Method for identifying interference in a radar system

Abstract

A method for identifying interference in a radar system (10) of a vehicle (1), wherein the following steps are carried out: receiving (101) at least one received signal (202) of the radar system (10); determining (102) detection information (210) from the received signal (202); carrying out an evaluation (103) of the detection information (210) by means of at least one neural network (401, 402); using (104) the result of the evaluation (103) as a prediction of interference (220) in the received signal (202).

Description

Field of the Invention

[0001] The present invention relates to a method for identifying interference in a radar system. Furthermore, the present invention relates to a corresponding radar system and a computer program. Background Art

[0002] It is known from the prior art that radar systems are used in vehicles to detect the surroundings of the vehicle. It is also known that other parameters of the detected objects, such as the distance, relative speed or angle of the target with respect to the vehicle, can also be determined in the radar system by means of signal processing. Such radar systems are now used in a very large number of vehicles. In addition, each vehicle may also have more than one radar system.

[0003] However, the increasing popularity of radar systems also has the following disadvantages: different radar systems may negatively affect each other. Such interference caused by interacting radar systems is also referred to as interference.

[0004] There is often also the problem here that interference cannot be reliably detected or eliminated in the radar system. Summary of the Invention

[0005] Therefore, the object of the present invention is to at least partially reduce the above-mentioned disadvantages. The object of the present invention is in particular to provide a more reliable detection of interference caused by interference in a radar system.

[0006] The above object is solved by a method according to the invention for identifying interference in a radar system of a vehicle, a radar system according to the invention for detecting target objects in the surroundings of a vehicle, and by a computer program product according to the invention. Other features and details of the present invention result from the description and the drawings. Here, the features and details described in connection with the method according to the invention of course also apply to the radar system according to the invention and the computer program according to the invention, and vice versa respectively, so that the disclosure regarding the individual aspects of the invention can always be mutually referred to.

[0007] The object is in particular solved by a method for identifying interference, in particular identifying interference and / or reducing interference, in a radar system of a vehicle.

[0008] It is in particular provided here that the following steps are carried out, preferably in the given order or in any order, and individual steps and / or all steps can also be carried out repeatedly:

[0009] - Receive at least one (in particular electromagnetic) received signal of the radar system, preferably by means of a radar sensor of the radar system and / or by means of at least one antenna (or alternatively in digital form by means of an analog-to-digital converter),

[0010] - Determine the (especially digital) detection information from the received received signal, preferably by means of a demodulator of the radar system and / or by means of a downconverter and / or by means of an analog-to-digital converter, wherein preferably the (optionally pre-demodulated and / or downconverted) detection information is provided in digital form by means of analog-to-digital conversion for subsequent steps, in particular by means of a processing device of the radar system.

[0011] - Carry out an evaluation of the (determined) detection information, preferably by means of at least one neural network and / or by taking into account an evaluation of the detection information that has already been carried out (hereinafter simply referred to as taking into account the previous evaluation) of the detection information that has been determined from the received signal in a previous step of the method.

[0012] - Use the result of the evaluation, in particular the output of the neural network, as a prediction (in particular a forecast) of interference, in particular interference, of the received signal.

[0013] The following advantages can be achieved here: By means of the evaluation - specifically, for example, by using a neural network and / or taking into account the previous evaluation - a more reliable and, if necessary, faster identification of interference can be achieved. This can be caused in such a way that not only the just-present interference is identified, but rather it is even possible to predict interference by means of the evaluation. For this purpose, the evaluation can be carried out in such a way that the tendency of recurrence and / or the recurrence pattern and / or the temporal correlation of the interference in the detection information is identified by means of the previous evaluation. Thus, if the neurons are temporally linked in the sense of a feedback loop (as in a recurrent neural network), the neural network can achieve this. In this way, information encoded in time can be determined in the detection information, which is specific for the interference and thus it is also possible to predict the interference.

[0014] The method according to the invention can be implemented at least in part as a computer-executed method, and thus implemented within the scope of digital signal processing of a radar system: the implementation of the evaluation and / or the use of the result and / or the determination of the detection information and / or the reception of the received signal. Specifically, for this purpose, a processing device of the radar system can be used, which for this purpose, for example, includes a processor and a memory. It can also be advantageous if only some steps are computer-executed or only some steps are implemented by the processing device, while other steps, such as "the reception of the received signal" and partly the determination of the detection information, are implemented by other dedicated components of the radar system. Thus, it is possible to significantly improve the efficiency of implementing the method if necessary.

[0015] The interference can in particular be in the form of interference and / or interference caused by another radar system, for example another vehicle. Thus, interference recognition is used to detect interference that obstructs the reception of signals. Such interference may be caused, for example, by the mutual influence and / or interference of radar signals of different radar systems. For this purpose, the method according to the invention can accordingly use the result of the evaluation in order to identify and / or predict interference in the detection information (for example, detection information in the form of time signals and / or radar spectra and / or spatial distribution of target objects), in particular interference caused by other radar systems. Thereby, the evaluation of the detection information can be significantly improved.

[0016] Another advantage achievable within the scope of the present invention is that, as another step of the method according to the invention, after determining the detection information, a Fourier transform is performed on the detection information in order to obtain information about the relative speed and / or distance of a target (i.e., a target object) in the vehicle's surroundings from the provided detection information. This information can then be used, for example, for object detection. The target may, for example, be a reflection and / or an object in the surroundings.

[0017] Optionally, it may be provided that providing the detection information includes demodulating and / or down-converting and / or analog-to-digital conversion of the received signal in order to obtain the provided detection information as a time signal. This enables reliable detection of the received signal.

[0018] For example, it may be provided that the at least one neural network or at least one neural network includes at least one recurrent neural network (RNN), and the recurrent neural network preferably takes into account the evaluation of detection information that is temporally prior. Here, the output of the neural network can be used as the result of the evaluation. It is feasible here that the output is determined by means of the output of the RNN, for example after the output of the RNN has been further processed by a decoder. The consideration of the temporally prior evaluation carried out thereby enables the determination of information encoded in time in the detection information in order to thereby provide a prediction of interference. For this purpose, for example, there may be context units for each hidden layer of the RNN, which process the output of the neurons and output it again in subsequent time steps. In addition, the RNN can be trained to perform a prediction of interference, so that the output of the RNN in the current iteration of the method step already indicates that interference is expected in a subsequent iteration or in one of the subsequent iterations. For this purpose, the RNN is trained, for example, using training data to be described in more detail. The training can be carried out, for example, by means of backpropagation through time (BPTT).

[0019] A feasible variant for the predictive training of a neural network or, specifically, for an RNN can be to train the network with the following training data, which consists of input data (input) and the associated output data (output). Within the scope of "supervised learning", the output data can include: the correct output expected under the associated input data. The correct output can be, for example, a prediction of the interference in the detection information that is only determined subsequently, for example in the form of an identification and / or segmentation in the current detection information. The detection information can exist, for example, as a two-dimensional image, in which the corresponding segmentation template can be superimposed by the output.

[0020] For training, for example, information determined successively in time (e.g., a sequence of detection information or information resulting from further processing of the detection information) E t-2 、E t-1 、E t+1 is used as the input data. The information is accordingly specific to the detection information determined successively in time. As the output data, the (e.g., manually established) identification (marking or segmentation) A t-2 、A t-1 、A t+1 of the interference in the input data can be used. Thus, the identification A t-2 in the detection cycle t - 2 indicates the interference in the information E t-2 , the identification A t-1 in the detection cycle t - 1 indicates the interference in the information E t-1 , and so on.

[0021] Training with the mentioned training data is sufficient to train the network to identify interference in the input data in the current detection cycle. To additionally perform prediction, the order of the sequence used for training can be changed. Specifically, the (detection) information E t-2 、E t-1 、E t+1 used for the input data can be moved forward in the sequence in terms of order or the identification A t-2 、A t-1 、A t+1 can be moved backward. In other words, thus, a new sequence can be defined as the new output data A' according to the assignment relationship A‘ t-x = A t-x-+1 . This achieves the following advantage: during training in one iteration (i.e., the detection cycle), the identification A t-X-+1 used as the output data does not correspond to the interference in the information E t-X in the current iteration, but already corresponds to the information E t-X+1 in a subsequent iteration. Since this identification A t-X-+1corresponding to the desired output of the network, so the network is trained to use information E t-X output a prediction for identification A t-X-+1 In this way, it is also possible to evaluate the detection information that is earlier in time.

[0022] Optionally furthermore, it is feasible that the at least one neural network includes at least one convolutional neural network (CNN), the convolutional neural network receives the detection information as input, and uses the output of the convolutional neural network as the input for a recurrent neural network. This can achieve, for example, reducing the data volume of the detection information to be processed by the RNN and / or for pre-evaluating to extract information about interference. Thereby, the efficiency of processing by the RNN can be improved. For this purpose, for example, the CNN is trained with the following training data, which consists of input data (input) and the associated output data (output). In the context of "supervised learning", the output data can include: the correct output desired under the associated input data. The input data is, for example, the unchanged detection information, and the output data is a reduced (scaled) version of the detection information or an indication of interference. In this way, the CNN is trained to provide an optimized input for the RNN as the output.

[0023] In another feasible solution, it can be stipulated that performing the evaluation of the detection information includes the following steps:

[0024] - Preprocess the detection information of the detection period, preferably by max-pooling, especially to reduce the data size of the detection information,

[0025] - Extract information about interference in the form of at least one interference in the received signal from the preprocessed detection information, especially by means of a convolutional neural network,

[0026] - Perform a prediction of the at least one interference for a detection period that is later in time by means of the extracted information and especially by means of an evaluation of the detection information that is earlier in time, preferably by means of an RNN.

[0027] This has the advantage that interference can be detected particularly reliably and quickly, even before the interference obstructs the detection information.

[0028] Optionally furthermore, it is stipulated that using the result of the evaluation, especially the output of the neural network, includes the following steps:

[0029] - Provide a prediction by means of an output frequency range in which interference will exist in the future.

[0030] What is utilized here is that interference - especially in the form of interference - only affects a limited frequency range. Therefore, it is possible to reliably characterize and, if necessary, even reduce interference by means of said frequency range. In order to train the network for this working principle, the following output data can be used for the training data, in which the frequency range of the interference in the input data has been manually entered.

[0031] Preferably, it can be provided that using the result of the evaluation, especially the output of a neural network, includes the following steps:

[0032] - Electronically output the result to an electronic device of the vehicle, preferably for a control device of the vehicle.

[0033] Thereby, the control device can become aware of the interference when it occurs and can react thereto. For example, discard the detected information affected by the interference.

[0034] Optionally, it can also be contemplated that the result of the evaluation includes a segmentation of the detected information, which segmentation indicates the predicted interference. For this purpose, for example, the following output data can be used as training data for the network, in which such (predicted) segmentation of the respective input data has been manually carried out.

[0035] According to another feasible solution, it can be provided that the at least one neural network is trained by setting (and especially pre-implementing) the following training steps:

[0036] - Store a plurality of successively determined detected information, especially a sequence of successively determined detected information,

[0037] - Provide output data, especially ground-truth data, by (especially manual) identification of interference, especially interference in the detected information,

[0038] - Train the neural network by means of the training data formed by the detected information and the output data, especially the ground-truth data.

[0039] Therefore, the input data can be formed from the sequence of detected information. The ground-truth data represents the desired output that the network should output when inputting the input data. Therefore, the desired output can specifically include a prediction of interference.

[0040] For this purpose, for example, it can be provided that providing the ground-truth data includes the following steps:

[0041] - Manually identify (especially the predicted) interference so that the at least one neural network is trained by training to predict interference as interference of the received signal.

[0042] According to another advantage, it can be provided that, for one detection cycle, a plurality of transmission signals of the radar system are successively emitted in at least one frequency range, respectively, in order to respectively receive the associated received signals, wherein the transmission signals are each implemented as at least one chirp having a frequency that varies in time within the frequency range. It is feasible here to provide other frequency ranges in which transmission signals can also be emitted.

[0043] Furthermore, within the scope of the present invention, it can be provided that the determination of the detection information is carried out (iteratively) for each detection cycle, and the determination of the detection information preferably respectively includes the following steps:

[0044] - Performing mixing of the respective transmission signal and the associated received signal in order to respectively obtain baseband signals,

[0045] - Determining the detection information from the obtained baseband signals, wherein the detection information is specific for object detection in the surroundings of the vehicle.

[0046] Thereby, the surroundings of the vehicle can be reliably detected by the radar system.

[0047] It is also conceivable that the result of the evaluation includes an indication of an interference frequency range in which interference is predicted in a subsequent detection cycle in time, wherein, preferably, using the result of the evaluation includes automatically and at least partially adjusting the frequency range. In this way, transmission signals can be emitted in a frequency range that is at least partially outside the predicted interference frequency range. Thereby, the at least one frequency range can be implemented as a frequency range that can be at least partially changed. This enables reliable reduction of interference because the interference frequency range is avoided.

[0048] The subject matter of the present invention is also a radar system for detecting target objects in the surroundings of a vehicle, the radar system having a processing device that is adjusted such that the processing device performs the following steps:

[0049] - Providing detection information from the received signals of the radar system, for example by digital reception and / or analog-to-digital conversion of the detection information,

[0050] - Performing an evaluation of the detection information, in particular by means of at least one neural network,

[0051] - Using the result of the evaluation as a prediction of interference with the received signals.

[0052] Thus, the radar system according to the present invention brings the same advantages as described in detail with reference to the method according to the present invention. In addition, the radar system can be adapted to implement the steps of the method according to the present invention. The steps of "receiving at least one received signal of the radar system" and / or "determining detection information" can be at least partially implemented, for example, by a radar sensor of the radar system, so as to digitally provide detection information for evaluation. These steps implemented by the radar sensor may also include demodulation and / or downconversion if necessary. Other steps - but also demodulation and / or downconversion if necessary - can be implemented by a processing device. In other words, the processing device can implement only some of the steps of the method according to the present invention - but optionally can also implement all the steps. Here, a processing unit may be provided, which, when executed by the processing device, causes the processing device to implement the steps of the method according to the present invention. The processing unit is implemented, for example, as a computer program according to the present invention. The processing device has, for example, a processor and / or a memory, and the processing unit is stored in the memory and can be read by the processor. The processing device is, for example, configured as a computer and / or a control device and / or the like of a vehicle.

[0053] The radar system is implemented, for example, as a 24 GHz radar system or a 77 GHz radar system. Alternatively or additionally, the radar system is configured as a continuous wave radar, in particular configured as an FMCW (Frequency Modulated Continuous Wave Radar), and the continuous wave radar can perform distance measurement and / or speed measurement.

[0054] In addition, the vehicle can be a motor vehicle and / or a car and / or an autonomous vehicle and / or an electric vehicle and / or the like. The detection information is used, for example, by vehicle-side components such as the vehicle's assistance systems and / or control devices, and the assistance systems and / or control devices advantageously provide at least partial automatic driving and / or automatic parking of the vehicle.

[0055] The subject matter of the present invention is also a computer program, preferably a computer program product. Here, it is stipulated that the computer program has instructions, which, when the computer program is executed by the processing device, cause the processing device to at least partially implement the steps of the method according to the present invention, and / or specifically implement the following steps:

[0056] - Provide detection information from the received signal of the radar system, for example, by digitally receiving and / or analog-to-digital conversion of the detection information,

[0057] - Perform an evaluation of the (provided) detection information, in particular by at least one neural network,

[0058] - Use the result of the evaluation as a prediction of the interference of the received signal.

[0059] Thus, the computer program according to the invention brings the same advantages as those described in detail with reference to the method according to the invention and / or the radar system according to the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Other advantages, features and details of the invention result from the following description, in which embodiments of the invention are described in detail with reference to the drawings. Here, the features mentioned in the description can be important for the invention individually or in any combination. In the figures:

[0061] Figure 1 Schematic views showing a vehicle having a radar system according to the invention and another vehicle traveling in front, respectively, in side view,

[0062] Figure 2 Schematic view showing the transmitted signal of the radar system according to the invention,

[0063] Figure 3 Schematic view showing the visualization of interference in the detected signal and in the resulting baseband signal,

[0064] Figure 4 Schematic view showing the visualization of the influence of interference in the baseband signal in the time domain and the frequency domain,

[0065] Figures 5 to 7 Schematic view showing the visualization of the method according to the invention. DETAILED DESCRIPTION

[0066] In the following figures, the same reference numerals are used for the same technical features of different embodiments.

[0067] In Figure 1 a vehicle 1 having a radar system 10 according to the invention is schematically shown. The arrow symbolizes the direction of movement of the vehicle 1 here. In addition, a target object 230 is shown in the surrounding environment 2 of the vehicle 1 in the form of another vehicle traveling in front. The radar system 10 can receive the reflections of the target object 230 here and detect the target object 230 and / or reconstruct the surrounding environment 2 based on these reflections. For this purpose, a received signal 202 is detected and signal processing of the received signal 202 is carried out (at least partially) by the processing device 15.

[0068] The described detection of the radar system 10 is shown in further detail below. First, a transmitted signal 201 can be generated by the radar system 10, which is radiated into the surrounding environment 2 of the vehicle 1. Such a transmitted signal 201 in Figure 2is shown exemplarily. The radar system 10 is advantageously a continuous wave radar and / or an FMCW radar (frequency modulated continuous wave radar), which in particular provides (preferably by means of the Doppler effect) distance measurement and / or speed measurement. Here, for example, a 24 GHz radar system or a 77 GHz radar system may be involved. Specifically, for example, a 77 GHz radar system may operate in the 77 GHz frequency band and thus provide Figure 5 the different frequency ranges 251, 252, 253 for the transmission signal 201 shown in between f = 76 GHz and 77 GHz. The frequency range specifically used for the transmission signal 201 can be variably determined by the radar system 10 from the possible frequency ranges 251, 252, 253, and thus the frequency range used varies for different detection periods 260.

[0069] To detect parameters of the target object 230 in the surroundings 2 of the vehicle 1, such as distance or speed or angle, the transmission signal 201 can be modulated, for example, with respect to the frequency f. Similarly, as Figure 2 shown, N frequency chirps with a varying frequency f within a duration T1 can be output continuously as the transmission signal 201 for detection. In such a chirp, the frequency f can change over time within the bandwidth B. For example, 128 chirps are emitted. The intermediate frequency of the chirp can be variably determined if necessary, that is, differently for different detection periods 260, and thus the intermediate frequency of the chirp can be within one of the possible frequency ranges 251, 252, 253. The duration of a corresponding chirp can be T1 / N, as also schematically shown by the double arrow in Figure 2 . For example, linear frequency modulation can be used, in which the frequency f changes linearly within the bandwidth B in a corresponding chirp. After the duration T1, the received signal 202 can be detected (for example, at 256 samples per chirp) within the time period T2 - T1 and evaluated by the processing device 15. Thus, the entire detection period 260 has a duration T2. The received signal 202 is shown in Figure 3 , where the received signal is in the HF (high frequency) band like the transmission signal 201.

[0070] The emitted transmission signal 201 can be reflected and / or backscattered by the target object 230 and thus received by the radar system 10. Subsequently, the received signal 202 received in this way can be demodulated and / or downmixed. In particular, this results in Figure 3The baseband signal 203 shown in [figure], wherein the frequency fb of the baseband signal is related to the signal propagation time of the reflected transmitted signal 201 and thus to the distance of the target object 230. Subsequently, the signal obtained from the received signal 202 (e.g., by analog-to-digital conversion and additional processing if necessary) can be converted into digital detection information 210. Until the end of the duration T1, the data thus obtained can be stored in an MxN matrix with M samples per chirp and N chirps. Here, an example starts from M = 256 and N = 128. Thus, a radar frame with raw data (the raw data having 128x256 pixels) can be obtained. Then this representation corresponds to the time-frequency space. The non-volatile memory unit of the processing device can be used for storage. With the help of this matrix, a spectrum can subsequently be determined by Fourier transform of the matrix (especially the detection information 210), which is specific for the relative velocity and / or distance of the target object 230 in the surrounding environment 2. Here, in particular, a two-dimensional spectrum is involved (corresponding to the two-dimensional matrix according to the detection information 210), so that different coordinates represent different parameters (such as distance and relative velocity). The detection information 210 or the information obtained therefrom can be used as an input for evaluation according to Figure 7 be used as an input for evaluation.

[0071] Especially when two spatially adjacent radar systems transmit in the same frequency range at the same time, interference may occur and specifically interference fringes may occur. An exemplary such interferer 240 is shown in Figure 3 with respect to time, which is near the corresponding transmitted signal and received signal 201, 202 or baseband signal 203 in terms of its interference frequency range. In Figure 4 the consequences of such an interferer 204 are visualized. Here, the interference 220 can appear in the time domain in the form of peaks and thus increase the spectrum in the frequency domain. This is problematic because, for example, false targets may be detected and the detection of the real target object 230 may be hindered.

[0072] In Figure 6 a method according to the invention is schematically visualized. The method is used to identify such interference 220 in the radar system 10 of the vehicle 1. Here, first, at least one received signal 202 of the radar system 10 is received 101. Subsequently, detection information 210 is determined 102 from the received signal 202. Then, the detection information 210 is evaluated 103 by at least one neural network 401, 402. Finally, according to step 104, the result of the evaluation 103 can be used as a prediction of the interference 220 of the received signal 202. These steps can be repeated for different detection periods 260.

[0073] In the evaluation 103, preprocessing of the detection information 210, such as max pooling, can be performed first. The detection information 210 corresponds to, for example, the raw data of a radar system having 128x256 pixels. To reduce the data volume, these data can be reduced to, for example, 32x32 pixels. Interferences 220 can also be identified in these reduced data. Subsequently, according to Figure 7 , these data can be used as the input to the CNN 401. The CNN 401 can extract information about the interference 220, where the output of the CNN 401 can be used as the input to the RNN 402. The horizontal dashed arrow represents the characteristic of the RNN 402, that is, the previously performed evaluation can be taken into account. Thus, the emission behavior of the interferer 240 over a longer time period can be considered. Subsequently, the output of the RNN 402 can be used as a prediction of the interfered frequency-time range in the next detection cycle 260. Here, segmentation 404 can be involved, and the segmentation can be trained with corresponding ground truth data 410. As an additional intermediate step, decoding 403 of the output is optionally also provided to obtain the segmentation 404.

[0074] The CNN 401 can have, for example, the following architecture, in which the input of the CNN 401 is first processed by alternately performing convolution and max pooling functions. In other words, (for example, 3) convolutional layers can be first provided, and each convolutional layer is connected to a pooling layer. Here, kernel sizes of 3x3x1 and / or 3x3x2 of the filter kernels can be used. Subsequently, the output of the CNN 401 can be passed to the RNN 402. This includes, for example, an LSTM (Long short-term memory) layer. Finally, the output of the RNN 402 can be provided to a decoder, which includes, for example, a fully connected layer and a deconvolution layer.

[0075] In response to the output of the predicted interference 220, the transmit signal 201 can be emitted in (at least partially) different frequency ranges 251, 252, 253 in the next detection cycle 260, and thus leave the interference frequency range. In Figure 5 , this adjustment of the frequency ranges 251, 252, 253 is visualized. Thereby, the influence of the interference 220 is at least reduced in the next detection cycle 260. Nevertheless, information about the interference 220 can continue to flow into the detection information 210 in a time-coded manner, so that the interference can continue to be predicted by the evaluation 103. It is also conceivable that, in order to detect this time-coded information about the interference 220, the received signal 202 is simultaneously detected in the interference frequency range.

[0076] The foregoing description of the embodiments describes the present invention only within the scope of examples. Of course, as long as it is technically meaningful, the various features of these embodiments can be freely combined with each other without departing from the scope of the present invention.

[0077] List of reference numerals

[0078] 1 Vehicle

[0079] 2 Surroundings

[0080] 10 Radar system

[0081] 15 Processing device

[0082] 101 First method step, receiving

[0083] 102 Second method step, determining

[0084] 103 Third method step, evaluating

[0085] 104 Fourth method step, using

[0086] 201 Transmitted signal

[0087] 202 Received signal

[0088] 203 Baseband signal in the time domain

[0089] 204 Baseband signal in the frequency domain

[0090] 210 Detection information

[0091] 220 Interference, disturbance

[0092] 230 Target object

[0093] 240 Jammer

[0094] 251 First frequency range

[0095] 252 Second frequency range

[0096] 253 Third frequency range

[0097] 260 Detection period

[0098] 401 CNN, Convolutional Neural Network

[0099] 402 RNN, Recurrent Neural Network

[0100] 403 Decoding

[0101] 404 Segmentation

[0102] 410 Ground truth data

[0103] f frequency

[0104] N Number of frequency chirps

[0105] T1 Duration

[0106] B Bandwidth

[0107] T2 Duration

Claims

1. A method for identifying interference in a radar system (10) of a vehicle (1), wherein, Perform the following steps: - Receive (101) at least one received signal (202) of the radar system (10), - Determine (102) detection information (210) from the received signal (202), - Perform an evaluation (103) of the detection information (210) by means of at least one neural network, - Use (104) the result of the evaluation (103) as a prediction of interference (220) for the received signal (202), wherein, for one detection period (260), a plurality of transmitted signals (201) of the radar system (10) are successively emitted in at least one frequency range (251, 252, 253) respectively in order to receive the respective received signals (202); and wherein the result of the evaluation (103) has an indication of an interference frequency range within which interference (220) is predicted in a detection period (260) that is temporally subsequent, wherein using (104) the result of the evaluation (103) includes automatically at least partially adjusting the frequency range (251, 252, 253), in which case the transmitted signal (201) is emitted in a frequency range (251, 252, 253) that is at least partially outside the predicted interference frequency range.

2. The method according to claim 1, wherein, the at least one neural network includes at least one recurrent neural network (402), and the recurrent neural network takes into account the evaluation (103) of the detection information (210) that is temporally prior and uses the output of the neural network as the result of the evaluation (103).

3. The method according to claim 2, wherein, the at least one neural network includes at least one convolutional neural network (401), the convolutional neural network obtains the detection information (210) as an input, and uses the output of the convolutional neural network as an input for the recurrent neural network (402).

4. The method according to any one of claims 1 to 3, wherein, performing the evaluation (103) of the detection information (210) includes the following steps: - Preprocess the detection information (210) of the detection period (260) in order to reduce the data size of the detection information (210), - Extract information about interference (220) in the form of at least one interference in the received signal (202) from the preprocessed detection information (210), - Perform a prediction of the at least one interference for a detection period (260) that is temporally subsequent by means of the extracted information.

5. The method according to claim 4, It is characterized in that The preprocessing is performed by max pooling.

6. The method according to claim 4, It is characterized in that The extraction is performed by a convolutional neural network (401).

7. The method according to claim 4, Characterized in that, Perform a prediction of the at least one interference for a detection period (260) that is temporally subsequent by means of the evaluation (103) of the detection information (210) that is temporally prior.

8. The method according to any one of claims 1 to 3, wherein, The result of the evaluation (103) is the output of the neural network.

9. The method according to any one of claims 1 to 3, characterized in that using (104) the result of the evaluation (103) comprises the following steps: - electronically outputting the result to an electronic device of the vehicle (1).

10. The method according to claim 9, characterized in that the result of the evaluation (103) is the output of the neural network.

11. The method according to claim 9, It is characterized in that electronically outputting the result to an electronic device of the vehicle (1) for a control device of the vehicle (1).

12. The method according to any one of claims 1 to 3, characterized in that the result of the evaluation (103) comprises a segmentation of the detection information (210), the segmentation indicating the predicted interference (220).

13. The method according to any one of claims 1 to 3, characterized in that the at least one neural network is trained by setting the following training steps: - storing a plurality of detection information (210) determined successively in time, - providing ground truth data (410) by identifying an interference in the detection information (210), - training the neural network by means of training data formed by the detection information (210) and the ground truth data (410).

14. The method according to claim 13, characterized in that providing the ground truth data (410) comprises the following steps: - manually identifying the interference so that the at least one neural network is configured by training to predict the interference as the interference (220) of the received signal (202).

15. The method according to any one of claims 1 to 3, characterized in that the transmitted signal (201) is respectively implemented as at least one chirp, the chirp having a frequency varying in time within the frequency range (251, 252, 253).

16. The method according to claim 15, characterized in that the determination (102) of the detection information (210) is carried out for each detection period (260), and the determination of the detection information respectively comprises the following steps: - carrying out a mixing of the respective transmitted signal (201) and the associated received signal (202) in order to respectively obtain a baseband signal (203), - determining the detection information (210) from the obtained baseband signal (203), the detection information (210) being specific for object detection in the surroundings (2) of the vehicle (1).

17. The method according to claim 15, characterized in that the at least one frequency range (251, 252, 253) is implemented as a frequency range that can be changed at least partially.

18. A radar system (10) for detecting a target object (230) in the surroundings (2) of a vehicle (1), the radar system having a processing device (15), the processing device being implemented for carrying out the following steps: - providing detection information (210) from a received signal (202) of the radar system (10), - carrying out an evaluation (103) of the detection information (210) by means of at least one neural network, - Use the result of the evaluation (103) in the use (104) as a prediction of the interference (220) on the received signal (202). Among them, For a detection period (260), successively emit a plurality of transmission signals (201) of the radar system (10) in at least one frequency range (251, 252, 253) respectively, so as to receive the respective received signals (202); and wherein the result of the evaluation (103) has an indication of the following interference frequency range, within which interference (220) is predicted in a subsequent detection period (260) in terms of time, and wherein using the result of the evaluation (103) in the use (104) includes automatically at least partially adjusting the frequency range (251, 252, 253), and in the case of such adjustment, emit the transmission signal (201) in the following frequency range (251, 252, 253), which is at least partially outside the predicted interference frequency range.

19. The radar system (10) according to claim 18,[ characterized in that it is provided with a processing unit, which, when executed by the processing device (15), causes the processing device to implement the steps of the method according to any one of claims 1 to 17.

20. A computer program product, which includes instructions that, when the computer program is executed by the processing device (15), cause the processing device to implement the following steps:[ - Provide detection information (210) from the received signal (202) of the radar system (10), - Implement the evaluation (103) of the detection information (210) through at least one neural network, - Use the result of the evaluation (103) in the use (104) as a prediction of the interference (220) on the received signal (202), Among them, For a detection period (260), successively emit a plurality of transmission signals (201) of the radar system (10) in at least one frequency range (251, 252, 253) respectively, so as to receive the respective received signals (202); and wherein the result of the evaluation (103) has an indication of the following interference frequency range, within which interference (220) is predicted in a subsequent detection period (260) in terms of time, and wherein using the result of the evaluation (103) in the use (104) includes automatically at least partially adjusting the frequency range (251, 252, 253), and in the case of such adjustment, emit the transmission signal (201) in the following frequency range (251, 252, 253), which is at least partially outside the predicted interference frequency range.

Images