Vehicle and method for issuing recommendations to a person driving a vehicle to take over vehicle control

Abstract

Vehicle (1) with an at least partially automated operating mode, where a recommendation system (2) comprising a data collection module (2.1), a prediction module (2.2) and a recommendation module (2.3), wherein the data collection module (2.1) is configured to collect vehicle data (D-FZG), environmental data (D-UMG) and / or environmental data (D-UMW); the prediction module (2.2) is configured to read in a driver profile from a set of driver profiles, wherein each driver profile comprises a machine learning model (3) or heuristic model individually trained for the respective driver profile, which is configured to read in the vehicle data (D-FZG), surrounding data (D-UMG) and / or environmental data (D-UMW) at least for a preceding route section and to output a predictive indication value (IND-PRÄ) as an output variable, wherein the predictive indication value (IND-PRÄ) is a numerical value; and the recommendation module (2.3) is configured to compare the predictive indication value (IND-PRÄ) with an indication threshold value (IND-SW) and to initiate the output of a recommendation in the vehicle (1) for a person driving the vehicle (4) to assume manual control of the vehicle if the predictive indication value (IND-PRÄ) lies in a first range compared to the indication threshold value (IND-SW), and to initiate the output of a recommendation in the vehicle (1) for a driver assistance system to assume at least partially automated control of the vehicle if the predictive indication value (IND-PRÄ) lies in a second range compared to the indication threshold value (IND-SW), characterized in that the recommendation system (2) further comprises a driver monitoring module (2.4) which is designed to monitor the person driving the vehicle (4) with the aid of at least one sensor and to determine a current control behavior (D-SV-AKT) and / or a current driver state (D-ZUS-AKT) from sensor data generated by the sensor and to assign the person driving the vehicle (4) to one of the driver profiles depending on the determined current control behavior (D-SV-AKT) and / or the current driver state (D-ZUS-AKT).

Description

[0001] The invention relates to a vehicle with an at least partially automated operating mode according to the type defined in more detail in the preamble of claim 1 and to a method for issuing recommendations to a person driving the vehicle according to the type defined in more detail in the preamble of claim 9.

[0002] Modern vehicles are increasingly being equipped with driver assistance systems. Driver assistance systems support the driver in taking over driving tasks. This can increase road safety, vehicle operation efficiency, and comfort for the driver. Various driver assistance systems are capable of at least partially automated, and in the future, autonomous, vehicle control. For example, a lane departure warning system can keep the vehicle in the lane currently occupied when driving on a highway, adaptive cruise control can maintain a safe distance from the vehicle in front, and a parking assistant can take over lateral guidance of the vehicle when reversing into a parking space, while the driver assumes longitudinal guidance.

[0003] To use a driver assistance system that enables at least partially automated vehicle control, certain driving conditions must be present that allow the driver assistance system to operate. For example, the sensors used to detect the surroundings must not be disrupted and must provide sufficiently usable sensor data, or a specific driving condition must exist, such as traveling on a specific road type within a specified speed range. If at least a partially automated assistance function is available, it typically has to be activated manually first. However, it can happen that the person driving the vehicle does not use it despite the assistance functions being available.

[0004] Reasons for this could include, for example, the loss of driving pleasure or the joy of self-driving, or a lack of trust in the safety of the corresponding driver assistance system and loss of control. The corresponding benefits in terms of safety, efficiency, and comfort that could potentially be generated by using the assistance function are thus lost.

[0005] DE 10 2017 208 504 A1 discloses a driving system and a method for activating a driving function for automated driving. The method discloses that it facilitates the operation of a driving system that offers an automated driving function. The method provides for checking the availability of the corresponding automated driving function after the driver of the vehicle has entered a request to use an automated driving function and, if available, offering it to the driver. If the offer is available, the driver of the vehicle is particularly easily able to accept the offer, i.e., activate the automated driving function. For this purpose, it may be sufficient for the driver to simply take their hands off the steering wheel to activate the automated driving function.This eliminates the need for the driver to first manually check the general availability of the automated driving function while using their vehicle and then, if available, to activate it by manually entering an operating action.

[0006] Furthermore, US 2020 / 0 264 608 A1 discloses support for deciding on a driving mode for an at least partially automated vehicle. For this purpose, a road network on a digital road map is divided into a plurality of route sections. The drivers of a plurality of vehicles are each assigned to a driver profile. For each route section, a statistical determination is made, depending on the driver profile, as to which SAE automation level is used when driving the respective route section. When driving a respective route section, the vehicle then recommends the most frequently used automation level for the selected driver profile for the respective route section.

[0007] In addition, DE 10 2018 214 204 A1 discloses a method and a device for creating a driving mode recommendation for a vehicle depending on an expected driving speed that differs in a manual and an automated driving mode.

[0008] Furthermore, US 2019 / 0 064 802 A1 discloses driving an automated vehicle in a mixed mode. A monitoring device is capable of switching the vehicle's driving mode between a manual mode, an autonomous mode, or another predefined mode.

[0009] In addition, WO 2019 / 047 596 A1 discloses a method and a device for changing a driving mode using a machine learning model.

[0010] The present invention is based on the object of providing an improved vehicle with an at least partially automated operating mode and a method for issuing recommendations to a person driving the vehicle to take over vehicle control, with the aid of which the proportion of use of an at least partially automated operation of the vehicle during a journey is increased.

[0011] According to the invention, this object is achieved by a vehicle having the features of claim 1 and a method for issuing recommendations to a person driving the vehicle for taking over manual vehicle control by the person driving the vehicle or for taking over at least partially automated vehicle control by a driver assistance system having the features of claim 10. Advantageous embodiments and further developments emerge from the dependent claims. A vehicle with an at least partially automated operating mode has a recommendation system comprising a data collection module, a prediction module, and a recommendation module, wherein the data collection module is configured to collect vehicle data, surrounding data, and / or environmental data; the prediction module is configured to read in a driver profile from a set of driver profiles, wherein each driver profile comprises a machine learning model individually trained for the respective driver profile, which is configured to read in the vehicle data, surrounding data, and / or environmental data at least for a preceding route section and to output a predictive indication value as an output variable,wherein the predictive indication value is a numerical value; and the recommendation module is configured to compare the predictive indication value with an indication threshold and to initiate the output of a recommendation in the vehicle for a person driving the vehicle to assume manual control of the vehicle if the predictive indication value lies within a first range relative to the indication threshold, and to initiate the output of a recommendation in the vehicle for a driver assistance system to assume at least partially automated control of the vehicle if the predictive indication value lies within a second range relative to the indication threshold. According to the invention, the recommendation system further comprises a driver monitoring module.which is designed to monitor the person driving the vehicle with the aid of at least one sensor and to determine a current control behavior and / or a current driver state from sensor data generated by the sensor and to assign the person driving the vehicle to one of the driver profiles depending on the determined current control behavior and / or the current driver state.

[0012] The vehicle according to the invention is capable of providing the driver with a suggestion to assume manual control of the vehicle, taking into account their driving style, precisely when the driver, in accordance with their personal vehicle usage preferences, is likely to want to assume control of the vehicle themselves, and of providing a recommendation to allow a driver assistance system to assume control of the vehicle when the corresponding driver assistance function is available and the driver is unlikely to want to drive the vehicle themselves. In other words, the vehicle provides the driver with a recommendation to assume at least partially automated vehicle control when the driver is very likely to comply with this request.This increases the degree of utilization of the at least partially automated vehicle control system during a journey with the vehicle.

[0013] If a situation arises in which at least partially automated vehicle control is possible, the driver is advised to hand over vehicle control to the driver assistance system, thus preventing this function from remaining unused on a route on which at least partially automated vehicle control is possible. However, it is possible that the driver does not want to hand over control of the vehicle despite at least partially automated driving functions being available, for example because the driver is experiencing a lot of driving pleasure on that section of the route. In such a case, the recommendation to hand over vehicle control to the vehicle is omitted, which increases the convenience of the driver because no unnecessary recommendations are issued.It is also possible for at least partially automated vehicle control to be active and for the vehicle to nevertheless recommend that the driver assume manual control. Depending on the driver's personal preferences, the vehicle determines which sections of the route the driver is likely to want to drive the vehicle themselves and issues appropriate recommendations. This ensures that the driver drives the vehicle themselves when driving on certain sections of the route that are likely to increase driving pleasure. This gradually increases the driver's acceptance of using at least partially automated driving functions, so that they use them more often when the at least partially automated driving function is available.

[0014] For example, the vehicle will issue a recommendation to take over at least partially automated vehicle control if a monotonous and boring section of road awaits the driver during the journey, so that the driver is happy to hand over vehicle control to the at least partially automated driver assistance system. The vehicle reads the driver's preferences from the driver profile and thus also recognizes upcoming situations in which the driver would otherwise like to control their vehicle manually, but currently does not want to do so due to changed conditions, for example, because the traffic volume is too heavy and / or the weather is adverse and / or a telephone conversation is in progress and / or there are other passengers in the vehicle.The vehicle thus recognises a driving situation ahead in which driving safety can be increased by at least partially automated control compared to manual control, whereupon the driver transfers control of the vehicle to the driver assistance system, as this increases the driver's sense of safety.

[0015] The vehicle can exhibit any degree of automation. The vehicle can therefore be controlled at least partially automatically, which corresponds to assisted vehicle control by the driver assistance system. For example, the driver assistance system can assume lateral or longitudinal control, with the remaining longitudinal or lateral control being assumed by the driver. However, the vehicle can also be controlled highly automated, fully automated, or even autonomously. In autonomous operating mode, no intervention by the driver is necessary, so the vehicle can generally perform its driving task even when the driver is absent or distracted.

[0016] The data collection module collects vehicle data, surrounding data, and / or environmental data. Vehicle data includes information relating to the vehicle itself, acquired with the aid of sensors, such as a vehicle's measured speed, vehicle acceleration, vehicle position, steering angle, accelerator pedal position, or the like. Environmental data includes information relating to the vehicle's immediate surroundings, such as a route (e.g., determined by evaluating images generated by a surrounding camera), a self-measured road surface condition, a road width, a measured ambient temperature, traffic regulations indicated by traffic signs, construction sites, and other obstacles in the roadway area (e.g., also detected in the images generated by the surrounding camera), precipitation detected with the aid of a humidity sensor, or the like.Environmental data includes information relating to the vehicle's surroundings, which is obtained from outside the vehicle or read from a memory. This includes, for example, the route read from a digital road map as well as the traffic regulations applicable to the route, traffic information obtained from a traffic service, a weather report, and the like. Environmental data can be used to verify environmental data. In addition, the corresponding data for a future section of the route can also be estimated. For example, a route can be read from a digital road map or determined for the next 200 meters from camera images, and the steering angle, acceleration values, and speed expected when driving on the corresponding section of the route can be predicted.To process this “heterogeneous” data, the prediction module can also use methods for semantic networks such as knowledge graphs and / or ontologies.

[0017] The vehicle data, ambient data, and / or environmental data can influence the driver's steering behavior and are therefore taken into account when determining the recommendation to adopt manual or at least partially automated vehicle control. For example, a sporty driver may want to control the vehicle manually on a particularly winding road. However, if slippery conditions are to be expected, for example due to wet conditions or temperatures below freezing, the driver may desire at least partially automated vehicle control. Another example of a driving situation in which at least partially automated vehicle control is desired by the driver is a monotonous driving situation such as a traffic jam or slow-moving traffic, or driving on a particularly long stretch of road with a constant speed limit.Taking into account in particular the vehicle data, surrounding data and environmental data, a wide variety of driving situations can be recognized and differentiated in a particularly differentiated manner.

[0018] In abstract terms, the driver profiles store the relationships of how a person with a certain driving style would manually drive their vehicle depending on the boundary conditions describing a particular driving situation. The boundary conditions are described by the vehicle data, surroundings data and / or environmental data. The boundary conditions are then processed by a machine learning model trained to the respective driving style. The person driving the vehicle initially determines which driver profile they feel they belong to, for example when purchasing the vehicle or at the start of each journey. All of the driver profiles can be specified by a vehicle manufacturer, for example. A certain number of standard driver profiles can be predefined. Examples of driver profiles can be: sporty, forward-looking, safe, novice driver, no longer young or.physically impaired drivers or the like.

[0019] The machine learning model reads vehicle data, surrounding data, and / or environmental data as input data, such as the expected brake pedal position for the upcoming section of road, steering angle, acceleration values, position values, traffic jam reports, elevation profiles, route layouts, weather information, and the like. The predictive indicator value is output as the output variable. This corresponds to a numerical value within a predefined range, for example, between 0 and 1 or -1 and 1. Differently trained machine learning models are then maintained for different driver profiles.

[0020] To train the various machine learning models, different drivers are subjected to a driving task for different route sections and indicate for each section whether manual or at least partially automated vehicle control is preferred. This allows the different machine learning models to learn, depending on the specific route and the boundary conditions described by the vehicle data, surrounding data, and / or environmental data, how the vehicle should preferably be controlled based on the driving style associated with the driver profile. Accordingly, the different machine learning models then output predictive indicator values ​​appropriate for the respective route sections.

[0021] The predictive indication value is then compared by the recommendation module with the indication threshold value and, depending on the result, the recommendation for manual vehicle control or at least partially automated vehicle control is issued for the upcoming route section.

[0022] The indication threshold can be specified as a fixed value. For example, the indication threshold can be 0.5. The first range can then be a range between 0.5 and 1, and the second range a range between 0 and 0.5. The first or second range can also include the value “0.5.” For example, the first range lies in the value range from 0.5 to 1 and the second range in the value range from 0 to 0.499, or vice versa. The first and / or second range do not necessarily have to be above or below the indication threshold. The first and second ranges can also be understood as a ratio. For example, the first range can correspond to a range ≥ 1 and the second range to a range between 0 and 1. The first range can also correspond to a positive range and the second range to a negative range, or vice versa.

[0023] The recommendation to assume control of the vehicle can be issued in a variety of ways in the vehicle, for example acoustically, visually, and / or haptically. For example, the vehicle's steering wheel can vibrate when a new recommendation is issued. The recommendation can then be reproduced acoustically as text, for example as the following spoken message: "There is a winding section of road ahead. Manual vehicle control is recommended." The corresponding message, optionally supplemented by pictograms, symbols, images, animations, or the like, can also be shown on any display device in the vehicle, for example in the instrument cluster or on the head unit display.

[0024] The vehicle can determine the upcoming route section in various ways. For example, the vehicle may have a navigation system with a programmed route. The route is then divided into sections. The corresponding upcoming route sections are then derived from the route still to be covered by the vehicle. If route guidance is inactive, the so-called most probable path, which the navigation system determines based on historical data, can also be used. If route guidance is inactive, the vehicle can determine a section of road immediately ahead as the upcoming route section. For example, the next 100 meters of driving distance can be determined as the next upcoming route section.If the upcoming section of road includes intersections, exits, access roads, or similar features, the next section of road ahead can also extend to such a road element. The vehicle is also capable of determining a future section of road ahead without a navigation system. To do this, the vehicle can evaluate the vehicle data and, for example, determine the route itself. To do this, the vehicle can identify the immediately ahead route using depth information sensors such as a stereo camera, a LIDAR, an ultrasonic sensor, and / or a radar system.

[0025] By using the driver monitoring module, the driver's driving behavior is gradually recognized, and a driver profile appropriate to the corresponding driving behavior is selected by the vehicle, which then outputs different predictive indicator values. This further increases comfort for the driver, as the degree of correspondence between the driver's actual driving style and the driving style assumed by the vehicle is increased, enabling the recommendation system to issue even more accurate recommendations for taking over vehicle control. This increases the likelihood that the recommendations issued by the vehicle for taking over vehicle control will actually be implemented by the driver.

[0026] To capture the current steering behavior, the driver's steering behavior and acceleration and braking behavior can be analyzed. For this purpose, the steering angles and pedal positions adjusted during the journey are recorded and analyzed by the driver monitoring module.

[0027] The driver's state describes, in particular, the driver's emotional mood, attention, and / or level of cognitive load. For this purpose, vital parameters of the driver are monitored, such as pulse rate, respiratory rate, skin conductivity, skin temperature, gaze direction, blink rate, gaze direction change rate, and the like. For example, an increased pulse rate, increased skin conductivity, and / or body temperature may indicate increased concentration and / or the experience of driving pleasure.

[0028] For example, a driver who remains calm in a hectic traffic situation and / or is stressed in a monotonous traffic situation is assigned to a sporty driver profile, and a person who is excited in a hectic traffic situation and / or relaxed in a monotonous traffic situation is assigned to a comfortable driver profile.

[0029] The prediction module can further be configured to estimate a predicted control behavior and / or a predicted driver state for an upcoming route section, i.e., an expected current control behavior and / or driver state likely to occur on the upcoming route section. Thus, the predicted control behavior can be compared with the current control behavior and / or the predicted driver state with the current driver state in order to evaluate the prediction quality of the prediction module, in particular of a respective machine learning model underlying the corresponding driver profile.

[0030] To assign a suitable driver profile to the driver, the implementation behavior of the vehicle's recommendation to take over vehicle control can also be used. If the driver follows the vehicle's recommendation more often than a specified threshold, the vehicle assumes that an appropriate driver profile has been selected. If, however, the driver does not follow the vehicle's recommendation to take over vehicle control or follows it less often than the specified threshold, a new driver profile is preferentially selected.

[0031] A predefined "standard driver profile" can be adapted to the actual driving style of the driver over the vehicle's service life, depending on the driver's actual steering behavior and / or driver state, and its predictive accuracy can also be improved through continuous training. For this purpose, according to a further advantageous embodiment of the vehicle, the driver monitoring module is further configured to determine a current indication value from the current steering behavior and / or the current driver state, and the prediction module is further configured to read the current steering behavior, the current driver state, and / or the current indication value, thereby further training the machine learning model.This further improves the predictive ability of the machine learning model, so that the recommendations issued by the recommendation system are implemented even more frequently by the person driving the vehicle. As already mentioned, the data collection module determines the input variables for the machine learning model, i.e., vehicle data, surrounding data, and / or environmental data, at least for the next section of the route ahead. Machine learning methods can also be used for this purpose, particularly to estimate the predicted control behavior and / or driver state.

[0032] The predicted control behavior and / or the predicted driver state can also be used as input to the machine learning model to determine the predictive indication value. By comparing it with the current control behavior and / or the current driver state, the prediction module can then check how well it predicted the predicted control behavior and / or the predicted driver state. The machine learning model learns accordingly and improves its prediction accuracy.

[0033] Similarly, a machine learning model trained for the respective driver profile can also be integrated into the driver monitoring module to determine the current indication value from the current data. The current indication value output by the driver monitoring module can then also be considered by the prediction module to check the predictive accuracy of the predictive indication value and further train the machine learning model accordingly.

[0034] A further advantageous embodiment of the vehicle further provides that the prediction module is configured to further train the machine learning model, taking into account the actual implementation of the recommendation module's recommendation to assume vehicle control. A corresponding result indicating whether the recommendation to assume vehicle control manually or at least partially automatically has actually been implemented by the person driving the vehicle can be transmitted back to the prediction module as feedback from a control unit, thereby further improving the learning capability of the machine learning model.

[0035] According to a further advantageous embodiment of the vehicle, the recommendation system is configured to receive fleet data, wherein the fleet data comprises an aggregated set of user-specific control behavior and / or driver states of the users of a large number of fleet vehicles, and to derive driver profiles from the fleet data and / or to update an existing driver profile, wherein, within defined limits, similar control behavior and / or driver states are assigned to the same driver profile. By collecting and using the fleet data, the corresponding machine learning models of the respective driver profiles are trained using particularly comprehensive data sets. This makes it possible to develop particularly comprehensive and differentiated driver profiles, so that a particularly appropriate machine learning model can be developed for each person driving the vehicle.The individual vehicle users can be assigned to these. By using these large data sets, the individual driver profile-specific machine learning models are also trained particularly accurately. This means that the prediction accuracy of the respective machine learning models for the respective driver profiles can be increased, so that the respective recommendations issued by the recommendation system are actually implemented wherever possible.

[0036] In addition to the user-specific steering behavior, the respective user-specific driver state can also be taken into account. This allows the steering behavior to be linked to the driver state, so that a driver profile-specific driver state is also available for a specific section of the route or vehicle data, ambient data, and / or environmental data. On a particularly steep and winding route with frequently changing and sharp steering inputs as well as strong acceleration and deceleration behavior, for example, a first driver state can indicate high cognitive load and anxiety, and a second driver state can indicate medium cognitive load and enjoyment. Accordingly, the person with the first driver state can be assigned to an anxious driver profile, and a person with the second driver state to a sporty driver profile.

[0037] A further advantageous embodiment of the vehicle further provides that the fleet data further includes the implementation behavior of the recommendations made by the recommendation modules of the fleet vehicles by the users of the fleet vehicles, and the prediction module is further configured to further train the machine learning model, taking into account the implementation behavior of the users of the driver profile assigned to the person driving the vehicle. This can be understood as a filter, so that before the machine learning model is further trained, the corresponding input data used for further training has actually led to an improvement in the implementation of the issued recommendations for a large number of users of the fleet vehicles. This prevents a deterioration of the machine learning model, which could lead to a reduced implementation of the corresponding recommendations.For example, depending on the vehicle data, surrounding data and / or environmental data or the predicted control behavior or driver states derived from them, route sections can be determined in which at least a specified number of users actually implement the issued recommendations, for example, at least 75% of users.

[0038] According to a further advantageous embodiment of the vehicle, the machine learning model includes a neural network. With the help of a neural network—but also comparable machine learning methods—the machine learning model is particularly reliably capable of determining a suitable predictive indication value depending on the input parameters.

[0039] A further advantageous embodiment of the vehicle further provides that a vehicle control unit is configured to automatically implement the recommendation made by the recommendation system regarding the assumption of vehicle control by the vehicle driver or the driver assistance system. This further improves comfort for the vehicle driver. Thus, when a corresponding recommendation is issued, the vehicle driver does not have to enter a manual control action each time the recommendation is followed. The vehicle can therefore automatically implement the corresponding recommendation, i.e., activate or deactivate at least partially automated vehicle control. The vehicle driver can, for example, set a configuration menu so that the recommendation should be automatically adopted under certain conditions.In certain situations, it may still be necessary for the driver to proactively respond to the issued recommendation before the recommendation is actually implemented. For example, at least partially automated vehicle control can always be activated automatically, while manual control can only be assumed after confirmation by the driver.

[0040] Preferably, the recommendation module is further configured to adjust the level of the indication threshold depending on a driver profile read in by the prediction module, the vehicle data, ambient data, and / or environmental data. This enables a particularly simple, rapid, and flexible response to decide which vehicle operating mode should be recommended. If the indication threshold is initially 0.5, for example, it can be increased or decreased depending on the driver profile. For example, it can be reduced to 0.3 for the "Sporty" driver profile if the recommendation for manual vehicle control is issued for indication values ​​that are closer to 1, and it can be increased to 0.8, for example, for the "Safety" driver profile.

[0041] Accordingly, the indication threshold can also be adjusted depending on the vehicle data, surrounding data, and / or environmental data. For example, the vehicle can analyze a section of road ahead and set the indication threshold to 0.7 for a monotonous section and to 0.33 for an exciting section.

[0042] A method for issuing recommendations to a person driving a vehicle as described above for taking over manual vehicle control by the person driving the vehicle or for taking over at least partially automated vehicle control by a driver assistance system comprises the following steps: - Collecting vehicle data, ambient data and / or environmental data through a data collection module; - Reading in a driver profile from a set of driver profiles by a prediction module, wherein each driver profile comprises a machine learning model individually trained for the respective driver profile, which is configured to read in the vehicle data, ambient data and / or environmental data at least for a route section ahead and to output a predictive indication value as an output variable, wherein the predictive indication value is a numerical value; - Determination of the predictive indication value for the upcoming route section by the prediction module; - comparing the predictive indication value with an indication threshold and initiating the issuance of a recommendation for the person driving the vehicle to assume manual control of the vehicle if the predictive indication value lies in a first range compared to the indication threshold, and initiating the issuance of a recommendation for the person driving the vehicle to assume at least partially automated control of the vehicle by a driver assistance system in the vehicle if the predictive indication value lies in a second range compared to the indication threshold, by a recommendation module; - monitoring the person driving the vehicle by a driver monitoring module using at least one sensor; - Determining a current control behavior and / or a current driver state from sensor data generated by the sensor; and - Assigning the person driving the vehicle to one of the driver profiles depending on the determined current control behavior and / or the current driver state.

[0043] With the aid of the method according to the invention, recommendations are issued to the person driving the vehicle regarding manual or at least partially automated vehicle control. The output of the respective recommendation is tailored to the respective driver profile, so that a recommendation for manual vehicle control is issued precisely when there is a particularly high probability that the person driving the vehicle actually wants to control their vehicle manually. Accordingly, a recommendation for at least partially automated vehicle control is issued when the person driving the vehicle desires at least partially automated vehicle control or at least has no objections.This ensures improved acceptance of driver assistance systems that enable at least partially automated vehicle control and thus increases the proportion of use of at least partially automated vehicle control when carrying out a journey.

[0044] Further advantageous embodiments of the vehicle according to the invention also emerge from the exemplary embodiments which are described in more detail below with reference to the figures.

[0045] Showing: Fig. 1 is a schematic representation of a vehicle according to the invention; and Fig. 2 a schematic representation of the information transfer paths of the quantities read in and output by a machine learning model running on a prediction module.

[0046] Fig. 1 shows a schematic front view of a vehicle 1 according to the invention. The vehicle 1 comprises a recommendation system 2, comprising a data collection module 2.1, a prediction module 2.2, a recommendation module 2.3 and a driver monitoring module 2.4.

[0047] The vehicle 1 has an at least partially automated operating mode. Depending on various boundary conditions, the at least partially automated operating mode is available and allows a person driving the vehicle 1 to control the vehicle 1 at least partially automatically. For example, this can be an adaptive cruise control system, a lane keeping assistant, a parking assistant, or even an autopilot, i.e., an autonomous control of the vehicle 1. The vehicle 1 according to the invention is capable of issuing a recommendation to the person driving the vehicle 4 as to whether, on a section of road ahead during a journey with the vehicle 1, the vehicle should be manually controlled by the person driving the vehicle 4 or whether the vehicle should be at least partially automated by a driver assistance system.The vehicle 1 is oriented towards the preferences of the person driving the vehicle 4, so that a recommendation for manual vehicle control is issued precisely when there is a high probability that the person driving the vehicle 1 would like to control the vehicle 1 themselves and a recommendation for at least partially automated vehicle control is issued when there is a high probability that the at least partially automated vehicle control will contribute to increased user comfort for the person driving the vehicle 4.

[0048] For this purpose, the data collection module 2.1 collects vehicle data D-FZG, environmental data D-UMG and environmental data D-UMW. The vehicle data D-FZG is information recorded by the vehicle 1 itself and relating to the vehicle 1, such as travel speed, position or orientation in space, acceleration values, pedal position and the like. The environmental data D-UMG is information recorded by the vehicle 1 regarding the surroundings of the vehicle 1, such as the route, applicable traffic regulations, ambient temperature, existing precipitation or the like. To record the route, the vehicle 1 can scan its surroundings with the help of environmental sensors such as a mono or stereo camera, a LIDAR, ultrasonic sensors and / or a radar system and determine the route from the corresponding sensor data. The environmental data D-UMW includes information received from the vehicle 1 from outside orinformation concerning the environment read from a data storage device, such as a route read from a digital road map, traffic regulations stored in the digital road map, in particular for the route section ahead, traffic information obtained from a traffic service, weather reports obtained from a weather service and the like.

[0049] In addition, the driver monitoring module 2.4 is able to monitor the person driving the vehicle 4 with the help of various sensors and to record the current steering behavior D-SV-AKT as well as the current driver state D-ZUS-AKT. The current steering behavior D-SV-AKT includes, for example, the steering behavior and / or the acceleration or braking behavior of the person driving the vehicle 4, derived from a steering angle sensor and / or a pedal position sensor. The current driver state D-ZUS-AKT describes, for example, the emotional mood, attention, degree of cognitive load due to the driving task, or the like of the person driving the vehicle 4. For this purpose, the person driving the vehicle 4 can be monitored with the help of various sensors, whereby vital parameters can be recorded. The corresponding variables taken into account for assessing the driver state can then be derived from the vital parameters.The vital parameters include, for example, a pulse rate, an eyelid blink rate, a facial expression of the person driving the vehicle 4 derived from visual monitoring by image recognition algorithms, for example whether the person driving the vehicle 4 is smiling, a skin conductivity, a skin temperature, a direction of gaze, a frequency of changing the direction of gaze and the like.

[0050] Furthermore, the recommendation system 2 can receive fleet data D-FLO, for example, also via the data collection module 2.1. The fleet data D-FLO includes the driving behavior of a large number of different users of the vehicles 1 in a vehicle fleet and / or their driver states. From this driving behavior and / or driver states, various driver profiles can be derived, which are representative of a respective individual driving style and the emotions experienced in the process.

[0051] Before using their vehicle 1, the driver 4 selects which driver profile they feel they belong to, for example, a sporty or a comfortable driver profile. Depending on the selected driver profile, the prediction module 2.2 then determines for an upcoming section of road whether the driver 4 would likely prefer to drive the vehicle 1 manually or at least partially automatically. The prediction module 2.2 processes the corresponding vehicle data D-FZG, ambient data D-UMG, and / or environmental data D-UMW expected at least for the next upcoming section of road and is thus able to realistically estimate the driving situation that will arise on the upcoming section of road.A respective driver profile is representative of the expected manual control behavior of the person driving the vehicle 4 depending on the expected driving situation on the upcoming section of road. As a result, the prediction module 2.2 is able to estimate the manual control behavior expected of the person driving the vehicle 4 on the upcoming section of road. Additionally or alternatively, the prediction module 2.2 can also estimate an expected driver state. By comparing the expected manual control behavior and / or the expected driver state with the expected driving situation on the upcoming section of road, the prediction module 2.2 then estimates whether manual or at least partially automated vehicle control is likely to be desired. In order to carry out these method steps described in abstract terms, the prediction module 2.2 comprises a . Fig. 2, which is individually trained to the respective driver profile. The machine learning model 3 reads in at least the vehicle data D-FZG, environmental data D-UMG and / or environmental data D-UMW as input data. As an output variable, the machine learning model 3 provides a predictive indication value, which is a simple numerical value. This can be a natural number, integer, rational number or even a rounded irrational number. The predictive indication value can be assigned to a permissible value range, for example a value range from -1 to 1 or from 0 to 1 or the like. The predictive indication value can be understood as a value that describes the affinity of the person driving the vehicle 4 to controlling the vehicle 1 through manual operation or through operation at least partially automated by a driver assistance system.

[0052] The driver profile-specific machine learning models 3 are initially trained by the vehicle manufacturer. For this purpose, predefined driving situations are driven by different people with different driving styles, and then it is checked whether manual or at least partially automated vehicle control is desired for the respective driving situation. The machine learning model 3 comprises, in particular, an artificial neural network. Through the learning process, the individual neurons of the artificial neural network form corresponding connections, allowing a predictive indication value (IND-PRÄ) appropriate to the driving style and driving situation to be determined.

[0053] The predictive indication value IND-PRÄ is then read by the recommendation module 2.3 and compared with an indication threshold value IND-SW. For example, the value range of the predictive indication value IND-PRÄ can be from -1 to 1, and the indication threshold value IND-SW can be 0. If the predictive indication value IND-PRÄ is then in a first range relative to the indication threshold value IND-SW, for example, in the range between 0 and 1, a recommendation to assume manual vehicle control is issued in vehicle 1. If, however, the predictive indication value IND-PRÄ is in a second range relative to the indication threshold value IND-SW, for example, between -1 and 0, a recommendation to assume at least partially automated vehicle control by a driver assistance system is issued in vehicle 1.If the predictive indication value IND-PRÄ and the indication threshold value IND-SW have the same value, a recommendation to take over vehicle control may not be issued, for example, or a preset standard recommendation may be issued.

[0054] Fig. Figure 2 shows a more detailed representation of the information transfer paths of the quantities read in and output by the machine learning model 3.

[0055] The recommendation issued in vehicle 1 can be haptic, acoustic, and / or visual. Output occurs on a corresponding output device 6. This can be, for example, an actuator, a loudspeaker, or a display device.

[0056] With the help of the driver monitoring module 2.4, a current control behavior D-SV-AKT of the vehicle driver 4 and / or a current driver state D-ZUS-AKT can be determined. A correspondingly trained machine learning model 3 can also be integrated into the driver monitoring module 2.4, which determines a current indication value IND-AKT from the current control behavior D-SV-AKT and / or the current driver state D-ZUS-AKT. These variables can be read in by the prediction module 2.2 to further train the machine learning model 3 corresponding to the driver profile, thereby improving the prediction quality for the actual affinity of the vehicle driver 4. Additionally or alternatively, a control unit 7 can determine the implementation behavior of the vehicle driver 4, i.e., whether the vehicle driver 4 has followed the recommendation issued by the recommendation system 2, and transmit it as feedback to the prediction module 2.2.The prediction module 2.2 can also use this information to further train the corresponding machine learning model 3.

[0057] Furthermore, the driver monitoring module 2.4 is capable of analyzing the driving behavior of the person driving the vehicle 4 by recording the current steering behavior D-SV-AKT and / or the current driver state D-ZUS-AKT, thereby reassigning the person driving the vehicle 4 to a corresponding driver profile. While the person driving the vehicle 4 can estimate which driver profile is most appropriate for them, the driver monitoring module 2.4 reviews this estimate and thus selects a driver profile that is even more appropriate for the actual driving behavior of the person driving the vehicle 4. Through the continuous further training of the machine learning model 3, the corresponding driver profile of the person driving the vehicle 4 is tailored even more accurately to them.Since this is also performed for all vehicles 1 in a fleet, a large number of particularly realistic driver profiles can be defined and distributed to the individual vehicles 1 in the fleet. This gradually improves the predictive quality of the machine learning models 3 assigned to the various driver profiles, so that the recommendations issued in the vehicles 1 are very likely to actually be implemented by the respective vehicle drivers 4.

Claims

[1] Vehicle (1) with an at least partially automated operating mode, where a recommendation system (2) comprising a data collection module (2.1), a prediction module (2.2) and a recommendation module (2.3), wherein the data collection module (2.1) is configured to collect vehicle data (D-FZG), environmental data (D-UMG) and / or environmental data (D-UMW); the prediction module (2.2) is configured to read in a driver profile from a set of driver profiles, wherein each driver profile comprises a machine learning model (3) or heuristic model individually trained for the respective driver profile, which is configured to read in the vehicle data (D-FZG), surrounding data (D-UMG) and / or environmental data (D-UMW) at least for a preceding route section and to output a predictive indication value (IND-PRÄ) as an output variable, wherein the predictive indication value (IND-PRÄ) is a numerical value; and the recommendation module (2.3) is configured to compare the predictive indication value (IND-PRÄ) with an indication threshold value (IND-SW) and to initiate the output of a recommendation in the vehicle (1) for a person driving the vehicle (4) to assume manual control of the vehicle if the predictive indication value (IND-PRÄ) lies in a first range compared to the indication threshold value (IND-SW), and to initiate the output of a recommendation in the vehicle (1) for a driver assistance system to assume at least partially automated control of the vehicle if the predictive indication value (IND-PRÄ) lies in a second range compared to the indication threshold value (IND-SW), characterized by , that the recommendation system (2) further comprises a driver monitoring module (2.4) which is designed to monitor the person driving the vehicle (4) with the aid of at least one sensor and to determine a current control behavior (D-SV-AKT) and / or a current driver state (D-ZUS-AKT) from sensor data generated by the sensor and to assign the person driving the vehicle (4) to one of the driver profiles depending on the determined current control behavior (D-SV-AKT) and / or the current driver state (D-ZUS-AKT). [2] Vehicle (1) according to claim 1, characterized bythat the driver monitoring module (2.4) is further configured to determine a current indication value (IND-AKT) from the current control behavior (D-SV-AKT) and / or the current driver state (D-ZUS-AKT) and the prediction module (2.2) is further configured to read in the current control behavior (D-SV-AKT), the current driver state (D-ZUS-AKT) and / or the current indication value (IND-AKT) and thereby further optimize the machine learning model (3) and / or the heuristic method [3] Vehicle (1) according to claim 1 or 2, characterized by that the prediction module (2.2) is further configured to further train the machine learning model (3) and / or the heuristic model taking into account an actual implementation of the recommendation made by the recommendation module (2.3) to take over vehicle control. [4] Vehicle (1) according to one of claims 1 to 3, characterized by , that the recommendation system (2) is configured to receive fleet data (D-FLO), wherein the fleet data (D-FLO) comprises an aggregated set of user-specific control behavior and / or driver states of users of a plurality of fleet vehicles, and to derive driver profiles from the fleet data (D-FLO) and / or to update an existing driver profile, whereby similar control behavior and / or driver states are assigned to the same driver profile within defined limits. [5] Vehicle (1) according to claim 4, characterized bythat the fleet data (D-FLO) further comprise the implementation behaviour of the recommendations made by the recommendation modules (2.3) of the fleet vehicles by the users of the fleet vehicles and the prediction module (2.2) is further configured to further optimise the machine learning model (3) and / or the heuristic model taking into account the implementation behaviour of the users of the driver profile assigned to the person driving the vehicle (4). [6] Vehicle (1) according to one of claims 1 to 5, characterized by that the machine learning model (3) comprises a neural network and / or was initially assigned on the basis of heuristic functions. [7] Vehicle (1) according to one of claims 1 to 6, characterized by that a vehicle control unit (5) is configured to automatically implement the recommendation made by the recommendation system (2) for the person driving the vehicle (4) or the driver assistance system to take over vehicle control. [8] Vehicle (1) according to one of claims 1 to 7, characterized by that the recommendation module (2.3) is further configured to adapt the level of the indication threshold value (IND-SW) depending on a driver profile read in by the prediction module (2.2), the vehicle data (D-FZG), the surrounding data (D-UMG) and / or the environmental data (D-UMW). [9] Method for issuing recommendations to a person driving a vehicle (4) of a vehicle (1) according to one of claims 1 to 8 for taking over manual vehicle control by the person driving the vehicle (4) or for taking over at least partially automated vehicle control by a driver assistance system, comprising the steps: - Collection of vehicle data (D-FZG), ambient data (D-UMG) and / or environmental data (D-UMW) by a data collection module (2.1); - Reading in a driver profile from a set of driver profiles by a prediction module (2.2), wherein each driver profile comprises a machine learning model (3) or heuristic model individually trained for the respective driver profile, which is set up to read in the vehicle data (D-FZG), surroundings data (D-UMG) and / or environmental data (D-UMW) at least for a route section ahead and to output a predictive indication value (IND-PRÄ) as an output variable, wherein the predictive indication value (IND-PRÄ) is a numerical value; - Determination of the predictive indication value (IND-PRÄ) for the upcoming route section by the prediction module (2.2); - comparing the predictive indication value (IND-PRÄ) with an indication threshold value (IND-SW) and initiating the output of a recommendation for the person driving the vehicle (4) in the vehicle (1) to assume manual vehicle control if the predictive indication value (IND-PRÄ) lies in a first range compared to the indication threshold value (IND-SW), and initiating the output of a recommendation for the person driving the vehicle (4) to assume at least partially automated vehicle control by a driver assistance system in the vehicle (1) if the predictive indication value (IND-PRÄ) lies in a second range compared to the indication threshold value (IND-SW), by a recommendation module (2.3); - monitoring the person driving the vehicle (4) by a driver monitoring module (2.4) with the aid of at least one sensor; - Determining a current control behavior (D-SV-AKT) and / or a current driver state (D-ZUS-AKT) from sensor data generated by the sensor; and - Assigning the person driving the vehicle (4) to one of the driver profiles depending on the determined current control behavior (D-SV-AKT) and / or the current driver state (D-ZUS-AKT).

Images