Compensation for interference through seasonal time series decomposition in radar systems

Abstract

Method (100) for compensating interferences for a radar system (10), wherein at least the following steps are performed: - Providing (110) a receiving matrix (111) comprising a received radar signal (302) and at least one received jamming signal (303), - Compensation (120) of the receive matrix (111) to obtain a compensated receive matrix (121), thereby at least partially reducing the influence of the at least one received interference signal (303) in the compensated receive matrix (121).

Description

[0001] The invention relates to a method for compensating interference for a radar system, as well as a computing unit for implementing the method. Furthermore, the invention relates to a radar system and a vehicle with such a radar system. Finally, the invention comprises a non-volatile storage medium and a computer program product.

[0002] Radar systems are known, for example, for use in vehicles to detect at least one object. These can be vehicles traveling ahead that are at a certain distance, or at different distances, from the vehicle and, in particular, are moving away from or approaching at a relative radial speed. To assist the driver, for example, in the context of at least partially autonomous driving, the distance and / or the (relative radial) speed of the at least one object can be determined, in particular using a range Doppler map.

[0003] FMCW radar systems and PMCW radar systems are particularly well-known. In FMCW (Frequency Modulated Continuous Wave) radar systems, the frequency is modulated during transmission, while in PMCW (Phase Modulated Continuous Wave) radar systems, the phase is modulated.

[0004] The signal generation in an FMCW radar system, which can also be applied according to the invention, is described, for example, in "V. Winkler, "Novel waveform generation principle for short-range FMCW-radars," in Proc. German Microw. Conf., 2009, pp. 1-4."

[0005] However, the known devices and methods from the prior art have various disadvantages. Ideally, a radar system, for example that of a vehicle, emits a transmission signal, which is then at least partially reflected by at least one object and subsequently received as a reception signal by the radar system. However, when receiving radar signals, it is not possible to distinguish between radar radiation emitted and subsequently reflected by the (own) radar system on the one hand, and other external interference signals on the other. External interference signals can, for example, be radar radiation transmitted directly by other radar systems, for example from other vehicles approaching in a different lane. This can be particularly serious when the frequencies are the same or at least overlap.Alternatively or additionally, it may also be reflected transmission radiation from other radar systems. This can lead to interference and / or superposition of the different radar signals, which distorts the received signal. In other words, only a received signal can be recorded in which the different signal components are not directly separable. In further processing and use of the data, this can lead to increased noise, a deteriorated signal-to-noise ratio, false detections, artifacts, e.g. in a range Doppler map, inaccurate determination of distance (and / or speed), and / or in the worst case, incorrect driving actions (or even accidents), if, for example, the distance to a vehicle ahead is estimated to be greater than it actually is.Depending on the strength and / or length of the interference signal, the reception matrix and / or range-Doppler matrix can also be influenced over a large length and / or particularly strongly.

[0006] It is therefore an object of the present invention to at least partially overcome at least one of the disadvantages described above. In particular, the object of the invention is to enable improved signal quality on the receiving side. In this case, the signal-to-noise ratio (SNR) can be improved. Furthermore, a more robust, reliable, and / or more accurate database can be created, for example to calculate a range Doppler map and, based thereon, to determine a distance and / or the speed of an object (to be detected). Furthermore, the object of the invention can be to achieve an improvement in the signal-to-interference ratio (SIR), which is measured, for example, in dB.The signal-to-noise ratio can be determined by forming a quotient by relating the SIR of a reception matrix without applying the method according to the invention to the SIR of a reception matrix compensated by the method according to the invention.

[0007] The above object is achieved by a method having the features of independent patent claim 1, by a computing unit having the features of independent patent claim 11, by a radar system having the features of independent patent claim 12, by a computer program product having the features of independent patent claim 13, by a non-volatile computer-readable storage medium having the features of independent patent claim 14 and by a vehicle having the features of independent patent claim 15. Further features and details of the invention emerge from the subclaims, the description and the drawings.Features and details described in connection with the method according to the invention naturally also apply in connection with the computing unit according to the invention and / or in connection with the radar system according to the invention and / or in connection with the computer program product according to the invention and / or in connection with the non-volatile computer-readable storage medium according to the invention and / or in connection with the vehicle according to the invention and vice versa, so that with regard to the disclosure of the individual aspects of the invention, reference is or can always be made to each other. In particular, advantages described in the context of the first, second, third, fourth, fifth and / or sixth aspect also apply to the first, second, third, fourth, fifth and / or sixth aspect.

[0008] According to a first aspect of the invention, there is provided a method for compensating interference for a radar system, wherein at least the following steps are carried out: - Providing a reception matrix comprising a received radar signal and at least one received interference signal, - Compensation of the reception matrix in order to obtain a compensated reception matrix, whereby the influence of the at least one received interference signal in the compensated reception matrix is ​​at least partially, preferably completely, reduced.

[0009] The radar system can comprise an FMCW and / or PMCW radar system. These may be particularly well-suited for use in vehicles, particularly due to their robustness, cost, and / or availability.

[0010] The method can preferably be computer-implemented and, in particular, at least partially, preferably completely, carried out by a computing unit.

[0011] Interference can result from a superposition, particularly superposition or summation, of reflected transmission signals (which are actually intended to be received) and (unwanted) interference signals.

[0012] The reception matrix can preferably comprise the time-dependent reception signals received by the radar system, in particular in matrix form, continuously along the columns. The reception matrix can preferably comprise, in particular, complex-valued entries, comprising, for example, 64 chirps and 256 (temporal) samples.

[0013] Provision can be made within the framework of the conventional use of the radar system (see also explanations below).

[0014] Compensation of the reception matrix can involve reducing (i.e., compensating) the influence of the interference signal. Parts of the reception matrix that are particularly strongly influenced by the interference signal can be identified and subsequently compensated. Compensation of the parts can be achieved, for example, by replacement, in particular by superposing trend components and / or seasonal components, which can preferably be determined by seasonal time series decomposition. This can achieve higher signal quality, especially since heavily "contaminated" parts are not used. This can significantly improve the SNR and / or SIR, especially without requiring additional hardware in addition to the hardware already in use.

[0015] Based on the compensated range Doppler matrix, a (compensated) range Doppler map can be calculated. Similarly, based on an (uncompensated) receive matrix, an (uncompensated) range Doppler map can be calculated. By comparing the results, the effect of the inventive method can be quantified, e.g., the improved SNR and / or SIR can be quantified.

[0016] Within the scope of the invention, it may be advantageous that the provision of a reception matrix comprises at least one, preferably all, of the following steps: - transmitting a transmission sequence comprising chirps indexed by k, wherein a transmission sequence comprises a total of K chirps, wherein in particular 0 ≤ k ≤ K-1 applies, wherein in particular k and K are each a natural number, - At least partial reflection of the transmission sequence on at least one object, which is in particular located at a distance, - Receiving a reception sequence, which in particular comprises a received radar signal which at least partially corresponds to the K chirps of the transmission sequence reflected by at least one object, wherein the reception sequence (further) comprises at least one received interference signal, wherein in particular the received interference signal at least partially comprises an interference signal which was emitted by an external interference source, wherein in particular the external interference source comprises an external radar system, wherein in particular the interference signal at least partially interferes with at least one chirp of the transmission sequence and / or the received radar signal, - Sampling the reception sequence to obtain a, in particular discrete two-dimensional, reception matrix, in particular with K columns and S rows, where in particular S corresponds to the number of samples and K to the number of chirps.

[0017] In this case, it can be provided (in particular in the context of FMCW radar imaging) that each chirp, in particular linearly, traverses a, in particular discrete, bandwidth fhub of transmission frequencies, wherein in particular a central frequency of the different K chirps increases, in particular linearly, wherein a, in particular discrete, bandwidth fbw,center between 1 MHz and 10 GHz, preferably between 10 MHz and 1 GHz, particularly preferably between 700 MHz and 900 MHz. In this case, a transmission sequence can essentially comprise a series of adjacent chirps, which are preferably transmitted one after the other. Each chirp can comprise ramp-like (transmission) frequencies, which are preferably transmitted by a radar system. It can be provided that the (frequencies of the) chirps are changed essentially in a sawtooth manner during transmission.

[0018] The transmission sequence is preferably at least partially reflected by at least one object, preferably by all objects within range. For the purposes of the invention, it can be assumed for simplicity that essentially the entire transmission sequence is reflected and can be received as a reception sequence. This can result in the transmission sequence essentially receiving the (transmission) chirps. In other words, the reflected radiation can be received in the form of a reception sequence, which can be converted by sampling, in particular an analog-to-digital conversion, into a reception matrix, which is preferably discrete.

[0019] It can be provided that the reception sequence is a mixture, in particular a superposition, of received radar signals, i.e. in particular the actually desired transmission signals (of the own radar system) reflected from an object to be detected, and the direct and / or reflected interference signals from an external interference source, in particular an external radar system. An external radar system can, for example, be comprised of another vehicle, in particular an oncoming vehicle. In extreme cases, it can, for example, be an identical radar system, which can lead to particularly extensive interference. In other words, another external radar system can interfere with the actual radar measurement.

[0020] The receive matrix can have a two-dimensional matrix. A fast-time domain can be provided along a y-direction or along the columns, while a slow-time domain can be provided along an x-direction perpendicular thereto or along the rows. It can be provided that essentially the data attributable to a chirp is entered along the fast-time domain (in particular "fast" change, hence "fast"). It can be provided that essentially the data attributable to different chirps is entered along the slow-time domain (in particular "slow" change, hence "slow"). It can thus be provided that a row essentially represents the same times within a chirp. A column can represent the course of a chirp. The receive matrix therefore preferably includes temporal data, in particular along orin the fast-time domain, preferably (also) along or in the slow-time domain.

[0021] The number of samples S can be, for example, 128, 256, or 512. The receive matrix can therefore have a sampled receive signal along a column, which is attributable in particular to an at least partial reflection of a corresponding chirp of the transmit sequence. For example, the receive matrix (as described above) can have, in particular, complex-valued entries, comprising 64 chirps and 256 samples. Accordingly, a receive matrix in this example can have 64 columns and 256 rows.

[0022] Within the scope of the invention, it is conceivable that the reception matrix is ​​complex-valued and the compensation of the reception matrix is ​​carried out for the real part of the reception matrix and / or the imaginary part of the reception matrix.

[0023] This allows for separate application of the method for the real part and / or the imaginary part. This allows for comparison of the results. This advantageously allows for verification of the results without the need for additional data or information. This can lead to more precise and / or robust detection, for example, when a more reliable decision can be made as to whether a particular column is contaminated, i.e., subject to particularly strong interference, and therefore needs to be replaced.

[0024] It is conceivable to perform a replacement whenever both the real and imaginary parts indicate a disturbance. If only the real or imaginary parts indicate a disturbance, it may be planned to perform a replacement to calculate a range Doppler map. By comparing this result with the result without compensation (e.g., using SNR or SIR), it can be determined whether a replacement is or was justified. Alternatively or additionally, it is conceivable to calculate the signal amplitude and apply the remaining processing steps to it. This can make it possible to combine all the information into a single value (a kind of compression), thereby advantageously saving computing power.

[0025] It can be provided within the scope of the invention that during the compensation the reception matrix is ​​processed in parts, in particular column by column, in that the compensation is carried out step by step for one column of the reception matrix, wherein in particular one column corresponds to one received chirp.

[0026] In this case, a chirp can essentially comprise (as described above) the at least partially reflected transmission signal of the radar system, and in particular also interference signals which are also detected within the chirp.

[0027] The receive matrix or its columns ("receive chirps") can correspond to the input of the process. Column-by-column processing allows for easy discarding and / or replacement of precisely these chirps or parts of the receive matrix. Dividing the data into columns also enables particularly high levels of parallelization during calculation.

[0028] It is further conceivable that the application of a seasonal time series decomposition is carried out for the compensation, wherein the application of a seasonal time series decomposition involves dividing the reception matrix into at least one time-dependent component, wherein in particular the at least one time-dependent component has at least one of the following time-dependent components: - a trend component T(t), - a seasonal component S(t), and / or - a residual component e(t).

[0029] The trend component can represent the general trend of the signal curve, in particular at the (discrete) time t. The seasonal component can represent the behavior within a specific, in particular moving, interval, in particular at the (discrete) time t. The remaining residual component, in particular at the (discrete) time t, can advantageously contain at least partial information about interference signals, which in particular are not attributable to the radar system's actual transmission or reception signal of interest (in particular its comparatively stronger influence on the trend component and / or seasonal component), but rather to the influence of an external interference source. This not only allows the detection of interference, in particular based on the residual component, but can also enable the replacement of contaminated parts with the trend component and / or seasonal component.

[0030] Seasonal time series decomposition can be considered a statistical method. It involves a partitioning of the reception matrix or the processed component of the reception matrix (reception matrix component), for example, a column of the reception matrix. The reception matrix can be represented within a preferred additive model as: Y(t)=T(t)+S(t)+e(t)

[0031] Alternatively or additionally, the use of a multiplicative model is conceivable: Y(t)=T(t)∗S(t)∗e(t)

[0032] For the problem of interference compensation in a radar system, the additive model may be particularly preferable, especially over a multiplicative model, especially since it can be assumed that the interference signal is superimposed (e.g., superposed) with the reflected transmitted signal. The seasonal component may be the only parameter that needs to be used, particularly either in the context of the moving-average filter and / or when using the LOESS filter.

[0033] It is also conceivable that when applying a seasonal time series decomposition to the reception matrix, a moving average filter is applied, in particular to extract the trend component and / or the seasonal component, wherein the trend component and / or the seasonal component are subsequently subtracted from the reception matrix (for example, column by column or per received chirp) to obtain the residual component.

[0034] The moving average filter or its application can involve the formation of a moving average and / or be defined in particular as follows: T(t)=1q∑j=−kkYt+j

[0035] Here, Y(t) can represent the reception matrix, particularly a discrete and / or time-dependent one, especially its (column-wise) entries or chirps. This (above) moving-average filter can be used for smoothing, in particular by calculating the moving average. As a result, the trend component T(t) can be determined or estimated. The moving-average filter can have a period q: q=2k+1

[0036] The period q can be a natural number, for example, 30. Thus, q (e.g., a total of 30) samples can be used to calculate the trend component, especially at a specific point in time. Specifically, the average value of 30, particularly consecutive, samples, in particular values ​​of Y(t), can be used to determine the trend component. This allows sufficient smoothing to be achieved with comparatively low computational effort.

[0037] The seasonal component (S), especially after determining T, can be determined by: S(t)=(Y(t)−T(t)) / R

[0038] Here, R can be defined as the average value of all values ​​of the term (Y(t) - T(t)), especially over a period q. For example, a specific value of the term (Y(t) - T(t)) can be normalized with the average value of a total of q (e.g., 30), especially consecutive, samples or (Y(t) - T(t)).

[0039] The residual component e can be determined, in particular after determining S, by: e(t)=Y(t)−T(t)−S(t)

[0040] Thus, the trend component and the seasonal component can be clearly determined from the reception matrix, e.g., a received chirp contained in a column. The residual component can be obtained by subtraction, and based on this, a disturbance in this chirp can be determined. If a disturbance is detected, e.g., above a threshold value, the chirp can be replaced by a combination of the trend component and / or the seasonal component in order to advantageously optimize the SIR and / or SNR or to enable the advantages of the invention.

[0041] The implementation of the moving average filter can be particularly simple and / or computationally efficient (calculable more quickly). Furthermore, visual validation, for example, by a validation algorithm and / or a human user, can be particularly fast and / or intuitive.

[0042] Within the scope of the invention, it is optionally possible to use a LOESS filter for the application of a seasonal time series decomposition, wherein at least one of the following steps is carried out, in particular in the following order: - Dividing the reception matrix into overlapping reception matrix intervals, for example columns or chirps, - for each reception matrix interval, determining a trend component by fitting a local weighted, in particular smoothed, regression curve to the values ​​of the reception matrix interval, by minimizing the sum of the squared residuals between the values ​​of the reception matrix and the regression curve, in particular this can be done piecewise or over the entire reception matrix interval, - subtracting the trend component from the reception matrix to obtain an intermediate component, the intermediate component comprising the seasonal component and / or the residual component, - dividing the intermediate component into overlapping intermediate component intervals, wherein in particular the overlap can be between 0 and the total length of the intermediate component, preferably between 0.1 and 0.5 of the total length of the intermediate component, particularly preferably between 0.2 and 0.3 of the total length of the intermediate component, - for each intermediate component interval, in particular repeating or applying the second step, determining a seasonal component by fitting a local weighted, in particular smoothed, (second) regression curve to the values ​​of the intermediate component interval, by minimizing the sum of the squared residuals between the values ​​of the intermediate component interval and the regression curve, - Calculate the residual component by subtracting the trend component and the seasonal component from the reception matrix.

[0043] The LOESS filter (locally estimated scatterplot smoothing) can lead to smoothing, especially locally.

[0044] Advantageously, the use of a LOESS filter can be more robust, especially compared to a moving average filter.

[0045] It may be intended to use only a moving average filter or a LOESS filter. This can be particularly computationally efficient. Alternatively, it may be intended to use both a moving average filter and a LOESS filter, particularly to enable validation of the results. This can also support the decision, within the context of a learning process, as to which method produces better results in a given situation.

[0046] Furthermore, it can be provided within the scope of the invention that a detection of interference in the reception matrix, for example in its columns or chirps, is carried out, in particular following the application of a seasonal time series decomposition, comprising at least one of the following steps: - Providing the remaining component, - Calculating an amplitude and / or a power of the residual component, - applying a detection step to the amplitude and / or power of the residual component in order to detect components of the reception matrix contaminated by interference and to determine their position in the reception matrix, wherein the detection step preferably comprises at least one K-means clustering, a CFAR method, adaptive thresholding, linear prediction and / or a machine-learning method.

[0047] The amplitude can be defined as |e(t)|.

[0048] Preferably, the amplitude, in particular only the amplitude, can be used for calculation. This can be advantageous for reducing the number of calculation operations.

[0049] The power can be defined as: |e(t)| 2 Alternatively or additionally, it may be provided to use the power, in particular to adequately take into account the power of the reflected transmission signal and / or if the effect or severity of the interference is comparatively not too strong.

[0050] In particular, the invention is based on the finding that the residual component optimally reproduces random fluctuations and / or noise, which are attributable, for example, to interference signals, particularly compared to the seasonal component and / or the trend component. The residual component may comprise uncorrelated and / or random components of the time signal, in particular of the reception matrix, which may preferably include interference based on interference signals.

[0051] The goal may be to detect contaminated components (i.e., components corrupted by the interference signal), such as columns or received chirps. By determining their position (e.g., column number) within the reception matrix, these components can be replaced.

[0052] K-means clustering may be particularly preferred in this case, as it enables precise classification into clusters with comparatively low computational effort. This allows for particularly good yet fast detection, especially through peak detection, which advantageously enables reliable detection of contaminated parts of the reception matrix. Furthermore, the (often limited) computing power, for example in vehicles, may be sufficient to perform these steps.

[0053] Alternatively or additionally, a machine learning method can be used, including the use of a deep neural network, a convolutional neural network, a GAN, and / or autoencoder. The use of deep neural networks can be particularly precise, especially when sufficient data, computing power, and / or sufficiently verified or labeled data are available.

[0054] At least parts or all of the steps can be carried out by a computing unit.

[0055] With regard to the present invention, it is conceivable that, after the application of the detection step, an at least partial replacement of the components of the reception matrix contaminated by interference is carried out in order to obtain the compensated reception matrix, wherein in particular the components contaminated by interference are replaced by corresponding components of the trend component and / or the seasonal component.

[0056] Particularly preferably, replacement is carried out by a superposition of trend component and seasonal component. For example, a chirp or a column of the reception matrix is ​​identified as contaminated and in particular unusable. The column can be deleted based on its position. Instead, a superposition of trend component and seasonal component, preferably of identical length, can be inserted at this position in the reception matrix. Accordingly, the compensated reception matrix can comprise the reception matrix, or in particular exclusively comprise it, particularly if no disturbance was detected. If a disturbance was detected, the compensated reception matrix can at least partially comprise the original reception matrix (or its uncontaminated columns) and at least partially comprise columns which comprise a superposition of trend component and seasonal component.It is also conceivable that in the case of strong interference signals, the entire original reception matrix is ​​replaced by a superposition of the trend component and the seasonal component. In this case, and / or if a large portion, such as at least 30%, preferably at least 50%, particularly preferably at least 70%, ideally at least 80%, of the parts of the reception matrix are contaminated, a warning can be issued that the reliability of the relevant information is reduced. Corresponding derived data, such as a distance and / or a speed (based, for example, on a calculated range Doppler map) can then be discarded. It can also be provided that a vehicle then carries out a driving reaction, for example braking and / or evasive maneuvering.

[0057] A complete replacement of the contaminated parts may be particularly preferred. Preferably, entire columns of the receive matrix can be replaced to obtain the compensated receive matrix. This can enable a particularly efficient and / or parallelizable calculation. Provision can be made to determine the position of the contaminated parts, in particular their column (or column index / number) and / or row (or row index / number). Subsequently, the corresponding values ​​of the receive matrix can be replaced.

[0058] Furthermore, it is conceivable that after the compensation, a calculation of a range Doppler map is carried out based on the compensated reception matrix, wherein the calculation comprises at least one, preferably all, of the following steps: - Fourier transformation of the compensated reception matrix along a fast-time domain to obtain a range-slow-time matrix, - Fourier transform of the range-slow-time matrix along a slow-time domain to obtain a range-Doppler map, - Determining detection results based on the range Doppler map, in particular comprising the detection of at least one object, for example by a CFAR detection and / or a histogram detection method, whereby preferably a precise detection of at least one object is enabled, wherein the at least one object can be particularly sharply defined in particular due to a compensation of interference, whereby preferably a precise determination of the distance and / or a speed of the object can be achieved.

[0059] The Fourier transformation can be carried out, in particular, discretely and / or one-dimensionally.

[0060] It can be provided that detection results are summarized in an object list, wherein, for example, the object list has an object speed, a distance and / or an object position of the at least one object.

[0061] Furthermore, it is conceivable that, following the determination of detection results, a comparison is carried out with existing or previously measured detection results, in particular to check, validate and / or falsify them.

[0062] It can further be provided that the method, or its method steps, are carried out at least partially in parallel on different computing cores of a computing unit, in order to preferably enable faster and / or real-time calculation.

[0063] The above object is further achieved according to a second aspect by a computing unit according to the invention, comprising means for carrying out the method according to one of the preceding claims. The computing unit can be designed as part of a control unit in a vehicle, which advantageously saves additional components, weight and / or costs. Alternatively or additionally, a separate computing unit can be provided, which is in particular specially configured for the calculation. This can enable reliable and / or robust calculation, in particular without influencing a control unit. The computing unit can have ASICs optimized for the computing operations. The computing unit can also be part of a computing unit of the (existing) radar system.

[0064] The above object is further achieved according to a third aspect by a radar system according to the invention comprising at least one transmitter for transmitting at least one transmission sequence, at least one receiver for receiving at least one reception sequence based on the transmission sequence, and a computing unit according to the second aspect.

[0065] Within the scope of the invention, the distance of the at least one object can be a radial distance between the object and the transmitter or receiver. Preferably, the transmitter and receiver are essentially co-located.

[0066] The above object is further achieved according to a fourth aspect by a computer program product according to the invention, comprising instructions executable by a computing unit which, when the program is executed by a computing unit, cause the computing unit to implement the method according to the first aspect.

[0067] The above object is further achieved according to a fifth aspect by a non-volatile computer-readable storage medium according to the invention comprising a computer program product according to the fourth aspect.

[0068] The above object is further achieved by a vehicle according to the invention, comprising at least one radar system according to the third aspect.

[0069] Further advantages, features, and details of the invention will become apparent from the following description, which describes several embodiments of the invention in detail with reference to the drawings. The features mentioned in the claims and in the description may be essential to the invention individually or in any combination.

[0070] The invention is shown schematically in the following figures: Fig. 1 radar system for detecting at least one object, Fig. 2 Vehicle with radar system and external jamming signal, Fig. 3 reception matrix with columns, Fig. 4 Method 100 comprising sampling 304, Fig. 5 Method 100 comprising a seasonal time series decomposition, Fig. 6 Representation of a seasonal time series decomposition, Fig. 7 possibilities of seasonal time series decomposition, Fig. 8 Application of a LOESS filter, Fig. 9 Detection 140, Fig. 10 Method 100 for calculating a (compensated) range Doppler map, Fig. 11 Calculate 160, Fig. 12 Representation of the components Y, T,S, e, |e|, Fig. 13 Range Doppler map including the application of a moving average filter, and Fig. 14 Range Doppler map including the application of a LOESS filter.

[0071] Fig. 1 shows a radar system 10 according to the invention for detecting the distance 3 and / or the (relative) speed of at least one object 1. The radar system 10 can have a transmitter 12 for transmitting 200 a transmission sequence 201 comprising (transmitted) chirps 202, as well as a receiver 13 for receiving 300 a reception sequence 301. The reception sequence 301 can have a received radar signal 302, comprising at least partially the transmission sequence 201 reflected by the object 1. The radar device 10 can further have a computing unit 20 according to the invention, which can have means 21 for carrying out the method according to the invention.

[0072] Fig. 2 shows a vehicle 30 according to the invention comprising a radar system 10 according to the invention (as shown for example in Fig. 1) for detecting the distance 3 and / or the (relative) speed of a real object 1. The radar device 10 can further comprise a computing unit 20 according to the invention, which can comprise means 21 for carrying out the method according to the invention. In this case, the vehicle 30 can further comprise a non-volatile computer-readable storage medium 40 according to the invention, which preferably comprises a computer program product 50 according to the invention. The radar system 10 can comprise a transmitter 12 for transmitting 200 a transmit sequence 201, comprising (transmitted) chirps 202, and a receiver 13 for receiving 300 a receive sequence 301. The receive sequence 301 can comprise a received radar signal 302, comprising at least partially the transmit sequence 201 reflected by the object 1. Furthermore, the receive sequence 301 can additionally also comprise the signal from an external interference source 60, such asan external radar system 61, in particular another vehicle. The received interference signal 303 can directly comprise the interference signal 210 emitted by the external interference source 60. Alternatively or additionally, the interference signal 210 can also be scattered or reflected by the object 1 or another object and subsequently received as a received interference signal 303 by the receiver 13, and in particular be included in the reception sequence. Thus, the reception sequence 301 can comprise the (actually desired) received radar signal and the (actually undesired) received interference signal 303, which are detected in a mixed or superimposed manner.

[0073] Fig. 3 shows a reception matrix 111, which is obtained in particular by sampling the reception sequence 301. The chirps converted (received) by an ADC can be included along a column of the reception matrix. Columns can include columns 112 not contaminated by interference signals. If an interference signal is superimposed on a specific received part of the reception matrix, e.g., a received chirp, a contaminated column 113 results. The contaminated columns 113 can form the components of the reception matrix 111 contaminated by interference. This can be detected within the scope of the invention. Subsequently, this column 113 can be replaced, in particular by a superposition of trend components and seasonal components, in order to obtain a compensated reception matrix 120.

[0074] Fig. Figure 4 shows a method 100 comprising providing or processing the receive sequence 301, which is converted into discrete time signals by sampling 304, for example, by an ADC. These can be sorted along columns of a matrix to obtain a receive matrix 111.

[0075] Fig. 5 shows a method 100 comprising providing 110 a reception matrix 111, as well as compensating 120 the reception matrix 111 in order to obtain a compensated reception matrix 121 by applying a seasonal time series compensation 130, whereby the influence of the at least one received interference signal 303 in the compensated reception matrix 121 is at least partially reduced.

[0076] Fig. 6 shows the step of applying a seasonal time series compensation 130, whereby the receiving matrix 111 (e.g. its columns) can be used as input, and a decomposition into a trend component T, a seasonal component S and a residual component e is carried out.

[0077] Fig. 7 shows a method 100 wherein the application of a seasonal time series decomposition 130 is achieved by a moving average filter 131 and / or a LOESS filter 132 to obtain the trend component T, a seasonal component S and a residual component e.

[0078] Fig. Figure 8 shows a more detailed explanation of the application of a LOESS filter 132. This involves subdividing 133 the reception matrix 111 into overlapping reception matrix intervals. Subsequently, for each reception matrix interval, a trend component T is determined by fitting 134 a local weighted, in particular smoothed, regression curve to the values ​​of the reception matrix interval by minimizing the sum of the squared residuals between the values ​​of the reception matrix 111 and the regression curve. Subsequently, the trend component T is subtracted 135 from the reception matrix 111 to obtain an intermediate component, wherein the intermediate component comprises the seasonal component S and residual component e. Subsequently, the intermediate component is subdivided 136 into overlapping intermediate component intervals.Subsequently, for each intercomponent interval, a seasonal component S is determined by fitting 137 a local weighted, in particular smoothed, regression curve to the values ​​of the intercomponent interval by minimizing the sum of the squared residuals between the values ​​of the intercomponent interval 111 and the corresponding regression curve. Finally, the residual component e is calculated 138 by subtracting the trend component T and the seasonal component S from the receiving matrix 111.

[0079] Fig. 9 shows a detection 140 of interference, which can take place within the scope of the method 100, in particular following the seasonal time series decomposition 130. In this case, the residual component e can be provided 141. Subsequently, an amplitude and / or a power of the residual component e is calculated 142. Finally, a detection step 143 is applied to the amplitude and / or power of the residual component e in order to detect components 144 of the reception matrix contaminated by interference and to determine their position in the reception matrix, wherein the detection step 143 preferably comprises at least one K-means clustering, a CFAR method, adaptive thresholding, linear prediction (LPC) and / or a machine learning method.

[0080] Figure 10 shows the method 100 in a preferred order. This involves providing 110 a reception matrix 111, compensating 120 the reception matrix, applying a seasonal time series compensation 130 (according to one or both possibilities of Fig. 7) in order to subsequently detect 140 interference in the reception matrix 111. Subsequently, the parts contaminated by interference (such as columns of the reception matrix 111) are replaced 150, in particular to obtain a compensated reception matrix 121. Based on this, a (compensated) range Doppler map 161 can then be calculated, which, compared to a calculation based on the original (uncompensated) reception matrix, is optimized and preferably has improved SNR and / or SIR.

[0081] Fig. 11 shows a more detailed representation of the calculation 160 of a range-Doppler map 161. This starts with the compensated reception matrix 121. This is subjected to a Fourier transformation 170 along a fast-time domain (in particular along the columns) to obtain a range-slow-time matrix. This, in turn, is subjected to a Fourier transformation 180 along a slow-time domain (in particular along the rows) to obtain a range-Doppler map 161. Detection results can then be determined 190 based on the range-Doppler map 161. In this case, the at least one object 1 can be determined, advantageously with particular precision due to the compensation performed.

[0082] Fig. Figure 12 shows, by way of example, a time-dependent (discrete) signal Y(t), in particular a reception matrix 111, or a column 112, 113 of a reception matrix 111, for example with 256 samples. The signal Y(t) is divided into a trend component, which is divided by a seasonal time series decomposition 130 into a trend component T, a seasonal component S, and a residual component e. The amplitude |e(t)| is also shown. The trend component shows the respective trend of the signal Y(t) at a specific time t.

[0083] Fig. 13 and Fig. 14 shows the effect of method 100. Range Doppler maps are shown. The right image shows a range Doppler map without interference. The left image shows a range Doppler map calculated from a receive matrix without using method 100. The result can have poor SNR (strong noise is noticeable) and / or SIR. The object actually visible in the center can hardly be distinguished from the surrounding parts. In the middle, a range Doppler map 161 is shown, which was calculated using method 100. The range Doppler map can be significantly improved and can more closely approximate the result without the influence of interference (which is shown on the right).

[0084] This shows Fig. 13 the use of a moving average filter 131.

[0085] This shows Fig. 14 the use of a LOESS filter 132. List of reference symbols 1 object 3 Distance 10 Radar system 12 channels 13 recipients 20 computing units 21 remedies 40 storage medium 50 computer program product 60 external interference source 61 external radar system 100 procedures 110 Providing a reception matrix 111 Reception matrix 112 (uncontaminated) columns of the reception matrix 113 contaminated column of the reception matrix 120 Compensation 121 compensated reception matrix 130 Application of a seasonal time series decomposition T Trend component S seasonal component the residual component 131 Moving Average Filter 132 LOESS filters 133 Dividing the reception matrix into reception matrix intervals 134 Determining a trend component (T) by fitting 135 Subtracting the trend component from the reception matrix to obtain an intermediate component 136 Dividing the intermediate component into intermediate component intervals 137 Determining the seasonal component (S) 138 Calculating the residual component (e) 140 Detecting interference (in reception matrix) 141 Providing the remaining component (e) 142 Calculating an amplitude and / or power 143 Applying a detection step 144 components of the receiving matrix contaminated by interference 150 Replacing components contaminated by interference 160 Calculating a Range Doppler Map 161 Range Doppler Map 170 Fourier transform of the reception matrix 180 Fourier transform of each of the N filtered data sets 190 Determining detection results 200 Send 201 transmission sequence 202 Chirp 210 Interference signal 300 Receive 301 Receive sequence 302 received radar signal 303 received interference signal 304 samples QUOTES CONTAINED IN THE DESCRIPTION

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

[0000] V. Winkler, “Novel waveform generation principle for short-range FMCW radars,” in Proc. German Microw. Conf., 2009, pp. 1-4

[0004]

Claims

[1] Method (100) for compensating interference for a radar system (10), wherein at least the following steps are carried out: - providing (110) a reception matrix (111) comprising a received radar signal (302) and at least one received interference signal (303), - Compensation (120) of the reception matrix (111) to obtain a compensated reception matrix (121), whereby the influence of the at least one received interference signal (303) in the compensated reception matrix (121) is at least partially reduced. [2] Method (100) according to one of the preceding claims, characterized by that the provision (110) of a reception matrix comprises at least one, preferably all, of the following steps: - transmitting (200) a transmission sequence (201) comprising chirps (202) indexed by k, wherein a transmission sequence (201) comprises a total of K chirps (202), wherein in particular 0 ≤ k ≤ K - 1 applies, wherein in particular k and K are each a natural number, - At least partially reflecting the transmission sequence (201) on at least one object (1), which is in particular located at a distance (3), - receiving (300) a receive sequence (301), which in particular comprises a received radar signal (302) which at least partially corresponds to the K chirps (202) of the transmit sequence (201) reflected on at least one object (1), wherein the receive sequence (301) comprises at least one received interference signal (303), wherein in particular the received interference signal (303) at least partially comprises an interference signal (210) which was emitted by an external interference source (60), wherein in particular the external interference source (60) comprises an external radar system (61), wherein in particular the interference signal (210) at least partially interferes with at least one chirp (202) of the transmit sequence (201) and / or the received radar signal (302), - sampling (304) of the reception sequence (301) in order to obtain a, in particular discrete two-dimensional, reception matrix, in particular with K columns and S rows, where in particular S corresponds to the number of samples and K to the number of chirps (202). [3] Method (100) according to one of the preceding claims, characterized by that the reception matrix (111) is complex-valued, and the compensation (120) of the reception matrix (111) is carried out for the real part of the reception matrix (111) and / or the imaginary part of the reception matrix (111). [4] Method (100) according to one of the preceding claims, characterized by that during the compensation (120) the reception matrix (111) is processed in parts, in particular column by column, in that the compensation (120) is carried out step by step for a column of the reception matrix (111), wherein in particular one column corresponds to a received chirp. [5] Method (100) according to one of the preceding claims, characterized by that for the compensation (120) the application of a seasonal time series decomposition (130) is carried out, wherein when applying a seasonal time series decomposition (130) the reception matrix (111) is divided into at least one time-dependent component, wherein in particular the at least one time-dependent component has at least one of the following time-dependent components: - a trend component (T), - a seasonal component (S), and / or - a residual component (e). [6] Method (100) according to claim 5, characterized by , that when applying a seasonal time series decomposition (130) to the reception matrix (111), a moving average filter (131) is applied, in particular to extract the trend component (T) and / or the seasonal component (S), wherein the trend component (T) and / or the seasonal component (S) are subsequently subtracted from the reception matrix (111) to obtain the residual component (e). [7] Method (100) according to one of the preceding claims 5 or 6, characterized by that a LOESS filter (132) is used for the application of a seasonal time series decomposition (130), wherein at least one of the following steps is carried out: - dividing (133) the reception matrix (111) into overlapping reception matrix intervals, - for each reception matrix interval, determining a trend component (T) by fitting (134) a local weighted, in particular smoothed, regression curve to the values ​​of the reception matrix interval, by minimizing the sum of the squared residuals between the values ​​of the reception matrix (111) and the regression curve, - subtracting (135) the trend component (T) from the reception matrix (111) to obtain an intermediate component, the intermediate component comprising the seasonal component (S) and the residual component (e), - dividing (136) the intermediate component into overlapping intermediate component intervals, - for each intermediate component interval, determining a seasonal component (S) by fitting (137) a local weighted, in particular smoothed, regression curve to the values ​​of the intermediate component interval, by minimizing the sum of the squared residuals between the values ​​of the intermediate component interval (111) and the regression curve, - Calculating (138) the residual component (e) by subtracting the trend component (T) and the seasonal component (S) from the reception matrix (111). [8] Method (100) according to one of the preceding claims 6 to 7, characterized bythat a detection (140) of interference in the reception matrix (111), in particular following the application of a seasonal time series decomposition (130), is carried out, comprising at least one of the following steps: - providing (141) the residual component (e), - calculating (142) an amplitude and / or a power of the residual component (e), - applying a detection step (143) to the amplitude and / or power of the residual component (e) in order to detect components (144) of the reception matrix contaminated by interference and to determine their position in the reception matrix, wherein the detection step (143) preferably comprises at least one K-means clustering, a CFAR method, adaptive thresholding, linear prediction (LPC) and / or a machine-learning method. [9] Method (100) according to claim 8, characterized bythat after the application of the detection step (143), an at least partial replacement (150) of the components (144) of the reception matrix (111) contaminated by interference is carried out in order to obtain the compensated reception matrix (121), wherein in particular the components (144) contaminated by interference are replaced by corresponding components of the trend component and / or the seasonal component. [10] Method (100) according to one of the preceding claims, characterized by that after the compensation (120) a calculation (160) of a range Doppler map (161) based on the compensated reception matrix (121) is carried out, wherein the calculation (160) comprises at least one, preferably all, of the following steps: - Fourier transformation (170) of the compensated reception matrix (121) along a fast-time domain to obtain a range-slow-time matrix, - Fourier transformation (180) of the range-slow-time matrix along a slow-time domain to obtain a range-Doppler map (161), - Determining (190) detection results based on the range Doppler map (161), in particular comprising the detection of at least one object (1), for example by a CFAR detection and / or a histogram detection method, whereby preferably a precise detection of at least one object (1) is enabled, wherein the at least one object (1) can be particularly sharply defined, in particular due to a compensation (120) of interference, whereby preferably a precise determination of the distance (3) and / or a speed of the object (1) can be achieved. [11] Computing unit (20) comprising means (21) for carrying out the method (100) according to one of the preceding claims. [12] Radar system (10) comprising at least one transmitter (12) for transmitting (200) at least one transmission sequence (201), at least one receiver (13) for receiving (300) at least one reception sequence (301) based on the transmission sequence (201), and a computing unit (20) according to the preceding claim. [13] Computer program product (50) comprising instructions executable by a computing unit (20) which, upon execution of the program by a computing unit (20), cause the computing unit (20) to implement the method (100) according to one of claims 1 to 10. [14] Non-transitory computer-readable storage medium (40) comprising a computer program product (50) according to the preceding claim. [15] Vehicle comprising at least one radar system (10) according to claim 12.

Images