to drive the vehicle in a safer manner.

The invention is based on the object of specifying a method, which is improved over the prior art, for influencing the behavior of at least one occupant in a vehicle.

The object is achieved according to the invention by a method for influencing the behavior of at least one occupant in a vehicle having the features of claim 1.

Advantageous refinements of the invention are the subject matter of the subclaims.

In a method for influencing the behavior of at least one occupant in a vehicle, a behavior-triggering context for an, in particular undesired, behavior is recognized according to the invention. On the basis of the existing behavior-triggering context recognized, in particular a nontrivial context, in particular a complex context, it is predicted that, without influencing the occupant, an undesired behavior of the occupant, namely the behavior whose behavior-triggering context was recognized, will take place. In the method according to the invention, therefore, it is advantageously not an undesired behavior that is already taking place that is recognized, but rather the context, i. H. Influencing factors that subsequently lead to undesirable behavior. This is used to predict that the occupant will carry out the undesired behavior in the future, for example after one or more time units, without influencing the occupant. As a result, the occupant can already be warned before the occurrence of the undesired behavior. According to the invention, an intervention is then carried out to influence the behavior of the occupant in order to suppress or change this future undesired behavior of the occupant, which is likely to occur without this influencing.

The method according to the invention thus in particular enables automatic recognition of the behavior-triggering context for behavior change and, for example, supportive behavioral therapy in the vehicle, i. H. to enable this change in behavior and, for example, supportive behavior therapy in the vehicle.

For example, recognized behavior-triggering properties of the behavior-triggering context are used in order to change the behavior of the occupant and / or to support behavior therapy, in the vehicle and / or outside the vehicle. This behavior therapy can be carried out or at least supported by a medical professional, for example.

For the intervention to influence the occupant, in particular the vehicle, in particular at least one component of the vehicle, is used in order to allow the occupant at least one behavioral alternative or several behavioral alternatives.

In order to influence his behavior, the occupant is informed, for example, that it has been predicted that undesired behavior of the occupant will occur, in order to thereby influence and thus change the behavior of the occupant.

Methods of computer-aided and / or machine learning are used in particular to identify the undesired behavior exercised by the occupant and / or to identify the context that triggers the behavior, in particular by means of at least one technical device in the vehicle.

For example, undesirable behavior on the part of the occupant is classified.

To carry out the method according to the invention, it therefore advantageously enables a technical device in the vehicle, For example, to recognize security-relevant wrong behaviors and / or at least one behavior marked as undesirable or harmful, for example by the user himself, i.e. the undesired behavior of the user, to learn the context of the undesired behavior that triggers the behavior and to predict when the behavior will occur. This gives the user the opportunity to change or correct his undesired behavior.

The solution according to the invention increases safety in road traffic in particular by recognizing and correcting the undesired behavior, in particular incorrect behavior that is relevant to safety, in particular of users acting as drivers of the respective vehicle. The solution according to the invention also increases comfort in the vehicle, since it supports the user in discarding harmful behavior or behavior that the user considers undesirable. Furthermore, the solution according to the invention can increase the effectiveness of behavior therapy, since it can support a medical specialist in the therapy. As a result, financial and time savings can be achieved for the user.

In particular, in the described method, in particular by means of computer-aided vision methods, camera-based, for example by means of one or more interior cameras of the vehicle, undesired or harmful behaviors that have been selected in particular by the user, i. H. the undesired behavior of the user, at a point in time t0, d. H. at a current point in time, recognized and the point in time t0 is marked. Advantageously, feature values, for example location, time, emotional state and / or the presence of other occupants in the vehicle, which occurred at time t0 and also before the undesired behavior, i.e. H. occurred before time t0, recorded, in particular features and / or actions immediately preceding in time. Advantageously, using methods of computer-assisted and / or machine learning, for example decision trees and / or neural networks, models are trained which recognize patterns in the feature values which are predictive of the undesired behavior. I. E. a certain combination of feature values and, for example, an additional change in feature value combinations over time will predict the undesired behavior as likely. This particular combination of characteristic values and, for example, the change over time of characteristic value combinations is the context that triggers the behavior.

If the model detects the undesired behavior on the basis of recorded current and past characteristic values, i. H. in particular a behavior marked by the user as undesirable or harmful is predicted, the vehicle can inform the user about this and alternative courses of action, i.e. H. Behavioral alternatives are recommended. These alternative modes of action can, for example, be predefined by a vehicle manufacturer, for example during the development of the vehicle, and / or can be adapted by the user according to his personal intentions.

Advantageously, important features and important feature values and, for example, additionally important temporal profiles of feature values, which are predictive of the undesired behavior, can be extracted from the model.

Advantageously, the method and thus the vehicle can support a behavior therapy in which the behavior-triggering context, a different routine corresponding to the personal intention, in particular a different behavior, as well as a reward for the different behavior and thus for avoiding the undesired behavior, adjusted by a medical professional and in the vehicle interaction, d. H. in the procedure described here.

Embodiments of the invention are explained in more detail below with reference to drawings.

• 1 eine schematische Darstellung eines Verfahrens zur Beeinflussung eines Verhaltens mindestens eines Insassen in einem Fahrzeug, und
• 2 eine schematische Darstellung eines Ablaufdiagramms des Verfahrens zur Beeinflussung eines Verhaltens mindestens eines Insassen in einem Fahrzeug.

Show:

• 1 a schematic representation of a method for influencing the behavior of at least one occupant in a vehicle, and
• 2 a schematic representation of a flow chart of the method for influencing the behavior of at least one occupant in a vehicle.

Corresponding parts are provided with the same reference symbols in all figures.

Based on 1 and 2 In the following, a method for influencing the behavior of at least one occupant in a vehicle is described, which in particular automatically recognizes a behavior-triggering context for behavior change and supportive behavior therapy 8th in the vehicle. It shows 1 a schematic representation of the process and 2 shows a flow chart of the method.

First briefly summarized in the method described here using a recognized present behavior-triggering context, in particular a nontrivial context, in particular a complex context, predicts that, without influencing the occupant, undesired behavior of the occupant will occur. It is thus advantageously not an undesired behavior that is already taking place that is recognized, but rather the context, ie influencing factors which subsequently lead to the undesired behavior. This is used to predict that the occupant will carry out the undesired behavior in the future, for example after one or more time units, without influencing the occupant. As a result, the occupant can already be warned before the occurrence of the undesired behavior. An intervention is then carried out for influencing the behavior of the occupant in order to suppress or change this future undesired behavior of the occupant which is likely to occur without this influencing.

For example, recognized behavior-triggering properties of the behavior-triggering context are used in order to change the behavior of the occupant and / or a behavior therapy 8th , in the vehicle and / or outside the vehicle. This behavior therapy 8th can, for example, be carried out by a healthcare professional 9 take place or at least be supported.

For the intervention to influence the occupant, in particular the vehicle, in particular at least one component of the vehicle, is used in order to allow the occupant at least one behavioral alternative or several behavioral alternatives.

In order to influence his behavior, the occupant is informed, for example, that it has been predicted that undesired behavior of the occupant will occur, in order to thereby influence and thus change the behavior of the occupant.

Methods of computer-assisted and / or machine learning are used in particular to identify the undesired behavior exercised by the occupant and / or to identify the context that triggers the behavior, in particular by means of at least one technical device 3 in the vehicle. This technical facility 3 is, for example, a control unit or another device for carrying out the method.

For example, undesirable behavior on the part of the occupant is classified.

For the method described here, an interior camera arranged in the vehicle or several such interior cameras are advantageously used. As a result, data about the occupant or the multiple occupants of the vehicle can be recorded in a time-resolved manner. In particular by means of computer-aided and / or machine vision methods, a position of the respective occupant and body movements, for example gestures, are recognized and localized. Furthermore, by means of these methods of computer-aided and / or machine vision, objects can also be recognized and / or object classes can be assigned and a position of the respective object can be recognized. Thus, relative positions of objects to occupants of the vehicle can also be recorded, with an interior camera designed as a mono camera, at least in the image plane, with an interior camera designed as a stereo camera by triangulating the respective positions in three-dimensional space. In this way, a variety of behaviors can be recognized in an automated manner, for example using a cell phone, smoking a cigarette, eating sweets, grinding or clenching one's teeth, biting nails and other, in particular undesirable, behavior.

In the manner described, the methods of computer-aided vision thus make it possible, camera-based, in particular by means of at least one interior camera of the vehicle, a variety of behaviors at a point in time t0, ie. h to recognize at a current point in time at which the respective undesired behavior takes place and to mark this point in time t0. In the method described here, this is carried out with at least one undesirable behavior of at least one occupant of the vehicle.

Advantageously, sensors arranged in the vehicle are also used 2 Feature values, ie sensor data and sizes of various features derived from sensor data. As sensors 2 For example, a position determination sensor, a clock, an interior microphone or several such interior microphones, one or more interior cameras, a thermometer, a rain sensor, a vehicle environment monitoring system and / or other sensors 2 be used. The feature values, ie the sensor data as well as variables derived from the sensor data, are then, for example, a location of the vehicle, a time, the occupant's current emotion, the occupant's action, an outside temperature, an amount of rain and / or a traffic flow and / or other characteristic values corresponding characteristics. The features to which these feature values belong are, for example, location, time, emotion, action, weather, traffic situation and / or other features.

In particular, these feature values and their temporal progression are recorded and recorded not only at the point in time t0, that is to say at the current point in time, but also in a period prior to this current point in time t0.

The features and thus also the feature values are advantageously assigned unambiguously to the respective occupant, for example a driver and passengers of the vehicle. Various methods are known for identifying the respective occupant. For example, the respective occupant is captured by image recordings, in particular by means of the at least one interior camera of the vehicle in combination with known methods of machine vision, in particular by face recognition, and / or by a personalized vehicle key and / or by a personal PIN code or other code, identified. The recorded sensor data, i. H. the feature values can thus be clearly assigned to a respective occupant in the vehicle.

The detection of emotions of the respective occupant takes place, for example, on the basis of facial, speech or physical signals by means of a corresponding sensor system of the vehicle, for example as in FIG US 2012/0130196 A1 described. For this purpose, the above-mentioned technical facility analyzes 3 of the vehicle, which is provided for carrying out the method described here, images from the at least one interior camera or the multiple interior cameras of the vehicle and / or data from vital sensors arranged in the vehicle in real time / online. A machine vision approach, in particular using neural networks, makes it possible to classify facial expressions and body movements of the respective occupant and to automatically recognize and classify emotions.

In addition to an analysis based on visual data, seat sensors in a respective vehicle seat can also, alternatively or additionally, be evaluated. Their data can, for example, indicate jerky, for example frightening, movements of the respective occupant.

As an alternative or in addition, for example, an analysis of audio data, which is recorded in particular by means of the at least one interior microphone, can be carried out. This enables, for example, moments of fright and / or stress to be recognized.

Furthermore, the already mentioned vital sensors can be used, for example, to qualify the well-being of the respective occupant, for example on the basis of abnormalities in the EKG, in the heart rate variability, in the systolic blood pressure, and / or in the oxygen saturation in the blood.

In the method described here, it can be provided, for example, that if a respective occupant of the vehicle, for example the driver and / or a passenger, behaves in an undesirable manner, for example a harmful or subjectively undesirable behavior, for example using a mobile phone, smoking a cigarette, sweets eating, grinding or clenching teeth, nail biting or similar behavior, for example, he selects the technical device via a menu function in vehicle settings or in another way 3 automatically recognizes the selected undesired behavior and marks the point in time t0 of the occurrence of the undesired behavior, ie the point in time at which the undesired behavior occurs.

Feature values which at this point in time t0 as well as before the occurrence of the undesired behavior, i. H. occurred before time t0, in particular directly preceding actions, e.g. other behaviors, for example using the radio or other behaviors, are recorded. Such recorded feature values are, for example, location, time, emotional state of the occupant, presence of other occupants in the vehicle and / or the traffic situation.

Using machine learning methods, for example with decision trees and / or neural networks, these recorded data in the technical facility 3 Models train which patterns in the feature values as well as the temporal progression of feature values recognize which are predictive of the undesired behavior. I. E. a certain combination or a set of combinations of feature values as well as the temporal course of feature values will predict the undesired behavior as likely. This specific combination or set of combinations of feature values and the time course of feature values that are predictive of the undesired behavior are thus a behavior-triggering context for the undesired behavior. In the method described here, the behavior-triggering context, which subsequently leads to the undesired behavior at least with a very high degree of probability, is thus first learned in the method described here.

If the model is sufficiently trained with examples of the undesired behavior and the data of the feature values, then it is possible to identify a behavior which the occupant has identified as undesirable or harmful, ie the undesired Behavior to be predicted by the model or to be expected with a certain, in particular high or very high probability.

Additionally or alternatively, the feature values and their temporal course can be evaluated algorithmically, for example by calculating feature value distributions, according to predetermined rules, in particular created in a development. If, for example, feature value distributions follow a predefined curve or feature values are above a predefined threshold value, an undesirable or harmful behavior, i.e. the undesired behavior of the occupant, can be predicted with a certain, in particular high or very high, probability using the model created with the algorithm will. However, this can only be done during development, in particular during development of the vehicle and / or the technical device 3 and / or the specific process sequence, known behavior-triggering contexts are taken into account. Individual, complex behavior-triggering contexts are recorded in the other way described above with the model created by means of machine learning.

If the model, which recognizes the behavior-triggering context of the undesired behavior, has been sufficiently trained in the manner described, then in the method described here, the model predicts in the technical facility 3 On the basis of current feature values, for example time, place, emotional state, other passengers traveling with them, weather and / or radio program, as well as the temporal course of the feature values continuously, whether the undesired behavior will occur, in particular whether the probability of the occurrence of the undesired behavior over a predetermined threshold is. In this case it is advantageously provided that the occupant is informed by the vehicle that he will carry out an undesired behavior, in particular a behavior selected by him as undesirable.

Furthermore, the occupant is advantageously given at least one alternative course of action, i. H. at least one behavioral alternative is recommended. Such alternative modes of action can, for example, be predefined during development and / or adapted or specified by the occupant according to his or her personal intentions. In the case of several courses of action it can be provided, for example, that the occupant can select one of the alternative courses of action so that it is always suggested in the future. The method thus advantageously leads to the occupant not following his unconscious undesired behavior, but his conscious plan, in particular in the form of the behavior alternative. In this way, the unconscious and undesirable behavior can be replaced by a planned and desired behavior.

In an advantageous embodiment, important features and important feature values are extracted from the learned model, which are predictive of the undesired behavior. This information about the behavior-triggering context can be made available to the user of the method even if he is not an occupant of the vehicle, i. H. also outside of the vehicle, help to recognize situations with the context that triggers the behavior and to adapt one's behavior accordingly in order to avoid the undesired behavior. For example, it is provided that this information is provided to the occupant / user via a so-called user interface of the vehicle, i.e. H. in particular via an optical and / or acoustic output unit, can be made accessible, for example displayed to him. Alternatively or additionally, it can be provided, for example, that this information is transmitted from the vehicle, for example via the user interface, to another, in particular personal, unit of the occupant / user, for example on a mobile phone, in particular smartphone, or another computer-aided device of the occupant / User.

In one possible embodiment it can be provided that, in particular from the system, in particular from the technical device 3 behavior therapy to carry out the procedure 8th of the occupant is supported. For example, important features, important feature values and the temporal course of important feature values that are predictive of the undesired behavior, i.e. the context that triggers the behavior, for example by means of data transmission via a cloud or a digital medical record, i.e. in particular via a corresponding server external to the vehicle, a treating medical professional 9 , for example to a doctor. I. E. this information on the behavior-triggering context is transmitted from the vehicle to the corresponding server external to the vehicle. You are then, for example, the medical professional 9 accessible and can therefore be used for behavioral therapy, for example 8th used outside the vehicle.

For example, it can also be provided that the behavior-triggering context, another routine corresponding to the personal intention of the occupant / user, ie a behavior alternative, as well as, for example, a reward for avoiding the undesired behavior by the medical professional 9 adapted and in the Interaction of the vehicle with the occupant should be taken into account during the procedure. For this purpose, there is advantageously an opportunity for the medical professional to interact 9 provided with the vehicle, for example the transmission of this by the medical professional 9 adapted information from the cloud or the digital medical record, ie in particular from the server external to the vehicle, into the vehicle, in particular into the technical device 3 to carry out the process, in particular by means of a corresponding data transfer. As an alternative or in addition, for example, an in particular manual interaction directly with the vehicle, in particular in the vehicle, can be provided so that this can be carried out by the medical specialist 9 adapted information in the vehicle, in particular in the technical device 3 to carry out the process.

In a possible embodiment of the method, as an alternative or in addition to the above-described selection and / or specification of the undesired behavior to be avoided by the respective occupant and / or already during development, it can be provided that, in particular by the technical device 3 , harmful or undesirable behavior, in particular incorrect behavior that is relevant to safety, for example gripping a steering wheel of the vehicle with only one hand, letting go of the steering wheel when the assistance systems are switched on, making calls while driving directly using a mobile phone and / or reaching under the steering wheel Maneuvering, as undesirable behavior of the respective occupant, in particular the driver, are automatically recognized. In this case, the further procedure takes place as described above, with the exception that the undesired behavior in question is now automatically recognized as such undesirable behavior and does not have to be specified first.

The method has so far been described on the basis of a single undesired behavior and thus on the basis of a model trained on it, in particular on its behavior-triggering context. As an alternative, it can of course also be provided that the method is carried out simultaneously in the manner described with regard to several undesired behavior of the same occupant and / or several different occupants. In particular, it is provided for this that, in particular by means of the technical device 3 , not just one, but several models for different behaviors, ie for different undesirable behaviors, are trained, in particular trained in parallel, and used accordingly to predict the respective undesired behavior in the manner described above. Different behaviors, ie different undesirable behaviors, can thus be analyzed at the same time. The method described above is thus a method for influencing the behavior of at least one occupant or several occupants in the vehicle in order to avoid at least one undesired behavior or several undesirable behaviors of the at least one or the respective occupant of the vehicle.

In the following, two examples of undesired behavior and how to avoid it are described to illustrate the method.

In the first example, the occupant of the vehicle acting as the driver is a frequent driver in the field and would like to eat less sweets while driving. The undesirable behavior is therefore the consumption of sweets. For example, the occupant now selects in the user interface of the vehicle that the consumption of sweets is to be recognized as undesirable behavior and that alternative actions, ie. H. Behavioral alternatives should be proposed.

If the occupant now guides objects recognized as candy by the vehicle, for example chocolate bars or chips, in the direction of their mouth, this is recognized as undesirable behavior, the behavior-triggering context is recorded and the model is trained in the manner described above. The occupant advantageously receives a warning from the vehicle that he is currently consuming sweets, and advantageously the additional information that the behavior-triggering context is being analyzed.

After the occupant has eaten candy several times and thus recognized the undesired behavior several times and recorded the behavior-triggering context, the model is sufficiently trained to recognize the behavior-triggering context. For example, the occupant only reaches for the candy between 2 p.m. and 4 p.m. when he is stressed and stuck in a traffic jam.

On the next trip in the afternoon, the vehicle is stuck in a traffic jam and the stress level increases. The likelihood of the behavior “eating sweets”, ie undesirable behavior, is high and the context that triggers the behavior is determined by the technical facility 3 recognized. The occupant now receives a message from the vehicle that the behavior “eating sweets”, ie the undesired behavior, is likely, and the behavioral alternatives of listening to an audio book, drinking water or planning a break at a rest area are suggested, for example. The occupant can select an alternative behavior, which in the future, for example, will always be included should be linked to the context that triggers the behavior. Furthermore, the vehicle occupant is advantageously given the information about the context that triggers the behavior.

In the second example, the technical facility recognizes 3 the vehicle automatically shows a safety-relevant incorrect behavior, ie undesirable behavior, for example constantly reaching under the steering wheel when maneuvering. The occupant acting as the driver advantageously receives a warning that he is currently gripping the steering wheel from below. He then also receives an explanation that if the vehicle stops suddenly, the wrist can break. As an alternative recommendation for action, ie as an alternative behavior, it is recommended to grip the steering wheel from above. In addition, the occupant is advantageously informed that the context that triggers the behavior is being analyzed. After the occupant has seen and understood the explanation, he grips the steering wheel when maneuvering from above, but falls back into the old unconscious and undesired behavior in certain situations, ie after the context that triggered the behavior has been established. This is the behavior-triggering context of the technical facility 3 recognized, the occupant is also shown the warning and the explanation again as a precaution.

Based on the schematic representation of the process in 1 the method is now described in summary again. In the case shown, the method includes an occupant input 1 regarding attitudes towards undesired behavior that should be recognized and avoided in the future. Furthermore, the internal and / or external sensors already described above are used for the method 2 and thus sensor data from these sensors 2 used. Sensors 2 are for example the at least one interior camera, the at least one interior microphone, the at least one thermometer, the at least one rain sensor, the at least one environment monitoring sensor system, at least one vital sensor, at least one seat sensor in at least one vehicle seat and / or the at least one position determination sensor and / or the at least a clock.

The technical facility 3 for example, a control unit of the vehicle is used to carry out the method. The procedure carried out by this technical facility 3 is executed includes a recognition section 4th , an information section 5 , a model training section 6th and a model usage section 7th . The by means of the occupant input 1 The settings made for undesired behavior that are to be recognized and avoided in the future flow into the recognition section 4th a. The sensor data of the internal and / or external sensors 2 flow into the recognition section 4th , the model training section 6th and the model usage section 7th a.

In the recognition section 4th takes place, in particular on the basis of the occupant input 1 settings made for undesired behavior that is to be detected and avoided in the future, and based on the sensor data of the internal and / or external sensors 2 , in the manner described above, the detection of the undesired behavior at the point in time t0, ie at the point in time at which the undesired behavior is currently taking place.

When the undesirable behavior in the detection section 4th detected, then takes place in the information section 5 the corresponding information to the occupant, as described above. In particular, the occupant is informed about the detected undesirable behavior, and alternative options for action, ie behavior alternatives, are indicated. If information on the behavior-triggering context is already known, this behavior-triggering context information, in particular extracted from the model, is communicated to the occupant.

Furthermore, if the undesirable behavior in the detection section 4th detected in the model training section 6th trains a model in the manner described above, in particular by means of the sensor data of the internal and / or external sensors 2 and the features derived therefrom, for example emotions and / or actions, for the period before time t0 to this time t0. The training takes place with methods of machine and / or computer-aided learning, for example neural networks and / or decision trees.

If the model is sufficiently trained, it will be used in the model usage section 7th used to predict the impending occurrence of the undesirable behavior. The prediction of the likely occurrence of the undesired behavior takes place here on the basis of the sensor data from the internal and / or external sensors 2 and the features derived from these using the trained model.

If this prediction is available, it takes place in the information section 5 the corresponding information to the occupant, as described above. In particular, the occupant is informed about the predicted undesired behavior and alternative options for action, ie behavior alternatives, are indicated. Furthermore, the occupants are shown the files extracted from the model Behavior-triggering context information communicated to the behavior-triggering context.

In one possible embodiment, external behavior therapy is also optional 8th provided, d. h outside the vehicle. This is done using the information from the information section 5 , in particular the information issued to the occupant, additionally to a medical professional in the manner described above 9 transmitted. The health professional 9 then transmits information for optimizing the model, in particular to the model usage section 7th . For example, it specifies options for action, ie behavior alternatives, and / or alternative rewards and / or adapts the context information to the context that triggers the behavior.

2 shows a flow chart of an exemplary embodiment of the method. This includes decision-making process steps whose decision to be checked can be answered with yes, y or no, N. After a start S. of the procedure are in a first procedural step VS1 Sensor data of the internal and / or external sensors 2 , for example the interior cameras and / or interior microphones and / or further internal and / or external sensors 2 of the vehicle.

In a second process step VS2 the occupant is identified for a model to be trained individually for him.

In a third process step VS3 , a decision-making process step, it is checked whether the occupant has selected an undesired behavior to be prevented and / or whether an undesired, in particular safety-relevant, behavior to be prevented is to be recognized, and whether the occupant is using the method and thus also the system, in particular the technical device 3 to carry out the procedure. If not, the process is stopped SP .

If yes, then in a fourth procedural step VS4 , a decision-making process step, checks whether the model has already been trained with sufficient accuracy for the undesired behavior.

If so, then Y is in a fifth procedural step VS5 , a decision-making process step, checks whether the undesired behavior is predicted.

If not, the process is stopped SP . If so Y, will be in a sixth step VS6 Information about the predicted undesired behavior, alternative options for action, ie behavior alternatives, and behavior-triggering context information of the behavior-triggering context extracted from the model are output to the occupant. Then the process is stopped SP .

If in the fourth step VS4 it is determined that the model has not yet been trained with sufficient accuracy for the undesired behavior, is carried out in a seventh method step VS7 , a decision-making process step, checks whether the undesired behavior was recognized at time t0. If not, the process is stopped SP .

If yes, then in the sixth process step already described VS6 the information about the now recognized undesired behavior, alternative options for action, ie behavior alternatives, and, if these are already available, behavior-triggering context information of the behavior-triggering context extracted from the model is output to the occupant. Furthermore, in an eighth process step VS8 the model continues to train on the basis of the sensor data for the period from time t0 to time t0. When both are done, the process is stopped SP .

Even if in the fourth process step VS4 it is established that the model has already been trained with sufficient accuracy for the undesired behavior (in 2 not shown) additional training of the model in the eighth process step VS8 be provided, in particular if it is determined that the undesired behavior in the fifth method step VS5 was not predicted, but was then recognized at time t0, because this indicates that the assumption in the fourth method step VS4 that the model has already been trained with sufficient accuracy for the undesired behavior was incorrect, so that further training of the model is required.

It could thus be provided, for example, that if in the fifth method step VS5 the undesirable behavior is not predicted, the procedure is not stopped, as in 2 shown, but with the seventh process step VS7 is continued in order to determine whether the undesired behavior occurs at time t0 anyway, in order to then use the model in the eighth method step VS8 continue to train and the occupant in the sixth step VS6 to output at least the information about the now recognized undesired behavior, alternative options for action, ie behavior alternatives, and, if these are already available, behavior-triggering context information of the behavior-triggering context extracted from the model.

List of reference symbols

1

Occupant input

2

sensor

3

Facility

4th

Recognition section

5

Information section

6th

Model training section

7th

Model usage section

8th

Behavior therapy

9

professional

J

Yes

N

no

S.

begin

SP

Procedure stopped

VS1 to VS8

Process step

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 112019001601 T5 [0002]
• US 2012/0130196 A1 [0034]

Claims (6)

Method for influencing the behavior of at least one occupant in a vehicle, characterized in that - a behavior-triggering context for a behavior is recognized, - on the basis of the existing behavior-triggering context that is recognized, it is predicted that, without influencing the occupant, undesired behavior of the occupant will occur, and an intervention for influencing the behavior of the occupant is carried out in order to suppress or change the undesired behavior of the occupant. Procedure according to Claim 1 , characterized in that recognized behavior-triggering properties of the behavior-triggering context are used in order to change the behavior of the occupant and / or to support a behavior therapy (8). Method according to one of the preceding claims, characterized in that at least one component of the vehicle is used for the intervention to influence the occupant in order to enable the occupant to have at least one behavioral alternative. Method according to one of the preceding claims, characterized in that methods of computer-assisted and / or machine learning are used to identify the undesired behavior and / or to identify the context that triggers the behavior. Method according to one of the preceding claims, characterized in that undesired behavior of the occupant is classified. Method according to one of the preceding claims, characterized in that the occupant is informed, in order to influence his behavior, that it has been predicted that undesired behavior of the occupant will occur.

Images