Procedures for informing vehicle drivers about an accident

Abstract

The invention relates to a method for informing the drivers of vehicles (1) about an accident (2), wherein at least one vehicle (1.1) is involved in an accident, the accident (2) is detected by the vehicle (1.1) and / or by at least one third vehicle (1.2) using sensors (3), a processing unit generates accident information (4) by processing corresponding sensor data, which is transmitted to the respective vehicle (1) to inform the drivers and is displayed there via output means (5). The method according to the invention is characterized in that - the vehicle involved in the accident (1.1) and / or the third vehicle (1.2) transmits the sensor data to a computing device (6) for processing; - the computing device (6) generates a synthetic accident representation (7) from the sensor data; - an abstraction indicator is stored in an internal vehicle computing unit (8) of a respective vehicle (1), which is transmitted to the computing device (6); - the computing device (6) determines a level of abstraction for the synthetic accident representation (7) depending on the received abstraction indicator and transmits the abstracted accident representation (7) back to the respective vehicle (1); and - in a given vehicle (1) the abstracted accident representation (7) is output as accident information (4).

Description

[0001] The invention relates to a method for informing vehicle drivers about an accident according to the type defined in the preamble of claim 1.

[0002] Accidents happen again and again in road traffic. If first aid has been provided to injured people or if emergency services are already on the scene, the rest of the traffic can be diverted around the accident site. For example, two lanes of a motorway can be closed and the backed-up traffic in the third lane can be diverted around the accident site. Such an accident can have a negative impact on a driver's ability to operate their vehicle. For example, the driver is distracted by the accident. This creates the potential for a rear-end collision. If an accident site is not yet sufficiently secured, passers-by may also see crashed vehicles and people. This sight can sometimes lead to considerable mental stress, even after the accident site has been passed.Some people even deliberately slow down to get a better view of an accident. These individuals are also referred to as "gawkers." This behavior further disrupts traffic flow. Some gawkers also take videos while driving past the accident scene, which particularly compromises the privacy of those already in a particularly difficult situation.

[0003] This creates the need to provide procedures and means to improve road safety after an accident and to protect the data protection of persons involved in an accident.

[0004] JP 2000-298 797 A discloses a system for transmitting accident information to vehicles. The document describes the provision of accident information to a central location, which then transmits the accident information to a large number of individual vehicles. The accident information can be provided acoustically and visually in the vehicles. In particular, the accident location is displayed on a digital road map, and a detour route is calculated for the respective vehicles.

[0005] CN 1 04 079 896 A also discloses an emergency procedure and system for vehicles. A vehicle registers an accident and transmits an accident signal to a central emergency response center. The central emergency response center, in turn, dispatches a video vehicle to film the accident scene. The video vehicle transmits a live video stream of the accident scene to the emergency response center. This allows emergency responders to gain a better understanding of the situation.

[0006] Furthermore, a method and a device for transmitting information between vehicles are known from US 2002 / 0 198 660 A1. This document describes the detection of the immediate vehicle surroundings using sensors. This enables vehicles to detect objects. The vehicles can also communicate with each other and exchange kinematic status information. This enables the vehicles to detect the relative position of static and dynamic objects in the surrounding area, as well as their direction and speed of movement. The exchange of sensor data between the vehicles allows the vehicles to see around corners or other obstacles. This allows the vehicles to anticipate other road users, thereby reducing the potential for accidents.Based on the transmitted location and direction information, the respective vehicles can also estimate whether a collision with the other detected road users is imminent. The vehicles can then derive control instructions for manual control or control commands for automated control to avoid a collision. The respective traffic situation can be displayed on an in-vehicle display.

[0007] Furthermore, DE 10 2021 001 795 A1 discloses a method for detecting traffic accidents in a vehicle and for processing accident data.

[0008] Furthermore, DE 10 2020 004 792 A1 discloses a method and a device for detecting and reporting parking accidents for vehicles.

[0009] Furthermore, DE 10 2016 216 522 A1 discloses a method for providing at least one piece of information about an accident between at least two road users.

[0010] Furthermore, DE 10 2019 215 762 A1 discloses an arrangement and a method for networked driving.

[0011] In addition, DE 10 2010 001 006 A1 discloses a method for providing at least one piece of information about an accident between at least two vehicles.

[0012] The present invention is based on the object of providing an improved method for informing vehicle drivers about an accident, with the aid of which road safety can be further increased.

[0013] According to the invention, this object is achieved by a method for informing vehicle drivers about an accident having the features of claim 1. Advantageous embodiments and further developments also arise from the dependent claims.

[0014] A generic method for informing the drivers of vehicles about an accident, wherein at least one accident vehicle is involved in an accident, the accident is detected by the accident vehicle and / or by at least one third vehicle with the aid of sensors, a computing device generates accident information by processing corresponding sensor data, which is transmitted to a respective vehicle for informing the drivers and is output there via output means, is further developed according to the invention in that - the accident vehicle and / or the third-party vehicle transmits the sensor data to the computing device for processing; - the computing device generates a synthetic accident representation from the sensor data; - an abstraction indicator is stored in an on-board computing unit of a respective vehicle, which is transmitted to the computing device; - the computing device determines a degree of abstraction for the synthetic accident representation depending on the received abstraction indicator and transmits the abstracted accident representation back to the respective vehicle; and - the abstracted accident representation is output as accident information in a respective vehicle.

[0015] By displaying the accident information in the respective vehicles, the drivers are informed of the accident before they pass the scene. This allows drivers to mentally prepare for what's happening at the scene. This means that drivers are less distracted by the accident as they pass the scene and are also less mentally stressed because they are already prepared for the accident.

[0016] By representing the accident in an abstract manner, the data protection of the people involved in the accident, be they victims, first responders, or emergency personnel, is ensured. However, if camera images or videos of the accident were recorded and displayed in the vehicles, the people and vehicles involved could be recognized and potentially identified by their license plates, which would compromise data protection. The abstraction indicator can be used to adjust the level of abstraction for the synthetic accident representation, making it possible to select the optimal level of detail for each driver for the accident representation displayed in the vehicle, depending on the specific situation. This allows the level of detail to be reduced for people with weak nerves, so as not to place excessive mental strain on them.For people with strong nerves or gawkers, the level of detail can be increased or even exaggerated.

[0017] Onlookers can enjoy the abstracted representation of the accident even before arriving at the scene, thereby satisfying their respective desire for sensation. When driving past the actual accident scene, such a person is then less inclined to slow down or even film the accident. This can prevent so-called onlooker traffic jams or at least mitigate their negative impact on traffic.

[0018] By mentally preparing drivers for the accident, it is also possible to reduce their mental strain and, in the best case scenario, also reduce the likelihood of subsequent accidents caused by increased mental strain and / or distraction. Traffic jams can occur, particularly in the case of accidents on country roads or highways. By providing the accident information, the drivers involved in the traffic jam can be informed of the cause of the traffic jam, which increases their well-being due to improved acceptance.

[0019] Accidents can be detected and recorded in a variety of ways. The accident can be registered by both the vehicles involved in the accident and by third-party vehicles passing by the scene of the accident. The telemetry data of the respective vehicles can therefore provide indications of the occurrence of an accident. For this purpose, acceleration values ​​and / or wheel speed values ​​measured using acceleration sensors can be analyzed by a respective computing unit. When the vehicle involved in the accident collides with static or dynamic objects in the vicinity, corresponding deformation forces act on the vehicle, which lead to a corresponding acceleration or deceleration of the vehicle and can be measured using the acceleration sensors. Vehicles passing by the accident site or vehicles approaching the end of a traffic jam can brake abruptly and excessively to avoid further rear-end collisions.This can also be detected by analyzing acceleration values ​​or the respective wheel speed.

[0020] The respective vehicles may also be equipped with positioning devices such as a navigation system. Such a navigation system enables a respective vehicle to determine its location based on signals and map data transmitted by a global navigation satellite system. This allows the respective vehicle to determine its location and thus also the location of the accident. The change in location also makes it possible to determine the vehicle's speed and whether the vehicle is stuck in a traffic jam. This is a particular indication of a full or partial road closure.

[0021] With the help of laser scanners such as LiDARs, radar sensor systems, and ultrasonic sensor systems, depth information can also be obtained. This allows vehicles to scan their surroundings. This allows a three-dimensional representation of the immediate surroundings to be generated in the respective vehicles. From this three-dimensional data, the geometry of the vehicles and / or people involved in the accident can also be derived. This allows the respective vehicles to determine how many vehicles are involved in an accident and what type of vehicle it is, for example, a car, truck, tractor, bus, or similar, as well as possibly how many people are involved at the accident scene outside of the vehicles and possibly whether other buildings or objects have been damaged.With a comparatively high sampling resolution, even subtle details such as the type and severity of the damage to the vehicles involved in the accident can be determined. It's also conceivable that vehicles passing by the accident site may deliberately increase their sampling resolution to capture more details.

[0022] A particularly differentiated assessment of the current situation is possible with the help of imaging techniques. For example, the vehicles can be equipped with one or more cameras, which, using proven image recognition algorithms or machine vision methods, allows a particularly comprehensive situation assessment. Characteristic features can be recognized in corresponding camera images or videos, such as characteristic shapes or colors, which allow conclusions to be drawn about the type of vehicle or emergency services involved. This can also be used to deduce, for example, whether a vehicle has been damaged and to what extent. For example, it can be determined whether a truck has overturned or how severely a car was deformed in a rear-end collision. Such visual features can be used to increase the level of detail of the synthetic, abstracted accident representation.For example, a vehicle involved in an accident can be shown in the color in which it is actually painted.

[0023] Based on visual characteristics, vehicles in the vehicle's surroundings can also be identified as emergency vehicles. The current location of a particular emergency vehicle and, if the respective emergency vehicle is still en route to the accident scene, its direction of movement and speed can be derived. This makes it possible to estimate a potential accident location, particularly if several emergency vehicles are identified. In particular, not only ground-based emergency vehicles but also airborne emergency vehicles such as drones, helicopters or aircraft can be detected, recognized and identified. The location as well as the direction of movement and speed of movement can be identified using a stationary ego vehicle if the respective emergency vehicle is detected again by sensors at two consecutive times. The same emergency vehicle can also be detected by several vehicles at the same time.This allows the position, direction of movement, and speed of movement to be determined particularly reliably using triangulation. A probable accident location can also be estimated taking traffic flow into account. For example, the formation or presence of a traffic jam indicates that an accident has occurred some distance away on the respective road. Traffic jams and slow-moving traffic can be detected by the respective vehicles themselves, and information about them can be obtained from a traffic service. In particular, the determined trajectory of an emergency vehicle is correlated with the traffic jam situation to determine the end of the traffic jam or the accident location.

[0024] Microphones can also be used to identify characteristic acoustic features, such as the sound of a horn, siren, or rotor noise. These characteristic features can be used to determine the presence of an emergency vehicle, and, taking into account the temporal change in volume and the time difference of an acoustic signal, the location and direction of movement or speed of movement can also be derived. The accident itself can also be recorded using sensors if relevant vehicles are present at the scene during the accident. Characteristic accident noises, such as the rubbing of metal on metal or on asphalt, can also be recorded and identified using acoustic recording devices such as microphones.

[0025] The computing device can be implemented as a central solution, for example as a central server or server network, in particular forming a cloud. The computing device could then also be referred to as a central computing device. The computing device can also be implemented as a distributed solution, wherein individual hardware and / or software components are distributed across different physical computing systems, for example distributed across different server farms, desktop computers, mobile devices, in particular in-vehicle control units, and the like. This approach is also referred to as edge computing. The computing device can take additional information into account in order to enrich the synthetic accident representation with details. For example, the current weather, the current time of day, the current lighting conditions, or the like can be recorded and used to enrich a corresponding synthetic scenery.For example, the vehicles can use the sensors to determine whether it's raining, the sun is shining, how cloudy it is, what the lighting conditions are, and so on. All of this information is correctly transferred to the synthetic accident representation, so that the synthetic accident representation reflects the corresponding weather conditions or lighting conditions. In particular, environmental objects at the accident scene are recorded and classified, such as the presence of certain trees, bushes, plants, rocks, mountains, rivers, lakes, buildings, and the like. The corresponding synthetic or virtual accident representation then contains a correspondingly designed virtual scenery. This makes the synthetic accident representation appear particularly realistic, even though it is made up of rendered computer graphics objects. Such computer graphics objects can also be imprinted with photorealistic textures.Such textures can be based on camera footage taken at the accident scene.

[0026] The sensor data transmitted by the numerous vehicles is centrally evaluated by the computer system. This enables the computer system to accurately assess the accident situation in the regions where each vehicle is located.

[0027] Data exchange between vehicles and with the computing device is possible in a proven manner. Such vehicles can be equipped with a vehicle-to-vehicle communication interface. These vehicles can be equipped with a telecommunications unit, also known as a telematics unit. Such a telematics unit enables the vehicles to be connected to the computing device via the internet. Data can be transmitted via mobile communications and / or Wi-Fi.

[0028] The respective vehicles can be equipped with a wide variety of output devices, particularly acoustic and visual output devices. Loudspeakers can be used as acoustic output devices, and display devices such as screens can be used as visual output devices.

[0029] The level of detail of the synthetic, abstracted accident representation in the respective vehicles is selected depending on the abstraction indicator. This allows the optimal level of detail to be determined for the respective situation and the mental stress tolerance of the different drivers.

[0030] An advantageous embodiment of the method according to the invention provides that the computing device determines the course of events by processing the sensor data and supplements the synthetic accident representation with the course of events, wherein an abstracted course of events and / or current accident status is output as accident information in a respective vehicle. "Course of events" refers to the events that caused the accident and the actual occurrence of the accident. "Current accident status" refers to the condition in which the vehicles involved in the accident are at the current time after the accident, for example, tipped over on their side, crushed, etc. The sensor data can show the course of events themselves. In this way, the accident events can be recorded by the vehicles involved in the accident themselves and transmitted to the computing device. The course of the accident can also be observed by third-party vehicles.If the course of the accident itself is recorded, it is particularly reliable to determine the course of the accident using the computer, as it was observed. However, the course of the accident can also be deduced solely from the recording of the accident participants after the accident has occurred. In particular, the degree of destruction of the respective vehicles as well as their resting place and resting position can be used as indicators. In particular, by taking into account local traffic conditions such as the layout of the road, the applicable speed limit, the location and extent of the damage to the vehicles involved in the accident, etc., it is possible to reconstruct the course of the accident. For example, if a first vehicle has a deformed front and a second vehicle has a deformed side, this is an indication that the first vehicle collided head-on with the side of the second vehicle. In addition, the respective vehicle model can be identified.Information linked to the vehicle model, such as vehicle mass, can be retrieved from a database. This information can be linked to the degree of deformation of the respective vehicle body, allowing the calculation of impact energy. From this impact energy, the respective speeds of the vehicles involved in the accident can be estimated. This allows, in particular, the course of the accident to be assessed.

[0031] If not only the current accident status but also the course of the accident is provided as accident information, the drivers of the vehicles can be mentally prepared even more comprehensively for the accident.

[0032] According to a further advantageous embodiment of the method according to the invention, if a high abstraction indicator is present, the accident information is output visually as text and / or acoustically as a read-out voice message, and if a low abstraction indicator is present, the accident information is output at least visually as an image or video. If the abstraction indicator assumes a high value, this means that the synthetic accident representation should be highly abstracted, and vice versa if the abstraction indicator is low. High abstraction is particularly possible with a purely text-based and / or acoustic description of the accident. A higher level of detail can be achieved in particular with a visual description of the accident as an image or video. A visual representation of the accident can be enriched in particular by simultaneous acoustic reproduction.

[0033] Depending on the level of the abstraction indicator, any desired level of detail can be selected. With a low level of detail, i.e. a high level of abstraction, accident vehicles can be replaced with symbolic images or pictograms. This means that the actual vehicle model, the actual vehicle design (e.g., a convertible or coupe), and the actual vehicle color cannot be immediately recognized. Instead of a 3D representation, a 2D representation of the accident or accident sequence can be used, for example, from the side or a bird's eye view. Other objects to be represented in the accident representation can also be idealized using pictograms. A highly abstracted text-based or acoustic accident representation could, for example, be: "Two people are lying in front of a heavily damaged red car, which is parked in front of a slightly damaged blue truck."

[0034] The synthetic accident representation can, however, also be very detailed. In particular, the actual vehicle model and the actual accident environment can be reproduced realistically. In particular, the accident events or the accident itself are reproduced as a virtual, rendered 3D representation. As already mentioned, high-resolution or even photorealistic textures can be applied to the corresponding virtual 3D object. Particularly preferably, a respective driver or vehicle occupant can interact with the video. This means that the video can be rewound or paused as required. The synthetic accident representation can also be implemented as a virtual 3D environment, for example, like the virtual environment of a three-dimensional video game. The respective driver or vehicle occupantVehicle occupants can then interact with this virtual 3D environment, for example, to adjust the position of a virtual camera within the virtual 3D environment—specifically, the position, the angle of view, and the focal length (i.e., the degree of zoom). The video playback speed can also be adjusted.

[0035] Even with a high level of detail, i.e., a low level of abstraction, no details are output that would allow the identification of the people and vehicles involved in the accident. In particular, the faces of people at the scene of the accident are pixelated, obscured by a black bar, or replaced with idealized standard facial features, such as a mask. License plates can also be pixelated, grayed out, or replaced with pseudo-license plates.

[0036] If the driver is a gawker, the description of the accident and / or the current accident status may be exaggerated, so that the gawker no longer has any need to slow down and inspect the accident closely when arriving at the scene.

[0037] To abstract the synthetic accident representation, certain stylistic devices, especially graphical ones, can also be used. For example, a corresponding image or video can be presented in a comic style, or a virtual 3D environment can be rendered using cel shading.

[0038] A further advantageous embodiment of the method according to the invention further provides that at least one physiological parameter of the driver is recorded in a respective vehicle using sensors, and the computing unit automatically determines the level of the abstraction indicator depending on the value of the at least one physiological parameter. In general, the abstraction indicator could be manually specified by a vehicle occupant. Conventional operating means, such as a touch-sensitive display device, can be used for this purpose. Thus, a person with weak nerves can specify a high abstraction indicator, while a person with strong nerves can specify a low abstraction indicator.

[0039] However, the vehicle or the vehicle's internal computing unit particularly preferably determines the level of the abstraction indicator depending on the physiological parameter itself. A wide variety of physiological parameters can be recorded, such as heart rate, blood pressure, oxygen saturation, skin conductance, skin temperature, respiratory rate, blinking rate, gaze direction change frequency, and the like. Such physiological parameters are indicators of a person's mental stress. Depending on the characteristics of the respective physiological parameters, the computing unit can thus determine the driver's mental stress. For example, a high heart rate, respiratory rate, skin conductance, or skin temperature are indications that the person in question is stressed.

[0040] There are two ways in which the abstraction indicator can be changed depending on the physiological parameters. Either the computing unit can increase the abstraction indicator if the least physiological parameter describes an increase in the driver's level of mental strain after the accident information has been output. This means that the synthetic accident representation is more abstracted if the driver's mental strain increases. This approach is particularly suitable for people with weak nerves. By abstracting the accident sequence or the accident itself more strongly, the person in question can create a greater mental distance from the accident and is therefore less mentally strained. This makes it easier for the driver to keep their cool when driving past the accident scene and to operate their vehicle appropriately. This even more reliably reduces the risk of further accidents such as rear-end collisions.

[0041] According to an advantageous alternative embodiment of the method according to the invention, the computing unit could also lower the abstraction indicator if the at least one physiological parameter describes an increase in the driver's level of mental stress after the accident information is output. This approach is particularly advantageous if the driver is a gawker. For example, the gawker could be cognitively stimulated by viewing the accident information, causing their heart rate or similar physiological parameters to increase accordingly. In this case, the abstraction indicator and thus the level of abstraction should be lowered in order to display more details. This "satisfies" the gawker, which reduces the risk that the gawker will reduce their driving speed when driving past the accident scene.In particular, accident information that has been provided once can be stored in the vehicle's computer unit, allowing the driver to retrieve the accident information again at a later time. This can satisfy, in particular, the desire of onlookers for relevant accident images or videos over a longer period of time.

[0042] According to a further advantageous embodiment of the method according to the invention, the computing device uses artificial intelligence, particularly in the form of deep neural networks, to process at least part of the sensor data. The use of artificial intelligence allows a wide range of information to be reliably obtained based on the sensor data. Artificial intelligence can be used for a wide variety of purposes. For example, artificial intelligence can be used in conjunction with computer vision to recognize characteristic visual features even more reliably. Artificial intelligence can also be used to reconstruct the course of events of an accident based on the degree of damage to a vehicle involved in an accident.In particular, KL models are used to reconstruct the physical course of the accident, i.e. in particular to estimate the accident energy and the trajectories of the vehicles involved in the accident and / or to correlate them with each other.

[0043] Artificial intelligence is particularly preferred for the computing device to create the synthetic accident representation itself. Corresponding AI models can then generate the relevant texts, voice messages, images, videos, or virtual 3D environments of the accident. So-called large language models can also be used for this purpose.

[0044] The computing device preferably uses a large language model to describe at least part of the synthetic accident representation with text, wherein at least part of the text is read in by a further AI model to generate image content and / or video content based on speech, wherein the computing device uses the image content and / or video content generated by the further AI model to enrich the synthetic accident representation. So-called large language models, also known as Large Language Models (LLM), are already very powerful today. Such large language models can process information in a differentiated manner and gain insights from it. As a result, such a large language model provides a description of the synthetic accident representation in the form of a text. This text can be displayed directly in the vehicle or output by a speech generator as a computer-read voice message.

[0045] Preferably, however, at least part of this text, or even the entire text, is read by an AI model to generate image or video content, and the correspondingly generated image or video content is displayed as accident information. Here, too, the degree of abstraction can be selected differently depending on the abstraction indicator. Furthermore, AI models are already capable of generating detailed and realistic image or video content solely by interpreting text. This allows both highly abstracted and detailed representations of the accident scene to be generated reliably, quickly, and easily. The corresponding AI models or neural networks can be executed directly on the computing device or on dedicated computing devices, particularly high-performance server farms.The computing device can then access these high-performance server farms using suitable APIs to utilize a corresponding image or video generation service.

[0046] In particular, the sensor data collected by the vehicles is processed and stored in a table format. This format is particularly easy to process thanks to a correspondingly large language model. Furthermore, the respective language model makes it easier to relate information to each other. For example, a first entry could be "A person lying at location X1" and a second entry could be "A person lying at location X2," which could automatically infer "two people lying" if positions X1 and X2 are within a certain distance of each other.In addition, it is advantageous to provide the language model with as much context information as possible in order to positively influence the result - for example: "the representation style should be like in a comic" and in the same way a negated influence on the text model should take place - for example: "the representation of animals and objects that are not related to road traffic is excluded".

[0047] According to a further advantageous embodiment of the method according to the invention, accident information is automatically output in a vehicle as soon as the respective vehicle has approached the accident scene to within a specified distance. The computing device determines the accident location or an estimated accident location from the sensor data transmitted by the vehicles. As already mentioned, the respective vehicles can determine their location with the aid of the positioning means, which is then compared with the accident scene. This makes it possible to detect when a respective vehicle has fallen below the specified distance from the accident scene. The specified distance can be specified as a fixed value, for example, 500 meters, one kilometer, or even fractions or multiples thereof.The specified distance can also be implemented adaptively, for example, depending on the time required to reach the accident scene, so that, for example, the accident information is output to the vehicle 30 seconds, one minute, 10 minutes, half an hour, or even fractions or multiples thereof before reaching the accident scene. Another option for adaptively determining the output time of the accident information is to output the synthetic, abstracted accident representation when the driver's own vehicle arrives at the end of a traffic jam.

[0048] Accident information can also be output only if it can be determined that the vehicle's navigation route passes the accident scene. However, it is also possible to output accident information only if the vehicle is within a location circle centered at the accident scene and within a radius corresponding to the specified distance. This allows accident information to be output even if no navigation guidance is active in the vehicle.

[0049] A further advantageous embodiment of the method according to the invention further provides that the accident information is supplemented by a degree of knowledge, which describes how completely the computing device has recorded the accident and / or the course of the accident based on the available sensor data. The more vehicles that record the accident or the course of the accident, especially if extensive sensor technology is or was used for this purpose, the more detailed the accident events can be reconstructed and displayed accordingly.

[0050] If, for example, only a single vehicle is involved in an accident and only this vehicle registers the accident using its own sensors, comparatively little information about the accident is available. The synthetic accident representation then describes, for example, only the accident location. Further information such as the actual damage to the accident vehicle, the pose the vehicle assumed after the accident, the vehicle model, the vehicle color, the lighting conditions at the accident scene, the weather at the accident scene, and the scenery of the accident scene (for example, whether mountains are visible in the background or trees are at the side of the road) must then be estimated or imagined by the computing device. In this case, a comparatively low degree of insight is determined, in the range between 0% and 10%.

[0051] If not only the current status of the accident but also the course of events is recorded—not only from the vehicles involved in the accident itself, but also from a large number of vehicles passing by—comparatively more information is available. Especially if the accident is recorded using cameras, the computer system can reconstruct and display the accident in detail. This makes it possible, in particular, to reconstruct the course of events and / or the current accident status and to present them in the synthetic accident representation as they actually occurred or are occurring. The degree of insight is correspondingly high, for example, in the range between 90% and 100%.

[0052] By specifying the level of insight when outputting accident information, the driver can be informed about the extent to which the synthetic accident representation matches reality. While this information is independent of the level of abstraction, it is particularly important at a low level of abstraction, i.e., a high level of detail. This means the driver is less surprised if the actual accident deviates from the synthetic accident representation output in the vehicle, even though the driver has selected a low abstraction indicator and thus a high level of detail. Accordingly, the risk that the driver, due to surprise, will perform rash maneuvers when controlling their vehicle is reduced. This could generally increase the potential for an accident, or that a gawker might be tempted to marvel at the accident.

[0053] Further advantageous embodiments of the method according to the invention for informing vehicle drivers about an accident also emerge from the exemplary embodiments which are described in more detail below with reference to the figures.

[0054] Showing: Fig. 1 a schematic plan view of a section of motorway on which an accident occurred; Fig. 2 a schematic representation of a highly abstracted synthetic accident representation shown on an in-vehicle display device; and Fig. 3 a schematic representation of a barely abstracted synthetic accident representation shown on the vehicle's internal display device.

[0055] Accidents can always happen in road traffic. Fig. Figure 1 shows, by way of example and representative, an accident 2 that occurred on a section of highway. In accident 2, the vehicles involved are a truck and a car. This is a rear-end collision, with the car hitting the truck from behind.

[0056] Accident 2 is already secured by emergency personnel 10. Various emergency vehicles 1.3 and support personnel 11 such as police officers, firefighters, and paramedics are present at the scene.

[0057] A traffic jam 12 has formed in the direction of travel in front of accident 2. One of the lanes of the motorway was not closed but is open, so that traffic can be diverted around accident 2. The sight of accident 2 can be disturbing for the drivers of vehicles 1 and lead to mental stress. This, in turn, can have a lasting negative impact on the well-being of the people involved and adversely affect the drivers' ability to drive, which potentially increases the risk of further accidents. In addition, onlookers may change their driving behavior in the area of ​​the accident scene, for example by slowing down. This creates an additional traffic obstruction. Onlookers also tend to film the accident with a camera, which puts the data protection of injured persons and of emergency personnel 11 at risk.

[0058] Using a method according to the invention, the drivers of vehicles 1 are informed about the accident, allowing them to mentally prepare for accident 2. If the drivers then drive past accident 2, they can mentally process the scene more easily, thereby reducing the potential for an accident. Furthermore, the onlookers' desire for sensation is satisfied early on, making them more likely to drive past accident 2 unchallenged.

[0059] For this purpose, the accident 2 is recorded by the accident vehicle 1.1 and / or by at least one third-party vehicle 1.2 using sensors 3. Corresponding sensor data is transmitted from the accident vehicle 1.1 or the third-party vehicle(s) 1.2 to a computing device 6 for processing. The computing device 6 is, in particular, the server or server network of a vehicle manufacturer. The vehicles participating in the method according to the invention can have been produced by one and the same vehicle manufacturer. In general, however, it is also possible for vehicles from different vehicle manufacturers to jointly carry out the method according to the invention. In this case, the computing device can also be operated by a service provider.

[0060] The computing device 6 in turn generates a synthetic accident representation 7 from the sensor data, which is stored in the Fig. 2 and Fig. 3. The synthetic accident representation 7 is output as accident information 4 via corresponding output means in a respective vehicle 1.

[0061] The synthetic accident representation 7 is abstracted, firstly to protect the data protection of the persons involved in the accident 2, and secondly to accommodate the mental resilience of the drivers. For this purpose, an abstraction indicator is maintained in an internal computing unit 8 of a respective vehicle 1, which is transmitted to the computing device 6 if necessary. The computing device 6, in turn, determines a degree of abstraction for the synthetic accident representation 7 for the respective vehicle 1 depending on the received abstraction indicator and abstracts the accident representation 7 accordingly. This accident representation 7, abstracted in this way, is transmitted back to the respective vehicle 1 and output accordingly. Acoustic and visual output means such as loudspeakers and / or display devices can be used as output means 5.

[0062] For reasons of clarity, the sensors 3, the computing unit 8 and a telecommunication unit 13 used for communication with the computing device 6 are shown only for a third vehicle 1.2 passing the accident 2.

[0063] According to one embodiment of the method according to the invention, the accident information 4 is automatically output depending on the distance of a respective vehicle 1 to the accident 2. For this purpose, a corresponding distance 9, indicated by a road area, is Fig. 1. This is an area on the side of the road where accident 2 occurred, at the end of traffic jam 12 and a sufficiently early area on the opposite lane. This ensures that the drivers of the respective vehicles 1 receive accident information 4 sufficiently early and are thus mentally prepared for accident 2.

[0064] The Fig. 2 and Fig. 3 show a different degree of abstraction of the accident representation 7. In the Fig. 2 and Fig. 3, a visual representation of the abstracted accident representation 7 is provided as an image or video, which is reproduced via a display device as output means 5.

[0065] Fig. Figure 2 shows the case where a high abstraction indicator is present, meaning that the accident representation 7 is highly abstracted. The accident vehicles 1.1 are indicated here by idealized symbolic vehicles. The arrangement of the idealized vehicles and impact symbols 14 indicates that this is a rear-end collision. In addition, a corresponding warning message can be displayed. The abstracted accident representation 7 could be Fig. 2 is merely indicated by a single image, which is illustrated by a solid line. However, it could also be an animation or a video, which is indicated by corresponding dashed lines. In such an animation, the car is shown driving into the rear of the truck.

[0066] The driver of the vehicle in which the synthetic, abstracted accident representation 7 is output recognizes early on, by viewing the accident information 4, that a rear-end collision between a car and a truck has occurred. He can mentally prepare accordingly.

[0067] Fig. 3 shows a less abstracted or more detailed accident representation 7. To indicate a high level of detail, Fig. 3 the accident scene in accordance with the representation in Fig. 1. In fact, however, it is preferably not a 2D representation from a bird's eye view, but a three-dimensional representation, in particular from the perspective of a vehicle passing by the accident 2. In the Fig. The representation shown in Figure 3 realistically depicts the visual characteristics of the accident vehicles 1.1, the emergency vehicles 1.3, and the emergency personnel 10. However, this is a computer graphic, i.e., a rendered representation. Under no circumstances are camera images or videos generated by a camera directly output as accident information 4, as this would compromise data protection and thus the information would not be presented in an abstract manner.

[0068] The corresponding video can be paused or rewound at will, as indicated by a bar 15. In addition, control panels 16 with which the driver can interact can be shown on a touch-sensitive display device. The abstracted accident representation 7 can be implemented as a virtual three-dimensional environment, similar to a video game. The driver can move within this virtual environment and thus view the accident 2 from different perspectives. For a detailed representation, high-resolution textures can be used for 3D objects. For more abstract representations, simplified 3D models can be used, in particular monochrome 3D models with a comparatively smaller number of polygons. In particular, the faces of people and vehicle license plates are falsified or distorted or replaced with idealized dummies.

[0069] By adapting the level of abstraction, the mental resilience of different drivers can be accommodated. A driver can determine for themselves which level of abstraction they prefer. According to an advantageous embodiment of the method according to the invention, however, the computing unit 8 determines the abstraction indicator and thus the level of abstraction independently. For this purpose, physiological parameters of the driver or other vehicle occupants can be recorded and monitored with the help of internal vehicle sensors. This allows the computing unit 8 to determine the level of mental strain on the driver or their stress level. After the accident information 4 has been output, a change in the mental state of the driver 1 is used as an indicator for changing the abstraction indicator. If the stress level increases, this can be an indicator that the driver is being stressed by an insufficiently abstracted accident representation 7.However, it could also be an indicator that a gawker is able to satisfy his or her desire for sensation. The physiological parameter can be evaluated once or multiple times, for example, immediately after the accident information 4 is issued and / or during or after passing the accident scene.

[0070] Instead of the Fig. 2 and Fig.The abstracted accident representation 7 shown in Figure 3 could also be formed solely by a text or a read-aloud voice message. Such a text or voice message could, for example, read in a different accident situation: "On the A81 motorway, an accident occurred at approximately 7:45 a.m. near the Böblingen Hulb exit / entrance towards Stuttgart while merging onto the motorway. Two vehicles were involved, one of which is now perpendicular to the center lane. On the A81 motorway, a complete closure of the motorway near the Böblingen Hulb exit / entrance towards Stuttgart can be expected, as a fire engine, an ambulance, and a rescue helicopter are currently approaching the probable accident scene."

Claims

[1] Method for informing the drivers of vehicles (1) about an accident (2), wherein at least one accident vehicle (1.1) is involved in an accident, the accident (2) is detected by the accident vehicle (1.1) and / or by at least one third vehicle (1.2) with the aid of sensors (3), a computing device (6) generates accident information (4) by processing corresponding sensor data, which is transmitted to a respective vehicle (1) for informing the drivers and is output there via output means (5), characterized by , that - the accident vehicle (1.1) and / or the third vehicle (1.2) transmits the sensor data to the computing device (6) for processing; - the computing device (6) generates a synthetic accident representation (7) from the sensor data; - an abstraction indicator is stored in an internal computing unit (8) of a respective vehicle (1), which is transmitted to the computing device (6); - the computing device (6) determines a degree of abstraction for the synthetic accident representation (7) as a function of the received abstraction indicator and transmits the abstracted accident representation (7) back to the respective vehicle (1); and - in a respective vehicle (1) the abstracted accident representation (7) is output as accident information (4). [2] Method according to claim 1, characterized by that the computing device (6) determines the course of an accident by processing the sensor data and supplements the synthetic accident representation (7) with the course of the accident, wherein an abstracted course of the accident and / or current accident status is output as accident information (4) in a respective vehicle (1). [3] Method according to claim 1 or 2, characterized bythat if a high abstraction indicator is present, the accident information (4) is output visually as text and / or acoustically as a read-out voice message and if a low abstraction indicator is present, the accident information (4) is output at least visually as an image or video. [4] Method according to one of claims 1 to 3, characterized by that in a respective vehicle (1) at least one physiological parameter of the driver is recorded with the aid of sensors and the computing unit (8) automatically determines the level of the abstraction indicator depending on a characteristic of the at least one physiological parameter. [5] Method according to claim 4, characterized by that the computing unit (8) increases the abstraction indicator if the at least one physiological parameter describes an increase in the driver's mental stress level after the accident information (4) is output. [6] Method according to claim 4, characterized bythat the computing unit (8) lowers the abstraction indicator if the at least one physiological parameter describes an increase in the driver's mental stress level after the accident information (4) has been output. [7] Method according to one of claims 1 to 6, characterized by that the computing device (6) uses artificial intelligence, in particular in the form of deep neural networks, to process at least part of the sensor data. [8] Method according to claim 7, characterized by in that the computing device (6) uses a large language model to describe at least part of the synthetic accident representation (7) with text, wherein at least part of the text is read in by a further Kl model to generate image content and / or video content based on speech, wherein the computing device (6) uses the image content and / or video content generated by the further Kl model to enrich the synthetic accident representation (7). [9] Method according to one of claims 1 to 8, characterized by that the accident information (4) in a vehicle (1) is automatically output as soon as the respective vehicle (1) has approached the accident site to within a specified distance (9). [10] Method according to one of claims 1 to 9, characterized by that the accident information (4) is supplemented by a degree of knowledge which describes how completely the computing device (6) has recorded the accident (2) and / or the course of the accident based on the available sensor data.

Images