DE102018133674A1 - Method and device for generating control signals for assisting occupants of a vehicle - Google Patents

Abstract

The invention relates to a method for generating control signals for assisting occupants of a vehicle, in which a context of the vehicle is ascertained, depending on the ascertained context, a rule of a rule-based data system (8) is selected, the rule-based data system (8) comprises a plurality of rules, each rule comprising a condition part and a result part, the condition part comprising conditions for the context of the vehicle, a control result of the selected rule is generated, a control signal is generated and output as a function of the generated rule result; on the basis of the condition part of the selected rule and the determined context, a user output is generated which makes the application of the selected rule plausible for the user (17), and the generated user output is output to the user (17). The invention further relates to a device for carrying out this method.

Description

The present invention relates to a method and an apparatus for generating control signals for assisting occupants of a vehicle.

Various assistance systems are known to support the occupants of a vehicle, in particular the driver of the vehicle. Depending on the context of the vehicle, information is made available to the driver, for example, which makes it easier for him to master the driving task. Such assistance systems include, for example, navigation systems, web guidance systems such as an automatic distance control or a lane change assistance system. Systems are also known which increasingly take on the driving task partially or completely autonomously.

Such systems are e.g. B. designed adaptively. You can learn from previous situations in particular. At the moment, adaptive systems are used in particular, which either consist rigidly of fixed expert rules or are based on techniques of artificial intelligence or machine learning. In the first case, fixed behavior patterns are defined in the design of the system and described with rules. In this case, the system is not actually adaptive. However, it can be adjusted by changing settings. Such systems are used, for example, in the area of infotainment and in the driver assistance area in the automotive environment. In the second case, data models are trained on largely historical data. The adaptation of the system is derived on the basis of these models.

From the US 2015/0294223 A1 For example, a system is known that uses prediction algorithms to create context information for an event. The event recording is transferred to a cloud storage system. In this system, conclusions are drawn from saved event logs. These conclusions are then made available to other mobile devices.

From the WO 2017/067853 A1 an assistance system for a vehicle user is known, in which recommendations for the use of the vehicle are given to the user. A self-learning system allows patterns to be recognized and recommendations to be issued based on these patterns.

From the DE 10 2007 025 352 A1 a driver information system for a motor vehicle is known in which a personalized adaptive cockpit is implemented. Depending on a user profile, the information displayed can be universally adapted. Contexts are determined from a large number of information recorded during operation of the vehicle, which take into account properties or actions of the driver or a user of the vehicle. Depending on these contexts, the cockpit is adapted.

From the US 2017/0190337 A1 a communication system for a vehicle is known which controls the output of messages to a vehicle occupant. The vehicle occupant is monitored by sensors. The output is then controlled as a function of parameters which have been obtained from this monitoring of the vehicle occupant.

A disadvantage of known solutions for adaptive learning systems is that the preferences of the user are obtained at most from recorded data about a situation or a context and about the behavior of the user. Knowledge is automatically obtained from this, by means of which the system is adapted.

Such an automatic adjustment is often incomprehensible to the user. The acceptance of the system by the user is thereby reduced.

The present invention is therefore based on the object of providing a method and a device of the type mentioned at the outset in which the control signals generated can be transparently reproduced for the vehicle occupants.

According to the invention, this object is achieved by a method having the features of claim 1 and an apparatus having the features of claim 14. Advantageous refinements and developments result from the dependent claims.

In the method according to the invention, a context of the vehicle is determined. Depending on the determined context, a rule of a rule-based data system is selected. The rule-based data system comprises a large number of rules, each rule having a condition part and a result part, the condition part comprising conditions for the context of the vehicle. A control result of the selected rule is then generated and then a control signal is generated and output as a function of the generated control result. Based on the condition part of the selected rule and the determined context, a user output is generated which shows the application of the selected one Usually made plausible for the user. The user output generated is finally output to the user.

The user issue for plausibility checking increases the acceptance of the process through a high degree of transparency. The user can also more easily understand the change in the vehicle that was generated by the generated control signal

According to the invention, the context of the vehicle means in particular the situation in which the vehicle is located. The situation of the vehicle can be derived from the surroundings of the vehicle. As an alternative or in addition, the situation of the vehicle can be derived from the states of devices of the vehicle. Finally, the context of the vehicle can be affected by data that is remote from the vehicle, i. H. were not obtained from the surroundings of the vehicle and were transferred to the vehicle.

The condition part describes in particular the situation in which the rule is valid. The result part corresponds to the preference of the user assumed by the rule in this situation or in this context. Such rules have the advantage that they are machine-interpretable and thus precisely define the behavior of the method or the system. However, they can also be converted into a natural language representation. This natural language representation is used according to the invention to explain the logic of a rule to the user and thus to make the behavior of the method and the system plausible.

According to an embodiment of the method according to the invention, the user output contains a verbal and / or textual output of the condition part and / or result part of the selected rule, the result of which was generated. For example, this issue can explain to the user why a seat heater was activated automatically. The underlying conditions and the result generated can be explained. This increases the acceptance of the process through a high degree of transparency. The user can also more easily understand the change in the vehicle that was generated by the generated control signal.

According to a further embodiment of the method according to the invention, a confidence value assigned to the selected rule is determined. This confidence value indicates the probability with which the rule result corresponds to the rule of the user's preference. The user output preferably contains a linguistic and / or textual output of the confidence value of the selected rule. For example, it can be output that an automatically executed rule corresponds to the user's preference with a probability of 80%. In conjunction with the information on which basis a rule was automatically executed, the output of the confidence value also leads to greater transparency and traceability of the procedure.

According to a further embodiment of the method according to the invention, the user output is generated as a function of the confidence value. For example, the user output, which makes the application of a rule plausible to the user, can only be generated if the confidence value is below a threshold value. This ensures that the user receives no user expenditure if the application corresponds to the rule with a high probability of his preference.

For example, the confidence value of a rule can be a relative value from 0 to 1. This confidence value describes the extent to which the method or the system is certain that the rule result corresponds to the actual preference of a user. In this way, the reliability of the rule is assessed.

According to a further embodiment of the method according to the invention, a rule, in particular a selected rule, of the rule-based data system is assigned a status value for a status. The status value indicates how the confidence value of this rule is to be interpreted for this rule. The status comprises in particular at least a first, a second and a third status value. The assigned rule is applied automatically for the first status value, the assigned rule is not applied automatically for the second status value, and the assigned rule is applied for the third status value depending on the confidence value of the assigned rule.

In this case, the user output can also contain a verbal and / or text output of the status value of the selected rule. This also increases the transparency and traceability of the process.

According to a further development of the method according to the invention, the control signal automates a vehicle function with a degree of automation, the degree of automation depending on the confidence value of the selected rule. A vehicle function is thus advantageously automated, the degree of automation being selected by means of the confidence value as a function of the probability of a user's preference.

According to a further embodiment of the method according to the invention, the user is identified and a degree of user experience is determined. The degree of automation then depends on the determined degree of user experience. In this way, the degree of automation can advantageously be adapted to the needs of the user.

According to a further embodiment of the method according to the invention, a user is identified and the rule-based data system is a user-specific rule-based data system which is stored outside the vehicle and is transmitted to the vehicle via an interface for the identified user. The transmitted rule-based data system is transformed as a function of the determined context, so that the transformed rule-based data system can be used in the determined context. The degree of automation can also be adjusted when transforming the rule-based data system. The rule of the transformed rule-based data system is then selected depending on the context determined. This advantageously supports a driver who uses different vehicles, since the user-specific rule-based data system of this user is loaded into the currently used vehicle and is adapted to the current context of this vehicle by means of a transformation.

According to a development of the method according to the invention, a hypothesis for a rule result of a rule of the data system is generated depending on the determined context and on the rule-based data system. The hypothesis generated is then output using a user interface. A user input for the output of the hypothesis is then recorded. The hypothesis output is evaluated by means of the recorded user input. The rule-based data system is then automatically adapted depending on the assessment of the hypothesis. Depending on the determined context and the adapted rule-based data system, the rule result of the rule of the data system is generated automatically. The control signal is then generated and output.

• Es wird ein regelbasiertes Datensystem verwendet, um Verhaltensmuster zu erlernen und situationsspezifische Hypothesen für das Systemverhalten abzuleiten. Des Weiteren wird eine kooperative Kopplung zwischen dem adaptiven System, d. h. insbesondere dem adaptierten regelbasierten Datensystem, und der Nutzerschnittstelle für eine nutzerzentrierte Bewertung der erzeugten Hypothesen verwendet. Schließlich wird ein Aufbau für das adaptive System gewählt, das komplexe Situationen über den Kontext erkennen kann und geeignete Hypothesen für ein Regelergebnis ableiten kann. Mittels der Lernfähigkeit, die über die Adaption des regelbasierten Datensystems bereitgestellt wird, kann das Verfahren an einen Nutzer adaptiert werden.

The method according to the invention, in particular according to this development, uses the following concepts to generate control signals which provide improved support for the vehicle occupants:

• A rule-based data system is used to learn behavior patterns and to derive situation-specific hypotheses for system behavior. Furthermore, a cooperative coupling between the adaptive system, ie in particular the adapted rule-based data system, and the user interface is used for a user-centered evaluation of the hypotheses generated. Finally, a structure is chosen for the adaptive system that can recognize complex situations via the context and can derive suitable hypotheses for a rule result. The method can be adapted to a user by means of the learning ability, which is provided via the adaptation of the rule-based data system.

In the method according to the invention, use is made of the user interface in order to validate hypotheses for a rule result by the user. If, in the method, a recurring hypothesis is increasingly confirmed by the recorded user input, the user interface can be used again in order to have the user generally evaluate a hypothesis. Thus, hypothetical assumptions about possible preferences of a user can become confirmed facts, which increase user acceptance with regard to the adaptive method, because the method is better adapted to the user. A cooperation of data-based learning or data-based derivation of hypothetical knowledge and a user interface is thus advantageously used in order to obtain an explicit user feedback. This avoids the disadvantages of methods that are only based on machine learning or algorithms for artificial intelligence.

In the method according to the invention, the knowledge management via the adapted rule-based data system and the knowledge application via the user interface are conceptually separated. Furthermore, the use of a rule-based data system advantageously enables the method according to the invention to be certified, since the rules of the rule-based data system are comprehensible. Such certifiability is not available, for example, in systems with neural networks, since it is not easy to predict how such a system will react to certain conditions.

By ascertaining the context of the vehicle, a situation-appropriate interaction can advantageously be achieved. By adapting the rule-based data system, the process results in a learning process of how and when the process should interact or communicate with the user in different contexts. For example, the method can adapt the rule-based data system in such a way that it only offers urgent and important information to a user in a very abbreviated form if the user is currently driving on the freeway or if it has been detected that he is in a driving situation that focuses on the driving situation required.

• Durch das Verfahren können individuelle situative Nutzerpräferenzen erlernt werden. Durch das Verfahren können verschiedenste Präferenzen eines individuellen Nutzers für unterschiedliche Situationen erlernt werden. Hierfür wird ein Hypothese erzeugt, ausgegeben und die Nutzereingabe hierzu erfasst. Aus dieser Nutzereingabe ergibt sich dann eine Adaption des regelbasierten Datensystems.

The following advantages also result from the method according to the invention:

• The process enables individual situational user preferences to be learned. The method enables the most varied preferences of an individual user to be learned for different situations. For this purpose, a hypothesis is generated, output and the user input for this is recorded. This user input then results in an adaptation of the rule-based data system.

In the method, the rule that is valid for this context can be identified on the basis of an underlying rule set each time the current context changes. Possible adaptations of the rule-based data system can be derived from this, which hypothetically correspond to the preferences of the user. The user interface can also be adapted in this way.

By means of the method, dedicated feedback of a user can be interpreted, processed and the rule-based data system can be adapted accordingly.

According to an embodiment of the method according to the invention, measured values from sensors and / or data from interfaces of the vehicle are recorded. In this case, the context of the vehicle is determined as a function of the recorded measured values and / or the recorded data. In particular, measured values and / or data on the current traffic conditions, the behavior of the driver, interaction of the driver with a further user interface, the current weather conditions, the current ambient light conditions, the current season and / or the current time of day are recorded by means of the sensors and / or interfaces . This list is only an example; all measurement values and data relevant for the vehicle and for vehicle guidance can be recorded and processed in a suitable manner.

According to an embodiment of the method according to the invention, it is also possible to integrate so-called fleet knowledge into the context of the vehicle via an interface of the vehicle. Fleet knowledge is understood to mean data that are available about a vehicle fleet, i. H. a large number of vehicles were collected and allow conclusions to be drawn about the context of a single vehicle. By means of the method, the rule-based data system can be adapted on the basis of data generated by a large number of vehicles. In this way it is also possible to add new rules to the rule-based data system. This advantageously makes it possible to monitor the parameterization of certain rules of the rule-based data system in a specific geographic region or under certain contextual influences, such as weather, climatic conditions, ambient light or time of day, across a vehicle fleet in order to propose a new rule or an initial parameterization of a rule adapt.

In the method according to the invention, an exact value of the condition part and the result part can be learned by adapting the rule-based data system, in particular for a suitable rule. The rule can be parameterized in this way in order to best represent the user preference. It is also possible to use data-driven methods to learn new rules. The hypothesis can thus include an assumption for a preference of the user for the rule result of the hypothesis in the context determined.

The confidence value can also change during the execution of the method. It is therefore part of the adaptation of the rule-based data system. The change can take place as a result of active interactions with a user. In particular, the confidence value can change after a certain user input.

The rules of the data model are assumptions about possible preferences of the user. The method thus derives possible hypothetical preferences for a context based on the rules and the hypotheses, which correspond to the actual preference of the user with a certain confidence, ie with a certain probability. Since the confidence value in the method is initially calculated on the basis of the previously recorded data about the context, that is, for example about situational user interactions, it is always hypothetical knowledge that is available about the user preferences. If, in the method according to the invention, a hypothesis has been derived for a context that most likely corresponds to the user preference, the method interacts with the user through the user interface in order to make confirmed facts from hypothetical assumptions. A user interface is used to give the user, for example, a recommendation that he can confirm. Furthermore, the user can be offered, for example, that a certain rule will always be applied automatically in the future. The user can agree to this automation by user input. By adapting the rule-based data system, there is certainty about the correctness of this rule and the user can be supported from now on with the execution of this rule. Thus, the user interface is used to request explicit feedback from the user for hypothetical assumptions and to adapt the rule-based data system ensure that it is increasingly adapted to the user. In the event that the user contradicts a rule by user input, the rule-based data system can be adapted in such a way that the confidence of the rule is reduced so that the rule can now be used less frequently. This advantageously increases user acceptance, since the rule-based data system can find out specific certainty about possible preferences of the user.

The user input can thus include confirmation or rejection of the hypothesis. If the hypothesis is rejected, in particular the confidence value of the rule belonging to the hypothesis is reduced. If the hypothesis is confirmed, the confidence value of the rule associated with the hypothesis is increased. In addition, the user input or multiple user inputs can induce activation or deactivation of the rule associated with the hypothesis.

If the user is identified in accordance with a development of the method according to the invention, the rule-based data system can also be adapted in a user-specific manner. In this way, it is advantageously possible to adapt the method quickly to a first-time user without data about interactions of this user having previously been recorded. The user-specific, rule-based data system can be loaded via an interface of the vehicle as soon as a user has been identified in the vehicle. The user-specific, rule-based data system can therefore also be used for a specific user in rental vehicles, second vehicles or follow-up vehicles. Such a rule-based data system can also be applied to more restricted or more potent vehicle platforms. In contrast to neural networks, rule sets can easily be restricted to smaller domains or embedded in larger domains. Even if the function of a subsequent vehicle platform differs, function-specific transformation models can be applied to the parameterization of the rule set, which adapt the rule-based data model to the new vehicle platform. The rule-based data system used by the method is thus advantageously particularly easy to transfer.

In this case, the process is designed to be very user-centered. This ensures the economic development and sustainability of the method according to the invention. It can offer the user added value over the entire period of use for years. By means of the method, the preferences of the user can be learned correctly in order to offer the user the control signals for a desired function or information in the desired manner of interaction at the right moment and in the right situation. This advantageously increases the acceptance of the method according to the invention.

According to a further embodiment of the method according to the invention, hypothesis-independent user inputs are also recorded which are independent of the output of a hypothesis. The hypothesis-independent user inputs are evaluated and the rule-based data system is automatically adapted depending on the result of the evaluation. In this way, the rule-based data system of the method according to the invention can be individualized. The user can additionally define the behavior of the method. This is particularly possible because the user can understand a rule. The user can therefore change rules in a targeted manner or add new rules that extend the method using the user interface.

The device according to the invention for generating control signals to support occupants of a vehicle comprises a context determination unit for determining the context of the vehicle. Furthermore, it comprises a result unit, which is designed to select a rule of a rule-based data system depending on the determined context, the rule-based data system comprising a multiplicity of rules, each rule having a condition part and a result part, the condition part conditions for the context of the vehicle. Furthermore, the result unit is designed to generate a rule result of the selected rule. Furthermore, the device comprises a plausibility check, which is designed to generate a user output on the basis of the condition part of the selected rule and the determined context, which checks the automatic execution of the rule for the user. Finally, the device comprises an output unit for generating and outputting a control signal as a function of the generated control result and for outputting the generated user output.

The device according to the invention is particularly designed to carry out the method according to the invention. It therefore has the same advantages as the method according to the invention.

The invention further relates to a computer program product comprising instructions which, when executed by at least one computer, cause the computer to execute the method according to the invention.

The method and the device according to the invention are particularly easy to handle for a developer. The rule-based Data system of the method or the device can be handled in particular in such a way that the developer can explicitly define how the method or the device should behave in the event of defined events. For example, the developer can define in advance that in the event of an event in which the user rejects a system adaptation, the system will no longer carry out the system adaptation for a defined period of time.

The method and the device can be certified by using a rule-based data system. This certifiability is possible even before the method is carried out or before the device is used. In this way it can be ensured that the device and the method do not show, learn or accept any behavior that is undesirable. This is particularly important for automated driving functions. This is guaranteed by using a rule-based data system, since the adaptation is explicitly defined in the set of rules. The adaptation of the rules is thus known before the procedure is carried out and can be certified and approved accordingly. The method and the device will not make any adaptation that have not been previously defined.

• 1 zeigt den grundsätzlichen Aufbau eines Ausführungsbeispiels der erfindungsgemäßen Vorrichtung und
• 2 zeigt den Ablauf eines Ausführungsbeispiels des erfindungsgemäßen Verfahrens.

In addition, the rule-based data system can be defined initially. It is not yet necessary for data about a user's interaction to be recorded. This means that even first-time users of the device and the method can benefit directly from the control signals generated. In addition, the status and the confidence of a rule can be defined initially, so that the method and the device can already generate control signals during the initial interaction with the user, without the user having to have used the device or the method for a long time beforehand and without data being collected. For example, expert knowledge can be integrated for the initial implementation of the method or the device. For example, rules defined by experts that describe the sensible use of vehicle systems can initially be initialized as rules with high confidence. In this way, even for an inexperienced user, helpful control signals can also be generated for functions that the user has not used before. For example, the activation of an assistance system can be recommended to a user if the use of this system is recommended in the current context. Such recommendations can also be derived from fleet data.

• 1 shows the basic structure of an embodiment of the device according to the invention and
• 2nd shows the sequence of an embodiment of the method according to the invention.

• Die erfindungsgemäße Vorrichtung wird allgemein von einem adaptiven System 20 und den Kopplungen dieses adaptiven Systems 20 mit weiteren Einrichtungen in einem Fahrzeug und außerhalb des Fahrzeugs bereitgestellt.

First, with reference to 1 the structure of the device according to the invention is described:

• The device according to the invention is generally based on an adaptive system 20 and the couplings of this adaptive system 20 provided with further facilities in a vehicle and outside the vehicle.

The vehicle has a large number of sensors 1 and interfaces 2nd with which measured values and data can be recorded. The sensors 1 can include, for example, measurements of the current weather conditions, the current ambient light conditions, the current season and the current time of day. You can also use sensors 1 be provided, which monitor the behavior of the driver and capture the driver's interactions with a user interface. Furthermore, the sensors 1 Provide measured values for driving parameters, such as the current speed, a possible slip of the wheels and further measured values, which are usually evaluated by driver assistance systems known per se. Via the interfaces 2nd data on the current traffic conditions around the vehicle or on the route of a vehicle as well as regional weather conditions can be received. The interfaces 2nd can also enable the transmission of data via the mobile network or the Internet in order to couple the vehicle with external services that are operated in a cloud or on a central server of the vehicle manufacturer. In this way, any data that are relevant to the context of the vehicle can be transmitted to the vehicle.

In addition, the interfaces 2nd pair with mobile devices inside the vehicle. Furthermore, the interfaces 2nd communication with other vehicles or infrastructure, such as traffic management systems and traffic lights, is possible.

Using the sensors 1 or the interfaces 2nd can also be a specific user 17th who is the occupant of the vehicle. For example, one of the sensors 1 designed as a user identification unit. Alternatively, the user is identified via an interface, for example by making an entry. In this way, all adaptations, which are described below, can be carried out user-specifically.

The sensors 1 are with an extraction unit 3rd coupled, by means of which the raw data of the measured values of the sensors 1 can be converted into data for higher-value characteristics. For example, using a sensor 1 the current speed is continuously recorded. The extraction unit 3rd can in this case continuously analyze this rapidly changing data and derive the information that the user has been driving at a constant speed since a certain point in time. The extraction unit 3rd enables the adaptive system 20 in addition to capturing raw data, capturing and recognizing complex situations in highly dynamic, heterogeneous data about the driver, the vehicle and the environment. The extraction unit 3rd can use artificial intelligence technologies (e.g. deep learning, time series analysis).

That by means of the extraction unit 3rd obtained data is sent to a context data determination unit 4th transfer. Likewise, those on the interfaces 2nd received data to this context data determination unit 4th transfer. In the context data determination unit 4th are all available from the sensors 1 and the interfaces 2nd transmitted data analyzed and aggregated if necessary. The context data acquisition unit 4th generates a continuously adapted description of the context in which the vehicle is located. In particular, the current situation in which the vehicle is located is determined. The determined context of the vehicle is sent to an adaptive module 5 of the adaptive system 20 transferred and on the other hand via a suitable interface to an external pattern learning module 16 .

The adaptive module 5 again comprises a hypothesis generation unit 6 and an evaluation unit 7 . The hypothesis generating unit 6 is formed depending on that of the context data determination unit 4th determined context and depending on a rule-based data system 8th with which the adaptive module 5 coupled, a hypothesis for a rule result of a rule of the data system 8th to create.

The rule-based data system 8th is a memory in which a large number of rules are stored in a structured manner. Each rule has a condition part and a result part, the condition part comprising conditions for the context of the vehicle. The rule-based data system 8th thus contains the set of rules that define how the adaptive system works 20 behaves in different situations. It also includes information for each rule regarding its confidence and status, as will be explained later. The rule-based data system 8th can be divided into subsystems that differ in the origin of their rules. The rule-based data system 8th is also with the external pattern learning module 16 coupled. About this external sample learning module 16 new rules can be defined and sent to the rule-based data system 8th be transmitted.

The hypothesis generating unit 6 accesses the rules of the rule-based data system 8th to. It can be the rule set of the data system 6 based on the current situation or context. The hypothesis generation unit iterates in each situation 6 about the set of rules of the rules-based data system 8th and identifies the appropriate rules for the context. The rule results of this set of identified rules represent a set of hypothesis about the preferences of the user for the current context or the current situation. Each of these identified and appropriate rules as well as their rule results are sent to a user interface controller in the form of a list of hypotheses 10th transfer.

The user interface control 10th comprises a hypothesis processing unit 12th and a result unit 13 . Furthermore, the user interface control 10th with a user interface model 11 coupled.

That from the hypothesis generating unit 6 received hypotheses are processed by the hypothesis processing unit 12th processed. The hypothesis processing unit 12th determines how the adaptive system works 20 should behave depending on a hypothesis. It decides on the basis of the confidence and status of each rule, on which the hypothesis is based, whether the rule represents a possible preference of an individual user. The hypothesis processing unit also decides 12th on this basis, whether a rule on which the hypothesis is based is automatically carried out, whether the application of the underlying rule is offered or recommended to the user and how the user is to be interacted with. The user interface controller decides 10th also about which user interface should be used to interact with the user. As user interfaces 18th can, for example, a graphical user interface 19th and a voice interface 21st be provided.

The hypothesis processing unit 12th can decide, for example, that the user has the automatic activation of the seat heating via the voice interface 21st is informed by the output of synthesized speech. The application of the rule on which the hypothesis is based can be made plausible to the user by specifying the reason why the seat heating is activated automatically. For example, per Language are output that the user has recently used the seat heating frequently if the outside temperature corresponded to the current temperature. The concrete decision logic for this and the data about how a rule is translated into natural language is in the user interface model 11 defined and modeled. An alternative to the voice-based information output for the user is, for example, the display of the information using a graphical user interface 19th that the seat heating has been activated. A button can also be displayed that can be used to undo the execution. If the user is recommended to execute a specific rule on which the hypothesis is based, the hypothesis processing unit can 12th the voice interface 21st use to actively offer the user the recommendation and to conduct a dialogue with the user about it. For example, if the user asks why the adaptive system 20 has concluded that he would like to use the seat heating, the hypothesis processing unit can 12th this using the voice interface 21st to explain. The user should make two successive recommendations of the adaptive system 20 may have ignored the hypothesis processing unit 12th For example, decide that no further rule will be executed for this user for a defined time.

The hypothesis processing unit 12th is thus set up to determine when and in which cycles from implicit feedback actions should be learned whether a certain rule for a current user 17th fits which was previously identified. For example, it can be defined for the function of the seat heating of the vehicle that only the first two minutes of each journey should be learned on the basis of implicit knowledge whether a rule for the seat heating function matches a user interaction pattern. If within the first two minutes a certain rule in a certain situation, which results from the context, is valid and the user 17th if the seat heating itself is activated, the rule is interpreted as appropriate and an adaptation of the rule is initiated. If for a particular rule in the hypothesis processing unit 12th cyclic learning is defined, the adaptation of the rule is initiated in a previously defined cycle as long as the conditions for this are met.

The user interface model 11 will be determined in advance. This defines how a specific rule or a specific hypothesis of a specific function or domain with a specific confidence and given status must be used. It is thus determined how the user interfaces 18th be used to communicate system behavior. It also defines how a rule or hypothesis can be meaningfully translated into natural language. This can be done in the user interface model 11 Various text modules can be saved or a natural language generator is used. Examples of text modules are the name of a sensor 1 , the condition value of the rule for a sensor 1 , the current actual value of the sensor 1 , the name of an actuator and the value to be set for the actuator. Furthermore, in the user interface model 11 be defined that the adaptive system 20 for example for a defined period of time with the execution of rules for a particular domain if the user 17th has repeatedly ignored the execution of rules of this domain. The domains can delimit different areas for functions of the vehicle from one another.

The user interfaces 18th generally include a set of available user interfaces to interact with the user. The user interfaces 18th are used by the user interface control 10th controlled. This way the user 17th a rule execution communicates, is offered or feedback from the user is requested or the user is informed about automation. Thereby the user 17th Appropriate user input is offered to influence the assessment of a hypothesis or the execution of a rule.

The user interfaces 18th can, in particular, controlled by the user interface controller 10th by the hypothesis generating unit 6 output generated hypothesis. A user input by means of an input unit can respond to such an output of the hypothesis 22 are recorded. The input unit 22 is also part of the user interfaces 18th .

The user input or generally the user feedback in response to the output of a hypothesis can take place in various ways. For example, the user can ignore a recommendation contained in the hypothesis that is output. He can also explicitly reject or question such a recommendation. In particular, a recommended automation can be questioned. Furthermore, the user can accept a recommendation contained in the hypothesis or ignore automation. The user can also undo automation. Finally, the user can provide dedicated feedback on a recommendation or automation, for example by adapting the condition part or the result part of a specific rule on which the hypothesis is based. For example, the user can enter that the seat heating should be activated at 13 ° C instead of 12 ° C and instead of the step 2nd , the stage 3rd should be activated.

The explicit feedback from the user 17th via the user input is sent to the hypothesis processing unit 12th the user interface control 10th transmitted which in turn it back to the adaptive module 5 transmits. In the adaptive module 5 this user feedback is the user input in the evaluation unit 7 processed. The valuation unit 7 is trained to carry out a further evaluation of the previously given hypothesis. The valuation unit 7 is supported by the data from the hypothesis processing unit 12th were transferred. The valuation unit 7 is especially designed to evaluate the previously issued hypothesis by means of the recorded user input.

The valuation unit 7 is trained to learn which rules or hypotheses with which parameterization for an individual user 17th fit and thus correspond to his preferences. The valuation unit 7 thus ensures the adaptivity of the adaptive system 20 .

The evaluation of a hypothesis is done by the evaluation unit 7 to the adaptation unit 9 transfer. Depending on the evaluation of an issued hypothesis, the adaptation unit 9 automatically the rules of the rule-based data system 8th adapt. By means of the adaptation unit 9 can thus be a hypothesis that was previously created by the hypothesis generating unit 6 generated and by the valuation unit 7 has been validated, used to make the rules-based data system 8th to adapt.

Automated learning and adaptation of the system 20 can take place on the basis of two information classes. The valuation unit 7 can evaluate a user's interactions and receive implicit feedback. Furthermore, the evaluation unit 7 receive explicit user feedback on hypotheses as described above.

The evaluation unit receives implicit feedback 7 about data when interacting with the user 17th are recorded. The valuation unit 7 receives from the context data determination unit 4th certain contextual data that provide a description of the situation the vehicle is in. It can now be recorded in which situation the user uses a function himself and how he sets it. This information is referred to as implicit feedback. The valuation unit 7 records, for example, when user-initiated seat heating is activated and checks which of the rules of the rule-based data system 8th conform to this pattern. If one of the rules matches this pattern in the implied feedback, the assessment unit can 7 conclude that this rule is more likely to be the user's preference 17th corresponds. Via the adaptation unit 9 can the valuation unit 7 in this case increase the confidence value of the corresponding rule. This would be the case, for example, after the seat heating had been activated several times by the user 17th in the always the same situation, the confidence value rose so that by means of the hypothesis generating unit 6 and the hypothesis processing unit 12th it is decided to generate an output that recommends the user to execute the corresponding rule or even to automate it for him.

An explicit user feedback is understood to mean that a hypothesis is issued by the user 17th makes a user input that indicates how the hypothesis is to be evaluated. The adaptive system 20 can learn in this way to what extent the rule belonging to this hypothesis corresponds to the user preference and then by means of the adaptation unit 9 the rule-based data system 8th adjust accordingly.

• Für die Adaption der Regel umfasst die Adaptionseinheit 9 eine Instanz, die definiert, wie basierend auf implizitem und explizitem Feedback des Nutzers 17 erlernt wird, ob eine Regel den Präferenzen dieses Nutzers 17 entspricht. Die Adaption wird angestoßen, wenn entweder das Muster der jeweiligen Regel in dem impliziten Feedback des Nutzers 17 festgestellt wurde und die Bewertungseinheit 7 den Adaptionsvorgang anstößt oder wenn der Nutzer 17 ein explizites Feedback durch eine Nutzereingabe über die Nutzerschnittstellen 18 zu einer bestimmten Regel gibt.

Details on how an adaptation of a rule is initiated and how a rule is then adapted are explained below:

• The adaptation unit includes for the adaptation of the rule 9 an instance that defines how based on implicit and explicit feedback from the user 17th is learned whether a rule is based on the preferences of this user 17th corresponds. The adaptation is initiated when either the pattern of the respective rule in the implicit feedback of the user 17th was determined and the evaluation unit 7 initiates the adaptation process or if the user 17th an explicit feedback through user input via the user interfaces 18th to a certain rule there.

The rules of the rule-based data system 8th include a condition part and a result part. The condition part includes conditions for the context of the vehicle in which the rule is executed. The likelihood that the rule will execute the preferences of the user 17th is based on a confidence value of the rule. This confidence value is determined by means of a confidence value determination unit 23 determined that with the rule-based data system 8th connected is. When a rule is adapted, this confidence value can be changed. The adaptation unit 9 differentiates between the different forms of user feedback. With an adaptation based on implicit knowledge, the confidence value is only slightly increased or decreased. In contrast, with explicit feedback, in which a hypothesis has been issued and a user input has been recorded, there is a much greater change in the confidence value, since a explicit user rating 17th to the hypothesis. The status of a rule can also be changed. The rule can be activated, deactivated or the status can be set in such a way that the rule is applied depending on the confidence value.

A rule can be set using the adaptation unit 9 For example, be adapted so that when a rule is repeatedly reset or undone, a rule is deactivated and the confidence value is set to zero.

If in the hypothesis the user 17th a rule execution was offered and he accepted it, the rule can also be assigned a high confidence value. In this case, for example, the hypothesis processing unit 12th decide that the user is to be offered automation of this rule immediately after the execution of an assigned rule, or that the status of the rule should be set to “activated”. For example, the voice output can be: "Would you like me to always activate the seat heating for you when it is colder than 12 ° C in the morning?"

Furthermore, it is possible that the user 17th a rule belonging to the hypothesis output is deactivated by the user input. From the valuation unit 7 is in this case by means of the adaptation unit 9 the status of the rule is set to "deactivated". If a deactivated rule has an increasing confidence value over time, the hypothesis processing unit can 12th , despite the rule status "deactivated", by means of the hypothesis generation unit 6 generate a hypothesis again and generate an output that the user 17th offers to remove the deactivation of this rule. The output via the voice interface 21st can be, for example, the following in this case: "I know that the seat heating should not be automated for you. However, since you have used the seat heating several times in this situation, I could resolve the deactivation of the rule for you and activate the rule. "

If through one of the interfaces 2nd the user has been identified, the rule-based data system 8th can also be customized. The explicit user feedback determines whether certain rules, for which hypotheses have been issued, relate to the individual preferences of the identified user 17th fit. The rules can then be adapted to specific users. If, for example, a user input has been recorded that a specific rule corresponds to the preference of a specific pattern, it can be derived from the situation previously recorded via the context which seat heating level a particular user has mostly set. Various heuristics and procedures can be used to derive the appropriate actuator value. This learning process for the parameterization of a rule can be applied analogously for the sensor values of the condition part. If no clear pattern for parameterization can be derived, a hypothesis can be generated and sent to the user via one of the user interfaces 18th be issued. With a user input, the user can 17th then give an explicit feedback, which is evaluated, whereupon the corresponding rule can be adapted user-specifically. For example, the hypothesis can be output as follows: “I have found that in situation s you usually activate the seat heating. Would you like me to automate this rule for you? And if so, at which level should I activate the seat heating for you? "

Furthermore, the adaptation unit 9 the rule-based data system 8th change in another way. For example, the adaptation unit 9 make sure the rules-based data system 8th contains no contradicting rules or logic. For example, it can perform a consistency check. Furthermore, the adaptation unit 9 can be used to ensure domain-specific requirements by identifying and then deactivating rules that contradict these requirements. For example, a climate domain might require that no rule activate seat heating when a child seat is mounted on the seat. Such an adaptation of the rule-based data system 8th by means of the adaptation unit 9 This also means that the system that is learning itself remains certifiable.

In addition to the adaptive system 20 A variety of other adaptive systems can be used for your own vehicle 30th other vehicles. The other adaptive system 30th stands for a large number of adaptive systems in a fleet that is used by different users. These other adaptive systems 30th can also send context data to the pattern learning module 16 transfer.

The sample learning module 16 is carried out, for example, on the vehicle manufacturer's server or in a cloud. It contains all context data from all adaptive systems 20 , 30th collected. Since the pattern learning module is carried out externally, ie independently of the vehicle, it can rely on much larger memory and computing resources. In this way, in the pattern learning module 16 further analyzes of the recorded data are carried out. Using the sample learning module 16 the recorded data, in particular the context data, are analyzed further in order to learn new rules. Methods of artificial intelligence and machine learning. Deep learning, for example, is a technology that places high demands on the computing power of a system and is suitable for analyzing the data. Here, the pattern learning module 16 either learn new rules based on the data of a single individual user or based on fleet data. In the case of the fleet data, for example, all situations or contexts in which users drove on a particular road can be examined in order to learn rules for a context for this particular road. Newly learned rules are either in a specific rule-based data system 8th of a user or to all rule-based data systems 8th transferred to the fleet. For the purpose of quality assurance, the sample learning module 16 newly learned rules before transferring them to a rule-based data system 8th of an adaptive system 20 , 30th checked, certified and if necessary released or deleted by experts.

With such an active learning process, certain situations can be learned in a targeted manner, in this way in the external model learning module 16 verify hypothetical knowledge. You can do this in the sample learning module 16 Rules with a low confidence value are generated on adaptive systems 20 , 30th of a vehicle and the development of the confidence value of this rule in the individual adaptive systems 20 , 30th to be watched.

After the rule-based data system 8th through the hypothesis output and the subsequent hypothesis evaluation using the adaptation unit 9 A result unit can be adapted 13 execute a rule when the user interface control 10th determined that no hypothesis should be issued. The result unit 13 generated depending on the determined context and the possibly adapted rule-based data system 8th a rule result of a rule of this data system 8th . This control result is sent to an output unit 14 transmitted, which generates and outputs a control signal depending on the control result generated. This control signal can be sent to the user interfaces 18th transmitted to output certain information. Furthermore, the control signal can be sent to a data bus 15 are transmitted, which enables the control of the various devices of the vehicle. For example, the various assistance systems of the vehicle can be controlled, the media player can be controlled, various climate functions can be carried out, lights can be controlled or control signals can be transmitted to external devices such as a smartphone in the vehicle.

According to a further exemplary embodiment, the result unit 13 , regardless of whether the rule-based data system 8th was adapted or not, depending on the determined context, a rule of the rule-based data system 8th choose. The results unit generates for this selected rule 13 a rule result. As described above, the control result is sent to the output unit 14 transmitted, which generates and outputs a control signal. Furthermore is in the output unit 19th a plausibility check 23 integrated. Using this plausibility check 23 On the basis of the condition part of the selected rule and the determined context, a user output can be generated which makes the automatic execution of a rule plausible for the user. This user output can then be spoken via the voice interface 21st or textually on a display. This user output contains a verbal and / or textual output of the condition part and / or result part of the selected rule. The confidence value that is assigned to the selected rule is also output. Alternatively, the user output can also be generated as a function of the confidence value. With a low confidence value, the user output for plausibility checking the application of the selected rule is more detailed than with a higher confidence value. In addition, the status value of the selected rule described above can be contained in the user output.

In one embodiment, the control signal automates a vehicle function. For example, functions of an assistance system can be activated that relieve the driver of partial areas of the driving task. It can also be used via the data bus 15 a control unit can be addressed which initiates a partially automated or fully automated journey of the vehicle. The control signal thus automates a vehicle function with a certain degree of automation. In one embodiment, the level of automation depends on the confidence value of the selected rule. For example, a high degree of automation, such as fully automatic driving of the vehicle, can only be activated by a rule that has an extremely high confidence value. On the other hand, a driving function with a low degree of automation, such as the automatic activation of light functions of the vehicle, can also be activated by a rule that has a low confidence value. In addition, the degree of automation of the vehicle function triggered by the control signal depends on the determined context of the vehicle. In the case of a highly complex driving task, the control signal becomes a lower degree of automation at a certain confidence value controlled than with an easier driving task such as on a highway.

Has z. B. via a sensor 1 the driver is also identified as a user, for example by means of the context data determination unit 4th the level of user experience is determined. The level of automation then also depends on the level of user experience. A higher level of automation can be selected for an experienced user than for a less experienced user. For certain driving functions, however, a high degree of automation can also be activated for a less experienced user in order to support the less experienced user.

• Bei einem sehr geringen Konfidenzwert der Regel, d. h. wenn der Konfidenzwert beispielsweise in den niedrigsten Bereich fällt, wird aktiv ein Vorschlag zum Ausführen einer Bedienhandlung an den Nutzer zusammen mit einer Nutzerausgabe erzeugt, welche die angewendete Regel für den Nutzer plausibilisiert, wie es vorstehend erläutert wurde.

As explained above, a rule is assigned a status value for a status. An assigned rule is applied to the third status value depending on the confidence value. For the degree of automation, this third status value can be further subdivided into different automation levels depending on the confidence value, at which the control signal controls various other outputs to the user:

• In the case of a very low confidence value of the rule, that is to say if the confidence value falls, for example, in the lowest range, a suggestion for performing an operating action on the user is generated together with a user output, which makes the applied rule plausible for the user, as was explained above .

If there is a low confidence level, i.e. H. If the confidence value falls, for example, in a second lowest range, a suggestion for performing an operating action on the user is actively generated without an additional user output, which makes the applied rule plausible for the user.

If there is an average confidence score, i.e. H. if, for example, the confidence value falls in a third lowest range, the control signal for automatically executing an operating action is generated together with a notification and a user output which makes the applied rule plausible for the user. The notification informs the user about the execution of the control signal, the user output makes the applied rule plausible.

With a high confidence value, i.e. H. if, for example, the confidence value falls in a fourth lowest range, the control signal for automatically executing an operating action together with a notification, but without a user output, which makes the applied rule plausible for the user, is actively generated.

With a very high confidence value, i.e. H. if the confidence value falls, for example, in a fifth lowest range or the highest range for the confidence values, the control signal for automatically executing an operating action without a notification and without a user output for plausibility checking of the applied rule is generated.

According to a further exemplary embodiment, the rule-based data system 8th designed user-specific. This user-specific rule-based data system 8th is stored in a memory external to the vehicle, for example in a cloud. This is in 1 through the dashed line of the adaptive system 20 shown. After a user uses a sensor, for example 1 has been identified, the user-specific rule-based data system 8th transmitted to the vehicle for the identified user. It can then optionally be performed by the adaptive module as explained above 5 , from the user interface control 10th and especially from the result unit 13 be used. In this case, the rule-based data system 8th however, depending on the determined current context, transformed so that the transformed rule-based data system 8th is applicable in the current context determined. The transformed rule-based data system 8th can be stored outside the vehicle, but preferably in an in-vehicle memory. If a user uses different vehicles, for example, the user-specific rule-based data system 8th loaded into a vehicle as soon as this user has been identified in the vehicle. The behavior learned in one vehicle is thus made available to the user in another vehicle. However, the rule-based data system is transformed 8th to adapt to the current context.

Became the rule-based data system 8th In winter, for example, a user trains in Germany with regard to the user's climate preferences and if the user then travels to a warm holiday area and rents a vehicle there, the rule-based data system becomes 8th loaded into this rented vehicle. However, since this rented vehicle is in a different context, namely in a warmer environment, the rules-based data system is transformed according to the invention 8th , which adapts to the new context. For example, the rule-based data system 8th be adapted to a different temperature sensation at a different humidity level.

For example, when transforming the rule-based data system 8th the degree of automation, as explained above, can be adjusted. Furthermore, the confidence value of one Hypothesis or a rule to be adapted during the transformation. Furthermore, a condition part of a rule can be adapted and / or expanded during the transformation. The adjustments include in particular an addition, a subtraction, an extension or compression of a certain value. When expanding the condition part, certain attributes can be added, for example, humidity-specific attributes.

After the transformed rule-based data system 8th a rule can be selected, as explained above, a rule result of the selected rule can be generated and a control signal can be generated and output as a function of the rule result generated. An additional user issue to check the plausibility of the selected rule can also be carried out.

• Im Schritt S1 wird der Kontext des Fahrzeugs ermittelt. Hierfür wird auf die Kontextdaten-Ermittlungseinheit 4 zurückgegriffen, welche wiederum auf die Extraktionseinheit 3, die Schnittstellen 2 und die Sensoren 1 zurückgreift. Im Schritt S1 kann auch der Nutzer identifiziert werden.
• Im Schritt S2 wird in Abhängigkeit von dem ermittelten Kontext und von dem regelbasierten Datensystem 8 eine Hypothese für ein Regelergebnis einer Regel des regelbasierten Datensystems 8 erzeugt. Hierfür wird auf die Hypothesen-Erzeugungseinheit 6 zugegriffen, welche wiederum Regeln aus dem regelbasierten Datensystem 8 lädt.
• Im Schritt S3 wird die Hypothese ausgegeben. Hierfür werden die Nutzerschnittstellen 18 verwendet.
• Im Schritt S4 wird zu der Ausgabe der Hypothese eine Nutzereingabe erfasst. Diese Nutzereingabe erfolgt erneut über die Eingabeeinheit 22 der Nutzerschnittstellen 18.

In reference to 2nd An exemplary embodiment of the method according to the invention is explained, which can be carried out with the device explained above:

• In step S1 the context of the vehicle is determined. For this, the context data determination unit 4th resorted to, which in turn on the extraction unit 3rd , the interfaces 2nd and the sensors 1 falls back. In step S1 the user can also be identified.
• In step S2 becomes dependent on the determined context and on the rule-based data system 8th a hypothesis for a rule result of a rule of the rule-based data system 8th generated. For this purpose, the hypothesis generation unit 6 accessed, which in turn rules from the rule-based data system 8th loads.
• In step S3 the hypothesis is put out. For this, the user interfaces 18th used.
• In step S4 a user input is recorded for the output of the hypothesis. This user input takes place again via the input unit 22 of user interfaces 18th .

The user input is sent to the evaluation unit 7 transfer which in step S5 evaluated the hypothesis.

Depending on the assessment of the hypothesis, the step S6 then the rule-based data system 8th by means of the adaptation unit 9 adapted. If the user interface control 10th by means of the hypothesis processing unit 12th has determined that validation of a hypothesis or user feedback is not required, the step S7 depending on the determined context and the adapted rule-based data system 8th generates a rule result of a rule of the data system. The user interface control 10th accesses the result unit 13 to.

In step S8 depending on the control result generated, a control signal is generated which is generated by the output unit 14 is appropriately spent.

According to a development of the method, the step S6 when adapting the rule-based data system, fleet knowledge is also accessed, which the rule-based data system from the sample learning module 16 is transmitted. In this way, new rules can be added to the rule-based data system 8th can be imported or the rules of the rule-based data system 8th are adapted based on data, which is via another adaptive system 30th won.

According to a development of the method, a user output is generated on the basis of the condition part of the selected rule and the determined context, which makes the application of the selected rule plausible for the user. This is done on the plausibility check 23 accessed. This user input is also by means of the output unit 14 appropriately spent. The user output contains a verbal and / or textual output of the condition part and / or the result part of the selected rule as well as the confidence value and the status value of the selected rule. All editions are transformed into a naturally linguistic message to make the application of the selected rule understandable to the user. This can be done using the user interface model 11 be accessed.

In one embodiment, the step S8 generates and outputs a control signal which automates a vehicle function with a degree of automation as stated above. The degree of automation depends on the confidence value of the selected rule. Furthermore, in one step S9 User outputs are generated and output, which make an applied rule plausible for the user, as explained above. Whether such user output is generated depends on the confidence value of the selected rule. Furthermore, depending on the confidence value, a notification can be issued as explained above, if necessary.

According to a further exemplary embodiment, the step S1 the user identified. After that, before the step S2 the user-specific rule-based data system stored outside the vehicle 8th For transfer the identified user to the vehicle and into the transformed rule-based data system 8th transferred. You can then optionally follow the steps S2 to S6 be performed. In step S7 is then dependent on the determined context and the transformed, possibly adapted rule-based data system 8th generates a rule result and then in step S8 generates and outputs the control signal.

Reference list

1

Sensors; User identification unit

2nd

Interfaces

3rd

Extraction unit

4th

Context data determination unit

5

Adaptive module

6

Hypothesis generating unit

7

Valuation unit

8th

rule-based data system

9

Adaptation unit

10th

User interface control

11

User interface model

12th

Hypothesis processing unit

13

Result unit

14

Output unit

15

Data bus

16

Pattern learning module

17th

Users

18th

User interfaces

19th

graphical user interface

20

adaptive system

21

Voice interface

22

Input unit

23

Plausibility check

24th

Confidence value determination unit

25th

Transformation unit

30th

another adaptive system

QUOTES INCLUDE IN THE DESCRIPTION

This list of documents listed by the applicant has been generated automatically and is only included 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 2015/0294223 A1 [0004]
• WO 2017/067853 A1 [0005]
• DE 102007025352 A1 [0006]
• US 2017/0190337 A1 [0007]

Claims (15)

Method for generating control signals for assisting occupants of a vehicle, in which a context of the vehicle is determined a rule of a rule-based data system (8) is selected as a function of the determined context, the rule-based data system (8) comprising a plurality of rules, each rule having a condition part and a result part, the condition part comprising conditions for the context of the vehicle , a rule result of the selected rule is generated, depending on the control result generated, a control signal is generated and output; on the basis of the condition part of the selected rule and the determined context, a user output is generated which makes the application of the selected rule plausible for the user (17), and the user output generated is output to the user (17). Procedure according to Claim 1 , characterized in that the user output contains a verbal and / or textual output of the condition part of the rule, the rule result of which was generated. Procedure according to Claim 1 or 2nd , characterized in that the user output contains a verbal and / or textual output of the result part of the selected rule, the rule result of which was generated. Method according to one of the preceding claims, characterized in that a confidence value assigned to the selected rule is determined, the confidence value indicating the probability with which the rule result of the selected rule corresponds to the preference of the user (17). Procedure according to Claim 4 , characterized in that the user output contains a verbal and / or textual output of the confidence value of the selected rule. Procedure according to Claim 4 or 5 , characterized in that the user output is generated depending on the confidence value. Procedure according to one of the Claims 4 to 6 , characterized in that the selected rule of the rule-based data system (8), the rule result of which has been generated, is assigned a status value for a status, the status value indicating how the confidence value of this rule is to be interpreted for this rule. Procedure according to Claim 7 , characterized in that the user output contains a verbal and / or text output of the status value of the selected rule. Procedure according to one of the Claims 4 to 8th , characterized in that the control signal automates a vehicle function with a degree of automation, the degree of automation depending on the confidence value of the selected rule. Procedure according to Claim 9 , characterized in that the user (17) is identified, a degree of user experience is determined and the degree of automation also depends on the determined degree of user experience. Method according to one of the preceding claims, characterized in that a user (17) is identified, the rule-based data system is a user-specific rule-based data system (8) which is stored outside the vehicle and via an interface (2) for the identified user (17) the vehicle is transmitted, the transmitted rule-based data system is transformed depending on the determined context, so that the transformed rule-based data system is applicable in the determined context, and the rule of the transformed rule-based data system is selected depending on the determined context. Procedure according to one of the Claims 9 to 11 , characterized in that the degree of automation is adapted during the transformation of the rule-based data system. Method according to one of the preceding claims, characterized in that, depending on the determined context and the rule-based data system (8), a hypothesis for a rule result of a rule of the data system (8) is first generated by means of a user interface (18) the hypothesis generated is output, a user input for the output of the hypothesis is recorded, by means of the recorded user input the output hypothesis is evaluated, the rule-based data system (8) is automatically adapted depending on the evaluation of the hypothesis and depending on the determined context and the adapted one rule-based data system (8) the rule result of the rule of the data system (8) is generated automatically. Device for generating control signals for supporting occupants of a vehicle, with a context determination unit (4) for determining the context of the vehicle, a result unit (13) which is designed to select a rule of a rule-based data system depending on the determined context, the rule-based data system comprising a plurality of rules, each rule having a condition part and a result part, the condition part comprising conditions for the context of the vehicle, and generating a rule result of the selected rule, a plausibility check unit (23) which is designed to to generate a user output based on the condition part of the selected rule and the determined context, which makes the automatic execution of the rule plausible for the user (17), and an output unit (14) for generating and outputting a control signal depending on the generated control result and for Spend the generated user output. Computer program product, comprising instructions which, when executed by at least one computer, cause the computer to execute the method according to one of the Claims 1 to 13 to execute.

Images