Methods for the localization of an object, for the supervised training of a learning algorithm, and for the at least partially automatic guidance of a vehicle, radar system, electronic vehicle guidance system, and data processing system

Abstract

The invention relates to a method for the radar-based localization of an object by means of a radar system (2), to a method for the supervised training of a learning algorithm for emulating a virtual antenna array, to a method for the at least partially automatic guidance of a vehicle (1), to a data processing system (3), to a radar system (2) for the localization of an object, to an electronic vehicle guidance system for the at least partially autonomous guidance of a vehicle, and to a computer program product. The method for localization comprises the following steps: - receiving first radar data (7) of a first antenna array of a first radar sensor (4) of the radar system (2); - receiving second radar data (8) of a second antenna array, different from the first antenna array, of a second radar sensor (5) of the radar system (2); - generating third radar data (10) by emulating a third antenna array, different from the first and second antenna arrays, depending on the first radar data (7) and the second radar data (8); and - localizing the object depending on the third radar data (10).

Description

[0001] Methods for locating an object, for supervised training of a learning algorithm, and for at least partially automatic driving of a vehicle, radar system, electronic vehicle guidance system and data processing system

[0002] Aspects of the invention relate to a method for radar-based localization of an object using a radar system, a method for supervised training of a learning algorithm for emulating a virtual antenna array, a method for at least partially automatic driving of a vehicle, a data processing system, a radar system for localizing an object, an electronic vehicle guidance system for at least partially autonomous driving of a vehicle, and a computer program product.

[0003] Radar sensor systems are used in many different vehicle applications, especially for functions at higher speeds. Typical functions include adaptive cruise control (ACC), cross traffic alert (CTA), and blind spot monitoring.

[0004] The performance demands placed on environmental sensor systems, such as radar systems, are constantly increasing. To meet these high demands, particularly for applications requiring long ranges, such as those involving high vehicle speeds, high angular resolution is especially desirable. This allows for the differentiation of objects located relatively close together, for example, in adjacent lanes. Furthermore, high accuracy in object localization, particularly regarding their angular position, is essential.

[0005] US patent 2022 / 0283265 A1 discloses a radar system comprising several essentially identical transmit-receive devices, each forming essentially identical, overlapping virtual antenna groups.

[0006] Fuchs, Jonas & Gardill, Markus & Lübke, Maximilian & Dubey, Anand & Lurz, Fabian. (2022). A Machine Learning Perspective on Automotive Radar Direction of Arrival Estimation. IEEE Access. PP. 1 -1. 10.1109 / ACCESS.2022.3141587. This article provides an overview of current progress and work in the field of deep learning-based estimation of arrival direction in the context of automotive radar.

[0007] It is an object of the invention to increase the angular resolution and / or the angular accuracy of the localization and / or to enlarge the detection range when localizing an object in the environment of a vehicle using radar.

[0008] This problem is solved by the subject matter of the independent claim. Advantageous further developments and preferred embodiments are the subject matter of the dependent claims.

[0009] The invention is based on the idea of ​​emulating a third antenna array located between two virtual antenna arrays using a machine learning algorithm.

[0010] According to one aspect of the invention, a method for radar-based localization of an object, particularly in the vicinity of a vehicle, is provided. The vehicle is, in particular, a motor vehicle, a passenger car, a truck, or other commercial vehicle. The method comprises the following steps:

[0011] In particular, initial radar data from a first antenna array of a first radar sensor of the radar system are received. Specifically, the initial radar data are provided to, and in particular received by, an evaluation unit of the radar system. The initial radar data can also be referred to as initial measurement data.

[0012] In particular, second radar data from a second antenna array of a second radar sensor of the radar system, which is different from the first antenna array and, in particular, does not overlap with it, is received. Specifically, the second radar data is provided to the evaluation unit and, in particular, received by the evaluation unit. The second radar data can also be referred to as second measurement data.

[0013] In particular, third radar data are generated by emulating a third antenna array that is different from the first and second antenna arrays, and in particular does not overlap, depending on the first radar data and the second radar data, especially by means of the evaluation unit.

[0014] In particular, the object is located based on the third radar data, especially using the evaluation unit. If necessary, the object is located based on the first radar data and / or the second and third radar data.

[0015] For example, to locate the object, at least one angle of the object, such as an angle of incidence, an elevation angle and / or an azimuthal angle, and / or a distance of the object to the vehicle is determined depending on the third radar data and, for example, the first radar data and / or the second radar data.

[0016] The method can, for example, be computer-implemented, i.e., purely computer-implemented. Unless otherwise specified, all steps of the computer-implemented method can be performed by a data processing system comprising at least one data processing device, in particular by a data processing system of the vehicle. The data processing device can also be referred to as an evaluation unit. Specifically, the at least one data processing device is configured or adapted to perform 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 it to execute the computer-implemented method.The terms "data processing system" and "at least one data processing device" can be used interchangeably.

[0017] All data processing devices of the at least one data processing device can be part of the vehicle. However, it is also possible that all data processing devices of the at least one data processing device are part of an external computing system outside the vehicle, for example, a backend server or a cloud computing system. It is also possible that the at least one data processing device comprises both at least one vehicle data processing device of the vehicle and at least one external data processing device of the external computing system. The at least one vehicle data processing device can, for example, comprise one or more electronic control units (ECUs), and / or one or more zone control units (ZCLIs), and / or one or more domain control units (DCLIs) of the vehicle, and / or the radar system.In particular, the evaluation unit of the radar system is part of the data processing system, for example, one of at least one data processing device.

[0018] In the event that 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, for example, 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.

[0019] Each embodiment of the computer-implemented method results in a corresponding embodiment of a method for locating an object that is not purely computer-implemented, by including corresponding steps for generating the first radar data, for example by means of the radar system, in particular by means of the first radar sensor, and / or the second radar data, for example by means of the radar system, in particular by means of the second radar sensor.

[0020] For example, the first antenna array has several first antenna elements, and the second antenna array has several second antenna elements. Specifically, the first and second antenna elements are spaced apart from each other; in particular, they do not overlap. For example, the distance between the first and second antenna arrays is between 15 centimeters and 2 meters. The antenna elements can also be referred to as elements.

[0021] In particular, the first and second radar sensors are spaced apart from each other and arranged in a non-overlapping manner. Optionally, the fields of view of the first and second radar sensors can overlap. Preferably, the radar sensors are MIMO (Multiple-Input-Multiple-Output) radar sensors. In particular, the radar sensors are monostatic MIMO radar sensors, and specifically not bisstatic MIMO radar sensors. For example, the behavior of the radar sensors can be described by virtual antenna arrays. For example, the first radar sensor has n transmitters and m receivers. For example, the first virtual antenna array has n x m virtual antenna elements. In particular, this also applies analogously to the second virtual antenna array. A virtual antenna element results, in particular, from a combination of a transmitter and a receiver. In particular, the radar data are received data of the, in particular virtual, antenna elements.Preferably, the virtual first antenna elements and the virtual second antenna elements, and in particular their virtual positions, are spaced apart from each other and, in particular, do not overlap. Virtual positions of the virtual antenna elements can be calculated, for example, by means of a convolution operation between the discrete, especially real, elements of which the array consists. A property of this convolution is that a single virtual array can be computed from many possible non-degenerate real antenna arrays.

[0022] A virtual element is, in particular, a receiving channel, i.e., a virtual receiving element. A physical receiving element can, for example, receive data from multiple transmitting elements. These are treated, for instance, as if they were separate virtual elements.

[0023] The first and second radar sensors are, in particular, primary radar systems, for example, frequency-modulated continuous wave radars, also known as FMCW radars (FMCW: frequency modulated continuous wave). The first and second radar sensors can, for example, be FMCW-MIMO radars (MIMO: multiple-input multiple-output), especially 4D-FMCW-MIMO radars.

[0024] For example, locating the object corresponds to at least partially determining its position, in particular its azimuthal angle and / or elevation angle and / or radial distance. The position can be determined in polar coordinates, Cartesian coordinates, or other coordinate systems.

[0025] Radar data, particularly depending on the type of radar sensor, can be of different types or generated in different ways. For example, radar data can be raw radar data, the temporal profile of a receiver signal from one or more receiving antennas of the radar sensor, specific parameters of such signals (e.g., pulse amplitude, pulse width, or pulse transit time), or digital data sampled using an analog-to-digital converter (ADC). Radar data can also be further preprocessed data, with preprocessing potentially including a Fourier transform, especially of the digital data sampled by the ADC.

[0026] For example, in FMCW radars, the radar data can correspond to the received signals sampled by the ADC converter and Fourier-transformed. In this case, the term radar frequency data is sometimes used.

[0027] By locating the object based on the third radar data, the object can be located more accurately and over a larger area within its surroundings. This could also be achieved, for example, by using an additional third radar sensor or a larger radar sensor with more antenna elements than the first and second radar sensors combined. This would require more space and increase costs. By emulating the third radar data, costs and space can be saved, for example, while maintaining the same resolution.

[0028] In one embodiment, the first antenna array is spaced apart from the second antenna array. The third antenna array is emulated as being located at a distance between the first and second antenna arrays.

[0029] The first, second, and third antenna arrays thus form a combined antenna array, specifically a virtual combined antenna array. In particular, the virtual elements of the combined antenna array do not overlap with respect to their virtual positions. For example, all elements of the combined antenna array are arranged in a row, for example along a line, specifically virtually. For example, the line is oriented parallel to a transverse axis of the vehicle.

[0030] Virtually creating the entire antenna array makes it possible to locate the object more accurately and over a larger area within its surroundings. In particular, additional or larger radar sensors can be eliminated, since the virtual antenna array behaves like the antenna array of a single radar sensor, only larger and with more antenna elements than the first and second radar sensors combined.

[0031] In one embodiment, when emulating the third antenna array, antenna elements of the third antenna array are emulated. A predetermined number of emulated antenna elements depends on the distance between the first and second antenna arrays.

[0032] This results in a uniform overall antenna array with consistent spacing between its elements. This enables a more realistic emulation of the entire array.

[0033] The number of emulated antenna elements is determined, for example, by the distance between the first and second antenna arrays and a predetermined further distance between elements of the antenna arrays. For example, the distance between the elements is half a wavelength of an emitted radar signal A / 2. For example, the distances between the elements of the first antenna array are equal to the distances of the second antenna array and, in particular, equal to the distances of the emulated third antenna array and, in particular, equal to the distances between an outermost element of the first or second antenna array and an outermost element of the emulated third antenna array.

[0034] In one embodiment, the first radar data is generated based on a portion of a first signal transmitted by the first radar sensor that is reflected by the object. Alternatively or additionally, the second radar data is generated based on a portion of a second signal transmitted by the second radar sensor that is reflected by the object.

[0035] For example, a portion of the first signal reflected by the object is received by the first radar sensor. Depending on the received portion of the first signal, the first radar sensor generates the first radar data. Before receiving the second radar data, a second signal may be transmitted by the second radar sensor, and a portion of the second signal reflected by the object is received by the second radar sensor. Depending on the received portion of the second signal, the second radar sensor generates the second radar data. Specifically, the generated first and second radar data are provided to the data processing system, for example, the evaluation unit. This embodiment may, if applicable, be a measurement method. In one embodiment, locating the object involves determining a radial distance of the object, particularly from the vehicle to the object.Consequently, in such embodiments, the distance of the object from the vehicle or from the radar sensors can be determined with increased accuracy or resolution.

[0036] In one embodiment, locating the object involves determining its azimuthal angle. Consequently, in such embodiments, the object's azimuthal angle relative to the vehicle or the radar sensors can be determined with increased accuracy or resolution.

[0037] In one embodiment, locating the object involves determining its elevation angle. Consequently, in such embodiments, the object's elevation angle relative to the vehicle or the radar sensors can be determined with increased accuracy or resolution.

[0038] In one embodiment, locating the object involves determining the angle of arrival, AoA (English: "angle of arrival"), also referred to as the direction of arrival, DoA (English: "direction of arrival"), of the object.

[0039] The angle of incidence can be understood, in particular, as a combination of the elevation angle and the azimuthal angle. Consequently, in such embodiments, the azimuthal angle and the elevation angle of the object relative to the vehicle or relative to the radar sensors can be determined with increased accuracy or resolution.

[0040] In one embodiment, locating the object involves determining its Cartesian coordinates. These Cartesian coordinates can, for example, be two-dimensional or three-dimensional.

[0041] In the case of two-dimensional Cartesian coordinates, these lie in a plane whose normal vector points in the direction of the radial distance for an azimuthal angle of 0° and an elevation angle of 0°, i.e., in the line of sight of the respective radar sensor or along the longitudinal axis of the vehicle. Consequently, in such embodiments, the Cartesian position of the object relative to the vehicle or to the radar sensors can be determined with increased accuracy or resolution.

[0042] In one embodiment, locating the object involves defining a two-dimensional or three-dimensional bounding box for the object.

[0043] For example, a two-dimensional boundary box is a rectangle, and a three-dimensional boundary box is a cuboid. Depending on the embodiment, however, other two-dimensional or three-dimensional geometric shapes can also be used as boundary boxes. Consequently, in such embodiments, the position of the object relative to the vehicle or the radar sensors can be determined with increased accuracy and resolution.

[0044] In one embodiment, the third antenna array is emulated using a trained learning algorithm. The trained learning algorithm can also be referred to as a trained machine learning model (MLM).

[0045] A trained MLM can be understood as an algorithm, specifically a computer-implemented algorithm, that can replicate concrete functions or, more broadly, functions possible through human cognitive processes. A trained MLM can also be described, for example, as a "trained function."

[0046] When training an MLM, its parameters are generally adjusted or updated. Training can be supervised, semi-supervised, or unsupervised. It can also include reinforcement learning, representation learning, and / or other established training methods. In particular, the MLM's parameters can be adjusted iteratively over several 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.

[0047] An MLM can, in particular, include an ANN, a support vector machine, a k-means cluster algorithm, a decision tree, and so on. Specifically, an ANN can be or include a deep neural network and / or a convolutional neural network, a CNN (especially a deep CNN), a recurrent neural network, a RNN (especially a recurrent CNN), a transformer network, and / or a generative adversarial network (GAN).

[0048] The MLM can, for example, be an ANN or be based on one. The first radar data and the second radar data are, in particular, input data fed to the MLM to emulate the third radar data. The learning algorithm is thus applied to the input data. The input data is specifically dependent on, includes, or consists of the first and second radar data.

[0049] One advantage of using MLM to emulate the third radar data is that it is not necessary to know in detail exactly how the first radar data and the second radar data must be processed to improve the accuracy or resolution of the localization, in particular how the third radar data can be estimated from the first radar data and the second radar data.

[0050] Another aspect of the invention relates to a method, in particular a computer-implemented method, for supervised training of a learning algorithm or a machine learning model for emulating a virtual antenna array, particularly in a method according to the invention for localizing an object. In particular, the method comprises the following steps:

[0051] Providing initial training radar data from the first antenna array of a radar system;

[0052] Providing second training radar data from a second antenna array of the radar system that differs from the first antenna array;

[0053] Providing third training radar data from a third antenna array, different from the first and second antenna arrays, using a third radar sensor;

[0054] Predicting radar data from the third antenna array by applying the untrained or partially trained learning algorithm to input training data containing the first training radar data and the second training radar data; and updating the untrained or partially trained learning algorithm depending on the predicted radar data using the third training radar data as an annotation (label).

[0055] During training, the learning algorithm learns to predict the third training radar data based on the first and second training radar data sets. To do this, the learning algorithm predicts radar data from the third antenna array based on the first and second training radar data sets. The predicted radar data is then compared with the third training radar data. This comparison involves evaluating a loss function that depends on the predicted radar data and the third training radar data, particularly on their deviation from each other. For example, the learning algorithm is updated based on the result of the loss function evaluation. Specifically, the learning algorithm, for example, if it is an ANN (active neural network), has weights that are updated based on the result of the loss function (backpropagation).

[0056] To generate the training radar data, for example, a first radar sensor with the first antenna array, a second radar sensor with the second antenna array, and a third radar sensor with the third antenna array can be used. The third radar sensor can, for example, cover the first and second radar sensors and an intermediate distance. In this case, for instance, only the large third radar sensor is used to generate the training radar data. Alternatively, the third radar sensor can cover only the intermediate distance, and the training radar data is generated using the first, second, and third radar sensors.

[0057] In particular, to generate the training radar data, the first and second radar sensors are positioned at a distance from each other. For example, the first and second antenna arrays do not overlap. For example, the third radar sensor is designed and arranged such that the third antenna array connects directly to or overlaps with the first and second antenna arrays. Elements of the third antenna array are arranged within the distance between the elements of the first and second antenna arrays. In one embodiment of the radar-based localization method, the third antenna array is emulated using a trained learning algorithm. The learning algorithm is or was trained according to the inventive method for supervised training of a learning algorithm.In particular, the learning algorithm has been trained according to the procedure for training the learning algorithm.

[0058] Another aspect concerns a method for at least partially automated vehicle control. In particular, a method according to one of the aspects mentioned above or an embodiment thereof is carried out, and, in particular, at least one control signal for at least partially automated vehicle control is generated depending on the object's location. Alternatively or additionally, assistance information to support a person driving the vehicle is generated, depending on the object's location. "Depending on the object's location" means, in particular, that the assistance information and / or the control signals are generated based on a result of the localization, for example, based on distance and / or azimuthal angle and / or elevation angle.Since the localization is improved by the methods according to the invention and their embodiments, the vehicle can also be guided more effectively.

[0059] Another aspect of the invention relates to a data processing system adapted to carry out a method according to the aspects mentioned above or an embodiment thereof. In particular, the data processing system carries out the corresponding method. The data processing system comprises at least one data processing device. The at least one data processing device is specifically adapted to carry out a method according to the invention.

[0060] The terms "data processing system" and "at least one data processing device" may be used interchangeably within the scope of this disclosure. In this disclosure, a data processing device may, for example, be understood as a device with processing circuits for processing data. A data processing device can thus perform arithmetic operations to process data. Indexed access to a data structure, such as a lookup table (LUT) or a database, may also be considered an arithmetic operation.

[0061] A data processing device may, in particular, comprise 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 comprise 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 comprise a physical or virtual cluster of computers or other devices of the aforementioned type.

[0062] A data processing device may also include one or more hardware and / or software interfaces, for example for receiving and / or providing data.

[0063] A data processing device may also include one or more storage devices. A storage device may be implemented as volatile memory, such as dynamic random access memory (DRAM) or static random access memory (SRAM), or as non-volatile data storage, such as read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), flash memory or flash EEPROM, ferroelectric random access memory (FRAM), magnetoresistive random access memory (MRAM), or phase-change random access memory (PCRAM).

[0064] Another aspect of the invention relates to a radar system for locating an object. The radar system comprises a first radar sensor, a second radar sensor, and a data processing system according to the invention. The first radar sensor is configured to generate first radar data from a first antenna array of the first radar sensor and to provide this data to the data processing system, for example, the evaluation unit. The second radar sensor is configured to generate second radar data from a second antenna array of the second radar sensor, which is different from the first antenna array, and to provide this second radar data to the data processing system, for example, the evaluation unit.

[0065] It is possible that the radar system comprises one or more pairs of additional first and additional second radar sensors, the antenna arrays of which are preferably spaced apart from each other. If applicable, the explanations given for the first and second radar sensors apply analogously to the additional first and second radar sensors. In particular, the radar system is configured to perform a method according to one of the aspects or embodiments mentioned above.

[0066] In one embodiment, the first radar sensor and the second radar sensor are configured as monostatic M1 MO radar systems and / or frequency-modulated continuous wave radars (FMCW). These enable improved object localization.

[0067] Another aspect of the invention relates to an electronic vehicle guidance system for at least partially automated vehicle control. The electronic vehicle guidance system comprises a radar system according to the aspect mentioned above or an embodiment thereof. The electronic vehicle guidance system is configured, for example, to guide the vehicle at least semi-autonomously or at least partially automatically, depending on the object's location. The at least one control signal can, for example, be provided to one or more actuators of the vehicle, including, for example, one or more brake actuators and / or one or more steering actuators and / or one or more drive motors of the vehicle. The one or more actuators can influence the longitudinal and / or lateral control of the vehicle in order to guide the vehicle at least partially automatically.

[0068] An electronic vehicle guidance system can also be understood as an advanced driver assistance system (ADAS), which supports the driver during partially automated or semi-autonomous driving. Specifically, the electronic vehicle guidance system can implement a partially automated 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 in its April 2021 version. The assistance information can be output via a vehicle output device, such as a display and / or an audio output system and / or a haptic output system.

[0069] In one embodiment, the data processing system is configured to generate at least one control signal for at least partially automatic vehicle control, depending on the object's location. Alternatively or additionally, the data processing system is configured to generate assistance information to support a person driving the vehicle.

[0070] Another aspect of the invention relates to a vehicle which has the vehicle guidance system according to the invention or its embodiments.

[0071] According to a further aspect of the invention, a computer program with instructions is specified. When the instructions are executed by a data processing system, the instructions cause the data processing system to carry out a method according to the invention.

[0072] The instructions 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.

[0073] According to another aspect of the invention, a computer-readable storage medium is specified that stores a computer program according to the invention.

[0074] The computer program and the computer-readable storage medium are each computer program products containing the commands.

[0075] Further features of the invention are evident 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 those subsequently mentioned in the description of the figures and / or shown in the figures alone, are not only usable in the combinations specified but also in other combinations without departing from the scope of the invention. Thus, embodiments that are not explicitly shown and explained in the figures but can be derived and generated from the explained embodiments by separate combinations of features are also to be considered as encompassed and disclosed by the invention. Embodiments and combinations of features that do not exhibit all the features of an originally formulated independent claim are also to be considered disclosed.Furthermore, embodiments and combinations of features, in particular those set out above, are to be considered disclosed which go beyond or deviate from the combinations of features set out in the cross-references of the claims.

[0076] The following are exemplary embodiments of the invention. The following are shown:

[0077] Fig. 1 shows a vehicle with an embodiment of a radar system according to the invention;

[0078] Fig. 2 shows a flowchart of an embodiment of a method according to the invention for radar-based localization of an object;

[0079] Fig. 3 shows a flowchart of an embodiment of a method according to the invention for training a learning algorithm;

[0080] Fig. 4 shows a flowchart of a further embodiment of the inventive method for training the learning algorithm; and

[0081] Fig. 5 shows a flowchart of a further embodiment of the inventive method for radar-based localization of the object.

[0082] In the figures, identical reference symbols denote functionally equivalent elements.

[0083] Fig. 1 shows a vehicle 1 with an embodiment of a radar system 2 according to the invention. The radar system 2 includes, for example, an embodiment of a data processing system 3 according to the invention. Optionally, the radar system 2 has a first radar sensor 4 and a second radar sensor 5, which are mounted at different locations on the vehicle 1, but in particular with overlapping fields of view. For example, the first radar sensor 4 and the second radar sensor 5 are both front radars, or both rear radars, or both radars of the vehicle 1 mounted on a left or right side of the vehicle 1. For example, the first radar sensor 4 and the second radar sensor 5 are configured as MIMO radars and / or as FMCW radars. For example, the first radar sensor 4 is configured to generate first radar data 7 and the second radar sensor 5 is configured to generate second radar data 8.

[0084] Fig. 2 shows a schematic representation of an embodiment of a flowchart of a method according to the invention for radar-based localization of an object using the radar system 2. For example, the data processing system 3 receives first radar data 7 from a first antenna array of the first radar sensor 4. For example, the data processing system 3 receives second radar data 8 from a second antenna array of the second radar sensor 5 that is different from, and in particular does not overlap with, the first antenna array. In particular, third radar data 10 are generated by emulating a third antenna array that is different from and different from the first and second antenna arrays, depending on the first radar data 7 and the second radar data 8. The object is localized depending on the third radar data 10 and in particular on the first radar data 7 and / or the second radar data 8, in particular by means of the data processing system 3.

[0085] For example, the first radar data 7 and the second radar data 8 are provided to a trained machine learning algorithm 9. In particular, the third radar data are generated using the trained machine learning algorithm 9.

[0086] Figure 3 shows a schematic representation of a flowchart for an embodiment of a method for training the learning algorithm 9. The machine learning algorithm 9 is, for example, an artificial neural network. For example, the machine learning algorithm 9 is provided with first training radar data 7a from the first antenna array of the first radar sensor 4 and second training radar data 8a from the second antenna array of the second radar sensor 5, which is different from the first antenna array. In particular, during the training procedure, the learning algorithm 9 is provided with third training radar data 11 from a third antenna array, which is different from the first and second antenna arrays. For example, radar data 10a of the third antenna array is predicted by applying the untrained or partially trained learning algorithm 9 to input training data.For example, the input training data includes the first training radar data 7a and the second training radar data 8a. The untrained or partially trained learning algorithm 9 is updated, if necessary, depending on the predicted radar data 10a, using the third training radar data 11 as an annotation. In particular, the predicted radar data 10a and the third training radar data 11 are fed to a loss function L. For example, an outcome A is determined using the loss function L. The outcome A of the loss function L is then fed to the machine learning algorithm 9. If necessary, weights of the machine learning algorithm 9 are updated depending on the determined outcome A of the loss function L. To train the learning algorithm 9 or to update the weights, methods known in the context of supervised training, such as the backpropagation algorithm, can be used.

[0087] For example, a vehicle 1 in Fig. 1 has a third radar sensor 6 for the supervised training method of the learning algorithm 9. In particular, the third radar sensor 6 generates the third training radar data 11. In one embodiment, the third radar sensor 6 generates the first training radar data 7a, the second training radar data 8a, and the third training radar data 11. In particular, the vehicle 1 does not have the third radar sensor 6 when carrying out the radar-based object localization method. For example, when carrying out the radar-based object localization method, the third radar data 10a are emulated such that they essentially assume values ​​that the third radar sensor 6 would provide if it were still arranged on the vehicle 1.If, for example, the training procedure is carried out only with the large third radar sensor 6, then the third radar data 10a are emulated in such a way that they essentially assume values ​​which the third radar sensor 6 would provide as third training radar data 11 if it were located on the vehicle 1.

[0088] Fig. 4 shows a flowchart of a further embodiment of the inventive method for training the learning algorithm 9. For example, the first antenna array is represented by a virtual antenna array having virtual elements 12. For example, the second antenna array is also represented by a virtual antenna array having second elements 13. For example, the third antenna array is represented by a virtual antenna array having elements 14. Optionally, the third elements 14 of the third antenna array are virtually spaced apart from each other. For example, the spacing is half a wavelength A / 2 of the emitted radar signals of the third radar sensor 6 or the emulated third radar sensor 6. For example, the third radar sensor 6 generates third sensor data, which contains the third training radar data 11.The third elements 14 may differ from the first elements 12 and the second elements 13, particularly with respect to a virtual position of the elements. For example, the third sensor data have boundary regions 15. For example, the boundary regions 15 are similar or identical to the first training radar data 7a and / or the second training radar data 8a with respect to their virtual position and / or their values. In particular, the learning algorithm 9 is trained such that the predicted radar data 10a correspond as closely as possible to the third training radar data 11, especially for a given example tuple of the first training radar data 7a and the second training radar data 8a.

[0089] Figure 5 shows a flowchart of another embodiment of the inventive method for radar-based object localization. For example, total radar data 16 can be generated from the first radar data 7, the second radar data 8, and the emulated radar data 10. In particular, in one embodiment of the method for radar-based object localization, the total radar data 16 is generated from the first radar data 7, the second radar data 8, and the emulated radar data 10. For example, the total radar data 16 emulates the third sensor data from the third large radar sensor 6. The third radar sensor 6 is therefore preferably only required for the training procedure. The third radar sensor 6 is not required for the method for radar-based object localization.The object localization method can therefore have the same accuracy, detection range, and / or angular resolution as if, for example, the third radar sensor 6 were used, even though only the first radar sensor 4 and the second radar sensor 5 are actually used. Thus, a smaller radar sensor or fewer radar sensors can be permanently installed while maintaining the same accuracy, detection range, and / or angular resolution.

[0090] The vehicle 1, which is used to generate the training radar data, is generally different from the vehicle 1, whose radar system 2 according to the invention performs the radar-based localization method according to the invention.

Claims

Patent claims 1. Method for radar-based localization of an object using a radar system (2), comprising the steps: Receiving first radar data (7) from a first antenna array of a first radar sensor (4) of the radar system (2); Receiving second radar data (8) from a second antenna array different from the first antenna array of a second radar sensor (5) of the radar system (2); Generating third radar data (10) by emulating a third antenna array different from the first and second antenna arrays depending on the first radar data (7) and the second radar data (8); and Locating the object depending on the third radar data (10).

2. Method according to claim 1, wherein the first antenna array is spaced apart from the second antenna array, wherein the third antenna array is emulated as being located at a distance between the first antenna array and the second antenna array.

3. Method according to one of claims 1 or 2, wherein when emulating the third antenna array, antenna elements (14) of the third antenna array are emulated, wherein a predetermined number of the emulated antenna elements (14) depends on the distance between the first antenna array and the second antenna array.

4. Method according to one of the preceding claims, wherein the first radar data (7) are generated depending on a part of a first signal transmitted by means of the first radar sensor (4) that is reflected by the object and / or the second radar data (8) are generated depending on a part of a second signal transmitted by means of the second radar sensor (5) that is reflected by the object.

5. A method according to any of the preceding claims, wherein locating the object includes determining a radial distance of the object; and / or determining an azimuthal angle of the object; and / or determining an elevation angle of the object; and / or determining an angle of incidence for the object; and / or determining Cartesian coordinates of the object; and / or determining a two-dimensional or three-dimensional boundary box for the object.

6. Method according to one of the preceding claims, wherein the third antenna array is emulated using a trained learning algorithm (9).

7. Method for supervised training of a learning algorithm (9) for emulating a virtual antenna array, comprising the following steps: Providing initial training radar data (7a) of an initial antenna array of a radar system (2); Providing second training radar data (8a) from a second antenna array of the radar system (2) that is different from the first antenna array; Providing third training radar data (11 ) from a third antenna array different from the first and second antenna arrays; Predicting radar data from the third antenna array by applying the untrained or partially trained learning algorithm (9) to input training data containing the first training radar data (7a) and the second training radar data (8a); and Updating the untrained or partially trained learning algorithm (9) depending on the predicted radar data using the third training radar data (11) as an annotation.

8. Method according to claim 6, wherein the third antenna array is emulated by means of a trained learning algorithm (9), wherein the learning algorithm (9) is trained according to a method according to claim 7.

9. Method for at least partially automatically driving a vehicle (1), wherein a method according to one of the preceding claims is carried out and, depending on the localization of the object, at least one control signal for at least partially automatically driving the vehicle (1) is generated; and / or Assistance information is generated to support a person driving the vehicle (1) while driving the vehicle (1).

10. Data processing system (3) adapted to perform a method according to any of the preceding claims.

11. Radar system (2) for locating an object, comprising a first radar sensor and a second radar sensor and a data processing system (3) according to claim 10, wherein the first radar sensor is configured to generate first radar data (7) of a first antenna array of the first radar sensor and to provide it to the evaluation unit; the second radar sensor is configured to generate second radar data (8) of a second antenna array of the second radar sensor, which is different from the first antenna array, and to provide it to the evaluation unit.

12. Radar system (2) according to claim 11, wherein the first radar sensor and the second radar sensor are configured as monostatic MIMO radar systems (2) and / or frequency-modulated continuous wave radars (FMCW).

13. Electronic vehicle guidance system for at least partially automatically guiding a vehicle (1) comprising a radar system (2) according to claim 11, wherein the electronic vehicle guidance system is configured to guide the vehicle (1) depending on the localization of the object.

14. Electronic vehicle guidance system according to claim 13, wherein the data processing system (3) is configured to generate at least one control signal for at least partially automatic guidance of the vehicle (1), depending on the localization of the object; and / or To generate assistance information to support a person driving the vehicle (1) in driving the vehicle (1).

15. Computer program product comprising instructions which, when executed by a data processing system (3), cause the data processing system (3) to perform a method according to any one of claims 1 to 9.

Images