Emergency call system in a motor vehicle

Abstract

The invention relates, in a first aspect, to a method for operating an emergency call system in a motor vehicle, comprising the steps: acquiring accident data, including information from the passenger compartment and / or the surroundings of the motor vehicle; and providing the accident data by means of the emergency call system, wherein the emergency call system has a natural-language conversation with an emergency call remote station on the basis of the accident data, in particular as a response to questions originating from the emergency call remote station.

Description

[0001] -2650 PIF

[0002] 1

[0003] Emergency call system in a motor vehicle

[0004] The present invention is based on an emergency call system (eCall system) in motor vehicles, as legally required in the European Union since April 2018. In the event of an accident, this system sends basic information such as the vehicle's position and the time of the accident to an emergency call center.

[0005] However, the information transmitted is usually static and limited, meaning the emergency call center receives no further details about the accident, the occupants' injuries, or the surrounding situation. This makes efficient coordination and execution of rescue measures difficult.

[0006] The invention therefore aims to extend the functionality of the emergency call system so that, in the event of an accident, detailed information about the accident, the parties involved, and the surrounding environment can be transmitted to the emergency call center. This should enable the emergency call center to coordinate rescue measures more efficiently.

[0007] The invention provides a solution according to the independent claims. Further developments of the invention are the subject of the dependent claims.

[0008] In a first aspect, a procedure for operating an emergency call system in a motor vehicle is provided, comprising the steps of capturing accident data, including information from the interior and / or surroundings of the motor vehicle, and providing the accident data by the emergency call system, wherein the emergency call system has a -2650 PIF

[0009] 2

[0010] natural language dialogue with an emergency call center based on accident data, especially in response to questions from the emergency call center.

[0011] The emergency call system can capture accident data using at least one sensor of the motor vehicle, in particular using at least one camera, and / or by accessing data from at least one control unit of the motor vehicle.

[0012] The accident data may include information on road users, injuries and / or vehicle damage.

[0013] The emergency call system can include an artificial neural network and, in particular, a large language model that processes the accident data as input.

[0014] The emergency call system can obtain accident data from a communication module of the motor vehicle, in particular from at least one V2X communication module, a mobile communication module and / or a WLAN module.

[0015] The emergency call system can obtain accident data from the vehicle's sensors, in particular acceleration sensors, radar sensors, ultrasonic sensors and / or infrared sensors.

[0016] The emergency call system can obtain accident data from cloud-based services, in particular from traffic information services, weather services and / or geoinformation services.

[0017] The emergency call system can answer questions from the emergency call center, in particular about the road users involved, the severity of injuries, vehicle damage, and the accessibility of the accident site. -2650 PIF

[0018] 3

[0019] Required rescue equipment, condition of vehicles, recovery requirements, traffic situation, weather conditions, possible landing sites for rescue helicopters, accident sequence and / or driving maneuvers shortly before the accident.

[0020] At least one control unit can be a navigation unit configured to provide map data and / or position data.

[0021] In another aspect, an emergency call system is provided for a motor vehicle, wherein the emergency call system is set up to capture accident data, including information from the interior and / or the surroundings of the motor vehicle, and to conduct a natural language dialogue with an emergency call center based on the accident data, in particular in response to questions originating from the emergency call center.

[0022] In yet another aspect, a motor vehicle is provided with an emergency system as described herein.

[0023] The present invention extends the functionality of an emergency call system in a motor vehicle to enable the transmission of detailed information about the accident, the parties involved, and the surrounding environment to the emergency call center, in particular an emergency call center, in the event of an accident. The terms emergency call center and emergency call center are used synonymously in the following.

[0024] For this purpose, the emergency call system first collects accident data using various sensors and control units of the vehicle. This accident data includes, in particular, information from the interior and surroundings of the vehicle, such as data on other road users involved, injuries, and vehicle damage. -2650 PIF

[0025] 4

[0026] The emergency call system is capable of conducting a natural language dialogue between itself and an emergency call center. The system can transmit accident data to the call center in natural language, particularly in response to questions posed by the call center. The emergency call system preferably comprises an artificial neural network, especially a Large Language Model (LLM), which processes the accident data as input. The language model enables natural language interaction with the call center and can process and transmit the recorded accident data accordingly.

[0027] The emergency call system is designed to answer questions from the emergency call center relating to various accident-related aspects, for example, at least partially to:

[0028] - Involved road users: The system can provide information on how many road users (vehicles, bicycles, pedestrians, etc.) are involved in the accident and require medical assistance.

[0029] - Severity of injuries and / or vehicle damage: Based on sensor data shortly before the accident and the vehicle's condition, the system can provide information on the type and severity of injuries (inside and outside the vehicle) as well as the extent of vehicle damage.

[0030] - Accessibility of the accident site and required rescue equipment: The system can provide information on the accessibility and nature of the accident site and assess which rescue equipment (e.g., hydraulic rescue tools, helicopters) is required. - Condition of the vehicles and recovery requirements: The system can provide information on whether the vehicles involved are still drivable and how many tow trucks are needed.

[0031] - Traffic situation, weather conditions and possible landing sites:

[0032] The system can provide information about the traffic situation on the way to -2650 PIF.

[0033] 5

[0034] Describe the accident site, weather conditions, and possible landing sites for rescue helicopters.

[0035] - Accident sequence and driving maneuvers: The system can provide information about what happened immediately before the accident (e.g., child on the road, overtaking maneuver) and which driving maneuvers were performed.

[0036] To capture this extensive accident data, the emergency call system can access various sensors and control units of the vehicle. These include cameras and navigation systems, as well as acceleration, radar, ultrasonic and infrared sensors, and communication modules for V2X, mobile communications and / or WLAN.

[0037] The emergency call system can also retrieve data from cloud-based services such as traffic information, weather and geoinformation services to supplement the environmental information for the accident site.

[0038] By recording and transmitting this detailed accident information to the emergency call center, the emergency call center receives detailed information to optimally request and dispatch the necessary emergency personnel and rescue equipment.

[0039] The core of the emergency call system is a Large Language Model (LLM), a powerful artificial neural network capable of processing and generating natural language interactions. The LLM can process accident data captured by the emergency call system as input and, based on this, conduct a natural language dialogue with the emergency call center. Through training on large text corpora, the LLM has developed a deep understanding of language, context, and semantic relationships. It can therefore semantically analyze the captured accident data and extract relevant information.

[0040] 6

[0041] extract the information and then transmit it to the emergency call center in understandable language.

[0042] Furthermore, the LLM is able to generate precise and natural-sounding answers to questions from the emergency call center regarding various accident-related aspects. To do this, it not only accesses the directly recorded accident data but also combines this data with its extensive knowledge base to answer the questions as informatively and helpfully as possible.

[0043] The LLM's capabilities enable the emergency call system to maintain a natural dialogue with the emergency call center and provide it with detailed, context-related information about the accident. This allows the emergency call center to efficiently coordinate the necessary rescue measures and provide the best possible support to the emergency services.

[0044] The emergency call system can be implemented, at least partially, either directly in the vehicle or in a remote computer system in order to make optimal use of processing power.

[0045] The Large Language Model (LLM) in the emergency call system can be trained in various ways to optimize its ability to process and generate natural language dialogues:

[0046] - General language comprehension: Initially, the LLM can be trained on large, general text corpora to develop a comprehensive understanding of language, grammar, semantics, and context. This includes, in particular, text collections from books, newspapers, websites, and / or other public sources.

[0047] - Domain-specific knowledge: In addition, the LLM can be adapted with datasets from the fields of vehicle engineering, traffic safety and emergency rescue, e.g. using -2650 PIF.

[0048] 7

[0049] a so-called "Retrieval-Augmented Generation" (RAG) approach. This includes technical documentation, accident databases, emergency call logs, and other relevant technical publications. Through this domain-specific training, the LLM has developed an in-depth understanding of vehicle technology, traffic flow, accident mechanisms, and / or rescue measures. They can therefore precisely interpret the collected accident data and generate informative responses based on it.

[0050] - Interactive training: To further improve the LLM's natural language interaction skills, they can also be trained in simulated dialogue situations. They are encouraged to respond to typical questions and statements from the emergency call center as naturally and helpfully as possible. Through this interactive training, the LLM can continuously refine their dialogue skills, context understanding, and response generation, ultimately enabling fluent and effective communication with the emergency call center.

[0051] - Continuous learning: Furthermore, the LLM-based emergency call system is able to learn from every actual interaction, especially with the emergency call center. By recording and analyzing dialogues, the system can continuously improve its understanding and dialogue skills. In this way, the LLM can increase its performance over time and better respond to the needs and requirements of the emergency call center.

[0052] The accident data recorded by the emergency call system in the vehicle can be fed into the Large Language Model (LLM) in the emergency call system in various forms, e.g., as structured data input. Some of the accident data is already in structured form, such as sensor data, parameter values, or recordings from PIF-2650.

[0053] 8

[0054] Control units. This structured data can be transmitted directly to the LLM as numerical or categorical inputs.

[0055] Examples include vehicle speed, acceleration values, sensor information on objects, distances, movements, and vehicle status data such as airbag deployment, door opening, and / or power supply. Other accident data may be in the form of textual and / or visual descriptions, such as recordings from audio and / or video systems, e.g., transcripts of voice recordings. This unstructured text information can be transmitted directly to the LLM as input.

[0056] The LLM is capable of semantically analyzing this natural language text data and extracting the relevant information. The described data can also be transmitted to the emergency call center by the emergency call system in structured and / or unstructured form, particularly in textual form. Furthermore, the emergency call system can also transmit multimodal input to the LLM, i.e., a combination of structured data and textual and / or visual descriptions.

[0057] Sensor data on vehicle movement and position can be used in conjunction with camera footage and transcripts of voice recordings. The LLM can integrate these diverse data sources and derive a comprehensive understanding of the accident. This flexible processing of various data formats allows the LLM-based emergency call system to optimally utilize the captured accident data to transmit detailed and meaningful information to the emergency call center. -2650 PIF

[0058] 9

[0059] The invention is now also described with regard to the figures. They show:

[0060] Fig. 1 schematically shows a method according to the invention, and Fig. 2 schematically shows an emergency system according to the invention.

[0061] Figure 1 shows a flowchart of the method according to the invention. In a first step S10, the accident data is determined by the emergency call system. For this purpose, the emergency call system uses, in particular, data from vehicle sensors and / or control units of the vehicle. In a second step S20, the accident data is provided by the emergency call system. The emergency call system can contact a remote emergency call center and transmit the accident data in natural language, as audio output, and / or in text form. The emergency call system preferably answers questions posed by the emergency call center. In particular, based on the accident data, the emergency call system describes the condition of the occupants, the surroundings, the vehicle, and / or other road users. The dialogue can be conducted via voice or, for example, as part of a text-based conversation (text chat). Furthermore, the emergency call system can transmit image and / or audio material to the emergency call center.

[0062] The NRS emergency call system comprises a sensor network 101, which includes various sensors and control units of the vehicle. These include, for example, cameras 102, radar sensors 103, ultrasonic sensors 104, accelerometers 105, and communication modules 106 for V2X (vehicle-to-X communication), cellular networks, and WLAN. Using this sensor system, the emergency call system acquires accident data from the interior and surroundings of the vehicle.

[0063] The recorded accident data is processed by the NRS emergency call system using a Large Language Model (LLM) processor 108, based on the -2650 PIF.

[0064] 10

[0065] an artificial neural network can conduct a natural language dialogue with the emergency call center 109.

[0066] Through this dialogue, the NRS emergency call system can transmit detailed accident data to the emergency call center 109 and answer questions from the emergency call center 109 on various accident-related aspects.

[0067] In addition, the NRS emergency call system can also retrieve data from cloud-based services 107 such as traffic information, weather and geoinformation services to supplement the environmental information for the accident site.

[0068] An example of a natural language dialogue between the emergency call system and the emergency call center could proceed as follows:

[0069] Emergency call center: "This is the emergency call center. We have received your emergency call. Can you please give me details about the accident?"

[0070] Emergency call system NRS (LLM processor): "Yes, gladly. According to the sensor data in my vehicle, there has been a rear-end collision. Another vehicle rear-ended me while I was stopped at a red light."

[0071] Emergency call center 109: "Understood. How many vehicles are involved in the accident in total?"

[0072] Emergency call system NRS (LLM processor): "From the camera images and the radar sensors, we can see that another vehicle besides my vehicle is involved in the accident. It is a car that rear-ended me."

[0073] Emergency Call Center 109: "Okay, thank you. And is there any information on injuries to the occupants?" 24-2650 PIF

[0074] 11

[0075] Emergency Call System NRS (LLM Processor): "Yes, according to the acceleration sensors and the information from the vehicle interior, the driver in my vehicle appears to be slightly injured. He is complaining of chest pain. In the other vehicle, we have indications of moderate injuries to the driver."

[0076] Emergency call center 109: "Understood. Can you also say something about the accessibility of the accident site?"

[0077] Emergency call system NRS (LLM processor): "Yes, according to the GPS data and map data from the navigation system, the accident is located at an intersection with a multi-lane road. The accident site is easily accessible, and there is also a possible landing site for a rescue helicopter nearby."

[0078] Emergency call center 109: "Is there anything else you can tell me?"

[0079] Emergency call system NRS (LLM processor): "Yes, according to the weather data I retrieved from a cloud service, there is currently light rain and some wind at the accident site. According to traffic information services and my camera data, the traffic situation on the access roads is currently calm, so the emergency services can arrive quickly."

[0080] The at least one communication module 106 serves in particular for communication with the emergency call center 109 and / or communication with the aforementioned cloud services 107.

Claims

24-2650 PIF Claims 1. Procedure for operating an emergency call system (NRS) in a motor vehicle, comprising the steps - Recording accident data, including information from the interior and / or surroundings of the vehicle, - Provision of accident data by the emergency call system (NRS), wherein the emergency call system conducts a natural language dialogue with an emergency call center (109) based on the accident data, in particular in response to questions originating from the emergency call center (NRS).

2. Method according to claim 1, wherein the emergency call system (NRS) captures the accident data by means of at least one sensor of the motor vehicle, in particular with at least one camera (102), and / or by accessing data from at least one control unit of the motor vehicle.

3. Method according to claim 1 or 2, wherein the accident data includes information on road users, injuries and / or vehicle damage.

4. Method according to one of the preceding claims, wherein the emergency call system (NRS) comprises an artificial neural network (108) and in particular a large language model that processes the accident data as input.

5. Method according to any of the preceding claims, wherein the emergency call system (NRS) obtains accident data from at least one communication module of the motor vehicle, in particular from 24-2650 PIF 13 at least one V2X communication module (106), one cellular module and / or one WLAN module 6. Method according to one of the preceding claims, wherein the emergency call system obtains accident data from further sensors of the motor vehicle, in particular from acceleration sensors (105), radar sensors (103), ultrasonic sensors (104) and / or infrared sensors.

7. Method according to any of the preceding claims, wherein the emergency call system (NRS) obtains accident data from cloud-based services (107), in particular from traffic information services, weather services and / or geoinformation services.

8. Method according to any of the preceding claims, wherein the emergency call system (NRS) is configured to answer questions originating from the emergency call center (109), in particular regarding road users involved, severity of injuries, vehicle damage, accessibility of the accident site, required rescue equipment, condition of vehicles, recovery requirements, traffic situation, weather conditions, possible landing sites for rescue helicopters, accident sequence and / or driving maneuvers shortly before the accident.

9. Method according to claim 2, wherein the at least one control unit is a navigation unit configured to provide map data and / or position data.

10. Emergency call system in a motor vehicle, wherein the emergency call system (NRS) is designed to capture accident data, including information from the interior and / or surroundings of the motor vehicle, and to conduct a natural language dialogue with an emergency call center. 24-2650 PIF 14 (109) based on the accident data, in particular in response to questions originating from the emergency call center (109).

11. Vehicle with an emergency call system according to claim 10.

Images