A method of operating radar system in which position information of detected target is determined by using machine learning, radar system, driver assistance system and vehicle

Abstract

A method of operating a radar system (14), in particular a radar system (14) for a vehicle (10), a radar system (14), a driver assistance system (12) and a vehicle (10). In the method, at least one electromagnetic radar signal (22) is transmitted from at least one transmitting antenna element (34) of the radar system (14). At least one electromagnetic echo signal (40) generated by at least one radar signal (22) reflected from at least one target (20) in a field of view of the radar system (14) is received by at least one receiving antenna element (38) and converted into received data (72) adapted for signal processing. The radar system (14) is operated as a MIMO radar comprising a plurality of transmit antenna elements (34) and a plurality of receive antenna elements (38), where the transmit antenna elements (34) and the receive antenna elements (38) generate a virtual antenna array having a plurality of virtual antenna elements for receiving echo signals (40) during MIMO operation of the radar system (14). For at least a portion of the virtual antenna element, amplitude information and phase information are determined from at least a portion of the received data, respectively, by performing at least one two-dimensional fast Fourier transform. At least one array data set is determined that includes at least amplitude information and phase information for at least a portion of the virtual antenna elements. At least a portion of the data of the at least one array data set is fed to at least one neural network (64), with which at least one piece of positional information of the at least one detected target (20) is determined.

Description

Technical Field

[0001] The present invention relates to a method of operating a radar system, in particular a radar system for a vehicle, wherein:

[0002] transmitting at least one electromagnetic radar signal from at least one transmitting antenna element of the radar system,

[0003] receiving, by at least one receiving antenna element, at least one electromagnetic echo signal generated by at least one radar signal reflected from at least one target in the field of view of the radar system and converting into received data suitable for signal processing,

[0004] determining at least one amplitude information and at least one phase information corresponding to at least one received echo signal based on at least a portion of the received data,

[0005] At least one position information is determined by using machine learning, wherein the at least one position information characterizes at least a direction of the at least one detected target relative to a reference system associated with the radar system.

[0006] Furthermore, the present invention relates to a radar system, in particular a radar system for a vehicle, comprising:

[0007] at least one transmitting antenna element for transmitting electromagnetic radar signals,

[0008] at least one receiving antenna element for receiving electromagnetic echo signals,

[0009] means for converting the electromagnetic echo signals into received data suitable for signal processing,

[0010] means for determining amplitude information and phase information from the received data,

[0011] Means for determining at least one position information by using machine learning, the at least one position information characterizing at least a direction of at least one detected object.

[0012] Furthermore, the present invention relates to a driver assistance system comprising at least one radar system, wherein the at least one radar system comprises:

[0013] at least one transmitting antenna element for transmitting electromagnetic radar signals,

[0014] at least one receiving antenna element for receiving electromagnetic echo signals,

[0015] means for converting electromagnetic echo signals into received data suitable for signal processing,

[0016] means for determining amplitude information and phase information from the received data,

[0017] Means for determining at least one position information by using machine learning, the at least one position information characterizing at least a direction of at least one detected object.

[0018] Furthermore, the present invention relates to a vehicle comprising at least one radar system, wherein the at least one radar system comprises:

[0019] at least one transmitting antenna element for transmitting electromagnetic radar signals,

[0020] at least one receiving antenna element for receiving electromagnetic echo signals,

[0021] means for converting the electromagnetic echo signals into received data suitable for signal processing,

[0022] means for determining amplitude information and phase information from the received data,

[0023] Means for determining at least one position information by using machine learning, the at least one position information characterizing at least a direction of at least one detected object. Background Art

[0024] US 20210156985A1 describes techniques and apparatus for radar angular ambiguity resolution. These techniques enable determining the angular position of a target from a spatial response with multiple amplitude peaks. Rather than considering only the peak with the highest amplitude, the techniques for radar angular ambiguity resolution select one or more frequency sub-spectra that emphasize amplitude or phase differences in the spatial response and analyze the irregular shape of the spatial response across a wide field of view to determine the target's angular position. In this way, each angular position of a target has a unique signature that the radar system can determine and use to resolve angular ambiguities. Using these techniques, a radar can have antenna array element spacing greater than half the center wavelength of the reflected radar signal used to detect the target. The radar system determines the target's angular position by detecting radar signals reflected from the target and determining which steering angle corresponds to the radar signal's angle of arrival. Digital beamforming is used to generate a spatial response that includes amplitude and phase information for different steering angles. An angle estimator receives the spatial response and estimates the target's angular position by analyzing the shape of the spatial response across the field of view. In some aspects, the angle estimator may use signal processing techniques, pattern matching techniques, or machine learning to determine the radar signal's angle of arrival. Example signal processing techniques may utilize algorithms to analyze the shape of the spatial response and determine differences that indicate the direction of the target.

[0025] The object of the present invention is to provide a method, a radar system, a driver assistance system and a vehicle, wherein the determination of position information of an object detected by means of the radar system, in particular the direction of the object, can be improved. Summary of the Invention

[0026] The purpose of the present invention is achieved by the following method:

[0027] A radar system operates as a multiple-input multiple-output radar including a plurality of transmit antenna elements and a plurality of receive antenna elements, wherein the transmit antenna elements and the receive antenna elements generate a virtual antenna array having a plurality of virtual antenna elements for receiving return signals during multiple-input multiple-output operation of the radar system,

[0028] For at least a portion of the virtual antenna elements, determining amplitude information and phase information, respectively, from at least a portion of the received data by performing at least one two-dimensional fast Fourier transform,

[0029] determining at least one array data set comprising at least amplitude information and phase information for at least a portion of the virtual antenna elements,

[0030] At least a portion of the at least one array data set is fed to at least one neural network, and at least one position information of at least one detected target is determined using the neural network.

[0031] According to the present invention, the radar system operates as a multiple-input, multiple-output (MIMO) radar. A MIMO radar includes multiple transmit antenna elements and multiple receive antenna elements. During MIMO operation of the radar system, the transmit and receive antenna elements create a virtual antenna array with multiple virtual antenna elements. This increases the number of receive antenna elements used as sampling points for echo signals. This also improves the resolution of position information.

[0032] The received data is transformed into at least one array data set using a two-dimensional fast Fourier transform (2D FFT). The at least one array data set includes at least amplitude information and phase information obtained using at least a portion of the virtual antenna elements. In this way, the received data is transformed into the amplitude / phase domain. The array data set represents a 2D FFT spectrum. In the 2D FFT spectrum of the ambiguity function on the virtual antenna array, a characteristic pattern of sidelobes can be observed. This aperture pattern can be referred to as an "aperture pattern." In the case of only one or two detected targets, the aperture pattern is relatively simple, but for a larger number of targets, the complexity of the aperture pattern increases. The aperture pattern is determined by mixing (superimposing) both the amplitude and phase information of each target.

[0033] The complex mixing of received data from the echo signals of two or more targets within a single range / Doppler bin creates unique aperture patterns. By identifying these unique aperture patterns, the position information of two or more targets, and specifically their respective directions, can be separated. Resolution is characterized by the ability to separate two or more targets within the same range / Doppler bin. This resolution is related to the arrangement of the antenna subarrays, specifically the transmit and receive antenna arrays. According to the present invention, more than one target can be resolved by exploiting ambiguity.

[0034] At least a portion of the array data set is fed to at least one neural network. Using the at least one neural network, at least one position information item is determined for the position of at least one detected target. Using the at least one neural network according to the present invention, the aperture can be highly sampled to simplify the sidelobe pattern. Furthermore, symmetry planes that may cause mirror blurring can be calculated.

[0035] Advantageously, for any given set of targets and their phase / amplitude relationships, the aperture pattern can be determined as a function of both amplitude and phase information. By learning which data from at least one neural network array dataset creates the corresponding target combination, the position, and particularly the orientation, of at least one target can be isolated. Even large initial training array datasets can be systematically created using appropriate simulation techniques.

[0036] According to the present invention, the amplitude information, phase information, and time information (if any) are considered as a single image, from which at least one piece of position information, particularly direction information, can be learned using at least one neural network. The image of the amplitude information and the image of the phase information, as well as the time information (if any), are paired and associated with unique position information (particularly unique direction information).

[0037] A target, as defined in the present invention, is an area or reflection point on an object from which a radar signal can reflect. An object may have one or more such targets. If an object has multiple targets, radar signals may also reflect differently from these targets, for example, in different directions. To make it easier to distinguish, targets detected by a radar system may be referred to as "detected targets."

[0038] Using the radar system, at least direction information characterizing the direction of the target can be determined. In addition, the radar system can be used to determine range information and / or speed information, which characterizes the range and / or speed of the target relative to the radar system and / or the vehicle.

[0039] The amplitude information, phase information and position information, in particular the direction information, may comprise analog or digital signals or data, real numbers, complex values ​​or digital sets of data, signals, numbers or values.

[0040] Depending on the device used for signal processing, the received data may include electrical signals or values, such as digital values ​​based on bits. In this manner, the received data can be processed by the electrical device used for signal processing. Additionally or alternatively, the received data may include optical signals or values, such as those based on qubits. In this manner, the received data can be processed by the optical device used for signal processing (e.g., a quantum processor).

[0041] Advantageously, the echo signal can be converted into received data by an analog-to-digital converter. In this way, the received data can be determined as digital data.

[0042] A reference system associated with the radar system may include reference points, reference lines and / or reference planes. If the radar system is used on a vehicle, the reference system may also be associated with the vehicle. In this case, the reference system may include reference points, reference lines and / or reference planes, such as virtual axes of the vehicle, such as the longitudinal axis, vertical axis or transverse axis of the vehicle. Advantageously, one reference system may be a spherical coordinate system. Azimuth and elevation angles may be used as position information to characterize the direction of a detected target. Additionally or alternatively, one reference system may be a Cartesian coordinate system. X, y and z coordinates may be used as position information to characterize the position of a detected target. A vector with x, y and z coordinates may describe the direction of a detected target.

[0043] The present invention can be used in radar systems for vehicles, particularly motor vehicles. Advantageously, the present invention can be used on land vehicles, particularly passenger cars, trucks, buses, motorcycles, etc., as well as aircraft and / or ships. The present invention can also be used in radar systems for vehicles that can operate autonomously or partially autonomously. However, the present invention is not limited to vehicles. It can also be used in stationary vehicles, robots, drones, and / or machines, particularly construction or transport machines, such as cranes, excavators, and the like.

[0044] The radar system can advantageously be connected to or be part of at least one control device of a vehicle or machine, in particular a driver assistance system, in order to enable autonomous or partially autonomous operation of the vehicle or machine.

[0045] The present invention can be used in a radar system that is designed as a front radar system, a corner radar system, a rear radar system, a roof radar system, a bottom radar system, or an interior radar system of a vehicle. Therefore, the radar system can be a front radar system, a corner radar system, a rear radar system, a roof radar system, a bottom radar system, or an interior radar system of a vehicle.

[0046] Radar systems can be used to detect targets such as stationary or moving objects, in particular vehicles, persons, animals, obstacles, road irregularities, in particular potholes or stones, road restrictions, open spaces, in particular parking spaces, precipitation or the like.

[0047] According to an advantageous embodiment, as at least one piece of position information, at least one piece of direction information can be determined, in particular an angle, such as an azimuth and / or elevation angle, and / or a vector, which characterizes the direction of at least one detected object relative to a reference system. In this way, the direction of the detected object can be determined.

[0048] Advantageously, the directional information may be an angle, in particular an angle of arrival (AoA). In this way, the direction of arrival of the echo signal may be detected.

[0049] Advantageously, an azimuth and / or elevation angle can be determined as directional information. In this way, the position of at least one object can be defined in a spherical coordinate system. Additionally or alternatively, a vector can be determined as at least one piece of positional information. In this way, the position of at least one object can be defined in a Cartesian coordinate system. Such a vector can include Cartesian coordinates, in particular x, y, and / or z coordinates.

[0050] According to another advantageous embodiment, the at least one piece of amplitude information and the at least one piece of phase information, as well as the time information (if any), can be implemented as complex values, in particular as complex vectors. This allows for very efficient combination of the amplitude information and the phase information, and also, if any, the time information. Consequently, mathematical algorithms can be used to efficiently process the amplitude information and the phase information, and, if any, the time information.

[0051] According to another advantageous embodiment, time information can also be determined for at least some of the virtual antenna elements. In this way, the 2D FFT pattern can also have a time component, as the movement of at least one target causes the phase relationship to change. In some cases, the pattern of both the amplitude information and the phase information can have time-dependent characteristics.

[0052] According to another advantageous embodiment, at least a portion of the data of the array data set can be fed to at least one neural network designed as a spiking neural network. In this way, the at least one neural network can be used to learn position information, including for moving targets that cause amplitude and / or phase information to vary over time. Spiking neural networks are well-suited to the temporal information content of typical radar return signals, both in terms of time and processing parameters such as angle, distance, and Doppler value. Spiking neural networks incorporate the concept of time into their structure. Using spiking neural networks can reduce the complexity and power consumption of radar systems. Spiking neural networks can be efficiently implemented in neuromorphic hardware.

[0053] Alternatively, a neural processing unit (NPU) combined with a convolutional neural network (CNN) can be used. The NPU can be hardware accelerated. Therefore, only the magnitude and phase images can be considered as a learning dataset for the neural network.

[0054] According to another advantageous embodiment, an antenna system comprising transmit antenna elements and receive antenna elements can be arranged and operated to create an undersampled system, and / or an antenna array comprising transmit antenna elements and receive antenna elements can be arranged and operated to create a sparse virtual antenna array. In this way, the ratio between the aperture and the number of required antenna elements is improved. The aperture of the virtual antenna array can be increased.

[0055] The virtual antenna elements are the sampling points of the return signal. The spread of the virtual antenna array defines the aperture of the antenna arrangement. The angular resolution is related to the physical distance between the widest-spaced virtual antenna elements relative to the wavelength of the radar signal.

[0056] Advantageously, the virtual antenna elements can lie in one plane. In this way, the virtual antenna array can be aligned more easily.

[0057] Advantageously, the four most distant virtual antenna elements can be arranged at the corners of a rectangle, particularly a square. This allows for a rectangular virtual antenna array. Consequently, the aperture can be clearly defined in two orthogonal directions, for example, in azimuth and elevation. The four most distant virtual antenna elements can be arranged at the corners of a square. Consequently, the aperture in the two orthogonal directions is equal. In this way, the angular resolution in the two orthogonal directions can be defined in a general manner.

[0058] Advantageously, the side length of the rectangle can in each case correspond to an integer multiple of half the wavelength of the radar signal. In this way, the distance between the virtual antenna elements arranged at adjacent corners of the rectangle each corresponds to an integer multiple of half the wavelength of the radar system.

[0059] By placing the virtual antenna elements far apart, an undersampled system can be achieved. Consequently, the aperture will exhibit spacing-dependent aliasing. The determined position information, particularly the angle of arrival, is then ambiguous.

[0060] According to another advantageous embodiment, the four most distant virtual antenna elements can be arranged at the corners of a rectangle (particularly a square), the side lengths of which each correspond to an integer multiple of half the wavelength of the radar signal, and at least one additional virtual antenna element can be arranged at a distance of approximately half the wavelength of the radar signal from one of the four most distant virtual antenna elements. In this way, position information of the at least one target can be determined with a reasonable aperture.

[0061] According to another advantageous embodiment, a two-step learning technique can be applied to learn the at least one neural network. In this way, the at least one neural network can be learned very efficiently.

[0062] Advantageously, a two-step learning technique can be initiated by feeding an idealized artificial array dataset to the neural network. The idealized array dataset can represent an idealized aperture pattern. In this way, the network can learn the learning principle. Then, a so-called starter enhancement can be performed. To this end, the complexity of the idealized artificial array dataset can be increased, in particular by adding noise and / or random variations. Finally, the neural network can be fed with a real recorded array dataset.

[0063] According to another advantageous embodiment, at least one neural network can be learned from both amplitude information and phase information via the array data set. In this way, even complex array data sets characterizing complex aperture patterns can be recognized using at least one neural network.

[0064] Furthermore, the object of the invention is achieved in a radar system in that the radar system comprises at least a part of the device for carrying out the method according to the invention.

[0065] According to the present invention, a radar system is designed as a MIMO radar. The radar system includes multiple transmit antenna elements and multiple receive antenna elements. The radar system includes means for controlling the transmit antenna elements and the receive antenna elements in MIMO operation to generate a virtual antenna array having multiple virtual antenna elements for receiving echo signals.

[0066] Furthermore, the device for carrying out the method according to the invention comprises means for carrying out at least one two-dimensional fast Fourier transformation in order to determine amplitude information and phase information from the received data.

[0067] Furthermore, the device for executing the method according to the invention comprises means for determining an array data set comprising at least amplitude information and phase information of the virtual antenna elements.

[0068] Furthermore, the device for carrying out the method according to the invention comprises means for implementing at least one neural network, to which the data of the array data set can be fed to determine position information of the detected object.

[0069] Advantageously, the device for performing the method according to the invention may comprise at least one neural network, in particular at least one spiking neural network. In this way, position information of the detected object can be determined from the data of the array data set.

[0070] At least a portion of the apparatus for performing the method according to the present invention may be implemented using software. In this manner, in particular, flow charts, in particular programs, algorithms, and / or implementation tables for performing the method may be stored in the radar system. Additionally or alternatively, at least a portion of the apparatus for performing the method according to the present invention may be implemented using hardware.

[0071] Furthermore, the object of the present invention is achieved in a driver assistance system in that the driver assistance system comprises at least a part of a device for carrying out the method according to the present invention.

[0072] According to the present invention, a driver assistance system comprises at least one radar system, in particular at least one radar system according to the present invention. Advantageously, the at least one radar system of the driver assistance system can include at least a portion of a device for performing the method according to the present invention. Since the at least one radar system is part of the driver assistance system, the device for the at least one radar system is also part of the driver assistance system. This applies similarly to the device of the vehicle comprising at least one driver assistance system and / or at least one radar system.

[0073] Furthermore, the object of the invention is achieved in a vehicle in that the vehicle comprises at least a part of the device for carrying out the method according to the invention.

[0074] The vehicle comprises at least one radar system. With the at least one radar system, the vehicle's environment and / or the vehicle's interior can be monitored.

[0075] Advantageously, the vehicle may comprise at least one driver assistance system, with which information obtained from the at least one radar system may be used for autonomous or at least partially autonomous operation of the vehicle.

[0076] Advantageously, the at least one radar system can be part of or connected to at least one driver assistance system. In this way, information obtained using the at least one radar system can be sent to a control unit of the at least one driver assistance system.

[0077] Additionally or alternatively, at least a portion of the device for carrying out the method according to the present invention may be implemented separately from the at least one radar system, for example together with a control device of the vehicle and / or a control device of a driver assistance system.

[0078] Otherwise, the features and advantages described in conjunction with the method according to the invention, the radar system according to the invention, the driver assistance system according to the invention, and the vehicle according to the invention, as well as their respective advantageous embodiments, apply mutatis mutandis, and vice versa. The individual features and advantages can of course be combined with one another, thereby producing further advantageous effects that go beyond the sum of the individual effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] The present invention and the above and other objects and advantages may best be understood from the following detailed description of the embodiments, but the invention is not limited to the embodiments, in which:

[0080] Figure 1 a top view of a vehicle having a driver assistance system including a radar system;

[0081] Figure 2 yes Figure 1 a side view of a vehicle;

[0082] Figure 3 Is a Figure 1 and Figure 2 Functional diagram of the vehicle's driver assistance systems;

[0083] Figure 4 According to the first example, in MIMO operation mode, Figures 1 to 3 A virtual antenna array generated by a transmitting antenna array and a receiving antenna array of a vehicle radar system;

[0084] Figure 5 According to the second example, in MIMO operation mode, Figures 1 to 3 A virtual antenna array generated by a transmitting antenna array and a receiving antenna array of a vehicle radar system;

[0085] Figure 6 According to the third example, in MIMO operation mode, Figures 1 to 3 A virtual antenna array generated by a transmitting antenna array and a receiving antenna array of a vehicle radar system;

[0086] Figure 7 Is used to operate from Figures 1 to 3 A flowchart of a method for a vehicle radar system;

[0087] Figure 8 is the amplitude array image, which is visualized using Figures 1 to 3Aperture pattern of multiple targets detected by a vehicle's radar system.

[0088] In the accompanying drawings, the same or similar elements are represented by the same reference numerals. The accompanying drawings are merely schematic representations and are not intended to depict specific parameters of the present invention. In addition, the accompanying drawings are intended only to depict typical embodiments of the present invention and are therefore not to be construed as limiting the scope of the present invention. DETAILED DESCRIPTION

[0089] exist Figure 1 A vehicle 10 in the form of a passenger car is shown in a top view. Figure 2 The vehicle 10 is shown in side view.

[0090] Vehicle 10 includes a driver assistance system 12 . Figure 3 A functional diagram of a vehicle 10 is shown having a driver assistance system 12. With the driver assistance system 12, the vehicle 10 can be operated partially autonomously or autonomously.

[0091] Driver assistance system 12 includes a radar system 14 and a control unit 16 .

[0092] The radar system 14 can be used to monitor the environment in front of the vehicle 10. The radar system 14 is connected to a control unit 16 so that data collected by the radar system 14 about the environment can be transmitted to the control unit 16. The control unit 16 of the driver assistance system 12 can control operating functions of the vehicle 10 based on the information obtained by the radar system 14.

[0093] Radar system 14 is located, for example, in the front area of ​​vehicle 10, for example in the front bumper. Radar system 14 can be used to monitor a surveillance area in front of vehicle 10 in the direction of travel, for example for object 18. Figures 1 to 3 , object 18 is shown as an example. Radar system 14 can also be arranged in different locations on vehicle 10 and can be oriented differently. A plurality of radar systems 14 can also be provided.

[0094] The radar system 14 may detect targets 20 such as stationary or moving objects 18 , such as vehicles, people, animals, plants, obstacles, the ground, the road, road irregularities (such as potholes or stones), road boundaries, (traffic) signs, signals, free spaces (such as parking spaces), precipitation, etc.

[0095] A target 20 in the sense of the present invention is an area or reflection point of an object 18 from which a radar signal 22 can be reflected. An object 18 can have one or more such targets 20. If an object 18 has multiple targets 20, the radar signal 22 can also be reflected differently from these targets, for example in different directions. Targets 20 detected by a radar system 14 can be referred to as detected targets 20 for ease of distinction. Figures 1 to 3 , for the sake of clarity, only two targets 20 of the object 18 are shown as an example.

[0096] Radar system 14 can be used to determine the distance 24, direction, and velocity of object 18 relative to a reference system of vehicle 10. The reference system is, for example, a spherical coordinate system. Azimuth angle Φ and elevation angle Θ are used as directional information to characterize the direction of detected target 20. The origin of the spherical coordinate system is located at the intersection of longitudinal axis 26 of vehicle 10 and vertical axis 28 of vehicle 10. Azimuth angle Φ = 0° is located on longitudinal axis 26 of vehicle 10.

[0097] The radar system 14 is designed as a multiple-input multiple-output (MIMO) radar. The radar system 14 comprises a control and evaluation device 30, a transmitting antenna array 32 having a plurality of transmitting antenna elements 34 for transmitting electromagnetic radar signals 22, and a receiving antenna array 36 having a plurality of receiving antenna elements 38 for receiving electromagnetic echo signals 40. As an example, Figure 3 Two transmit antenna elements 34 and two receive antenna elements 38 are indicated in FIG.

[0098] Transmit antenna element 34 and receive antenna element 38 generate a virtual antenna array 42 having a plurality of virtual antenna elements 44 for receiving return signals 40 during multiple-input multiple-output operation of radar system 14. Virtual antenna array 42 is implemented as a sparse virtual antenna array.

[0099] exist Figure 4, a first example of a virtual antenna array 42 is shown. Virtual antenna array 42 includes six virtual antenna elements 44. Virtual antenna elements 44 are located in a plane. The four virtual antenna elements 44 that are furthest apart are arranged at the corners of a square. In this manner, a square virtual antenna array 42 is implemented. The horizontal extension of virtual antenna array 42 defines a horizontal aperture 46 of the antenna arrangement comprising transmit antenna array 32 and receive antenna array 36. The vertical extension defines a vertical aperture 48 of the antenna arrangement. The spacing between virtual antenna elements 44 arranged at adjacent corners of the square corresponds to an integer multiple of half the wavelength of radar system 14. The fifth virtual antenna element 44 is arranged on the connecting line between virtual antenna element 44 in the lower left corner and virtual antenna element 44 in the upper left corner of the square, at a spacing 50 from the virtual antenna element 44 in the lower left corner that is approximately half the wavelength of radar signal 22. Sixth virtual antenna element 44 is arranged on the connecting line between lower left virtual antenna element 44 and lower right virtual antenna element 44 at a distance 52 from the lower left virtual antenna element 44 by approximately half the wavelength of radar signal 22 .

[0100] exist Figure 5 In FIG, a second example of a virtual antenna array 42 is shown. Figure 4 Unlike the virtual antenna array 42 depicted in FIG, in the second example, the sixth virtual antenna element 44 is arranged on the connecting line between the virtual antenna element 44 in the upper left corner and the virtual antenna element 44 in the upper right corner, and the interval 54 is approximately half the wavelength of the radar signal 22 from the virtual antenna element 44 in the upper left corner.

[0101] exist Figure 6 , a third example of a virtual antenna array 42 is shown. Figure 4 Unlike the virtual antenna array 42 shown, the third example includes only five virtual antenna elements 44. Four of the virtual antenna elements 44 are located at the corners of the square, as shown in FIG. Figure 4 The fifth virtual antenna element 44 is arranged on the connecting line between the lower left virtual antenna element 44 and the upper right virtual antenna element 44 at an interval 56 of about half the wavelength of the radar signal 22 from the lower left virtual antenna element 44 .

[0102] Furthermore, the control and evaluation device 30 comprises means for converting the electromagnetic echo signal 40 into received data 72 suitable for signal processing, for example an analog-to-digital converter 58 .

[0103] The control and evaluation device 30 comprises a Fourier transformation device 60 for performing a two-dimensional fast Fourier transformation 74 in order to determine amplitude information and phase information from the received data 72. In conjunction with the sparse virtual antenna array 42, an undersampled system can be created.

[0104] Furthermore, the control and evaluation device 30 comprises a data set device 62 for determining an array data set 80 comprising amplitude information and phase information obtained with the virtual antenna elements 44 .

[0105] Furthermore, the control and evaluation device 30 comprises a neural network 64, to which the data of the array data set 80 can be fed in order to determine position information characterizing the direction of the detected target 20. The neural network 64 is designed as a spiking neural network 64. The spiking neural network 64 is well matched to the temporal information content properties of the echo signal 40.

[0106] Furthermore, the control and evaluation device 30 comprises a storage medium 66 , in which data of the array data set 80 , position information and learning data of the neural network 64 can be stored.

[0107] At least part of the means for executing the method for operating radar system 14 can be implemented by software. In storage medium 66 of control and evaluation device 30, for example, flow charts, such as programs, algorithms and / or implementation tables for executing the method can be stored.

[0108] Use the following Figure 7 The flowchart in describes the method of operating the radar system 14 in more detail.

[0109] A series of measurements having a plurality of radar measurements is performed using radar system 14 .

[0110] In processing step 68, radar signal 22 is transmitted for each measurement using each transmit antenna element 34 in accordance with the MIMO operating mode. Transmit antenna elements 34 and receive antenna elements 38 create a virtual antenna array 42 having a plurality of virtual antenna elements 44 during the MIMO operating mode. If object 18 is present in the field of view of radar system 14, radar signal 22 is reflected at target 20 of object 18. Electromagnetic echo signal 40 generated by radar signal 22 reflected from target 20 is received by virtual antenna element 44.

[0111] In process step 70 , the received echo signal 40 is converted using the analog-to-digital converter 58 into received data 72 suitable for further signal processing.

[0112] Each antenna element in the virtual antenna array 44 has a complex number representing each of the amplitude information and phase information corresponding to the echo signal in a single range Doppler unit, and represents information about the number and direction (azimuth and elevation) of the target sampled by the receiver array. This complex number information is then transformed using a two-dimensional Fourier transform 74. The resulting aperture pattern (example Figure 8 ) is represented as amplitude information and phase information, and can also be implemented as complex amplitude / phase values.

[0113] In process step 78, for each measurement, an array data set 80 is determined. Array data set 80 includes complex amplitude / phase values ​​76 having amplitude information and phase information for all virtual antenna elements 44. Furthermore, array data set 80 for each measurement includes time information that characterizes the time-sequential position of the measurement within a series of measurements within a single range-Doppler bin. The time information can be used to characterize the movement of targets 20 relative to one another and can be used to separably identify multiple targets within the same range-Doppler bin.

[0114] Figure 8 An amplitude array image is shown as a grayscale representation that visualizes the aperture pattern of amplitude information of an exemplary array data set 80 for a plurality of detected targets 20 as related to the angle of arrival (AoA) of received echo signals 40. The angle of arrival is characterized by an azimuth angle Φ and an elevation angle Θ. The amplitude information is defined according to the linear grayscale shown next to the amplitude array image.

[0115] Can be based on Figure 8 The amplitude array image is used to generate a phase array image (not shown) as a grayscale representation that visualizes an aperture pattern for phase information that is correlated with an aperture pattern of arrival angles of the received echo signals 40 .

[0116] After determining array dataset 80, the data of array dataset 80 is fed into neural network 64. The neural network can operate on both scalar values ​​and complex values. Using neural network 64, position information in the form of azimuth angle Φ and elevation angle Θ is determined for each detected target 20. The amplitude information, phase information, and time information are treated as a single image, from which directional information is learned using neural network 64. The image of amplitude information and the image of phase information are paired with corresponding time information and are associated with unique directional information.

[0117] The azimuth angle Φ and elevation angle Θ of each detected target 20 are transmitted to the control unit 16 of the driver assistance system 12. The driver assistance system 12 controls the operating functions of the vehicle 10 based on the position information of the detected targets 20.

[0118] Optionally, a two-step learning technique can be performed to train the neural network 64 prior to regular operation of the radar system 14. The two-step learning technique can be initiated by feeding the neural network 64 an idealized artificial array dataset 80. The idealized array dataset 80 represents an idealized aperture pattern based on amplitude and phase information. In this way, the neural network 64 can learn the learning principle. Then, a so-called starter enhancement can be performed. To this end, the complexity of the idealized artificial array dataset 80 can be increased, for example, by adding noise and / or random variations. Finally, the neural network 64 can be fed with a real recorded array dataset 80.

Claims

1. A method of operating a radar system (14), in particular a radar system (14) for a vehicle (10), wherein: transmitting at least one electromagnetic radar signal (22) from at least one transmitting antenna element (34) of the radar system (14), receiving at least one electromagnetic echo signal (40) by at least one receiving antenna element (38) and converting it into received data (72) suitable for signal processing, the at least one electromagnetic echo signal (40) being generated by at least one radar signal (22) reflected from at least one target (20) in the field of view of the radar system (14), determining at least one amplitude information (76) and at least one phase information (76) corresponding to the at least one received echo signal (40) based on at least a portion of the received data (72), determining at least one position information (Φ, Θ) by using machine learning (64), wherein the at least one position information (Φ, Θ) characterizes at least a direction of at least one detected target (20) relative to a reference system associated with the radar system (14), It is characterized by: The radar system (14) operates as a multiple-input multiple-output radar including a plurality of transmit antenna elements (34) and a plurality of receive antenna elements (38), wherein the transmit antenna elements (34) and the receive antenna elements (38) generate a virtual antenna array (42) having a plurality of virtual antenna elements (44) for receiving return signals (40) during multiple-input multiple-output operation of the radar system (14), determining amplitude information (76) and phase information (76) from at least a portion of the received data (72) by performing at least one two-dimensional fast Fourier transform (74) for at least a portion of the virtual antenna elements (44), respectively, determining at least one array data set (80) comprising at least the amplitude information (76) and the phase information (76) of at least a portion of the virtual antenna elements (44), At least a portion of the data of the at least one array data set (80) is fed to at least one neural network (64), and at least one position information (Φ, Θ) of the at least one detected target (20) is determined using the at least one neural network (64).

2. The method according to claim 1, characterized in that As at least one position information, at least one direction information, in particular an angle, such as an azimuth (Φ) and / or elevation (Θ), and / or a vector, can be determined, which characterizes the direction of at least one detected target (20) relative to the reference system.

3. The method according to claim 1 or 2, characterized in that The at least one amplitude information item and the at least one phase information item and, if applicable, the time information item are implemented as complex values ​​(76), in particular as complex vectors.

4. The method according to any one of the preceding claims, characterized in that Additional time information is determined for at least a portion of the virtual antenna elements (44).

5. The method according to any one of the preceding claims, characterized in that At least a portion of the data of the array data set (80) is fed to at least one neural network (64) designed as a spiking neural network.

6. The method according to any one of the preceding claims, characterized in that The antenna system (32, 36) having the transmit antenna elements (34) and the receive antenna elements (38) is arranged and operated in such a way as to create an under-sampling system, and / or the antenna array (32, 36) having the transmit antenna elements (34) and the receive antenna elements (38) is arranged and operated in such a way as to create a sparse virtual antenna array (42).

7. The method according to any one of the preceding claims, characterized in that The four most distant virtual antenna elements (44) are arranged at the corners of a rectangle, in particular a square, whose side lengths (46, 48) each correspond to integer multiples of half the wavelength of the radar signal (22), and at least one additional virtual antenna element (44) is arranged at a distance (50, 52; 54; 56) of approximately half the wavelength of the radar signal (22) from one of the four most distant virtual antenna elements (44).

8. The method according to any one of the preceding claims, characterized in that A two-step learning technique is applied to learn the at least one neural network (64).

9. The method according to any one of the preceding claims, characterized in that The at least one neural network (64) is learned from the amplitude information (76) and the phase information (76) via an array data set (80).

10. A radar system (14), in particular a radar system (14) of a vehicle (10), comprising: at least one transmitting antenna element (34) for transmitting an electromagnetic radar signal (22), at least one receiving antenna element (38) for receiving an electromagnetic echo signal (40), means for converting the electromagnetic echo signal (40) into received data (72) suitable for signal processing, means for determining amplitude information (76) and phase information (76) from received data (72), means for determining at least one position information (Φ, Θ) by using machine learning (64), the at least one position information (Φ, Θ) characterizing at least the direction of at least one detected target (20), It is characterized in that The radar system (14) comprises at least a part of a device for carrying out the method according to any one of claims 1 to 9.

11. A driver assistance system (12) comprising at least one radar system (14), wherein: The at least one radar system (14) comprises: at least one transmitting antenna element (34) for transmitting an electromagnetic radar signal (22), at least one receiving antenna element (38) for receiving an electromagnetic echo signal (40), means for converting the electromagnetic echo signal (40) into received data (72) suitable for signal processing, means for determining amplitude information (76) and phase information (76) from received data (72), means for determining at least one position information (Φ, Θ) characterizing at least the direction of at least one detected target (20) by using machine learning (64), It is characterized in that The driver assistance system (12) comprises at least a part of a device for carrying out the method according to any one of claims 1 to 9.

12. A vehicle (10) comprising at least one radar system (14), wherein: The at least one radar system (14) comprises: at least one transmitting antenna element (34) for transmitting an electromagnetic radar signal (22), at least one receiving antenna element (38) for receiving an electromagnetic echo signal (40), means for converting the electromagnetic echo signal (40) into received data (72) suitable for signal processing, means for determining amplitude information (76) and phase information (76) from received data (72), means for determining at least one position information (Φ, Θ) by using machine learning (64), the at least one position information (Φ, Θ) characterizing at least the direction of at least one detected target (20), It is characterized in that The vehicle (10) comprises at least a part of a device for carrying out the method according to any one of claims 1 to 9.

Images