Method and device for determining a target distance to a vehicle in front during partially automated or automated driving using a driver assistance system in a hands-off domain

Abstract

The invention relates to a method for determining a target distance (20) to a vehicle ahead during partially automated or automated driving using a driver assistance system (51) in a hands-off domain, wherein at least one currently existing state category (10, 10-x) is obtained and / or recorded, and wherein the target distance (20) is determined taking into account a frequency distribution (15) for the target distance (20) corresponding to the at least one currently existing state category (10, 10-x) and a distance-dependent merging probability (16). Furthermore, the invention relates to a device (1) for determining a target distance (20) to a vehicle ahead during partially automated or automated driving using a driver assistance system (51) in a hands-off domain.

Description

[0001] The invention relates to a method and a device for determining a target distance to a vehicle in front during semi-automated or automated driving using a driver assistance system in a hands-off domain.

[0002] Currently, there are Level 2 driver assistance systems that allow a driver to take their hands off the steering wheel in defined scenarios (or domains); the driver assistance system then takes over driving the vehicle independently. Such a scenario or domain could be, for example, a highway without significant curves, because it can be assumed that there are no people on the road and the system does not need to navigate tight bends. In this domain (Operational Design Domain, ODD), the driver is therefore allowed to take their hands off the steering wheel; however, when the vehicle leaves this domain, the driver must put their hands back on the steering wheel.

[0003] The described driver assistance system considers not only lateral vehicle control but also longitudinal vehicle control and interaction with objects. In previous implementations, developers define a longitudinal dynamics control (expert model) to implement a desired system behavior dictated by the system requirements. This longitudinal dynamics control is primarily determined by the desired (target) distance and the vehicle dynamics. Distance categories are selected, for example, as time intervals that are adjustable in increments (e.g., levels 1-5, where 1 represents a very small time interval, 5 a very large time interval, etc.). However, the optimal vehicle behavior from a developer's perspective may not correspond to the expectations of an average driver and therefore does not take into account the environment or current scenario in which the vehicle is located.However, this can play an important role in driver acceptance, as well as in a (subconscious) assessment of the system dynamics and, consequently, the perception of comfort. For example, a greater distance may be perceived as more comfortable in heavy rain and on a wet surface than on a dry surface. The multitude of possible influencing factors, however, is currently difficult to represent.

[0004] DE 10 2020 211 539 A1 describes a method and a driver assistance system for a single-track or multi-track motor vehicle, in which an object detection sensor in front of the vehicle detects objects ahead and the vehicle's speed is regulated by a cruise control system such that, upon detection of an object ahead, a predetermined distance between the vehicle and the object ahead is established. While following the object ahead, the driver can change the distance between the vehicle and the object ahead by operating the accelerator pedal or a rotary throttle. When the driver assistance system is deactivated, at least one or more vehicle systems are monitored, and if one or more of these systems is operated, the system infers the driver's level of distraction and automatically activates the driver assistance system depending on the level of distraction.

[0005] DE 10 2018 201 306 A1 describes a method for controlling the distance of a vehicle to a vehicle driving ahead, wherein in a limiting step the distance is limited to a preselected target distance value, and in a setting step the target distance value is set using a change in the accelerator pedal angle.

[0006] DE 10 2018 222 176 A1 discloses a method for predicting whether a foreign vehicle will change lanes in the vicinity of an ego vehicle, based on observations of the ego vehicle's surroundings, comprising the following steps: • the observations are fed to a plurality of individual classifiers; each individual classifier determines, according to an individual rule, from at least a subset of the observations, an individual probability that the foreign vehicle will change lanes; • the driving situation in which the ego vehicle finds itself, as a whole, is classified into one of several discrete classes using a situation classifier; • a set of weighting factors is determined, assigned to the class into which the situation classifier has classified the driving situation, which specifies the relative weighting of the individual classifiers for this driving situation;• The individual probabilities are combined with the weighting factors to obtain an overall probability that the other vehicle will change lanes. Furthermore, a method for training the weighting factors and an associated computer program are described.

[0007] From DE 10 2012 000 896 A1 a method for supporting a driver of a motor vehicle by means of a driver assistance system is known, comprising: - Capturing an environmental parameter describing the environment of the motor vehicle, - Determining a hazard parameter describing the probability of an object in the vicinity being struck, depending on the size of the environment. - Triggering a reaction from the driver assistance system to counteract an impending collision with the object, depending on the severity of the hazard. To enable improved support for a motor vehicle driver, the following steps are necessary: - Determining an object-specific characteristic of the object, - Determining the hazard magnitude is additionally planned depending on the object-specific characteristic of the object.

[0008] DE 10 2021 001 461 A1 discloses a further method and a further device for determining and providing geolocated frequencies of lane changes performed by vehicles.

[0009] The invention is based on the objective of providing a method and a device for determining a target distance to a vehicle in front during semi-automated or automated driving using a driver assistance system in a hands-off domain.

[0010] The problem is solved according to the invention by a method with the features of claim 1 and a device with the features of claim 8. Advantageous embodiments of the invention are set forth in the dependent claims.

[0011] In particular, a method for determining a target distance to a vehicle in front during partially automated or automated driving using a driver assistance system in a hands-off domain is provided, wherein at least one currently existing state category is maintained and / or recorded, and wherein the target distance is determined taking into account a frequency distribution for the target distance corresponding to the at least one currently existing state category and a distance-dependent merging probability.

[0012] Furthermore, in particular a device for determining a target distance to a vehicle in front during partially automated or automated driving using a driver assistance system in a hands-off domain is provided, comprising a control unit, wherein the control unit is configured to maintain and / or detect at least one currently existing state category, and to determine the target distance taking into account a frequency distribution for the target distance corresponding to the at least one currently existing state category and a distance-dependent probability of rejoining the vehicle.

[0013] The method and device enable the determination of a target distance (which can also be referred to as the set distance) to a vehicle in front, depending on the current scenario. The current scenario is defined using state categories. Furthermore, the method and device take into account the probability of merging (e.g., merging incidents per kilometer). It has been shown that in a hands-off environment, maintaining the safest possible driving condition is just as important as perceived comfort. For this, it is necessary, on the one hand, that the vehicle is driven with the greatest possible (target) distance to the vehicle in front. However, it has also been shown that an excessively large (target) distance to the vehicle in front increases the likelihood of merging.Merging vehicles are those that cut in from adjacent lanes between the vehicle in front and the vehicle in front. Such a merging vehicle negatively impacts the potential time-to-collision and thus increases the risk. Furthermore, excessively large gaps can be perceived as insecure and uncomfortable for the driver. Therefore, it is intended that at least one currently existing state category be maintained and / or recorded. This at least one state category defines, in particular, a current scenario, that is, a context in which the vehicle is currently located. For example, this at least one state category includes vehicle speed, weather conditions, etc.The target distance is determined taking into account a frequency distribution for the target distance corresponding to at least one currently existing state category and a distance-dependent merging probability. This allows the selection of the optimal target distance for a given current scenario, considering both driver comfort and safety requirements. The frequency probability and the merging probability were determined using statistical methods, particularly based on fleet data.

[0014] The acquisition of at least one condition category (e.g., weather conditions, traffic density, etc.) can be achieved, for example, through environmental sensing and / or perception using dedicated vehicle sensors. Alternatively or additionally, the at least one condition category (e.g., vehicle speed, etc.) can also be retrieved and / or provided by the vehicle's control system. Furthermore, alternatively or additionally, the at least one condition category (e.g., weather conditions, traffic density, etc.) can also be retrieved and / or provided by a data service, such as a traffic management system.

[0015] It is specifically intended that the target distance, once determined, is transmitted to the driver assistance system and implemented by it. The driver assistance system then regulates the distance, specifically to the target distance. This also provides a method for operating a vehicle.

[0016] A hands-off domain refers in particular to a domain in which semi-automated or automated driving is possible without the driver having their hands on the steering wheel (hands-off state).

[0017] Parts of the device, in particular the control unit, can be configured individually or collectively as a combination of hardware and software, for example, as program code executed on a computing unit, in particular a microcontroller or microprocessor. However, it can also be provided that parts are configured individually or collectively as an application-specific integrated circuit (ASIC) and / or a field-programmable gate array (FPGA) and / or a graphics processing unit (GPU) and / or a digital signal processor (DSP). The control unit can, in particular, comprise at least one computing unit and at least one memory.

[0018] Furthermore, a method for providing frequency distributions corresponding to state categories for a target distance and a distance-dependent merging probability is provided for a method according to one of the described embodiments, wherein fleet data of a vehicle fleet are collected during manual driving, wherein the fleet data are classified according to state categories, wherein for each of the state categories an associated manually driven distance to a vehicle in front is extracted, and wherein for each state category, a frequency distribution is determined and provided based on the extracted distances and a statistical evaluation, and wherein, based on a statistical evaluation of the collected fleet data, the merging probability is determined and provided for each of the state categories.The frequency distribution(s) are thus derived primarily from swarm data of manual journeys in the respective scenarios (or state categories) of the vehicle fleet. The probability of lane merging can be related to at least one state category and thus to a scenario defined by it. Determining the values ​​or characteristics of the state categories can be done manually and / or using statistical methods and / or data analysis, for example, using machine learning methods that perform a classification. The state categories themselves can also be defined using a machine learning method specifically designed for this purpose.

[0019] In one embodiment, a target distance determined from the frequency distribution is increased by a predetermined amount. This allows, in particular, consideration of the time a driver needs to put their hands back on the steering wheel. The predetermined amount can be determined, for example, based on empirical test series to ascertain the required time and / or through simulations.

[0020] In one embodiment, the system takes into account the dependence of the merging probability on the currently existing at least one state category. This allows for consideration of the fact that the merging probability is also dependent on the currently existing state category in most cases. For example, the merging probability depends, in particular, on traffic density. Furthermore, the merging probability will be lower on a single-lane rural road in each direction than on a three-lane highway, and so on. This allows for a more accurate consideration of current scenarios.

[0021] In one embodiment, the frequency distribution is stored in the form of a characteristic map dependent on the state category. Starting from the at least one currently existing state category, the corresponding frequency distribution is retrieved or generated from this map. This allows for a simple implementation of the method. It may also be provided that for values ​​of the at least one state category for which no frequency distribution exists in the characteristic map, interpolation and / or extrapolation is performed to generate a frequency distribution.

[0022] In one embodiment, the probability of lane merging is taken into account based on a predefined limit value for each state category. This allows for the consideration of empirical knowledge. For example, the predefined limit value specifies the maximum permissible probability of lane merging for each state category, such as a limit value for the number of lane merging vehicles per kilometer. The target distance can then be chosen to ensure that this predefined limit value is not exceeded. Since the number of lane merging vehicles per kilometer depends on the current scenario or on the at least one state category representing that scenario, a limit value is provided for each scenario or each of the at least one state categories.

[0023] In one embodiment, the system optimizes a cost function to determine the target distance, taking into account both the target distance and the probability of lane merging. Costs are assigned to the target distance and the resulting probability of lane merging based on the cost function. To find the optimal target distance, the cost function is minimized by varying the target distance. It may also be possible to consider the previously described predefined limit value.

[0024] In one embodiment, the at least one state category comprises at least one of the following: speed, traffic density, occupied lane, number of lanes, rainfall amount, time of day, road surface condition (e.g., dry, wet, or icy, etc.), road category (e.g., urban, rural, and / or motorway, etc.), vehicle type of a preceding vehicle (e.g., sedan, van, or truck, etc.), selected driver profile (e.g., sporty, normal, or eco-friendly, etc.), and / or driver type (e.g., experienced, inexperienced, old, young, etc.). This list is not exhaustive. Combinations of the state categories or of values ​​and / or characteristics of the state categories define, in particular, a scenario in which the vehicle finds itself.

[0025] Further features for the design of the device result from the description of embodiments of the method. The advantages of the device are the same in each case as in the embodiments of the method.

[0026] Furthermore, a vehicle is created comprising a driver assistance system equipped for at least partially automated longitudinal guidance with controlled distance to a vehicle in front, and a device according to one of the described embodiments, wherein the driver assistance system controls the distance to the target distance determined by means of the device.

[0027] The invention is explained in more detail below with reference to preferred embodiments and the figures. These show: Fig. 1 a schematic representation of an embodiment of the device for determining a target distance to a vehicle in front during semi-automated or automated driving using a driver assistance system in a hands-off domain; Fig. 2a, Fig. 2b Schematic representations to illustrate the invention; Fig. 3 a schematic representation to illustrate embodiments of the method and the device; Fig. 4a, Fig. 4b further schematic representations to clarify the invention; Fig. 5 a schematic representation to illustrate embodiments of the method and the device; Fig. 6 a schematic flowchart of embodiments of the method.

[0028] The Fig. Figure 1 shows a schematic representation of an embodiment of the device 1 for determining a target distance 20 to a vehicle ahead during partially automated or automated driving using a driver assistance system 51 in a hands-off domain. The device 1 is, in particular, arranged in a vehicle 50. The vehicle 50 comprises the driver assistance system 51, which can perform at least longitudinal vehicle guidance in a hands-off domain. The method is described in more detail below with reference to the device 1.

[0029] The device 1 comprises a control unit 2. The control unit 2 comprises a computing unit 2-1 and a memory 2-2. The computing unit 2-1 is configured to perform the necessary calculations for carrying out process steps and can access data stored in the memory 2-2 for this purpose.

[0030] The control unit 2 is configured to obtain and / or record at least one currently existing state category 10 and to determine the target distance 20 taking into account a frequency distribution 15 for the target distance 20 corresponding to the at least one currently existing state category 10, 10-x, and a distance-dependent merging probability 16. The determined target distance 20 is supplied to the driver assistance system 51, which sets the target distance 10 by steering and / or regulating.

[0031] The at least one currently existing condition category 10 (e.g., a road type) can be detected and / or provided, for example, by means of a sensor 52 of the vehicle 50. Furthermore, the at least one currently existing condition category 10 (e.g., a vehicle speed) can also be queried and / or provided by a vehicle control unit 53.

[0032] The Fig. 2a and Fig. Figure 2b shows schematic diagrams to illustrate the invention. Fig. Figure 2a shows frequency distributions 15 for two different scenarios, which can be expressed using the state category "weather". The frequency distributions 15 were determined primarily from fleet data and represent values ​​used by drivers during manual driving. A frequency 21 is shown as a function of the target distance 20. The left frequency distribution 15 occurs, for example, in sunny weather for a vehicle fleet, while the right frequency distribution 15 occurs in rainy weather (with otherwise identical conditions and state categories). It can be seen that the frequency distributions 15 are shifted relative to each other and each exhibits a maximum at a different target distance 20. This can be explained by poorer visibility and the drivers' perception of safety, as drivers generally maintain a greater distance from the vehicle in front during rainy weather. Fig. Figure 2b shows a merging probability 16 as a function of the target distance 20. Here, the merging probability 16 is greater with increasing target distance 20 in good weather than in rainy weather. This can also be explained by the drivers' perception of safety, since changing lanes in rainy weather is perceived as less pleasant than in good weather. The procedure described in this disclosure takes into account both the frequency distribution 15 and the merging probability 16, both in particular considering the at least one state category 10 ( Fig. 1) or the currently existing scenario.

[0033] It may be provided that a target distance 20 determined from the frequency distribution 15 is increased by a predetermined surcharge.

[0034] It may be provided that a dependency of the reversal probability 16 on the currently existing at least one state category 10 is taken into account. In particular, the reversal probability 16 assigned to each of the at least one state category 10 is then considered, as is the case with reference to the Fig. 2b above has already been described.

[0035] It can be provided that the frequency distribution 15 is stored in the form of a characteristic map 17 dependent on at least one state category 10, from which the corresponding frequency distribution 15 is retrieved or generated based on the at least one currently existing state category 10. In particular, a frequency distribution 15 can be stored in the characteristic map 17 for each combination of values ​​of individual state categories 10. When a given combination of values ​​of the state categories 10 occurs, the associated frequency distribution is retrieved from the characteristic map 17. A simple example for two values ​​of the state category 10 "weather" is shown in the Fig. 2a shown.

[0036] The Fig. Figure 3 shows a schematic representation to illustrate embodiments of the method and the device. In one embodiment, the merging probability 16 is taken into account based on a limit value 18 specified for each state category 10. The left side shows the scenario "fair weather," the right side "rainy weather." In particular, the target distance 20 is selected such that a maximum target distance 20 is achieved while adhering to the limit value 18 specified for the merging probability 16. In fair weather (left side of the Fig. 3) The specified limit value 18 is reached at a smaller target distance 20 than in bad weather, since the probability of merging 16 decreases with a larger target distance 20 (see also Fig. 2b).

[0037] In another embodiment, it is provided that, to determine the target distance 20, a value of a cost function is optimized in which the target distance 20 and the merging probability 16 are taken into account. This is also shown schematically in the Fig. Figure 3 illustrates this. A region 30 is marked, which contains a selected target distance 20 for manual driving ("hands-on"). A further region 31 is marked for a hands-off domain, in which optimization based on the cost function is to be performed. Specifically, the cost function is minimized for this purpose. An upper limit 19 of this region 31 can be determined, for example, by experts and / or by means of data analysis. The region 31 for optimization can, in principle, also encompass the entire (shown) frequency distribution 15 or at least that part of the frequency distribution 15 within which a predominant number of the target distances 20 lie. The respective cost factors for the target distance 20 and the merging probability 16 can be determined, for example, based on empirical test series and / or by simulation.

[0038] It may be stipulated that at least one condition category 10 includes at least one of the following: a speed 10-1, a traffic density 10-2, a lane in use 10-3, a number of lanes 10-4, a rainfall amount 10-5, a time of day 10-6 (e.g., day, night), a road surface condition 10-7, a road category 10-8, a vehicle type 10-9 of a preceding vehicle, a selected driver profile 10-10, and / or a driver type 10-11. In particular, a current scenario is defined by several condition categories 10 or respective values ​​and / or characteristics of these condition categories 10.

[0039] The Fig. 4a and Fig. Figure 4b shows further schematic representations to illustrate the invention. Fig. 4a and Fig. 4b basically correspond to the Fig. 2a or 2b. The Fig. Figure 4a shows frequency distributions 15 for three different scenarios, which can be expressed using the state category "traffic density", each schematically illustrated by the number of vehicles shown. Fig. Figure 4b shows the corresponding merging probabilities 16. In general, it can be observed that a merging probability 16 increases with increasing traffic density for a larger target distance 20.

[0040] The Fig. Figure 5 shows a schematic representation to illustrate embodiments of the method and the device 1. Fig. 5 corresponds to the Fig. 3, wherein in the Fig. 5 the three forms of traffic density (see Fig. 4a and Fig. 4b) specify the frequency distributions 15. It can be seen that when taking into account the limit value 18, with decreasing traffic density a certain target distance 20 moves further and further towards the upper end of the respective frequency distribution 15, since the merging probability 16 also changes.

[0041] The Fig. Figure 6 shows a schematic flowchart of embodiments of the method for determining a target distance to a vehicle ahead during partially or fully automated driving using a driver assistance system in a hands-off domain. The method is started, for example, when it is detected that the vehicle is in a hands-off domain.

[0042] In process step 100, at least one currently existing state category is obtained and / or recorded.

[0043] In a process step 101, the target distance is determined taking into account a frequency distribution for the target distance corresponding to the at least one currently existing state category and a distance-dependent reversal probability.

[0044] In process step 102, the specified target distance can be transmitted to the driver assistance system.

[0045] In a process step 103, it may be provided that the driver assistance system sets the target distance by regulating a distance to the vehicle in front to the target distance.

[0046] Further embodiments of the method have already been described in more detail above with reference to the device.

[0047] The frequency distributions and the retracement probabilities are or were provided in particular by means of a procedure for providing frequency distributions corresponding to state categories for a target distance and a distance-dependent retracement probability, as already described in the general description. Reference symbol list 1 Device 2 Control unit 2-1 Computing Equipment 2-2 storage 10 condition category 10-1 Speed ​​(of the vehicle) 10-2 traffic density 10-3 occupied lane 10-4 Number of tracks 10-5 rainfall 10-6 Daytime 10-7 a road surface condition 10-8 Street category 10-9 Vehicle type 10-10 selected driver profile 10-11 Driver type 15 Frequency distribution 16 Merging probability 17 Characteristic map 18 specified limit value 19 upper limit 20 target distance 30 area (manual rides) 31 Area (search area for optimization) 50 vehicles 51 Driver assistance systems 52 Sensors 53 Vehicle control

Claims

[1] Method for determining a target distance (20) to a vehicle ahead during partially automated or automated driving using a driver assistance system (51) in a hands-off domain, where at least one currently existing state category (10,10-x) is preserved and / or recorded, and wherein the target distance (20) is determined taking into account a frequency distribution (15) for the target distance (20) corresponding to the at least one currently existing state category (10,10-x) and a distance-dependent rejoining probability (16). [2] Method according to claim 1, characterized by , that a target distance (20) determined from the frequency distribution (15) is increased by a predetermined surcharge. [3] Method according to claim 1 or 2, characterized by, that a dependence of the shearing probability (15) on the currently existing at least one state category (10,10-x) is taken into account. [4] Method according to any of the preceding claims, characterized by , that the frequency distribution (15) is stored in the form of a characteristic map (17) dependent on the state category (10,10-x), from which the corresponding frequency distribution (15) is retrieved or generated starting from the at least one currently existing state category (10,10-x). [5] Method according to any of the preceding claims, characterized by , that the shearing probability (16) is taken into account on the basis of a limit value (18) specified for each state category (10,10-x). [6] Method according to any of the preceding claims, characterized by, that to determine the target distance (20) a value of a cost function is optimized in which the target distance (20) and the merging probability (16) are taken into account. [7] Method according to any of the preceding claims, characterized by , that the at least one condition category (10) includes at least one of the following: a speed (10-1), a traffic density (10-2), a lane occupied (10-3), a number of lanes (10-4), a rainfall amount (10-5), a time of day (10-6), a road surface condition (10-7), a road category (10-8), a vehicle type (10-9) of a preceding vehicle, a selected driver profile (10-10) and / or a driver type (10-11). [8] Device (1) for determining a target distance (20) to a vehicle ahead during semi-automated or automated driving using a driver assistance system (51) in a hands-off domain, comprising: a control device (2), wherein the control device (2) is configured to obtain and / or record at least one currently existing state category (10,10-x), and to determine the target distance (20) taking into account a frequency distribution (15) for the target distance (20) corresponding to the at least one currently existing state category (10,10-x) and a distance-dependent rejoining probability (16). [9] Vehicle (50) comprising a driver assistance system (51) configured for at least semi-automated longitudinal guidance with controlled distance to a vehicle in front, and a device (1) according to claim 8, wherein the driver assistance system (51) controls the distance to the target distance (20) determined by means of the device (1). [10] Method for providing frequency distributions (15) corresponding to state categories (10,10-x) for a target distance (20) and a distance-dependent merging probability (16) for a method according to one of claims 1 to 8, wherein fleet data of a vehicle fleet are collected during manual journeys, where the fleet data are classified according to condition categories (10,10-x), wherein for each of the condition categories (10,10-x) a corresponding manually driven distance to a vehicle in front is extracted, and wherein for each state category (10,10-x) a frequency distribution (15) is determined and provided based on the extracted distances and a statistical evaluation, where, based on a statistical evaluation of the collected fleet data, the probability of merging (16) is determined and provided for each of the condition categories (10,10-x).

Images