Segment-based driver analysis and individualized driver assistance

Abstract

Method for steering a vehicle over a route to be travelled (2) by - Driving a test track (1) by a driver, - Acquisition of geometric test data of the test track (1), - Recording of driving dynamics test data from the driver driving on the test track (1), - Dividing a test data space formed from the geometric test data into clusters by automatically dividing the driven test track into segments, - Defining vehicle dynamics data for each cluster based on the driver's vehicle dynamics test data that can be assigned to a respective cluster, - Assigning route segments (5) of the route to be traveled (2) to the clusters based on geometric data of the route to be traveled and - Control at least one vehicle component of the vehicle in one of the track sections (5) according to the specified vehicle dynamics data of the respective cluster.

Description

[0001] The present invention relates to a method for controlling a vehicle along a route. Furthermore, the present invention relates to a corresponding driver assistance system and a motor vehicle with such a driver assistance system.

[0002] Driver assistance systems and driving functions are generally not very individualized and therefore offer the driver little scope for an emotional driving experience, or generate low acceptance due to interventions that are not easily understood. Individualized driver assistance systems, on the other hand, enable improved customer acceptance and thus offer added value and increased motivation to configure the vehicle accordingly.

[0003] In established methods, a vehicle dynamics analysis of a driver's behavior and driving skills is usually performed by assigning the driver to individual classes, such as "sporty," "cautious," etc. This assignment can be done, for example, using acceleration diagrams, in which longitudinal and lateral acceleration are plotted two-dimensionally. This allows for a rough description of the limits of driving behavior and skill, but detailed preferences in vehicle handling cannot be taken into account.

[0004] Document US 2021 / 0012590A1 discloses a monitoring device for providing vehicle telemetry data. The monitoring device includes a sensor for detecting vibrations caused by vehicle and engine movements in vehicle parts in order to generate corresponding vibration data. Characteristics of the vehicle or engine can be extracted from the vibration data.

[0005] Furthermore, publication WO 2019 / 134 110 A1 discloses a method and a system for autonomous driving. A driving situation map is generated from sensor data. A deep learning algorithm is trained to generate a driving command based on this driving situation map.

[0006] German patent application DE 10 2017 209 258 A1 discloses a method and a device for monitoring the driving stability of a vehicle on a planned route. The current speed, steering angle, and position of the vehicle are determined. Furthermore, the driver type is identified. Based on this data, a speed signal and a steering angle signal are predicted.

[0007] Document US 2016 / 0026182 A1 discloses personalized driving capabilities for autonomous vehicles. This involves loading at least part of a driving profile onto the vehicle, which is then used to control it.

[0008] The publication DE 10 2019 219 534 A1 discloses a computer-implemented method for using machine learning to determine control parameters for a control system, in particular for a motor vehicle, especially for controlling the driving operation of the motor vehicle, wherein the method comprises: providing a set of driving trajectories; deriving reward functions from the driving trajectories using a method of inverse reinforcement learning; deriving driver-type-specific clusters based on the reward functions; determining control parameters for each driver-type-specific cluster.

[0009] Document WO 2016 / 109540A1 concerns a method for the autonomous operation of a vehicle using personalized driving profiles. Data relating to the vehicle's movement are recorded by sensors and used for autonomous driving. Machine learning and clustering can be employed in this process.

[0010] Document DE 10 2019 205 892 A1 describes a method for operating a motor vehicle and a motor vehicle designed to carry out such a method. The motor vehicle comprises at least one vehicle system designed to control the motor vehicle at least semi-autonomously, wherein a predefined driving behavior mode is activated in the vehicle system. This mode includes at least one control parameter that defines driving behavior in a predefined geographical area and is taken into account during the at least semi-autonomous control of the motor vehicle. The system considers the current geographical position of the motor vehicle, determines the predefined geographical area in which this current geographical position is located, and determines which predefined driving behavior mode is assigned to this determined geographical area.The detected driving behavior mode is activated in the vehicle system, whereupon the motor vehicle is controlled at least semi-autonomously, taking into account at least one control parameter of the activated driving behavior mode.

[0011] The publication DE 10 2016 117 136 A1 relates to a method for determining the driving behavior of a driver of a motor vehicle, comprising the steps of: determining a plurality of characteristics during operation of the motor vehicle, wherein the characteristics characterize the individual driving behavior of the driver, providing classifiers for assigning the characteristics to at least two predetermined driver profiles, assigning the characteristics to the at least two predetermined driver profiles using the provided classifiers, and selecting one of the at least two predetermined driver profiles based on the assignment of the characteristics to the at least two predetermined driver profiles.

[0012] German patent application DE 10 2009 034 096 A1 discloses an adaptive vehicle control system that classifies a driver's driving style based on characteristic cornering maneuvers and road and traffic conditions. The system includes several vehicle sensors that detect various vehicle parameters. A maneuver identification processor receives the sensor signals to identify a characteristic vehicle maneuver and provides a maneuver identification signal for that maneuver. The system also includes a traffic and road condition detection processor that receives the sensor signals and provides traffic condition signals and road condition signals to identify traffic conditions.A style characterization processor receives the maneuver identifier signals, sensor signals from the vehicle sensors, and the traffic and road condition signals, and classifies a driving style based on the signals to classify the style of the driver operating the vehicle.

[0013] Document DE 10 2016 216 335 A1 discloses a system for generating at least one second trajectory for a first segment of a road, comprising a first interface for receiving first data representing at least one first trajectory. The first data were acquired while the first segment of the road was being traveled by at least one vehicle driven by a human. The first interface is also configured to receive second data representing environmental conditions at the time the first trajectory was acquired, and third data representing vehicle-related characteristics present at the time the first trajectory was acquired.The system also includes a first data processing module, which performs clustering of several first trajectories based on associated second and / or third data, a database for retrievable storage of the clustering results, and a second interface for receiving a request to transmit a second trajectory and for the corresponding transmission of the requested second trajectory.

[0014] Document DE 10 2018 001 634 A1 discloses a method for operating a vehicle, wherein, during at least one test drive of a test vehicle of the same vehicle type as the vehicle, at least longitudinal and lateral dynamics of the test vehicle are recorded and stored. Based on the recorded test vehicle sensor data, at least one set of test vehicle parameters for optimizing the driving behavior of the test vehicle is determined and stored along with the test vehicle sensor data. During use of the vehicle, at least longitudinal and lateral dynamics of the vehicle are recorded, and the vehicle's driving behavior is analyzed and classified based on the recorded sensor data.Depending on the results of the analysis and classification of driving behavior, a vehicle user is given recommendations for adjusting vehicle parameters influencing the vehicle's driving behavior, derived from at least one stored set of test vehicle parameters.

[0015] The object of the present invention is to propose a driver assistance system that enables more individualized support for the driver.

[0016] According to the invention, this problem is solved by the subject matter of the independent claims. Advantageous embodiments of the invention are set forth in the dependent claims.

[0017] According to the present invention, a method for controlling a vehicle along a route is provided. Such a method is implemented, for example, in a driver assistance system. However, the method can also be used to control a vehicle autonomously. In one case, the driver receives individualized support tailored to the driver, and in the other case, the driver is completely relieved of control by having the vehicle driven individually according to the driver's habits.

[0018] First, a driver drives a test track. Subsequently, this driver can also drive on other test tracks to gather relevant test data. The test track(s) should ideally have as many different features as possible to collect a wide variety of test data.

[0019] Geometric test data is collected from the test track. This data collection can be performed using the vehicle's sensors. As the vehicle traverses the test track, these sensors record geometric data such as track width, curvature, and similar characteristics. Alternatively, this geometric data can be provided by a database and loaded into the vehicle as needed. This geometric test data forms the basis for classifying future routes or track sections.

[0020] When the driver traverses the test track, driving dynamics test data is recorded. This test data includes, for example, accelerations, speeds, and similar parameters. Naturally, this driving dynamics test data is linked to the geometric test data. This is the only way to individually determine driver behavior, for example, when cornering. The driving dynamics test data is acquired and stored by the vehicle's sensors. If necessary, this driving dynamics test data, along with the geometric test data, is transmitted to an external processing unit for processing, and the processed data is then sent back to the vehicle. The driving dynamics test data provides information about the driver's behavior in specific situations or on specific sections of the track.

[0021] The collected or partially provided test data, i.e., the geometric test data and, if applicable, other objective situational data, typically form a point cloud in a test data space, which is usually multidimensional. Each parameter of the geometric test data and the other situational data constitutes its own dimension in the test data space. The test data space or the point cloud is then divided into clusters. A cluster analysis can be performed to discover similarity structures in the dataset. The cluster analysis is based on objectively measurable quantities such as road width, curvature, incline, gradient, temperature, traffic volume, weather (wet, dry, etc.), surface (asphalt, cobblestones), and reference values ​​from a planning process (speeds, accelerations, yaw angles, etc.), but not on the subjective characteristics of a driver.Clustering serves as an automated subdivision of driven routes into scenarios / segments. It is intended as a tool to objectively provide areas for extracting driving style. In this way, the test data space or test data can be divided into a large number of clusters, which can then serve as the basis for subsequent classification.

[0022] In a further step, the vehicle dynamics data for each cluster are defined based on the respective cluster's vehicle dynamics test data. This means that the clustering process has identified scenarios in which a driver is highly likely to behave in a reproducible manner, and the driver's behavior is now determined and subsequently applied based on this data. For example, when negotiating a left-hand bend on a road with a hard shoulder, the driver tends to drive closer to the outer edge of the lane than to the center line. This has been demonstrated in several test drives. On average, the driver maintains a distance of one meter from the outer edge of the lane in such situations. Therefore, this distance to the outer edge of the lane is defined as a dynamic data set for precisely this cluster.If the vehicle is driving autonomously or with driver assistance, it will attempt to maintain precisely this individual distance of one meter to the outer edge of the road when negotiating a left turn. This allows the driver to be supported in exactly the same way they would decide for themselves.

[0023] In a further step, segments of the route to be driven are assigned to clusters based on geometric data of the route. In other words, a new, unknown route that the driver intends to drive is automatically divided into segments. Each segment, i.e., each section of the route, is assigned to a cluster. This assignment is based, among other things, on geometric data available prior to the route and, if necessary, on further data describing specific scenarios (route width, curvature, slope, gradient, temperature, traffic volume, weather, surface conditions, planning reference values, etc.). For example, the geometric data might be obtained from a navigation system or an external vehicle database. The geometric data of the route, such as coordinates, curvature, and route width, can be used to assign each point on the route to a cluster.For example, a centroid method can be used to assign each point along the route to a specific cluster. The point is assigned to the cluster to which it is closest in the data space. In this way, the entire route can be divided into sections or segments, each assigned to one of the clusters.

[0024] In a final step of the process, at least one vehicle component is controlled within a specific section of the route according to the defined vehicle dynamics data of the respective cluster. This means that at least one vehicle component is controlled accordingly, either providing driver assistance or enabling autonomous vehicle control. Because the route section to be traversed is uniquely classified, i.e., assigned to a cluster, precisely those defined vehicle dynamics data of the respective cluster can be used for control. For example, if the route section falls into the "straight section" cluster and the driver has consistently driven such "straight sections" at a speed of 90 km / h during test drives, then, with individual driver assistance, the vehicle will also be controlled to maintain a speed of 90 km / h on the "straight section."

[0025] Previously, there was only a basic driving profile selection. This optional feature allows you to always choose the appropriate driving profile: for example, Normal, Eco, Sport, Comfort, or Individual. The engine and transmission settings, as well as the behavior of certain driver assistance systems, are then adapted to the driving conditions. However, these settings do not learn based on driving scenarios and do not specifically adjust the assistance according to the respective driving situation. Furthermore, they do not take into account individualized driving styles that consider factors such as lane choice, braking points, etc.

[0026] However, the solution presented above, according to the invention, makes it possible to selectively extract features that describe customer behavior based on scenarios and implement them in an individualized assistance system. This offers the advantage of potentially higher acceptance of the driving dynamics support. Additionally, the function could be offered as an optional feature to an existing function, thereby increasing the product's added value without significant additional costs.

[0027] In an advantageous embodiment of the method according to the invention, the geometric test data includes position data, curvature, track width, combinations of track segment types (e.g., a curve followed by a straight section, etc.), and / or the number of lanes of the test track. This means that the geometry of the test track is sufficiently known. Further information can be obtained from this geometric test data, such as the length of a straight section, the length of a curve, the gradient, and the like. Furthermore, this test data also makes it possible to analyze the interaction of track segments.

[0028] For example, a driver will approach a left-hand bend differently if the preceding section of road was straight or a right-hand bend. Similarly, accelerating out of a bend will be affected if the road immediately follows a straight section rather than another bend. The number of lanes also influences driving behavior. Generally, drivers travel faster on a two-lane road than on a single-lane one. All these geometric parameters typically influence an individual driver's behavior and are therefore primarily used for cluster analysis.

[0029] According to another embodiment, additional situational data regarding weather, time of day, traffic density, and / or surroundings can be recorded while driving on the test track. This situational data is incorporated into the test data space and the clustering process. Current situational data for the route being driven is recorded, and the assignment of route segments to clusters also depends on this current situational data. This means that not only geometric data of the test track(s) can be used for clustering, but also other data that can influence a driver's behavior. Such situational data describes the situation or scenario in which the upcoming journey is embedded. Specifically, the weather can play a role if the road surface is wet or snow-covered.Furthermore, the time of day can also play a role, for example, in predicting traffic jams. Traffic density can also be determined via additional information channels such as radio and the like. In addition, the surrounding environment can be significant for the driving situation. Driving behavior changes, for instance, if the route is not in open countryside but in a tunnel. Furthermore, situational data relating to the environment also includes, for example, traffic signs at the roadside. For example, a driver in an urban area with a speed limit of 50 km / h will typically drive at 55 km / h, while another driver in the same situation will usually drive at exactly 50 km / h.This situation-specific individual behavior can also be taken into account by recording the driving dynamics test data (here the speed) for the "local area" section of the route and using it for individual control.

[0030] In general, the situational data mentioned above can be considered as separate dimensions within the test data space. This allows for the creation of specific scenarios or clusters, such as "cornering on a wet road," "cornering in a built-up area," "cornering on a rural road," and the like. For the transfer from the test track to the actual route, this means that relevant current situational data for driving on the route can / must be recorded or acquired. For example, if a weather forecast or a rain sensor indicates that it is currently raining, the algorithm can access cluster elements that contain the situation "rain" or "wet road." This allows the road segments to be assigned to situation-specific clusters based on the current conditions.Accordingly, the vehicle can be controlled using the driving dynamics data from these clusters.

[0031] In another embodiment, the acquisition of driving dynamics data for multiple driving modes and the division of the test data space into clusters can also be performed depending on the driving modes. For the route to be driven, one of the driving modes is selected, and the route segments are assigned to clusters based on the selected driving mode. In this case, the driver can choose their driving style, such as Sport, Comfort, Eco, etc., while the algorithm still provides driver-specific assistance or takes over control. For this to work, the driver must specify their preferred driving mode as an additional parameter during the test drives. For example, they might want to complete the test route in the shortest possible time, i.e., in a "sporty" mode.In another scenario, the driver wants to drive safely or comfortably on the test track and enters this as a parameter during the test drive. In the application case for a future route, the algorithm will then, when the corresponding driving mode is selected, divide the route segments into clusters of that specific driving mode and control the vehicle accordingly.

[0032] Furthermore, the vehicle dynamics data may include lane position, safety distance, acceleration, speed, and / or jerk. This means that the vehicle dynamics data can encompass one or more parameters, both during test drives and for controlling a future or current route. One of these parameters is lane position, which a driver typically chooses individually when driving a route. Even with the same geometry, lane position can vary depending on the situation. A similar principle applies to the safety distance to the vehicle ahead. Each driver chooses the appropriate safety distance individually, depending on the situation. Older drivers, in particular, tend to maintain a greater safety distance than younger drivers.This driver-specific safety distance varies again depending on the situation. Other parameters such as acceleration, speed, and jerk are also generally very driver-specific. For example, every driver chooses their own acceleration when starting from a traffic light. Drivers are more likely to accept driver assistance systems that precisely match this individual acceleration.

[0033] In another embodiment, it can be provided that when assigning route segments to clusters, the geometric data of the route and / or the current situational data are individually weighted. For example, it might be less relevant for a driver to consider the time of day when controlling the vehicle than the surroundings of the route currently being traveled. This avoids the simple averaging of numerous influencing parameters, as is currently the case with conventional systems that, for example, only offer "sporty" driving modes or similar settings.

[0034] In a further advantageous embodiment, the assignment of route segments to clusters is based on machine learning. This machine learning can, in particular, take place unsupervised. In a specific implementation of the machine learning, a neural network is used. Such a network can, for example, be trained with the test data and used to classify the route segments.

[0035] Furthermore, the test data space can also contain vehicle data, and the control of at least one vehicle component can be vehicle-specific. This means that the driver can obtain vehicle dynamics test data specific to a particular vehicle. If the driver drives the same test track in a different vehicle, the vehicle dynamics test data may differ. Therefore, it can be advantageous to store vehicle data, and especially the vehicle type, along with the test data. This allows the driver to use their test database for multiple vehicles and, for example, select a specific vehicle type for the current control operation.

[0036] Furthermore, the aforementioned problem is solved according to the invention by a driver assistance system for controlling a vehicle on a route to be driven, comprising a detection device for capturing geometric test data of the test track and for capturing driving dynamics test data when a driver drives on a test track, a classification device for dividing a test data space formed by the geometric test data into clusters, and a data processing device for determining driving dynamics data for each cluster based on those driving dynamics test data that can be assigned to a respective cluster.The classification system is designed to assign sections of the route to be driven to clusters based on geometric data of the route, and the driver assistance system includes a control unit for controlling at least one vehicle component in one of the route sections according to the defined driving dynamics data of the respective cluster. This means that the clustering is based on objectively verifiable parameters such as road width, curvature, incline, gradient, temperature, traffic volume, weather (wet, dry, etc.), road surface (asphalt, cobblestones), and planning reference parameters (speeds, accelerations, yaw angles, etc.), but not on the basis of a driver's subjective perceptions.

[0037] Such a driver assistance system therefore has a data acquisition unit, which in turn includes, for example, appropriate sensors. Furthermore, the data acquisition unit can also include suitable interfaces for recording test data. The classification unit of the driver assistance system can be based on vector machines, neural networks, or similar technologies. In any case, the classification unit for automated classification has a processor and corresponding memory modules. Likewise, the data processing unit of the driver assistance system is equipped with a corresponding processor and memory modules. The classification unit and the data processing unit may be implemented in a single unit. In addition, the driver assistance system has a control unit with which vehicle components can be controlled. Appropriate drivers may be required for this purpose.

[0038] The driver assistance system is preferably capable of executing the procedure described above and its further developments. This results in the same advantages for the driver assistance system as for the corresponding procedures.

[0039] Furthermore, a motor vehicle can be equipped with the aforementioned driver assistance system. This allows the vehicle to be individually controlled or supported according to the specific characteristics of the driver.

[0040] The invention also includes combinations of the features of the described embodiments.

[0041] The following describes exemplary embodiments of the invention. This is illustrated by: Fig. 1 a diagram of two routes; Fig. 2 an acceleration diagram for one of the two routes from Fig. 1 for a specific driver; Fig. 3 an acceleration diagram of the other route from Fig. 1 for the same driver; Fig. 4 one of the two routes in segmented form; and Fig. 5 the other of the two routes in segmented form.

[0042] The embodiments described below are preferred embodiments of the invention. In these embodiments, the described components each represent individual features of the invention that can be considered independently of one another. Each of these features further develops the invention independently and can therefore be considered part of the invention individually or in a combination other than that shown. Furthermore, the described embodiments can also be supplemented by other features of the invention already described.

[0043] In the figures, functionally identical elements are each provided with the same reference symbols.

[0044] The present invention aims to enable a driver and a vehicle to perform a driving task in symbiosis. For example, the driver should be granted as much freedom as possible, while the assistance system monitors the action and state space in a task-specific manner and intervenes as needed.

[0045] For this purpose, driver-specific characteristics and attributes should be identified as precisely and situationally as possible, which can be identified according to a given task and made available for a future assistance system.

[0046] As mentioned at the beginning, the classification of drivers into categories, e.g., sporty, cautious, is known from current technology. However, this classification is very rudimentary, and there is a need for a more precise, individualized representation of the driver in an assistance system.

[0047] It is therefore proposed that the driver be analyzed in geometrically similar route segments. Driver-specific characteristics can be derived from the difference between the actual driving behavior and a planned, objectively describable trajectory, and these characteristics can then be translated into objectively describable classes with respect to the respective route segment.

[0048] For example, preferred driving lines, individually selected safety distances, accelerations, speeds, and jerks, based on the analyzed road geometry, are intended to provide a driver-specific "fingerprint" that allows for local mapping and goes beyond classification into global categories. Furthermore, unknown road segments can be evaluated using this road segment analysis. This extrapolation allows future driving preferences to be mapped onto unknown road segments, enabling tailored support.

[0049] In one implementation, driver data is recorded in different modes (e.g., Sport, Comfort, Eco, Race). However, it is also possible that only a single mode is recorded and used as the data basis.

[0050] Specifically, track data and boundary data (data from the track's or vehicle's environment) can be stored. The data can be clustered according to factors such as track geometry (curvature), track width, reference speed, reference acceleration, and the future track alignment. Clustering can also be based on a selection of these parameters or on additional parameters.

[0051] The driver's behavior can now be analyzed within these clusters. From this, key performance indicators (KPIs) describing driving behavior or style can be derived in relation to the geometric segments. For a support algorithm, unknown route data can be analyzed according to the formed clusters, and driving behavior in these segments can be predicted.

[0052] This allows a driver fingerprint to be created, which can be passed to an "ADAS" (Advanced Driver Assistance System) for an individually adapted support function.

[0053] In connection with the Fig. Sections 1 to 5 now explain a concrete example of how a method according to the invention for controlling a vehicle on a track can be implemented. First, a driver drives a test track 1. In Fig. This test track 1 is represented in a Cartesian coordinate system with x-coordinates and y-coordinates. Furthermore, in Fig. A second track, 2, is also shown, representing the route to be driven. The vehicle, or rather the driver assistance system, is to be "trained" using test track 1, and based on the data obtained, track 2 is to be driven assistively or autonomously.

[0054] During the test drive on test track 1, geometric test data for test track 1 is recorded. However, this geometric test data can also be recorded before the test drive, for example, by providing the test data to the control system from a database.

[0055] During the test drive on test track 1, driving dynamics test data is also recorded from the driver. This driving dynamics test data characterizes the driver. It represents the aforementioned driver-specific fingerprint and relates, for example, to speeds, accelerations, safety distances, and the like in relation to the geometric test data of test track 1.

[0056] The Fig. 2 and Fig. Figure 3 shows acceleration diagrams that can be achieved, for example, on tracks 1 and 2 with one and the same driver. Specifically, the acceleration a is shown in both figures. y in the y-direction to the right and the acceleration a xThe diagram shows the x-axis upwards. Driving along tracks 1 and 2 generates numerous acceleration measurement points, which are depicted in the figures. These result in typical point clouds that characterize the driver's behavior with respect to acceleration. The two point clouds are intended to demonstrate that a two-dimensional data space can be obtained with respect to driver behavior. Additionally or alternatively, other data mentioned above, such as acceleration, track width, and the like, can also be recorded on the tracks. Accordingly, the recorded data results in a multidimensional data space or a multidimensional point cloud.

[0057] In the present example, this results in Fig. In the middle at the top, there is an area 3 containing data points that occur more frequently than the other points. These points lie around the lateral acceleration value of 0 m / s². 2 The longitudinal acceleration a x The acceleration diagram of Fig. 2 refers to route 2 of Fig. 1. This route 2 has longer straight sections and only a few curves. On the straight sections, the driver accelerates, so that the longitudinal acceleration a x a corresponding positive value. Since there are few sharp curves, high lateral accelerations occur. y only with low frequency.

[0058] Fig. Figure 3, however, shows the acceleration diagram for test track 1. In the point cloud of Fig. Area 4, identified as 3, represents a cluster of acceleration values. This area 4 is located in the upper right of the diagram, i.e., with small positive longitudinal acceleration values ​​and positive lateral acceleration values. This is because test track 1 is more winding and, due to its circular design, has more right-hand turns when driven clockwise. These right-hand turns result in the following in the diagram: Fig. 3 positive lateral acceleration values ​​a y .

[0059] For the method according to the invention, it is sufficient if, for example, test track 1 is initially driven to obtain the test data. Of course, it is preferable to have more data available as a basis. In principle, the data from track 2 could also be used as test data. However, for the purpose of explaining the invention, the test data from test track 1 are used to extrapolate the driver's behavior to track 2.

[0060] First, test track 1 is divided into five segments. Each segment represents a section of the track and is therefore assigned to a portion of the geometric test data. Each segment is assigned to one of the predefined clusters A to J. The entire test data space is, for example, divided into ten different clusters A to J. In principle, the number of clusters can be chosen arbitrarily. Therefore, fewer or more than ten clusters can be selected into which the test data space is divided.

[0061] The segmentation is based on a clustering approach, as implemented in the multidimensional test data space (excluding vehicle dynamics data). This point cloud can be visualized as shown in the... Fig. 2 and Fig. 3, but with correspondingly more dimensions. Each point on the test track is assigned to one of the clusters A to J. This results in a multitude of connected point groups on the track, each assigned to one of the clusters. Geometrically, each point group represents a segment of this test track 1.

[0062] For example, all points on a straight section of track before a sharp curve are assigned to cluster A. This section of track thus forms segment 5.

[0063] Now it must be determined which vehicle dynamics data should correspond to a cluster. For example, all vehicle dynamics data from the points in the cluster could be averaged. Alternatively, a median value of the respective parameter (e.g., speed or acceleration) could be chosen as a fixed value for the cluster.

[0064] The division into clusters thus leads to segments on the route that are geometrically similar and driven with similar dynamic characteristics. These clusters can then be used to classify and assign unknown route segments. The results of the test track driving analysis can then be transferred to the assistance system. This allows for a transfer of the test track analysis to the actual route 2.

[0065] The geometric data of the new route 2 to be driven are known. If necessary, they are determined via a navigation system or similar device. Based on this geometric data, clustering into segments can be performed. For this purpose, the clusters that were previously generated in the test data space are used. The clustering results in the following for the route 2 to be driven: Fig.The 5 segments. The route 2 to be driven is therefore segmented into sections, similar to test track 1. Each segment 5 represents a specific scenario, e.g., a longer straight stretch approaching a curve, initial speed 50 km / h, braking deceleration -2 m / s². 2 etc. The driver assistance system can then, for example, specify the speed and acceleration values ​​that are specified for the associated cluster when driving on route 2 in the respective segment.

[0066] Segment-based analysis offers the possibility of subdividing the driver, or rather their identification, into subsets. This allows for the integration of more aspects into the description and the weighting of key performance indicators to varying degrees. In other words, depending on the selected class, such as comfortable or sporty driving, individual, driver-specific characteristics can be mapped (such as lane guidance in a poorly visible scenario, safety margins, and track width, etc.). These characteristics are not averaged for every route but are evaluated individually for the described scenario. This enables more natural driver support that is perceived less as the imposition of expert-defined guidance behavior. In this way, comprehensible support can be achieved with high driver acceptance.

[0067] According to another embodiment, certain modes for driving dynamics support (such as Sport, Comfort, Eco, ...) can be provided in combination with the classification of the driver's skill level (e.g. sporty, cautious).

[0068] According to the invention, all the aforementioned embodiments offer the advantage of selectively extracting features that describe driver behavior based on scenarios and implementing them in a customized assistance system. This leads to greater acceptance of the driving dynamics support. Furthermore, the function could be offered as an additional feature to an existing function, thereby increasing the product's added value without significant additional costs. Reference symbol list 1 test track 2 route 3 area 4 area 5 segments

Claims

[1] Method for steering a vehicle over a route to be travelled (2) by - Driving a test track (1) by a driver, - Acquisition of geometric test data of the test track (1), - Recording of driving dynamics test data from the driver driving on the test track (1), - Dividing a test data space formed from the geometric test data into clusters by automatically dividing the driven test track into segments, - Defining vehicle dynamics data for each cluster based on the driver's vehicle dynamics test data that can be assigned to a respective cluster, - Assigning route segments (5) of the route to be traveled (2) to the clusters based on geometric data of the route to be traveled and - Control at least one vehicle component of the vehicle in one of the track sections (5) according to the specified vehicle dynamics data of the respective cluster. [2] Method according to claim 1, wherein the geometric test data include position data, a curvature, a track width, a combination of track section types and / or a number of lanes of the test track (1). [3] Method according to claim 1 or 2, wherein when driving on the test track (1) additional situation data regarding weather, time, traffic density and / or environment are recorded, the situation data are included in the test data space and the cluster formation, current situation data for driving on the route to be driven are recorded, and the assignment of the route segments (5) to the clusters also takes place depending on the current situation data. [4] Method according to one of the preceding claims, wherein the acquisition of the driving dynamics data for several driving modes of the driver and the division of the test data space into clusters also takes place depending on the several driving modes, one of the several driving modes is selected for the route (2) to be driven, and the assignment of the route segments (5) to the clusters takes place depending on the selected driving mode. [5] Method according to any of the preceding claims, wherein the vehicle dynamics data include a lane guidance, a safety distance, an acceleration, a speed and / or a jerk. [6] Method according to one of the preceding claims, wherein when assigning route segments (5) of the route (2) to the clusters the geometric data of the route (2) to be driven and / or the current situation data are individually weighted. [7] Method according to one of the preceding claims, wherein the assignment of route segments (5) of the route (2) to the clusters is based on machine learning. [8] Method according to one of the preceding claims, wherein the test data space also contains vehicle data about the vehicle, and the control of the at least one vehicle component is specific to the vehicle. [9] Driver assistance system for controlling a vehicle on a route to be driven (2) with - a recording device for recording geometric test data of the test track (1) and for recording driving dynamics test data when a driver drives on a test track (1), - a classification device for dividing a test data space formed by the geometric test data into clusters by automatically dividing the driven test track into segments, as well as - a data processing device for defining vehicle dynamics data for each cluster based on the driver's vehicle dynamics test data that can be assigned to a respective cluster, wherein - the classification device is designed to assign route segments (5) of the route to be traveled (2) to the clusters based on geometric data of the route to be traveled and - the driver assistance system includes a control device for controlling at least one vehicle component of the vehicle in one of the route segments (5) according to the specified vehicle dynamics data of the respective cluster. [10] Motor vehicle with a driver assistance system according to claim 9.

Images