DE102022117683A1 - Method, system and computer program for training a neural network designed to operate a vehicle and for operating a vehicle with a neural network - Google Patents

Abstract

The invention relates to a method (100), system, and a computer program for training (103) a neural network (DNN1, DNNZ, DNN2) designed to operate a vehicle (10, 11, 12, ... 20, 21, 22). and a method, a system and a computer program for operating a vehicle (10, 11, 12, ... 20, 21, 22) with a neural network (DNN1, DNNZ, DNN2). In the method (100), a neural network (DNN1, DNNZ, DNN2) designed to operate a vehicle (10, 11, 12, ... 20, 21, 22) is trained (103) depending on one or more first errors (FBFF) of a vehicle functionality of one or more vehicles (10, 11, 12, ... 20, 21, 22) and depending on one or more second errors (PBFF) of the driver of the same vehicle (10, 11, 12, ... 20, 21, 22) or a driver of one or more further vehicles (20, 21, 22), the training (103) relating to the interaction of a vehicle (10, 11, 12, ... 20, 21, 22) and a respective driver of the vehicle (10, 11, 12, ... 20, 21, 22).

Description

The invention relates to a method, a device, a neural network, a computer program product for training a neural network designed to operate a vehicle, and a method, system and a computer program for operating a vehicle with a neural network. The operation of the vehicle can in particular relate to a driver assistance system, an active safety system, a system for at least partially automated driving or parking, and/or a system for signaling to the surroundings of the vehicle.

It is the object of the invention to provide an improved method, a device, a system and a computer program for training a neural network designed to operate a vehicle and/or for operating a vehicle, in particular a driver assistance system, an active safety system, a system for at least partial automated driving or parking, or a system for signaling the vehicle's surroundings.

According to a first aspect of the invention, a method for operating a vehicle is proposed. The method includes training a neural network designed to operate a vehicle depending on one or more first errors of a vehicle functionality of one or more vehicles and depending on one or more second errors of the driver of the same vehicle or a driver of one or more other vehicles. The training can be carried out with regard to the interaction of a vehicle and a respective driver of the vehicle.

Preferably, the method includes providing the trained neural network for operating the first vehicle and/or one or more second vehicles, and/or operating the first vehicle and/or one or more second vehicles depending on the output layer of the (such) trained neural network network.

A first error can be a (e.g. reproducible or non-reproducible) error in the function or a specific error (as incorrect execution of the vehicle function). The error in the vehicle functionality can be a technical error, caused by technology and/or at least partially caused by the environment of the vehicle and/or at least partially caused by a driving action by the driver.

The second error of the driver of the same vehicle or a driver of one or more additional vehicles can be a driver error, in particular also to be understood as an error made by the driver (in particular also to be understood as caused or permitted) in connection with the execution of a driver action.

Preferably, the at least one second error of the driver of the vehicle (driver error) can be a typically human error, e.g. a careless error, an attention error, a memory error, a skill error, perception error, etc. The driver error can be unintentional, erroneous, other than The driver's action is desired and/or expected to produce a different result. In particular, a “driver error” is collectively referred to in this document as a part of a driver action that or the result of which does not correspond to the motivation or expectations of the respective driver. This can be a technically correct or flawless process recorded or carried out by the vehicle (e.g. steering, braking, accelerating, performing maneuvers, etc.).

In the method, parameters of the (first and/or second) error can be determined and/or taken into account. These can identify qualitative variables (e.g. type of driver) and/or quantitative variables (e.g. a level of misconduct).

Training in relation to the interaction of a vehicle and a driver particularly concerns a combination of one or more errors in the vehicle system and one or more drivers of the vehicle. The combination can, for example, include or be any sequence, sequence and/or accumulation and/or chaining of vehicle system errors and driver errors. The chaining of errors can in particular include more than two or more than three chained errors. The method can have a particular advantage because conventional or rule-based methods can hardly deal with a chain of errors.

Furthermore, the combinations of one or more first errors of the vehicle system and the second error of the driver can occur as an occurrence of one or more first errors of the vehicle system and the second error of the driver of the vehicle at least partially within a (comparatively short) time interval (e.g. within 0 ,5; 1, 2, 5, 8 seconds) or with a time overlap.

Alternatively or additionally, the one or more first errors of the vehicle system and the second error or errors of the driver of the vehicle can form an error combination if they occur in connection with the same maneuver or a sequence of maneuvers or maneuver phases.

Training in relation to the interaction of a vehicle and a driver is to be understood in particular in such a way that it involves one or more (selectively) any aspects and/or consequential errors and/or consequences of the occurrence of an error in a vehicle system (the vehicle system, a similar vehicle system and/ or another vehicle system) and/or the error of the (respective) driver.

The interaction of a vehicle and a respective driver of the vehicle preferably includes an effect that is an (initially possible, probable, in particular predicted) first error of one (of) the vehicle system and an (initially possible, probable, in particular predicted) second error of the driver of the vehicle as a whole, especially in a (certain) combination, can or will (possibly, probably, with a certain probability) have.

The training of the neural network can concern a possible consequence (also to be understood as a result or consequence) of a (error) combination of one or more first errors of the vehicle system and the second error of the driver. Different measures of consequence can be taken into account. These can, for example, indicate the type and/or intensity of the consequence (e.g. limitation of comfort, risk, disadvantage, ...).

In particular, the first error and/or the second error relates to vehicle control.

In particular, the first error and/or the second error relates to vehicle guidance in the sense that, due to the driver error, there is an undesirable (e.g. direct or indirect) effect on vehicle guidance, in particular on longitudinal guidance and/or lateral guidance, parking or maneuvering Execution of maneuvers with the vehicle and/or signaling of the vehicle to its surroundings (e.g. honking, light signaling, etc.) has or could have resulted (e.g. at least with a certain probability). The vehicle guidance can in particular include or be the signaling of the vehicle to its surroundings.

The first errors and/or second errors can therefore include errors that led to a consequence (e.g. a certain consequence or one that exceeds a certain level), or driver errors that could have at least potentially led to a consequence (e.g. with a certain level of probability). can.

For example, first errors and/or second errors may be errors that can or could only lead to an (e.g. undesirable consequence) under certain circumstances and/or potentially.

In particular, a first (driver) error is characterized by the fact that (e.g. in the at least partially manual and/or at least partially assisted, in particular automated, guidance of the vehicle) by the driver (directly or indirectly, e.g. through insufficient control and/or or setting and/or monitoring of a vehicle system) a result is achieved that differs qualitatively, in particular in principle, from its goal.

In the context of this document, the term vehicle control can also include parking or maneuvering (e.g. parking, exiting, maneuvering) as well as the transition between driving and parking. The term maneuver can include, for example, a lane change maneuver, evasive maneuver, overtaking maneuver, turning, exiting, entering (e.g. expressways). The first driver error relating to at least one vehicle control during a first driver action can be a driver error that affects the control of the vehicle in some way, for example, occurs as part of the control of the vehicle and / or affects the control of the vehicle.

Preferably, the first errors and/or second errors are determined from a plurality of vehicles and/or a plurality of trips and/or a plurality of driving situations and/or (driving) contexts and/or at different locations.

In this case, first data in connection with the first error and/or with the second error can be determined in a first data or second data that is particularly temporally and/or spatially and/or causally related.

• - Parameter der Fahrsituation, bei der, vor der, oder nach der der jeweilige erste Fehler und/oder zweite Fehler passiert ist;
• - Parameter des Fahrkontexts, bei dem, vor dem, oder nach dem der jeweilige erste Fehler passiert ist;
• - Parameter betreffend die Position des Fahrzeugs, bei der, vor der, oder nach der der jeweilige erste Fehler und/oder zweite Fehler passiert ist; Insbesondere kann es sich dabei um eine absolute Position (z.B. Geo-Position) und/oder eine relative Position zu (bestimmen) Objekten, Spurenmarkierungen, Kreuzungen, Fahrbahnbegrenzungen, Kurven(-Mittelpunkten) handeln; und/oder
• - Parameter betreffend den ersten oder zweiten Fehler, z.B. betreffend die Art und/oder die Intensität des ersten Fehlers und/oder zweiten Fehlers; und/oder
• - Parameter der Konsequenz des jeweiligen ersten Fehlers und/oder zweiten Fehlers. Insbesondere kann der jeweilige Parameter auch verschiedene, bestimmte Kombinationen eines ersten Fehlers und eines zweiten Fehlers kennzeichnen.

For example, first errors and/or second errors are recognized as such in a (large) number of vehicles and first data and/or second data characterizing one or more of the following parameters are determined:

• - Parameters of the driving situation in which, before or after which the respective first error and/or second error occurred;
• - Parameters of the driving context in which, before or after the respective first error occurred;
• - Parameters relating to the position of the vehicle at which, before or after which the respective first error and/or second error occurred; In particular, this can be an absolute position (e.g. geo-position) and/or a relative position to (determined) objects, lane markings, intersections, road boundaries, curve (center points); and or
• - Parameters relating to the first or second error, for example relating to the type and/or intensity of the first error and/or second error; and or
• - Parameters of the consequence of the respective first error and/or second error. In particular, the respective parameter can also identify different, specific combinations of a first error and a second error.

One or more of these parameters can relate to a time interval in which the error occurred and/or to a time interval preceding this time interval.

In particular, one or more (possibly further) parameters that affect the course of vehicle control (and, for example, applied before, during or after the first error and/or second error) are determined and taken into account. One or more such parameters, in particular a change characteristic of one or more parameter values (e.g. characterizing an interaction with the operating interface) can be determined and taken into account over time.

The first data, in particular first data sections, and/or the second data, in particular second data sections, can include the aforementioned parameters or can be determined depending on them.

The first data and/or the second data can be preprocessed and/or stored. These data or data sets can be used to train the neural network. In comparison to the direct use of such data, a neural network trained with this data can be advantageous.

The first data and/or second data can be determined depending on statistical accumulations and/or patterns of occurrences and/or non-occurrences of first driver errors of certain types and/or first driver errors or can include information about such accumulations or patterns.

As part of the method, first data and/or second data can be determined (e.g. in advance) in relation to a plurality (e.g. hundreds, thousands, millions) of first and/or second errors. These can be determined using a backend server, stored and/or processed by it. The first driver actions can have taken place in one or in several different (possibly identical) vehicles. Based on the data, frequency information relating to the frequency of a specific driver error in the plurality of first driver actions can be determined.

The one or more of the second errors (driver errors) are an error by the driver, which in particular relates to a specific driving action or a driving action of a specific type. The driver error can be an incorrect, missing (e.g. forgotten) driving action and/or a driving action that is inappropriate for the situation, road location, etc. The driving action may be an action by the driver with respect to the vehicle or an interaction by the driver with the vehicle.

The second error can be a driver error with characteristics of so-called human oversight or error. Examples of different types of driver error can include: a forgotten turn signal, a turn signal in the wrong direction, a turn signal not turned off; a missing or too late look in a certain direction, e.g. towards an exit, priority road or in the direction of a road user, a missing, incorrect action by the driver, particularly exaggerated or understated in terms of intensity, in particular an inappropriate reaction to a situation, e.g the surroundings of the vehicle and/or within the vehicle, a driving situation and/or an action by another road user.

The second errors can in particular be a driver error due to a distraction, e.g. by another road user, something happening in the vicinity of the vehicle, or due to an overlook or almost overlook or a false assumption about the movement of another road user.

A first error and/or a second error may relate to a (preferably the same) sequence of driver actions and/or maneuver phases. Such a sequence may include or be a vehicle speed reduction, a blinking, a turn or merging in the blinking direction, etc. The first error and/or the second error may, for example, be a missing part of such a sequence Sequence and/or a non-compliant sequence, inappropriate in itself and/or for the driving situation and/or incorrect sequence or inappropriate time intervals between individual parts of the sequence.

Alternatively or additionally, first errors and/or second errors can include or be a deviation (e.g. of approximately 20%, more than 30%, 50%) that exceeds (e.g. a certain limit or standard value). This can include, for example, lateral guidance of the vehicle, e.g. a position within the lane or leaving the lane or roadway. The driver error can relate to the (e.g. direct or indirect) execution of a driver action (e.g. at least partially carried out manually and/or at least partially carried out automatically).

Alternatively or additionally, detection of error-free, in particular up to a certain tolerance, error-free and/or precisely executed vehicle guidance (on the part of the vehicle functionality and/or the driver) can be carried out. The first determined first data and/or second data (alternatively, preferably additionally to be taken into account) can relate to error-free, in particular up to a certain tolerance, error-free and/or precisely executed driver action.

The one or more first driver errors and/or the absence of errors and/or the precision can (even only) relate to the actions of the vehicle functionality and/or the driver, in particular the driver's contribution to the control of the vehicle (e.g. to the overall result from the interaction of the driver with one or more systems of the vehicle). The driver can also be the (acting) user of an at least partially automated vehicle or the user in certain situations.

The one or more second driver errors and/or the absence of errors and/or the precision can (even only) relate to the actions of the vehicle functionality, in particular the contribution of the vehicle functionality to the control of the vehicle (e.g. to the overall result from the driver's interaction with the functionality).

The recognition of an error-free, in particular up to a certain tolerance, error-free and/or precisely executed driver action and/or the corresponding first data and/or second data can be taken into account according to one or more features of the method.

The term “operating” can be interpreted in particular as controlling, changing an operating mode, and/or varying one or more operating parameters (in particular also to be understood as causing corresponding, in particular certain, changes). For example, operating the vehicle includes controlling, changing the operating modes and/or varying one or more parameters of a vehicle system (also to be understood as vehicle functionality). In particular, the operation can include or be the effecting or adapting of a (e.g. specific or expedient) measure.

For example, depending on the output layer of the neural network (also to be understood as: directly or indirectly controlled by the output layer of the neural network), a measure is effected, changed or suspended. This can include or be a selection of the measure, an initiation, a deactivation, in particular a (e.g. temporary) suspension of the measure related to a specific case and/or determining or adjusting one or more parameters of the measure.

The neural network can be trained and/or operated for predictive control, changing an operating mode, and/or varying one or more operating parameters or causing corresponding, in particular specific, changes, in particular a change in one or more vehicle systems in the vehicle.

The term neural network is understood to mean (at least) one or more neural networks. The neural network may comprise or be a region (e.g. some specific layers) of a larger neural network. In particular, the neural network can be understood as a sum of several neural networks, in particular trained and/or interconnected depending on one another (e.g. directly or indirectly interconnected to operate the first vehicle and/or one or more second vehicles).

The at least one neural network can be a neural network of a driver assistance system, a system for active safety, a system for at least partially automated driving or parking, or a system for signaling to the surroundings of the vehicle, or a neural network for Operate one or more of these systems (or functionalities) of the first vehicle and / or the one or more second vehicles.

Depending on the output layer of the neural network, a vehicle system (in particular also to be understood as vehicle functionality), in particular a driver assistance system tem and/or an active safety system and/or a system for at least partially automated driving or parking and/or a system for signaling to the surroundings of the vehicle are operated, in particular controlled. As an alternative or in addition to causing an action (e.g. an intervention in the chassis, lights, horn of the vehicle), a property of the vehicle system (e.g. its reaction or reaction to events that have not yet occurred) can be changed at least temporarily.

For example, depending on the output layer of the neural network, a sensitivity and / or reaction of the first vehicle and / or the one or more second vehicles (e.g. corresponding driver assistance systems, active safety systems, systems for at least partially automated driving) and / or a system for signaling of the vehicle to its surroundings can be changed, in particular adapted, to certain events of the first vehicle and / or the one or more second vehicles (e.g. in certain situations). This can be done proactively.

The term “neural network” can alternatively or additionally also be understood to mean another form of artificial intelligence (in particular comprising a neural network or a comparable structure). In particular, training and/or operating the at least one neural network is (also) to be understood as meaning an appropriate application of other types of machine learning. For example, the at least one neural network, the training and/or operation of it, can include or be an appropriate application of one or more circumstances or combinations thereof listed below: supervised learning, unsupervised learning, reinforcement learning, multitask learning, classification and regressions trees, support vector machines, logical and relational learning, probabilistic graphical models, rule learning, instance based learning, latent representation, bio-inspired approaches.

One or more steps of the method can be carried out, in particular end-to-end, using a correspondingly trained and/or operated neural network. Of course, this does not exclude the use or co-use of other means (e.g. one or more computer-implemented calculation formulas, rule-based algorithms, etc.). For example, at least part of one or more steps of the method (at least in part) can be carried out using other means. Preferably, a significant part of the processing of the data described in the present document and/or a significant part of the effect of the method can be carried out by means of the at least one neural network.

The method can in particular be carried out in such a way that controlling, changing an operating mode, and/or varying one or more operating parameters of the first vehicle and/or the one or more second vehicles is carried out before (preferably shortly before) the driver error is made, in particular before this is at least largely done or before the consequence of the driver error occurs.

For example, the neural network can be trained and/or operated to support the (first or possibly second) driver of the first vehicle and/or the one or more second vehicles by a driver assistance system, active safety system, at least partially automated system Driving or parking and/or system for signaling the vehicle to its surroundings at least temporarily (e.g. in connection with a specific, concrete second driver action) to activate, deactivate and/or change (e.g. to intensify and/or adapt).

The method also does not require any direct or indirect control of the first vehicle and/or one or more second vehicles depending on first data and/or second data and/or a result of an aggregation of the first data with the second data.

The term “training” is understood to mean in particular machine learning, non-rule-based learning and/or deep learning or the execution of machine learning, non-rule-based learning and/or deep learning. The training can be (at least partially or phased) monitored or (at least partially or phased) unsupervised training.

Training the neural network is in particular (also) to be understood as setting up the neural network. Operating the neural network is in particular controlling the neural network, in particular the input layer (one or more inputs) of the neural network. Furthermore, changing the operating mode of the at least one neural network can also be counted as operation.

Training the at least one neural network can be end-to-end learning (also in German: end-to-end learning) and/or reinforcement learning (also in German: a reinforcement learning or reinforcement learning). Several, in particular a predominant number, or (essentially) all intermediate steps necessary for striving for one or more goals can be mapped or integrated in the (at least partially trained) neural network.

The training can take place for one or more predetermined, predeterminable and/or method-determined goals.

• - Verbesserung einer qualitativen und/oder quantitative Kennzahl, die den Erfolg einer Fahrzeugführung und/oder des resultierenden Fahrens kennzeichnet; und/oder
• - eine verbesserte Übereinstimmung eines Ergebnisses einer Fahrzeugführung mit der Motivation und/oder der Erwartung des jeweiligen Fahrers; und/oder
• - eine Reduktion der Anzahl und/oder der Schwere und/oder des Risikos von Fahrerfehlern; und/oder
• - eine Reduktion eines Risikomaßes, z.B. eines auf die Fahrt, pro Fahrstrecke resultierenden Risikomaßes; und/oder
• - eine Reduktion der Konsequenzen der Fahrerfehler (z.B. eines oder mehrerer nach dem Fahrerfehler bzw. infolge eines möglichen, wahrscheinlichen und/oder prädizierten ersten Fehlers und/oder zweiten Fehlers aufgetretenen Konsequenz).

The goal of training (in particular also to be understood as the goal of machine learning) can include or be:

• - Improvement of a qualitative and/or quantitative key figure that characterizes the success of vehicle management and/or the resulting driving; and or
• - an improved correspondence between a vehicle driving result and the motivation and/or expectations of the respective driver; and or
• - a reduction in the number and/or severity and/or risk of driver errors; and or
• - a reduction of a risk measure, for example a risk measure resulting from the journey, per route; and or
• - a reduction in the consequences of the driver errors (e.g. one or more consequences that occurred after the driver error or as a result of a possible, probable and/or predicted first error and/or second error).

A target value can be specified for one or more goals (e.g. those mentioned above). For example, a first target value can be specified when training the neural network and/or a second target value can be specified during operation (when using the trained neural network during operation).

For example, the invention includes a first method or part of the method for training the neural network and/or a second method or part of the method in which the trained neural network (for operating the vehicles in commercial operation) is operated. The operation includes, for example, active user operation, with the neural network being used according to a feature of the invention. The method can also include both of these methods, which can be carried out, for example, simultaneously, one after the other or alternately.

The training can be carried out in several phases, over a long period of time and/or over several executions or operating cycles of the method. The method can include a change, in particular a long-term, slow update of the neural network comprising a large number of operating cycles, depending on a large number of the parameters described in this document. The training of the at least one neural network can be done at least partially, in particular only or predominantly automatically, in the course of operating an already at least partially trained neural network and/or a separate process.

Training may include facilitation. Moderation can be carried out depending on the parameters determined in the process. For example, the moderation can take place depending on a measure of the failure to meet a target value (e.g. with different types of driver actions, contexts, driving situations and/or by different drivers).

The method can be carried out according to one or more predetermined criteria by selecting and/or excluding the sections of the first data to be processed for training the at least one neural network. Alternatively or additionally, the neural network can (also) be trained with other data, the training process being operated, in particular controlled, depending on the first data.

A selection of one or more sections of the first data and/or the second data, an exclusion of one or more sections and/or division of the first data and/or the second data into several (different) categories can be carried out. These (e.g. sections assigned to different categories) can each be used differently to train and/or operate the at least one neural network.

The selection, exclusion and/or division into several categories can depend on the achievement corresponding to the first data and/or the second data (including, in the context of the present document, in particular data sections based on the first data or second data). of the target or target value occurs during one or more first driver actions.

• - einer Kennzahl, die den Erfolg einer Fahrzeugführung, insbesondere des Zusammenwirkens kennzeichnet; und/oder
• - einer Übereinstimmung eines Ergebnisses der Fahrerhandlung mit der (z.B. für die Art der Fahrerhandlung typischen und/oder im Verfahren anhand beliebiger Mittel ermittelter, insbesondere geschätzter) Motivation und/oder der Erwartung des jeweiligen Fahrers bei der Fahrzeugführung, insbesondere des Zusammenwirken; und/oder
• - einer Anzahl und/oder Schwere der ersten Fehler und/oder zweiten Fehler bei einem oder mehreren ersten Fahrerhandlungen; und/oder
• - einer Konsequenz, insbesondere eines qualitativen und/oder quantitativer KonsequenzMaßes, eines oder mehreren ersten Fahrerfehler (z.B. eines oder mehreren nach dem Fahrerfehler bzw. infolge des Fahrerfehlers aufgetretenen Vorkommnisse).

For example, the training, in particular the selection, exclusion and/or splitting of the first data and/or second data can be carried out depending on:

• - a key figure that characterizes the success of vehicle control, in particular the interaction; and or
• - a correspondence of a result of the driver's action with the motivation and/or expectations of the respective driver when driving the vehicle (e.g. typical for the type of driver's action and/or determined in the process using any means, in particular estimated), in particular the interaction; and or
• - a number and/or severity of the first errors and/or second errors in one or more first driver actions; and or
• - a consequence, in particular a qualitative and/or quantitative consequence measure, of one or more first driver errors (eg one or more incidents that occurred after the driver error or as a result of the driver error).

• - Ermitteln von Trainingsdaten, insbesondere auf Basis der ersten Datenabschnitte und/oder der zweiten Datenabschnitte und/oder weiterer Daten; und/oder
• - Ermitteln der Backpropagation; und/oder
• - Auswahl oder Anpassen eines Parameters des Trainiervorgangs.

Training (in particular to be understood as the process of training) can include or be:

• - Determining training data, in particular based on the first data sections and/or the second data sections and/or further data; and or
• - Determine backpropagation; and or
• - Select or adjust a parameter of the training process.

One or more aspects of training described above may be varied depending on one or more first errors and/or second errors, depending on the first data and/or second data associated with the respective errors.

The first data and/or second data are each, for example, in a temporal and/or causal connection with one or more first errors and/or second errors of a specific type in one or more sections of the journey. For example, it is the first data from a time interval (e.g. of approximately 10, 20, 30 seconds) before the first error and/or second error, and/or from a time interval in which the driver error occurs (e.g. while driver errors are being committed ), and/or in a time interval (e.g. approx. 10, 20, 30 seconds) after the driver error.

• - Daten eines oder mehrerer Bedienelemente zur Ausführung der Fahrerhandlung, insbesondere zum Führen des Fahrzeugs (z.B. Parameter des Lenkwinkels, Fahrpedals, Bremspedals) im Zusammenhang mit der Fahrerhandlung;
• - Daten zum Zustand des jeweiligen Fahrers;
• - Daten einer oder mehreren vom Fahrer des Fahrzeugs abweichenden Handlungen des jeweiliger Fahrers, die in einem zeitlichen und/oder kausalen Zusammenhang mit der Fahrerhandlung ausgeführt hat.

The first data and/or second data may include, be or be determined depending on:

• - Data from one or more control elements for carrying out the driver's action, in particular for driving the vehicle (e.g. parameters of the steering angle, accelerator pedal, brake pedal) in connection with the driver's action;
• - Data on the condition of the respective driver;
• - Data on one or more actions by the respective driver that deviate from the driver of the vehicle and which were carried out in a temporal and/or causal connection to the driver's action.

The first data can in particular (also) include data that is determined dependently, in particular on the basis of the first data.

Preferably, first data and/or second data can (also) be determined which were (also) valid in a time interval that preceded the occurrence of the driver error (e.g. characterizing an event a few seconds before the driver error). The first data can initially be stored in a memory, for example in a ring memory or in a comparable memory structure, and (later) made available and/or further processed depending on the occurrence or detection of a driver error.

For example, the first data is (also) characteristic of one or more pieces of information about one or more technical parameters of the first vehicle and/or the settings and/or operating logic of the first vehicle, in particular a driver assistance system, that are active at the time of the driver's action active safety and/or system for at least partially automated driving of the (respective) first vehicle. The first data (determined for training the neural network) can be recorded, selected, processed and/or during training depending on the technical parameters of the first vehicle and/or depending on the settings and/or the operating logic of the first vehicle that were active at the time of the driver's action be used.

Furthermore, a (initially possible, probable or predicted) third error of one (of) the vehicle functionality can be considered or taken into account in the method, which occurs or can occur when operating the vehicle with the neural network (initially possible, probable or according to a prediction). . This can have the same characteristics as a first error. Furthermore, a (initially possible, probable or predicted) fourth error of a driver of the vehicle can be considered or taken into account in the method, which occurs when operating the vehicle with the neural network (initially possible, probable or according to a prediction). diction) occurs or can occur. This can have the same characteristics as a second error.

In the method, a prediction of a (possible and/or probable) third error and/or fourth error and/or a combination of a third error and a fourth error can take place and/or (thereafter) a consequence of the third and/or fourth error can be avoided or . be reduced. Alternatively or additionally, the method may include executing, suppressing and/or varying a measure. The measure can be aimed at improving the interaction of the vehicle (e.g. the vehicle functionality) with the driver of the vehicle. The measure can be used to reduce a consequence measure and/or probability measure of a third error and/or fourth error and/or a combination of a third error with the fourth error. This can be done in particular without explicitly predicting the third error and/or the fourth error or their parameters. For example, the neural network can be trained (end to end) to control the (e.g. most suitable, optimized) measure. For example, training is carried out to determine or adapt one or more parameters of the measure.

An objective technical effect can be achieved in the method, in which a prediction of a (possible and/or probable) error and/or execution, suppression and/or variation of a measure can be carried out, even without knowledge of (typically arbitrarily diverse and complex) logical or causal connections between certain causes and certain errors exist or are made possible. In particular, the logical or causal question about the (typically very complex and/or at least partially subjective) causes of a human error does not need to be answered.

The method can have the advantage that (at least partially) predictive operation, in particular control, of the first vehicle or the second vehicle is carried out. For example, the vehicle operated by a neural network (fully or at least trained to a certain extent) can already react to something and/or be set appropriately before a certain (e.g. otherwise recognizable) occurrence and/or a possible consequence (second) driver error is determined or can be determined. A driver error is preferably predicted (e.g. a few seconds or fractions of a second) before its expected occurrence, in particular before the second driver action. This alone can save (valuable) time for determining, selecting and executing the measure. In particular, a time lead (so-called “lead time”) can be gained over a possible third error and/or fourth error. For example, a less intensive or surprising adjustment to the operation of the vehicle or the at least one measure can be initiated as a measure that occurs in the event of an already recognizable third and/or fourth error (or a possible combination or consequence a third error and a fourth error).

The method may include determining an input from the driver of the vehicle regarding a desired reaction to a (his, initially potential, possible) third error and / or a possible combination of a third error and a fourth error.

In the input, the driver can, for example, determine (also to be understood as selecting or specifying) how to react to a possible third error and/or a possible combination of a third error and a fourth error (e.g. by automatically correcting the possible driver error and /or by issuing a notice regarding the possible driver error). A measure dependent on the driver's input can then be determined and/or effected (e.g. in the event of the driver error occurring and/or in the event of an increased probability that the driver error will occur). In other words, the driver may be able to decide in advance how he should be helped in the event of an (initially potentially possible, specific) third error and (initially potentially possible, specific) fourth error of a specific functionality of the vehicle.

For example, the at least one neural network (e.g. to be used during operation) can be understood as part of a neural network with data based on the at least one (e.g. at least partially trained) neural network. For example, when training the at least one neural network, the data will be generated on any computing unit, in particular the computing unit of the backend, for example in order to then use it on one or more other computing units, in particular in (e.g. respective, specific) vehicles and/or in the backend ( e.g. for one or more vehicles) to make it operable (also to be understood: executable). Training the at least one neural network can include or be setting up the at least one neural network depending on the data from at least one corresponding neural network or one trained for the corresponding use.

In the context of the method, the use of a recurrent neural network is particularly advantageous. The topology can be set up, in particular via direct feedbacks, and/or via one or more indirect feedbacks, of the neural network (also to be understood as an interconnection of the at least two neural networks).

In particular, data based on the first data and/or second data is used as backpropagation for training and/or as an input variable for operating the at least one neural network. The backpropagation can be determined depending on the parameters achieved in several different time intervals (e.g. under different circumstances) or the corresponding data sections, in particular depending on the comparison of the resulting parameters, with a target value and/or provided for training the at least one neural network . For example, the backpropagation of the at least one neural network will (also) be determined depending on the extent of one or more deviations from the one or more (each selected) target values.

The operation of the vehicle can involve, in particular, a dynamic and/or temporary (e.g. related to the time interval of a possible second driver error) determining or adjusting the parameters, functional logic, and/or the behavior of the vehicle, in particular the driver assistance system, active safety system and/or or at least partially automated driving. For example, a dynamic control that can be updated in short time intervals, in particular only or essentially only for certain roadway locations and/or situations or roadway locations and/or situations that are characterized by certain features, can take place.

The training of the neural network and/or the operation of the first vehicle and/or the one or more second vehicles can, for example, be designed or optimized for the most anticipatory effect possible (for the one or a specific driver). This can create a mode of operation that is pleasant to human perception and is practically useful.

Training and/or operating the at least one neural network can result in particular advantages (e.g. regarding the effectiveness of the method), because a trained neural network (according to the features of the invention) is very complex and otherwise typically not rule-based, comprehensible or predictable can handle processes.

In the context of this document, the term “driving situation” can be understood, for example, as a specific situation characterized by an arrangement, action or interaction of road users or by certain driving parameters of road users (particularly at the viewing level of certain objects). In particular, the meaning of the term "driving situation" therefore differs from the meaning of the term "traffic situation" that is often used colloquially, which rather includes general categories such as "free traffic", "dense traffic", "slow-moving traffic", "traffic jams", " End of traffic jam” etc. corresponds.

A driving situation can be described by parameters of the driving situation. One or more parameters of the driving situation can characterize, in particular represent, a specific pattern (also to be understood as a data pattern), e.g. a pattern characterizing an arrangement and/or speed of the objects and/or a pattern of the parameters of the driving situation. The driving situation can also be characterized by a spatial pattern of the so-called open spaces in the area surrounding the vehicle or by corresponding parameters.

• - eine (bestimmte) räumliche Verteilung der Verkehrsteilnehmer und/oder der Bewegungsparameter der Verkehrsteilnehmer, insbesondere ein Anordnungsmuster der Verkehrsteilnehmer in der Umgebung des (realen) Fahrzeugs,
• - eine (bestimmte) räumliche Verteilung unbeweglicher Objekte in der Umgebung des Fahrzeugs,
• - eine relative Position und/oder Bewegungsparameter zu bestimmten Arten von Spurmarkierungen, Verkehrszeichen, Ampeln (nicht notwendigerweise zu bestimmten Ampeln, etc.),
• - eine Information über die Vorfahrt des Fahrzeugs, insbesondere gegenüber bestimmten Verkehrsteilnehmern und/oder Verkehrsteilnehmern, die tatsächlich oder zumindest potenziell aus bestimmten Richtungen, z.B. einer querenden Straße von rechts oder von links kommen oder kommen können,
• - eine Information zu einer, z.B. einen Grenzwert überschreitenden, Handlung eines Verkehrsteilnehmers in der Umgebung des Fahrzeugs, wie z.B. ein Hupen, Lichthupen, Drängeln, Überholen des Fahrzeugs, ein Überholversuch und dergleichen.

The at least one driving situation can preferably be characterized by one or more features listed below (or parameters of the driving situation corresponding to these):

• - a (certain) spatial distribution of the road users and/or the movement parameters of the road users, in particular an arrangement pattern of the road users in the area surrounding the (real) vehicle,
• - a (certain) spatial distribution of immovable objects in the area surrounding the vehicle,
• - a relative position and/or movement parameter to certain types of lane markings, traffic signs, traffic lights (not necessarily to specific traffic lights, etc.),
• - information about the right of way of the vehicle, in particular towards certain road users and/or road users who are actually or at least potentially coming from certain directions, e.g. a crossing street from the right or from the left,
• - Information about an action by a road user in the vicinity of the vehicle, for example exceeding a limit value, such as honking, flashing lights, jostling, overtaking the vehicle, attempting to overtake and the like.

Furthermore, the driving situation can be characterized by one or more parameters in connection with relevant traffic rules, traffic signs, right-of-way, traffic lights and/or traffic light phases.

The at least one driving situation can preferably be a driving situation that exceeds certain limit values or a driving situation that is characterized by parameters that exceed certain limit values. For example, the at least one driving situation can be a driving situation with an undesirable or dangerous approach to an object or a road user, an acceleration value that exceeds a limit value, an undesirable arrangement with other road users, etc.
The driving situation can be a special situation (for example, which occurs comparatively rarely) or a dangerous driving situation, for example a driving situation for which an increased risk is determined or assumed.

For example, the driving situation can take into account or include one or more parameters of the (relevant in connection with this) traffic rules, e.g. traffic signs, right-of-way, traffic light phases. For example, the driving situation parameter(s) relating to or based on environmental sensor data and/or information transmitted to the vehicle (e.g. from another road user, using a Car-2-Car or Car-to-X) can be taken into account. The surrounding sensor data can be data processed in a specific way by a sensor that detects the surroundings, in particular a sensor system of at least one vehicle.

A temporal and/or spatial change characteristic, in particular a gradient, such as a temporal and /or spatial gradient of the respective parameter value applies or is taken into account.

When operating the first vehicle and/or the one or more second vehicles (with the neural network trained completely or at least to a certain extent), the data can be taken into account on the basis of one or more parameters relating to driving operation. These can be used (e.g. together with other data) to control the input (the input layer) of the neural network.

The data based on one or more parameters relating to driving operation can (entirely or selectively) identify the same or similar variables as the respective first data and/or second data and/or the same or corresponding features in relation to the second vehicle or on the second driver of the second vehicle. The data based on one or more parameters relating to driving operation are understood in particular as predicted data (e.g. valid before and/or during a possible second driver action or for the time interval of the second driver action).

The data based on one or more parameters relating to driving operation can be used to control the at least one neural network (trained depending on the first data). For example, the data is read in at the input layer of the at least one neural network based on one or more parameters relating to driving operation (with or without a processing step).

One or more similarity measures between the first data and/or the second data on the one hand, and the data based on one or more parameters relating to driving operation on the other hand can be determined and taken into account. For example, the input layer of the neural network is (also) dependent on a similarity measure between the (e.g. current) data based on one or more parameters relating to driving operation with the (e.g. at least with regard to one or more features, parameters, patterns, type of Driver action corresponding) first data and / or second data that were used to train the at least one neural network.

Alternatively or additionally, one or more patterns of the parameters of the second data can be determined and compared with one or more patterns in the first data, and the result of the comparison can be taken into account. For example, the input layer of the neural network is (also) controlled depending on the result of the comparison.

Preferably, the method can be carried out entirely or predominantly automatically, dynamically, and/or driver-specific or driver group-specific. In the simplest case, driver groups can be taken into account, which are characteristic of drivers who predominantly drive the vehicle manually or drivers who predominantly drive the vehicle automatically. The procedure can take into account that Fah Drivers from different driver groups cause different types of driver errors or driver errors with different probabilities or with different consequences (e.g. when operating a specific driver assistance system). In particular, the method (due to the described training of the neural network) can have the objective technical effect that (at least as a result) typical and / or probable drivers from different driver groups have different types of driver errors or driver errors that (at least potentially) lead to certain consequences can lead to each case being appropriately taken into account.

The method may include or be a method for training the at least one neural network and/or a method for operating the (first and/or one or more second) vehicles depending on a neural network trained according to one or more features described in the present document.

According to a further embodiment, the neural network is trained and/or operated to avoid, in particular (almost) exclude, one or more first errors of the vehicle system of the vehicle and one or more second errors of the driver of the vehicle, and/or a probability measure and/or To reduce the consequence level of the combination of at least one first error and at least one second error.

The avoidance and/or reduction of the probability measure and/or the consequence measure of the combination of one or more first errors and one or more second errors can be carried out depending on one or more predictions (e.g. according to one or more described in the present document).

According to a further embodiment, the method includes determining first data depending on the detection and / or the evaluation of a first error of a vehicle system of a first vehicle, determining the second data depending on the detection and / or evaluation of a second error of a driver first vehicle, and then training the neural network depending on the first data and depending on the second data.

The evaluation can be carried out with regard to the (respective) parameters, the cause and/or the consequence of the first errors and/or with regard to the (respective) parameters, the cause and/or the consequence of the second errors. Depending on the result of the evaluation, the selection, exclusion and/or processing of the first data and/or the second data can then be carried out. The training data for training the neural network can be determined (also to be understood: generated) depending on the result of the processing. Alternatively or additionally, training (e.g. the process of training) can take place depending on the result of the processing.

According to a further embodiment, a target value for training the neural network is dependent on one or more values predetermined or conditioned by the functional reliability and/or operational safety and/or a selected standard value of the operation of the vehicle, in particular one or more limit values for one or more probability measures and /or Consequence measures selected, determined or adjusted in connection with the occurrence of an error.

The functional safety and/or safety in use and/or the standard value of the operation of the vehicle can be a functional safety and/or safety in use and/or corresponding standard value related to one or more vehicle systems, each of which includes a part of the vehicle system or a part of form this.

For example, the corresponding values can be determined or adjusted in particular depending on values specified (e.g. by law and/or according to a standard and/or according to the goal of the manufacturer and/or operator of the vehicle).

For example, it can be the functional safety and/or usage safety and/or the standard value for the operation of the vehicle and/or the functional safety and/or usage safety and/or the standard value for the operation of a vehicle functionality of the vehicle for at least partially automated driving and/or or act parking.

For example, a target value can be selected depending on a parameter relating to the functional safety and/or the safety of use and/or the standard value of a (corresponding or related) functionality.

For example, a standard value can be selected. For example, the target value can depend on a value predetermined or conditioned by the functional safety and/or safety of use and/or the standard value, in particular depending on limit values for one or more probability measures and/or consequence measures be determined. The probability measures and/or consequence measures can relate to errors of a certain type and/or certain errors. A target value can be adjusted by a certain (e.g. multiplicative, additive or logarithmic) factor from the desired requirement to fulfill the functional safety and/or operational safety and/or standard value.

According to a further embodiment, the neural network is trained and/or operated to predict a driver error, in particular a late, insufficient, inappropriate and/or incorrect reaction of the driver of the vehicle to unavailability, failure and/or behavior of the vehicle Vehicle functionality of the vehicle, and/or to predict an undesirable functioning of a vehicle functionality of the vehicle in response to a driver error, in particular a late, insufficient, inappropriate and/or incorrect driver action by the driver of the vehicle.

It can be predicted whether and/or when the driver error or a combination of driver errors will (at least probably) happen. Alternatively, one or more parameters of a combination of errors and/or a consequence of the combination of the at least one first error and at least one error can be predicted. For example, it can be predicted whether or when a third error and/or a fourth error by the driver can or will happen (possibly with what probability), or which performance feature of the vehicle functionality or driver action a third error and/or a fourth error can occur of the driver (if necessary with what probability).

According to a further embodiment, the neural network is trained and/or operated to predict a reaction of the driver of the vehicle to an unavailability, a failure, a behavior, and/or a late, insufficient, inappropriate and/or incorrect action Vehicle functionality of the vehicle, and/or to predict a reaction of the vehicle functionality of the vehicle to unavailability, failure, behavior, and/or late, insufficient, inappropriate and/or incorrect driver action by the driver of the vehicle.

According to a further embodiment, the neural network is trained and/or operated to predict a consequence of the combination of unavailability, failure, behavior and/or a late, insufficient, inappropriate and/or incorrect action of the vehicle functionality vehicle and a late, insufficient, inappropriate and/or incorrect driver action by the driver of the vehicle, and/or to predict a consequence from a combination of a driver error by the driver of the vehicle and an action of the vehicle functionality of the vehicle.

According to a further embodiment, the first data and/or the data from one or more vehicles and/or the neural network is trained depending on one or more first parameters of a first error in a vehicle functionality Vehicle and / or depending on one or more second parameters of the second error of a driver of the vehicle.

For example, the first data can be determined depending on the detection of the second error and/or the second data can be determined depending on the detection of the first error (e.g. determined as underlying data within a corresponding vehicle, transmitted within the vehicle, transmitted to the backend), one subject to selection and/or exclusion and/or used for training.

For example, the first parameter of a first error and/or a second parameter of a second error can be dependent on a parameter of the driving situation in which this error occurred and/or depending on a determinable position (absolute and/or relative position), where this error occurred can be determined or depends on it.

Alternatively or additionally, the first parameter of a first error and/or a second parameter of a second error are determined depending on further parameters (e.g. dependent on the respective vehicle, the context of the trip and/or a time-dependent greeting and/or a space-dependent variable) and/or in procedures are processed.

According to a further embodiment, the first data characterize variables that are temporally and/or causally related to the occurrence of the first error in the vehicle functionality of the vehicle. Analogous to this, the second data characterize variables that are temporally and/or causally related to the occurrence of the second error by the driver of the vehicle.

Preferably, a first error and/or second error corresponding to a specific criterion can be detected and the first data (in particular a first data section of the first data) and/or the second data (in particular a second data section) can then be determined.

The criterion for selecting the first data and/or the second data relating to a vehicle can also be varied, in particular depending on data, in particular first data and/or second data from one or more second vehicles.

According to a further embodiment, the training of the neural network is dependent on one or more first absolute and/or relative first positions at which the respective first error occurred and/or at which the respective first data applied, and/or depending on one or several second absolute and/or relative second positions at which the respective second error occurred and/or at which the second data applied.

According to a further embodiment, the training data for training the neural network will include one or more of the following pieces of information and/or determined depending on this and/or used to train the neural network: first position data and/or second position data and/or parts of map information from the Environment concerning the first position(s) and/or the second position(s).

The map information can relate to a section of a digital map and/or one or more layers of the digital map.

According to a further embodiment, the control of the input of the neural network when operating a vehicle is carried out depending on one or more parameters relating to the driving operation of the vehicle.

According to a further embodiment, one or more of the parameters relating to the driving operation of the vehicle are characteristic of an absolute and/or relative position at which the vehicle is located and/or at least likely to be in the near future, and/or relating to map information Surroundings of the vehicle, in particular regarding the current and/or preceding section of the digital map, and/or for parameters of a current driving situation and/or for a driving situation predicted in the near future; and/or for parameters of the current driving context and/or for a driving context predicted in the near future.

According to a further or second aspect of the invention, a method for operating a vehicle is proposed. The method includes operating a first vehicle and/or one or more second vehicles depending on the output layer of a trained neural network, the neural network depending on one or more first errors of a vehicle functionality of one or more vehicles and depending on one or more second errors the driver of the same vehicle or a driver of one or more other vehicles is or is being trained. The training can relate to the interaction of a vehicle and a respective driver of the vehicle.

The method according to the first aspect or steps of the method according to the first aspect can be combined with the method according to the further or second aspect. In the preferred embodiment, these can form the process (overall process). The method (overall method) can include: detecting a first driver error relating to vehicle guidance in a first driver action by a first driver of a first vehicle, and (thereafter or depending on the detection of the driver error) determining first data in connection with the first driver error; and (thereafter) depending on the first data, training a neural network, and (thereafter) operating the first vehicle and/or one or more second vehicles depending on the output layer of the trained neural network.

According to a further or third aspect of the invention, a system for operating a vehicle is proposed, wherein the system is designed to train a neural network designed to operate a vehicle, the training being dependent on one or more first errors in vehicle functionality of one or more vehicles and subject to one or more second errors by the driver of the same vehicle or a driver of one or more additional vehicles. The training can relate to the interaction of a vehicle and a respective driver of the vehicle.

According to a further or fourth aspect of the invention, a system for operating a vehicle is proposed, wherein the system is designed to operate a first vehicle and / or one or more second vehicles depending on the output layer of a trained neural network, the neural network being dependent of one or more first errors of a vehicle functionality of one or more Vehicles and is or will be trained depending on one or more second errors of the driver of the same vehicle or a driver of one or more other vehicles. The training can relate to the interaction of a vehicle and a respective driver of the vehicle.

The system according to the third aspect or parts of the system according to the third aspect can be combined with the system according to the fourth aspect. In particular, the system according to the third aspect is designed to form a system (as an overall system) together with the system according to the fourth aspect or to be operated as a system (overall system).

In the preferred embodiment, the system (e.g. as an overall system) can be designed to detect a first driver error relating to (first) vehicle guidance in a first driver action by a first driver of a first vehicle and to determine first data in connection with the first driver error and for training a neural network depending on the first data and further for operating the first vehicle and / or one or more second vehicles depending on the output layer of the trained neural network.

The system preferably comprises one or more first vehicles, at least one second vehicle and/or a backend (also to be understood as a cloud, a server, a portal). One or more first vehicles are preferably designed for a direct or indirect operative connection with the at least one second vehicle and/or the backend. The one or more first vehicles can be designed to determine parameters characterizing one or more (first) driver errors, e.g. type of driver errors, type of corresponding driver actions, etc.

Furthermore, the one or more first vehicles can be designed to transmit the corresponding data (e.g. when executing the active connection) to the backend. The backend can be designed to process the data from a plurality of the first vehicles, in particular to determine one or more frequency information (e.g. for several different groups of driver actions and driver errors). A measure can be implemented (at least partially in the backend and/or at least partially in the second vehicle) based on the processed data (in particular in relation to the first data), in particular based on the frequency information.

In particular, the system can also include an operating system of the vehicles and/or a data interface set up to carry out the method, e.g. hardware-related interfaces. The system, e.g. the majority of the first and/or second vehicles and/or the backend can be operated by the same provider, operator, service provider and/or a standard or convention designed in accordance with the invention. The one or more first vehicles and the one or more second vehicles may be offered directly or indirectly by the same manufacturer or operator.

According to a fifth aspect of the invention, a neural network for operating a vehicle is proposed. The neural network is (completely or at least to a certain extent) dependent on one or more first errors of a vehicle functionality of one or more vehicles (e.g. dependent on the first data or first data sections) and dependent on one or more second errors of the driver of the same Vehicle or a driver of one or more other vehicles (e.g. depending on the first or second data or first or second data sections). The training can relate to the interaction of a vehicle and a respective driver of the vehicle. The neural network is in particular a neural network designed to operate a vehicle.

For example, the at least one neural network (e.g. to be used during operation) can be understood as features of or with data based on the at least one (e.g. at least partially trained) neural network.

For example, when training the at least one neural network, the data is generated on any computing unit, in particular the computing unit of the backend, for example in order to then use it on one or more other computing units, in particular in (e.g. respective, specific) vehicles and/or in the backend (e.g. for one or more vehicles) operable (also to be understood: executable). Training the at least one neural network can include or be setting up the at least one neural network depending on the data from at least one corresponding neural network or one trained for the corresponding use.

According to a sixth aspect of the invention, a computer program product comprising a computer program and/or data for executing the computer program is proposed, wherein the computer program product consists of is to carry out the method described in this document when it is executed on at least one computing unit.

The computer program product can be designed as an update of a previous computer program, which includes the parts of the computer program or the corresponding program code, for example as part of a functional expansion, for example as part of a so-called “remote software update”. The computer program product includes in particular a medium readable by the data processing device, on which the program code is stored, or at least an encrypted file. According to a further aspect, a program product is described that includes an authorized access right to stored data of the computer program product.

Claims (18)

Method (100) for training (103) a neural network (DNN1, DNNZ, DNN2) designed to operate a vehicle (10, 11, 12, ... 20, 21, 22) depending on one or more first errors (FBFF) a vehicle functionality of one or more vehicles (10, 11, 12, ... 20, 21, 22) and depending on one or more second errors (PBFF) of the driver of the same vehicle (10, 11, 12, ... 20, 21 , 22) or a driver of one or more further vehicles (20, 21, 22), the training (103) relating to the interaction of a vehicle (10, 11, 12, ... 20, 21, 22) and one respective driver of the vehicle (10, 11, 12, ... 20, 21, 22). Procedure (100) according to Claim 1 , wherein the neural network (DNN1, DNNZ, DNN2) is trained (103) and / or operated (104), - one or more first errors (FBFF) of the vehicle system of the vehicle (10, 11, 12, ... 20, 21, 22), and/or - to avoid one or more second errors (PBFF) of the driver of the vehicle (10, 11, 12, .. 20, 21, 22) and/or a probability measure and/or consequence measure of the combination at least of a first error (FBFF) and at least one second error (PBFF). Procedure (100) according to Claim 1 or 2 , comprising: - determining (102) first data depending on the detection and/or evaluation of a first error (FBFF) of a vehicle system of a first vehicle (10, 11, 12); - Determining (101) second data depending on the detection and/or evaluation of a second error (PBFF) of a driver of a first vehicle (10, 11, 12); - Training (103) of the neural network (DNN1, DNNZ, DNN2) depending on the first data and depending on the second data. Method (100) according to one of the preceding claims, wherein a target value for training (103) of the neural network (DNN1, DNNZ, DNN2) depends on one or more functional safety and / or usage safety and / or a selected standard value of the operation of the vehicle (20, 21, 22) predetermined or conditional values, in particular one or more limit values for one or more probability measures and / or consequence measures in connection with the occurrence of an error is selected, determined or adjusted. Method (100) according to one of the preceding claims, wherein the neural network (DNN1, DNNZ, DNN2) is trained (103) and/or operated (104) to: - Prediction (104) of a driver error, in particular a late, insufficient, inappropriate and/or incorrect reaction by the driver of the vehicle (20, 21, 22) to unavailability, failure and/or behavior of the vehicle functionality of the vehicle (20 , 21, 22); and or - Prediction (104) of an undesirable functioning of a vehicle functionality of the vehicle (20, 21, 22) in response to a driver error, in particular to a late, insufficient, inappropriate and/or incorrect driver action by the driver of the vehicle (20, 21, 22 ). Method (100) according to one of the preceding claims, wherein the neural network (DNN1, DNNZ, DNN2) is trained (103) and/or operated (104) to: - Prediction (104) of a reaction of the driver of the vehicle (20, 21, 22) to unavailability, failure, behavior, and/or a late, insufficient, inappropriate and/or incorrect action of a vehicle functionality of the vehicle (20 , 21, 22); and or - Prediction (104) of a reaction of the vehicle functionality of the vehicle (20, 21, 22) to unavailability, failure, behavior, and/or late, insufficient, inappropriate and/or incorrect driver action by the driver of the vehicle (20 , 21, 22). Method (100) according to one of the preceding claims, wherein the neural network (DNN1, DNNZ, DNN2) is trained (103) and/or operated (104) to: - Predict (104) a consequence from the combination of one unavailability, failure, behavior and/or delay, insufficient, inappropriate and/or incorrect action of the vehicle functionality of the vehicle (20, 21, 22), and a late, insufficient, inappropriate and/or incorrect action by the driver of the vehicle (20, 21, 22); and/or - prediction (104) of a consequence of a combination of a driver error by the driver of the vehicle (20, 21, 22) and an action of the vehicle functionality of the vehicle (20, 21, 22). Method (100) according to one of the preceding claims, wherein: determining and/or processing first data and/or second data from one or more vehicles (10, 11, 12); and/or training (103) the neural network (DDN1, DDNZ, DNN2); is executed depending on: - one or more first parameters of a first error (FBFF) of a vehicle functionality of the vehicle (20, 21, 22); and or - one or more second parameters of the second error (PBFF) of a driver of the vehicle (10, 11, 12). Method (100) according to one of the preceding claims, wherein: - the first data is characteristic of the variables that are temporally and/or causally related to the occurrence of the first error (FBFF) of the vehicle functionality of the vehicle (10, 11, 12, ... 20, 21, 22); and or - the second data is characteristic of the variables that are temporally and/or causally related to the occurrence of the second error (PBFF) of the driver of the vehicle (10, 11, 12, ... 20, 21, 22). Method (100) according to one of the preceding claims, wherein the training (103) of the neural network (DDN1, DDNZ, DNN2) is carried out depending on: - one or more first absolute and/or relative first positions at which the respective first error (FBFF) occurred and/or at which the respective first data applied; and or - one or more second absolute and/or relative second positions at which the second error (PBFF) occurred and/or at which the second data applied. Procedure (100) according to Claim 10 , wherein the training data for training (105) of the neural network (DNN1, DNNZ, DNN2) includes one or more of the following pieces of information and/or determined depending on this information and/or when training (105) of the neural network (DNN1, DNNZ, DNN2 ) used: - the first position data; and/or - the second position data; and/or - a part of map information from the surroundings of the first positions and/or second positions. Method (100) according to one of the preceding claims, wherein controlling the input of the neural network (DNN1, DNNZ, DNN2) when operating (104) a vehicle (10, 11, 12, ... 20, 21, 22) depends on one or more parameters relating to the driving operation of the vehicle (10, 11, 12, ... 20, 21, 22) is carried out. Method (100) according to one of the preceding claims, wherein one or more parameters relating to the driving operation of the vehicle (10, 11, 12, ... 20, 21, 22) are characteristic of: - an absolute and/or relative position at which the vehicle (10, 11, 12, ... 20, 21, 22) is located and/or is at least expected to be located in the near future; and or - map information regarding the surroundings of the vehicle (10, 11, 12, ... 20, 21, 22), in particular regarding the current and / or ahead section of the vehicle (10, 11, 12, ... 20, 21, 22); and or - at least one parameter of a current driving situation and/or a driving situation predicted for the near future and/or - at least one parameter of the current driving context and/or for a driving context predicted in the near future. Methods (100) for operating a vehicle (10, 11, 12, 20, 21, 22) include operating (104) a first vehicle (10, 11, 12) and/or one or more second vehicles (20, 21, 22) dependent on the output layer of a trained neural network (DNN1, DNNZ, DNN2), the neural network (DNN1, DNNZ, DNN2) depending on one or more first errors (FBFF) of a vehicle functionality of one or more vehicles (10, 11, 12, ... 20, 21, 22) and depending on one or more second errors (PBFF) of the driver of the same vehicle (10, 11, 12, ... 20, 21, 22) or a driver of one or more other vehicles (20, 21, 22) is or is being trained (103), the training (103) being based on the interaction of a vehicle (10, 11, 12, ... 20, 21, 22) and a respective driver of the vehicle ( 10, 11, 12, ... 20, 21, 22). System for operating a vehicle (10, 11, 12, ... 20, 21, 22,) wherein the system is designed to train (103) a vehicle (10, 11, 12, ...) to operate (104). 20, 21, 22) trained neural network (DNN1, DNNZ, DNN2), wherein the training (103) depends on one or more first errors (FBFF) of a vehicle functionality of one or more vehicles (10, 11, 12, ... 20, 21, 22) and depending on one or more second errors (PBFF) of the The driver of the same vehicle (10, 11, 12, ... 20, 21, 22) or a driver of one or more other vehicles (20, 21, 22) can be carried out, and the training (103) relates to the interaction of a vehicle (10, 11, 12, ... 20, 21, 22) and a respective driver of the vehicle (10, 11, 12, ... 20, 21, 22). System for operating a vehicle (10, 11, 12, ... 20, 21, 22), the system being designed to operate (104) a first vehicle (10, 11, 12) and/or one or more second vehicles (20, 21, 22) depending on the output layer of a trained neural network (DNN1, DNNZ, DNN2), wherein the neural network (DNN1, DNNZ, DNN2), depending on one or more first errors (FBFF) of a vehicle functionality of one or more Vehicles (10, 11, 12, ... 20, 21, 22) and depending on one or more second errors (PBFF) of the driver of the same vehicle (10, 11, 12, ... 20, 21, 22) or one The driver of one or more other vehicles (20, 21, 22) is or is being trained (103), the training (103) being based on the interaction of a vehicle (10, 11, 12, ... 20, 21, 22) and of a respective driver of the vehicle (10, 11, 12, ... 20, 21, 22). Neural network (DNN1, DNNZ, DNN2) for operating a vehicle (10, 11, 12, .. 20, 21, 22), the neural network (DNN1, DNNZ, DNN2) depending on one or more first errors (FBFF) a vehicle functionality of one or more vehicles (10, 11, 12, ... 20, 21, 22) and depending on one or more second errors (PBFF) of the driver of the same vehicle (10, 11, 12, ... 20, 21 , 22) or a driver of one or more other vehicles (20, 21, 22) is trained (103), the training focusing on the interaction of a vehicle (10, 11, 12, ... 20, 21, 22) and of a respective driver of the vehicle (10, 11, 12, ... 20, 21, 22). Computer program product comprising a computer program and / or data for executing the computer program, wherein the computer program product is designed, when executed on at least one computing unit, one or more steps (101, 102, 103, 104) of the method (100) according to one or more of the Claims 1 until 14 to carry out.

Images