Method for driving a vehicle, computer-implemented training method for training a machine learning model, data processing system and computer program product

Abstract

To enable at least partially automated driving of a vehicle (10), position data representing the current position of the vehicle (10) in a predefined reference coordinate system is received, a current geographic area (7a, 7b, 7c) in which the vehicle (10) is located is determined by applying a trained machine learning model (MLM) (1) to input data that depends on position data, and a traffic control for the vehicle (10) is determined based on the current geographic area (7a, 7b, 7c). At least one control signal for driving the vehicle (10) is generated depending on the traffic control, and / or assistance information to assist a driver of the vehicle (10) in driving the vehicle (10) is generated depending on the traffic control.

Description

[0001] The present invention relates to a method for at least partially automating the driving of a vehicle, to a computer-implemented training method for training a machine learning model for use in such a method, to a corresponding data processing system, and to a corresponding computer program product.

[0002] Many driver assistance functions, or other functions for at least partially automated vehicle operation, require knowledge of country-specific traffic regulations, such as general or implied speed limits. This applies, for example, to adaptive cruise control (ACC) functions or other functions that involve longitudinal vehicle control.

[0003] For this purpose, conventional systems retrieve a country code for the vehicle's current position from a digital map based on GPS data. If the digital map is stored in the vehicle itself, a disadvantage is the required storage space. Using a simplified map would result in reduced accuracy and therefore reduced reliability of the country code. If the digital map is stored externally, for example in a cloud storage system, its availability may be limited, for instance in the event of connection problems.

[0004] An objective of the present invention is to provide an improved concept for at least partially automatic driving of a vehicle based on a traffic control system that depends on the current geographical position of the vehicle and does not require map data at runtime.

[0005] This objective is achieved through the respective subject matter of the independent claims. Further embodiments and preferred configurations are the subject matter of the dependent claims.

[0006] The invention is based on the idea of ​​providing a trained machine learning model, MLM, which is capable of determining the current geographical area in which the vehicle is located, based on the current position of the vehicle.

[0007] According to one aspect of the invention, a method for at least partially automating the control of a vehicle, in particular a motor vehicle, for example a land vehicle such as a car, truck, van, or motorcycle, is provided. The following steps are performed by the data processing system: Position data representing the current position of the vehicle in a predefined reference coordinate system is received. A current geographical area in which the vehicle is located is determined by applying a trained machine learning model (MLM) to input data that depends on, includes, or consists of the position data. A traffic control strategy for the vehicle is then determined based on this current geographical area.At least one control signal for driving the vehicle is generated depending on the traffic control system and / or assistance information to assist a driver of the vehicle in driving the vehicle is generated depending on the traffic control system.

[0008] It is possible that the method according to the invention is purely computer-implemented. In this case, all steps of the computer-implemented method can be executed by the data processing system, which includes at least one data processing device, in particular a data processing system of the vehicle. Specifically, the at least one data processing device is configured or adapted to execute the steps of the computer-implemented method. For this purpose, the at least one data processing device can, for example, store a computer program containing instructions which, when executed by the at least one data processing device, cause the at least one data processing device to carry out the computer-implemented method. The terms "data processing system" and "at least one data processing device" can be used synonymously.

[0009] All data processing devices of at least one data processing device can be integrated into the vehicle. However, it is also possible that all data processing devices of at least one data processing device are part of an external computing system located outside the vehicle, for example, a mobile electronic device, a backend server, or a cloud computing system.

[0010] It is also possible that the at least one data processing device comprises at least one vehicle data processing device and at least one external data processing device included by the external computing system. The at least one vehicle data processing device may, for example, include one or more electronic control units (ECUs), and / or one or more zone control units (ZCUs), and / or one or more domain control units (DCUs) of the vehicle.

[0011] If the at least one data processing device comprises two or more data processing devices, certain steps performed by the at least one data processing device can be understood as different data processing devices performing different steps or different parts of a step. In particular, it is not necessary for each data processing device to perform the steps completely. In other words, the execution of the steps can be distributed among the two or more data processing devices.

[0012] From each execution of the computer-implemented procedure, a corresponding execution of the procedure that is not purely computer-implemented is obtained by including corresponding steps of at least partially automatic driving of the vehicle by controlling one or more actuators of the vehicle depending on the at least one control signal.

[0013] The at least one control signal can, for example, be provided to one or more of the vehicle's actuators, including one or more brake actuators and / or one or more steering actuators and / or one or more drive motors. The one or more actuators can influence the longitudinal and / or lateral steering of the vehicle in order to steer the vehicle at least partially automatically based on the at least one control signal.

[0014] The assistance information can be output via a vehicle output device, for example a display and / or an audio output system and / or a haptic output system.

[0015] The position data can be received, for example, by a GNSS (Global Navigation Satellite System) receiver in the vehicle, such as a GPS, GLONASS, BeiDou, or Galileo receiver. The position data can include, for example, the current position in the form of latitude and longitude. However, other formats can also be used.

[0016] The current geographic area is, in particular, a predefined geographic area, for example, one of a multitude of predefined geographic areas. A geographic area is defined, in particular, by its boundary or limit, which is given, for example, by a closed curve. A position enclosed by the boundary is considered to lie within the respective geographic area, while a position not enclosed by the boundary is considered to lie outside the respective geographic area. For positions that lie exactly on the boundary, a predefined convention can be used to decide whether the position lies within the respective geographic area or outside and, for example, within another geographic area. A geographic area can correspond to a country or a state.However, this is not absolutely necessary for the method according to the invention.

[0017] A trained MLM can be understood as an algorithm, particularly a computer-implemented algorithm, that can reproduce functions that are concretely or more broadly possible through human cognitive effort. A trained MLM can also be referred to as a "trained function." A trained MLM can be implemented in software and / or hardware.

[0018] When training an MLM, its parameters are generally adjusted or updated. Training can be supervised, partially supervised, or unsupervised. It can also include reinforcement learning, representation learning, and / or other known training methods. In particular, the MLM's parameters can be iteratively adjusted over multiple training steps. Specifically, a predefined loss function can be minimized during training. If the MLM is an artificial neural network (ANN), a backpropagation algorithm can be used to adjust the parameters.

[0019] In particular, an MLM can include an ANN, a support vector machine, a k-means clustering algorithm, a decision tree, and so on. Specifically, an ANN can be or include a deep neural network and / or a convolutional neural network, CNN (in particular a deep CNN) and / or a recurrent neural network, RNN (in particular a recurrent CNN), and / or a transformer network and / or a generative adversarial network, GAN.

[0020] In the present context, the trained MLM is, for example, stored on a storage device of the data processing system and / or implemented in hardware on the data processing system.

[0021] Training the MLM is not necessarily part of the method according to the invention. In other words, the MLM may have been trained before the method according to the invention is carried out. In some embodiments of the method according to the invention, training the MLM, or partially training the MLM, may be part of the method according to the invention. For this purpose, for example, a computer-implemented training method according to the invention may be used, as described further below in this disclosure.

[0022] In any case, when applied to the input data, the trained MLM has been trained to predict a geographic area based on positional data as input. This can be achieved, for example, by supervising and training the MLM to perform a specific classification task.

[0023] Thus, according to the inventive method, the geographical area is determined solely based on the position data, without the need for map data for comparison. However, this does not preclude the use of a map or map-like dataset defining the geographical areas for training the MLM. Once the MLM has been trained accordingly, it can be deployed in the vehicle or, conversely, in an entire fleet of vehicles. The required storage space for the trained MLM, particularly on a storage device of the data processing system, can be significantly reduced compared to the storage space required for an accurate digital map. Furthermore, it is not necessary for the data processing system to communicate with an external cloud server to retrieve the respective map data. Such communication could be affected by connection problems or similar issues.

[0024] According to some embodiments, the MLM comprises an ANN, in particular an ANN for classification, especially for multi-class classification, for example a multi-layer perceptron.

[0025] According to some embodiments, when the MLM is applied to the input data, a classification task is performed according to which one geographic area from a multitude of predefined geographic areas is predicted depending on the input data, and the current geographic area corresponds to the predicted geographic area.

[0026] In other words, the multitude of predefined geographic areas corresponds to several classes of the multi-class classification. Various MLM designs, particularly ANN designs, for multi-class classification are known and can be used in the method according to the invention. This results in a fast, accurate, and reliable prediction of the current geographic area.

[0027] According to some embodiments, determining the current geographical area involves determining a country code that corresponds to the geographical area, and the traffic regulation is determined depending on the country code.

[0028] A country code is an identification code, for example a numeric or alphanumeric code, that identifies countries or similar geographical areas. Since traffic regulations often apply to entire countries, the country code is particularly suitable for specifying the current geographical area in the context of the present invention.

[0029] In particular, in certain embodiments, each geographical area of ​​the multitude of geographical areas can be assigned a corresponding country code, and the MLM can directly or indirectly predict the respective country code for the current geographical area.

[0030] Various conventions can be used for the country code, including, for example, the two-letter code according to ISO 3166-1 alpha-2 (for example, FR for France, IE for Ireland or DE for Germany), the three-letter code according to ISO 3166-1 alpha-3 (for example, FRA for France, IRL for Ireland or DEU for Germany) or the three-digit code according to ISO 3166-1 numeric (for example, 250 for France, 372 for Ireland or 276 for Germany).

[0031] According to some embodiments, the traffic regulation corresponds at least to a speed limit for the current geographical area.

[0032] A speed limit, for example a general or implicit speed limit, is a speed limit that is not necessarily indicated by a traffic sign, but applies generally in the relevant geographical area or on specific types of roads within that geographical area, such as motorways, urban streets, rural roads, etc. Consequently, in some implementations, at least one speed limit may include different speed limits for different types of roads.

[0033] Therefore, the method according to the invention is particularly suitable for such speed limits as traffic control, since they cannot usually be determined based on environmental sensor systems that perceive the vehicle's surroundings.

[0034] According to some implementations, traffic control is determined based on a database.

[0035] In particular, the database may contain a specific traffic regulation for each of the numerous predefined geographical areas, for example, for the respective MLM outputs, especially the country code. Based on the MLM output indicating the current geographical area, for example, the country code of the current geographical area, the data processing system can then retrieve the corresponding traffic regulation from the database.

[0036] The database can be stored, for example, on a storage device of the data processing system, such as a storage device in the vehicle. This enables fast and reliable access to the traffic control information. Furthermore, the database can be updated promptly when traffic regulations change, either centrally or, if stored on a storage device in the vehicle, via an over-the-air (OTA) or wired update.

[0037] According to some implementations, the assistance information includes a warning message indicating the traffic control.

[0038] The warning may, for example, include a reference to the traffic regulation or part of the traffic regulation, such as a speed limit, that is relevant or valid for the current position of the vehicle.

[0039] This can help the driver of the vehicle to act in accordance with traffic regulations, for example, at least one speed limit.

[0040] According to some embodiments, at least one control signal is generated to guide the vehicle in order to limit the current speed of the vehicle according to at least one speed limit.

[0041] In other words, if one or more actuators are controlled based on at least one control signal, the vehicle is at least partially automatically controlled so that its speed does not exceed the speed limit applicable to the vehicle's current position. For example, it may be provided that the vehicle's speed is not actively controlled automatically until its current speed is less than the speed limit, but that acceleration beyond the speed limit is automatically prevented. However, this does not preclude exceeding the speed limit in emergency situations or similar circumstances.

[0042] According to some embodiments, at least one control signal is generated to guide the vehicle in order to automatically control a current speed of the vehicle, with the controlled speed being limited according to the at least one speed limit.

[0043] In other words, if one or more actuators are controlled based on at least one control signal, the vehicle's speed is automatically and actively controlled to ensure that it does not exceed the speed limit applicable to the vehicle's current position. This can be, for example, part of an ACC (Adaptive Cruise Control) function. However, this does not preclude exceeding the speed limit in emergency situations and similar circumstances.

[0044] According to some embodiments, sensor data representing an environment, in particular an external environment, of the vehicle are received by an environmental sensor system of the vehicle, and at least one control signal and / or assistance information are generated depending on the sensor data.

[0045] For example, the sensor data can be analyzed to determine the current road type of the road the vehicle is on, and the currently applicable speed limit is determined based on the road type and traffic regulations. This further improves the reliability of the process.

[0046] An environmental sensor system can be understood, for example, as a sensor system capable of generating sensor data or sensor signals that reproduce, represent, or map the environment of the environmental sensor system. In particular, the ability to capture or detect electromagnetic or other signals from the environment cannot be considered a sufficient condition for a sensor system to be designated as an environmental sensor system. For example, cameras, especially those operating in the visible spectrum or in the infrared spectrum, lidar systems, radar systems, or ultrasonic sensor systems can be considered environmental sensor systems.

[0047] The sensor data can therefore include, for example, one or more camera images, thermal images, lidar point clouds, lidar depth images, radar data, ultrasound data and / or corresponding images generated based on the radar or ultrasound data, according to the design of the environmental sensor system.

[0048] According to some embodiments, the environmental sensor system includes a camera and / or a lidar system.

[0049] According to a further aspect of the invention, a computer-implemented training method for training a multi-level language (MLM) for use in a method for at least partially automated driving of a vehicle, in particular a method according to the invention, is provided. A plurality of training data sets are received, each training data set comprising training position data representing a respective training position in the reference coordinate system, and annotation data specifying a geographical area in which the respective training position is located. The MLM is supervised trained, particularly in an untrained or partially trained state, based on the plurality of training data sets to perform a classification task.

[0050] Unless otherwise specified, all steps of the computer-implemented training procedure can be performed by another data processing system, which includes at least one other data processing device. In particular, this at least one other data processing device is configured or adapted to perform the steps of the computer-implemented training procedure. For this purpose, the at least one other data processing device can, for example, store another computer program containing further instructions which, when executed by the at least one other data processing device, cause it to perform the computer-implemented procedure.

[0051] If the at least one additional data processing device comprises two or more additional data processing devices, certain steps performed by the at least one additional data processing device can be understood as different additional data processing devices performing different steps or different parts of a step. In particular, it is not necessary for each additional data processing device to perform the steps completely. In other words, the execution of the steps can be distributed among the two or more additional data processing devices.

[0052] In particular, to train the MLM in a supervised manner based on the multitude of training datasets to perform the classification task, the following steps can be executed iteratively for each of the multitude of training datasets. For a given iteration, the MLM is applied to a respective training dataset from the multitude of training datasets to predict a geographic area from a multitude of predefined geographic areas. A predefined loss function is evaluated depending on the predicted geographic area and the area specified by the respective annotation data of the training dataset.Depending on the result of the loss function evaluation, particularly the corresponding value of the loss function, the MLM is updated and a subsequent iteration is performed with a different training dataset from the multitude of training datasets if a predefined termination criterion or convergence criterion concerning the value of the loss function is not met. If the termination criterion or convergence criterion is met, the iterations are terminated.

[0053] If the MLM includes an ANN, updating the MLM involves updating the respective network parameters of the ANN, in particular weighting factors and / or bias factors of the ANN. Known algorithms, for example a backpropagation algorithm, can be used for this purpose.

[0054] The loss function can be a conventional loss function used for multi-label or multi-class classification tasks, for example, a categorical cross-entropy loss.

[0055] Once the training is complete, the MLM is therefore able to predict a corresponding geographical area based on entered position data, even if the position data is not identical to the training position data of one of the many training datasets.

[0056] In some embodiments, geographic area data is obtained, comprising an assignment of each geographic area from a multitude of predefined geographic areas to each reference position from a multitude of reference positions in the reference coordinate system. The multitude of training datasets is generated based on this geographic area data.

[0057] In particular, a corresponding training dataset can be generated for each of the reference positions, containing the respective reference position as the training position data of the respective training dataset and the associated geographic area according to the geographic area data. Alternatively or additionally, further training position data that is not identical to any of the reference positions can be used to generate further training datasets, whereby the geographic area is determined based on the geographic area data, for example by interpolation or by identifying the reference position that is closest to the respective further training position data, and using the geographic area that is assigned to the next reference position for the further training dataset.

[0058] With such designs, a very large number of training data sets can be generated in a simple way.

[0059] According to some embodiments, for each training dataset, at least a subset of the multitude of training datasets receives the respective training position data, and a reference position from the multitude of reference positions corresponding to the respective training position is determined. The annotation data of the respective training dataset is determined such that it specifies the geographical area assigned to the respective reference position determined for the respective training position data.

[0060] In particular, the reference position determined for the respective training position data corresponds to the reference position that is closest to the respective training position data.

[0061] According to some embodiments, at least for a further subset of the multitude of geographical areas, the spatial density of the reference positions of a respective geographical area within a predefined boundary area of ​​the respective geographical area is greater than the spatial density of the reference positions of the respective geographical area outside the boundary area of ​​the respective geographical area.

[0062] The further subset can be identical to the subset or differ from the subset mentioned above.

[0063] Spatial density corresponds, in particular, to the number of reference positions per unit area. The boundary of a given geographic area can, for example, be defined as an area within a predefined distance from the boundary of that geographic area.

[0064] In such implementations, the training focus is increased within the boundary regions. Thus, the trained MLM will generate more accurate or reliable predictions in the boundary regions, where potential inaccuracies could lead to incorrect predictions of the geographic area based on the input positional data.

[0065] Further embodiments of the computer-implemented training method according to the invention arise directly from the various embodiments of the method according to the invention for driving a vehicle and vice versa. In particular, individual features and corresponding explanations as well as advantages concerning the various embodiments of the method according to the invention for driving a vehicle can be transferred accordingly to embodiments of the computer-implemented training method according to the invention and vice versa.

[0066] According to some embodiments of the method for at least partially automatic driving of a vehicle, the MLM is trained by using a computer-implemented training method according to the invention.

[0067] In other words, in such embodiments, the steps for training the MLM are also part of the procedure for at least partially automated driving of the vehicle.

[0068] According to a further aspect of the invention, a data processing system is provided which is set up to carry out an inventive method for at least partially automatic driving of a vehicle and / or an inventive computer-implemented training method.

[0069] The terms "data processing system" and "at least one data processing device" can be used synonymously in this disclosure. In this disclosure, a data processing device, also referred to as a computing device, can be understood, for example, as a device with processing circuitry for data processing. A data processing device can therefore perform computational operations to process data. Indexed access to a data structure, such as a lookup table (LUT) or a database, can also be considered a computational operation. Data processing that is partially or completely implemented in hardware can likewise be considered a computational operation.

[0070] In particular, a data processing device may contain one or more computers, one or more microcontrollers, and / or one or more integrated circuits, for example, one or more application-specific integrated circuits (ASICs), one or more field-programmable gate arrays (FPGAs), and / or one or more systems on a chip (SoCs). A data processing device may also contain one or more processors, for example, one or more microprocessors, one or more central processing units (CPUs), one or more graphics processing units (GPUs), and / or one or more signal processors, in particular one or more digital signal processors (DSPs).The data processing device may also include a physical or virtual network of computers or other units of the aforementioned type.

[0071] A data processing device may also include one or more hardware and / or software interfaces, for example for receiving and / or providing data.

[0072] A data processing device can also include one or more storage devices. A storage device can be volatile data storage, for example, dynamic random access memory (DRAM) or static random access memory (SRAM), or non-volatile data storage, for example, read-only memory (ROM), programmable read-only memory (PROM), erasable read-only memory (EPROM), electrically erasable read-only memory (EEPROM), flash memory or flash EEPROM, ferroelectric random access memory (FRAM), or magnetoresistive random access memory.It can be implemented as MRAM (magnetoresistive random access memory) or as phase-change random access memory, PCRAM (phase-change random access memory).

[0073] According to another aspect of the invention, an electronic vehicle guidance system is provided which includes a data processing device that is configured to carry out a method according to the invention for at least partially automatic guidance of a vehicle.

[0074] An electronic vehicle control system can be understood as an electronic system designed to control a vehicle fully automatically or autonomously, and in particular, without the need for manual intervention or control by a driver or user of the vehicle. The vehicle automatically performs all necessary functions, such as steering, braking, and / or acceleration maneuvers, as well as monitoring and recording road traffic and reacting accordingly. Specifically, the electronic vehicle control system can implement a fully automatic or fully autonomous driving mode according to Level 5 of the SAE J3016 classification. An electronic vehicle control system can also be implemented as an Advanced Driver Assistance System (ADAS), which assists a driver in semi-automated or semi-autonomous driving.In particular, the electronic vehicle guidance system can implement a semi-automatic or semi-autonomous driving mode according to levels 1 to 4 of the SAE J3016 classification. Here and in the following, SAE J3016 refers to the corresponding standard dated April 2021.

[0075] At least partially automated driving of the vehicle can therefore include driving the vehicle in a fully automated or fully autonomous driving mode according to level 5 of the SAE J3016 classification. At least partially automated driving of the vehicle can also include driving the vehicle in a semi-automated or semi-autonomous driving mode according to levels 1 to 4 of the SAE J3016 classification.

[0076] According to some embodiments, the electronic vehicle guidance system includes one or more actuators and / or the GNSS receiver and / or the environmental sensor system.

[0077] Further embodiments of the electronic vehicle guidance system according to the invention arise directly from the various embodiments of the method according to the invention for at least partially automatic vehicle control and vice versa. In particular, individual features and corresponding explanations as well as advantages relating to the various embodiments of the method according to the invention can be transferred accordingly to corresponding embodiments of the electronic vehicle guidance system according to the invention. In particular, the electronic vehicle guidance system according to the invention is designed or programmed to carry out the method according to the invention. In particular, the electronic vehicle guidance system according to the invention carries out the method according to the invention.

[0078] According to another aspect of the invention, a computer program containing instructions is provided. When these instructions are executed by a data processing system, they cause the data processing system to carry out a method according to the invention for at least partially automating the driving of a vehicle.

[0079] The commands can be provided, for example, as program code. The program code can be provided, for example, as binary code or assembly language, and / or as source code in a programming language such as C, and / or as a program script, such as Python.

[0080] According to another aspect of the invention, a further computer program containing additional instructions is provided. When these additional instructions are executed by a data processing system, they cause the data processing system to perform a computer-implemented training method according to the invention.

[0081] The additional commands can be provided, for example, as program code. This program code can be provided, for example, as binary code or assembly language, and / or as source code in a programming language such as C, and / or as a program script, such as Python.

[0082] According to another aspect of the invention, a computer-readable storage medium is provided which stores a computer program and / or another computer program according to the invention.

[0083] The computer program, the further computer program, and the computer-readable storage medium are respective computer program products that contain the commands and / or the further commands.

[0084] Further features of the invention will become apparent from the claims, the figures, and the description of the figures. The features and combinations of features mentioned above in the description, as well as the features and combinations of features mentioned below in the description of the figures and / or shown in the figures, may be encompassed by the invention not only in the combinations specified, but also in other combinations. In particular, the invention may also encompass embodiments and combinations of features that do not have all the features of an originally formulated claim. Furthermore, the invention may encompass embodiments and combinations of features that go beyond or deviate from the combinations of features set out in the references to the claims.

[0085] The invention is explained in detail below with reference to specific exemplary embodiments and corresponding schematic drawings. Identical and functionally equivalent elements may be designated by the same reference numerals in the drawings. The description of identical or functionally equivalent elements is not necessarily repeated with reference to other figures.

[0086] The figures show: Fig. 1 a schematic flowchart of an exemplary embodiment of a method according to the invention for at least partially automatic driving of a vehicle; Fig. 2 a schematic flowchart of an exemplary embodiment of a computer-implemented training method according to the invention; and Fig. 3 a schematic representation of a vehicle with an exemplary embodiment of an electronic vehicle guidance system according to the invention.

[0087] In Fig. Figure 1 is a flowchart of an exemplary embodiment of a method according to the invention for at least partially automatic driving of a vehicle 1 (see Figure 1). Fig. 3) shown.

[0088] In step 100, a data processing system 11 receives position data representing the current position of the vehicle 10 in a predefined reference coordinate system, and in step 120 determines a current geographical area 7a, 7b, 7c (see Fig. 2), in which the vehicle 10 is currently located, by means of an MLM 1 (see Fig. 2 and Fig. 3) is applied to input data that depends on the position data. A traffic control for the vehicle 10 is determined by the data processing system 11 based on the current geographical area 7a, 7b, 7c, for example from a database 15, and the data processing system 11, for example a vehicle control system 16 of the data processing system 11, generates at least one control signal for driving the vehicle 10 and / or assistance information to assist a driver of the vehicle 10 in driving the vehicle 10 depending on the traffic control.

[0089] The method can, for example, be carried out by an electronic vehicle guidance system 14 according to the invention. The electronic vehicle guidance system 14 includes at least the data processing system 11 according to the invention or consists of a data processing system 11 according to the invention. The data processing system 11 is configured to execute the steps of the method according to the invention and can, for example, store the trained MLM 1 for this purpose. Fig. Figure 3 shows a schematic representation of a vehicle 10 with a further exemplary embodiment of an electronic vehicle guidance system 14 according to the invention.

[0090] In some embodiments, the vehicle 10, in particular the electronic vehicle guidance system 14, has a GNSS receiver 12 which is configured to generate the position data and provide it to the data processing system 11.

[0091] In some embodiments, the vehicle 10, in particular the electronic vehicle guidance system 14, has an environmental sensor system 13, for example a camera, which is configured to generate sensor data representing the external environment of the vehicle 10. The at least one control signal and / or the assistance information are generated by the data processing system 11 depending on the sensor data.

[0092] Fig. Figure 3 shows a schematic flowchart of an exemplary embodiment of a computer-implemented training method according to the invention for training the MLM 1 for use in a method according to the invention for at least partially automatic driving of a vehicle, for example a method such as is described with reference to Fig. 1 and / or Fig. 2 is described.

[0093] A multitude of training datasets 2, 4 are received, each comprising training position data representing a specific training position 2 in the reference coordinate system, and annotation data specifying a geographic area 7a, 7b, 7c in which the respective training position 2 is located. The MLM 1 is supervised trained on this multitude of training datasets 2, 4 to perform a classification task.

[0094] For example, geographic area data 3 can be provided, for instance, in the form of a shape file that defines a variety of predefined geographic areas 7a, 7b, 7c with respect to their shapes, in particular their boundaries or limits 9. Alternatively, a separate shape file can be provided for each of the geographic areas 7a, 7b, 7c or for predefined groups of geographic areas 7a, 7b, 7c. The geographic area data 3 include a respective mapping of each of the respective geographic areas 7a, 7b, 7c to each of a variety of reference positions 8.

[0095] For a given training iteration, the corresponding training position data are received. The training position data can correspond to one of the reference positions 8. Otherwise, a reference position 8 is determined from the multitude of reference positions 8 that corresponds to the respective training position 2, for example, by determining the reference position 8 that is closest to the training position data 2. The annotation data 4 of the respective training dataset 2, 4 are determined such that they specify the geographic area 7a, 7b, 7c which is assigned to the respective reference position 8. The MLM 1 is then applied to the training dataset 2, 4 to predict a geographic area 7a, 7b, 7c.A predefined loss function 6 is evaluated depending on the predicted geographic area 7a, 7b, 7c and the geographic area 7a, 7b, 7c specified by the corresponding annotation data 4 of the training dataset 2, 4, for example by calculating a categorical cross-entropy loss. Depending on the result of the evaluation of the loss function 6, in particular the corresponding categorical cross-entropy loss, the MLM 1 is updated, for example by means of a backpropagation algorithm.

[0096] In some embodiments, the spatial density of the reference positions 8 of a respective geographic area 7a, 7b, 7c within a predefined boundary of the corresponding geographic area 7a, 7b, 7c is greater than the spatial density of the reference positions 8 of the respective geographic area 7a, 7b, 7c outside the boundary. This increases the robustness of the trained MLM 1 in predicting a geographic area 7a, 7b, 7c based on positional data that specifies a position near a boundary 9, where the risk of determining the wrong geographic area is increased because two geographic areas 7a, 7b, 7c are adjacent there.

[0097] As explained, the invention represents an improved concept for at least partially automatic driving of a vehicle 1 based on a traffic control system that depends on the current geographical position of the vehicle 1, without requiring map data at runtime.

[0098] In particular, some implementations offer further advantages such as reduced latency, increased accuracy, no need for frequent map data updates, and / or less support required for the map data. Furthermore, no application programming interface (API) is needed to retrieve map data from an external server, etc. QUOTES INCLUDED IN THE DESCRIPTION

[0000] This list of documents cited by the applicant was automatically generated and is included solely for the reader's convenience. The list is not part of the German patent or utility model application. The DPMA accepts no liability for any errors or omissions. Cited non-patent literature

[0000] ISO 3166-1

[0030]

Claims

[1] Method for at least partially automated driving of a vehicle (10) wherein a data processing system (11) - Position data representing the current position of the vehicle (10) in a predefined reference coordinate system are received; - a present geographical area (7a, 7b, 7c) in which the vehicle (10) is located is determined by applying a trained machine learning model, MLM, (1) to input data which depends on the position data; - a traffic regulation for the vehicle (10) is determined based on the current geographical area (7a, 7b, 7c); and - at least one control signal for driving the vehicle (10) is generated depending on the traffic control and / or assistance information is generated to assist a driver of the vehicle (10) in driving the vehicle (10) depending on the traffic control. [2] Method according to claim 1, wherein when applying the MLM (1) to the input data a classification task is performed according to which one of a plurality of predefined geographical areas (7a, 7b, 7c) is predicted depending on the input data and the actual geographical area (7a, 7b, 7c) corresponds to the predicted geographical area (7a, 7b, 7c). [3] Method according to one of the preceding claims, wherein the position data are received from a GNSS receiver (12) of the vehicle (10). [4] Method according to any of the preceding claims, wherein determining the current geographical area (7a, 7b, 7c) includes determining a country code corresponding to the geographical area (7a, 7b, 7c) and determining the traffic control depending on the country code. [5] Method according to any of the preceding claims, wherein the traffic control corresponds at least to a speed limit for the present geographical area (7a, 7b, 7c). [6] Method according to claim 5, wherein - the assistance information includes a warning message indicating the traffic regulation; and / or - that at least one control signal is generated to steer the vehicle (10) in order to limit the current speed of the vehicle (10) in accordance with at least one speed limit; and / or - that at least one control signal is generated to guide the vehicle (10) in order to automatically control a current speed of the vehicle (10), the controlled speed being limited according to the at least one speed limit. [7] Method according to one of the preceding claims, wherein sensor data representing an environment of the vehicle (10) are received by an environment sensor system (13) of the vehicle (10) and the at least one control signal and / or the assistance information are generated depending on the sensor data. [8] Method according to any of the preceding claims, wherein the MLM (1) comprises an artificial neural network. [9] Method according to any of the preceding claims, wherein the traffic control is determined based on a database (15). [10] Computer-implemented training method for training an MLM (1) for use in a method according to any of the preceding claims, wherein - a multitude of training datasets (2, 4) are received, each training dataset (2, 4) containing training position data representing a respective training position (2) in the reference coordinate system, and annotation data (4) specifying a geographic area (7a, 7b, 7c) in which the respective training position (2) is located; and - the MLM (1) is supervised trained based on the large number of training data sets (2, 4) to perform a classification task. [11] Computer-implemented training method according to claim 10, wherein - geographic area data (3) are obtained, which include an assignment of each geographic area (7a, 7b, 7c) of a plurality of predefined geographic areas (7a, 7b, 7c) to each of a plurality of reference positions (8) in the reference coordinate system; and - the large number of training datasets (2, 4) are generated based on the geographic area data (3). [12] Computer-implemented training method according to claim 11, wherein for each training data set (2, 4) at least a subset of the plurality of training data sets (2, 4) - the respective training position data are received; - a reference position (8) is determined from the plurality of reference positions (8) that corresponds to the respective position data (2); and - the annotation data (4) of the respective training dataset (2, 4) are determined in such a way that they specify the geographical area (7a, 7b, 7c) which is assigned to the respective reference position (8). [13] Computer-implemented training method according to one of claims 11 or 12, wherein at least for a further subset of the plurality of geographical areas (7a, 7b, 7c) a spatial density of reference positions (8) of a respective geographical area (7a, 7b, 7c) within a predefined boundary area of ​​the respective geographical area (7a, 7b, 7c) is greater than a spatial density of the reference positions (8) of the respective geographical area (7a, 7b, 7c) outside the boundary area of ​​the respective geographical area (7a, 7b, 7c). [14] Method according to any one of claims 1 to 9, wherein the MLM (1) is trained using a computer-implemented training method according to any one of claims 10 to 13. [15] Data processing system (11) adapted to perform a method according to one of claims 1 to 9 or 14 and / or a computer-implemented training method according to one of claims 10 to 13. [16] Computer program product which includes instructions which, when executed by a data processing system (11), cause the data processing system (11) to perform a method according to any one of claims 1 to 9 or 14 and / or a computer-implemented training method according to any one of claims 10 to 13.

Images