Radar system with improved rejection of jamming radiation effects

Abstract

The invention relates to a method for a radar system and to a radar system for robust environmental detection even in the presence of other potentially interfering radar systems. Here, the radar system comprises a transmitting device with a transmitting antenna for transmitting a transmitting signal, which comprises one or more sequences of K individual signals, a receiving device with a receiving antenna for receiving the transmitting signal reflected on an object, wherein for each of the K individual transmitting signals at least I digital reception values are formed for each receiving antenna in the receiving device, and a signal processing device for processing the received signal, wherein a further transformation of the received signal from a plurality of transmitting and / or receiving antennas is included in the digital signal processing device, which is referred to as a first forward transformation. After the first forward transformation in multiple dimensions, recognizable useful signal components are at least partially extracted, followed by a one- or multi-dimensional transformation, which is at least partially an inverse transformation of the forward transformation and is referred to as an inverse transformation. After the inverse transformation, recognizable interferences are at least partially suppressed, followed by a one- or multi-dimensional second forward transformation, wherein the results of the first forward transformation and the second forward transformation are used to detect the useful signal and thus to detect objects in the environment.

Description

Technical Field

[0001] This invention relates to a radar method for a radar system and a radar system for use in a driver assistance system of a motor vehicle. The radar system is preferably configured to perform the method according to the invention to improve the suppression of the negative effects of interference radiation, particularly from radar systems on other motor vehicles and commonly referred to as interference. Background Technology

[0002] Motor vehicles are increasingly equipped with driver assistance systems. These systems use sensor systems to detect the environment and, based on the identified traffic conditions, deduce automatic vehicle responses and / or provide instructions to the driver, particularly warnings. Here, comfort functions and safety functions are distinguished.

[0003] As a comfort feature, FSRA (Full Speed ​​Range Adaptive Cruise Control) plays an important role in current development. In this mode, if traffic conditions permit, the vehicle adjusts its own speed to the driver's desired speed; otherwise, its own speed automatically adapts to traffic conditions.

[0004] Safety features now come in various forms. In this case, one set of features reduces braking or stopping distance in emergency situations, up to autonomous emergency braking. Another set is lane change functionality: it warns the driver or intervenes in steering when the driver wants to make a dangerous lane change, i.e. when a vehicle in the adjacent lane is in the blind spot (called BSD – “Blind Spot Detection”) or is rapidly approaching from behind (LCA – “Lane Change Assist”).

[0005] But now drivers are no longer just being assisted; their tasks are increasingly being performed autonomously by the vehicles, meaning that drivers are being replaced more and more; this is known as autonomous driving.

[0006] For systems of the types described above, radar sensors are used, often fused with sensors from other technologies, such as camera sensors. One advantage of radar sensors is their reliable operation even in adverse weather conditions, and in addition to measuring the distance to an object, they can directly measure its radial relative velocity via the Doppler effect. Currently, 77 GHz and 79 GHz are commonly used as transmission frequencies.

[0007] The aforementioned functions require robust detection quality and high availability, which should be guaranteed even in the presence of interference radiation from other radar systems. Given the already high and continuing-growing prevalence of radar systems in motor vehicles (currently up to eight radar sensors per vehicle), effectively suppressing interference through signal processing is becoming increasingly important and challenging. This is because, for example, achieving overall avoidance by selecting different frequency bands is only achievable to a limited extent or not at all in the currently predominantly used 76-77 GHz band, as many sensors use a large portion or even the entire 1 GHz wideband for frequency modulation to achieve high accuracy and separation in distance measurements. However, currently known methods for suppressing interference through signal processing are limited in their performance. Summary of the Invention

[0008] The purpose of this invention is to provide a vehicle radar system with improved interference suppression achieved through improved digital signal processing.

[0009] This objective is essentially achieved by the method according to claim 1 and the radar system according to the co-claims. Advantageous embodiments of the invention are defined in the dependent claims. The core idea here is to first extract the strong useful signal (from the self-emitted signal reflected on the object) before interference suppression. For this purpose, a multidimensional spectral transformation is used; that is, the extraction of the strong useful signal occurs in the frequency domain, where the useful signal is more easily identified and distinguished than before the spectral transformation, due to the sensitivity gain of the spectral transformation and the interference converted into noise in that spectral domain. Subsequently, the multidimensional spectral transformation is at least partially undone / restored by an inverse transformation so that the interference suppression method can be subsequently applied.

[0010] The advantage of this invention is that, through this new and highly effective interference suppression, the radar system can maintain robust detection quality and high availability even in the presence of frequent and strong interference radiation.

[0011] The radar system involved in the method according to the invention for robust environmental detection even in the presence of other potential interference radar systems includes: 1) a transmitting device having one or more transmitting antennas operating serially or in parallel for transmitting a transmitted signal comprising one or more sequences of K single signals, the single signals preferably being identical or similar; 2) a receiving device having one or more receiving antennas operating serially or in parallel for receiving a transmitted signal reflected on an object, wherein, for each of the K individual transmitted signals, at least I digital received values ​​are formed in the receiving device for each receiving antenna, and the received values ​​are capable of containing (not exhaustively listed) the following components: a) a received signal from its own transmitted signal reflected on the object, hereinafter referred to as a useful signal; b) a received signal from a transmitted signal from other radar systems, hereinafter referred to as interference; and c) noise generated in the sensor itself, hereinafter referred to as inherent noise; and 3) a signal processing device for processing the received signal. The key feature of this method is that, in a digital signal processing apparatus, in addition to additional processing steps that can further transform the received signals from multiple transmit and / or receive antennas, it also a) uses K individual transmit signals and N RX At least I×K×N obtained from each receiving antenna RX The process involves: a) receiving a value, or performing a multidimensional signal transformation using the value derived therefrom, the multidimensional signal transformation hereinafter referred to as the first forward transformation, and the multidimensional signal transformation being suitably achievable in a multi-stage manner by performing a one-dimensional transformation in a corresponding sequence; b) after the multidimensional first forward transformation, at least partially extracting identifiable useful signal components; c) subsequently performing a complete or only partial one-dimensional or multidimensional transformation, the one-dimensional or multidimensional transformation being at least partially the opposite of the forward transformation, hereinafter referred to as the inverse transformation; d) after the inverse transformation, at least partially suppressing identifiable interference; e) subsequently performing a complete or only partial, one-dimensional or multidimensional second forward transformation, wherein preferably, identifiable interference is also at least partially suppressed during the multidimensional forward transformation; f) using the results of the first and second forward transformations to detect useful signals, and thereby detect objects in the environment.

[0012] Preferably, after the first forward transform, the extraction of identifiable useful signals is performed as follows: in the multidimensional numerical field obtained by the forward transform, values ​​that have or may have significant useful signal components are extracted, that is, they are cached and / or set to zero for use in the subsequent inverse transform, wherein, if necessary, more values ​​are set to zero than are cached, especially in order to save storage space.

[0013] Advantageously, the value containing, or at least possibly containing, the significant useful signal artifacts caused by the signal window (function) used in the positive transform can simply be set to zero, but not stored.

[0014] Furthermore, after the multidimensional positive transformation, the interference can be mapped to noise in at least one dimension, and the values ​​distinguished from the noise are identified as useful signals, wherein statistical methods are preferably used.

[0015] Advantageously, the inverse transform may not be exactly the inverse of the forward transform, especially because different signal windows are used, and / or different numerical scaling is used, and / or all values ​​that are not part of the forward transform are used in the inverse transform, and / or different number domains are used, especially the real and complex domains.

[0016] Preferably, the first positive transform and / or the second positive transform are implemented in a multi-level manner through multiple one-dimensional transforms, in which identifiable interference is also suppressed at least partially.

[0017] According to an advantageous design of the invention, interference is identified by distinguishing the values ​​affected by interference from the useful signal and / or inherent noise, preferably using statistical methods.

[0018] In addition, interference can be suppressed by reducing or setting the amplitude of the affected value to zero.

[0019] Advantageously, the result of the first forward transform can be combined with the result of the second forward transform in order to detect the useful signal in the following manner: the value extracted after the first forward transform, which is cached and set to zero for the subsequent inverse transform, is added to the result of the second forward transform, wherein preferably, in the case that interference suppression also changes the level of other signal components, the corresponding weighting is taken into account when adding.

[0020] According to an advantageous design of the invention, when detecting useful signals, especially to distinguish different categories for determining confidence levels, the categories may include: a) useful signals that have been identified after the first positive transform, b) useful signals that are identified only after the second positive transform but are unlikely to be interference suppression artifacts, and c) useful signals that are identified only after the second positive transform and are likely to be interference suppression artifacts. For this purpose, it is preferable to use an estimate of the energy of the useful signal removed due to interference suppression and a level / level / power ratio with other useful signals included after the second positive transform.

[0021] A key advantage of this invention is that a) a central buffer is provided, which preferably contains, in compressed form, two-dimensional data of a transformation region formed after performing a one-dimensional transformation on I received values ​​for each of K individual transmitted signals; b) the computation steps are performed accordingly in sequence, i.e., in a loop of one-dimensional data, and require at most one-dimensional, and thus significantly smaller, additional memory; and c) the data in the central buffer changes continuously throughout the computation.

[0022] Furthermore, multiple receiving antennas may be provided, and digital beamforming may be performed on the signals from the multiple receiving antennas, wherein a) extraction of the useful signal and / or at least one interference suppression is performed in the region obtained by digital beamforming after the first forward transform, b) if necessary, compressed data stored in a central buffer is not subjected to digital beamforming, c) if necessary, digital beamforming may be canceled by inverse operation, and beamforming and its inverse operation may be performed multiple times, and d) preferably, digital beamforming and its inverse operation are performed by means of discrete Fourier transform in the form of fast Fourier transform, especially when the receiving antennas are arranged in an equidistant grid.

[0023] Advantageously, for data whose content has not changed due to the extraction of useful signals or the suppression of interference, the inverse transform and / or the second forward transform may not be performed.

[0024] Furthermore, the forward transform, extraction of identifiable useful signals, inverse transform, and suppression of interference can be performed multiple times, especially in an iterative manner.

[0025] Preferably, the parameters of the transmitted signal—such as time interval, frequency position, and / or phase position—are varied in a random or pseudo-random manner to ensure that the radar system decorrelates relative to other radar systems to at least the extent that the interference is mapped as noise in at least one dimension after a multidimensional positive transformation.

[0026] An advantageous design of the invention is characterized by: a) in each of the K individual transmitted signals, the transmitted frequency varies linearly with at least approximately the same slope, wherein the frequency position of the transmitted signal—characterized by the starting frequency of the transmitted signal—varies at least approximately constant or linearly; b) where necessary, the spacing between the transmitted signals is at least approximately constant except for a small portion of random or pseudo-random and / or linear variations; c) in the receiving device, the received signal is mixed with the current transmitted frequency and filtered, in particular, using a low-pass filter; and d) the forward and inverse transformations are performed using the Discrete Fourier Transform, preferably calculated using the Fast Fourier Transform.

[0027] Furthermore, interference can be caused by or assumed to exist by at least one radar system that is also at least locally linearly frequency modulated. In particular, for each transmitted signal, pulse compression can be performed on the received value in the time or frequency domain, wherein the pulse compression is based on a filtered signal with a linear frequency variation, the slope of which is preferably estimated by the identified interference.

[0028] Another advantageous design of the invention is characterized by: a) within a single transmitted signal, the amplitude, frequency, and / or phase vary, for example, in the form of an OFDM signal or a signal with pseudo-random binary phase modulation; b) the spacing between the transmitted signals is at least approximately constant; c) in the receiving device, the received signal is down-converted using a constant frequency; d) each transmitted signal and I received values ​​from the receiving antenna are generated by Fourier transform and subsequently divided by the spectrum of the transmitted signal; e) the forward and inverse transforms are performed using the Discrete Fourier Transform, preferably calculated using the Fast Fourier Transform. Attached Figure Description

[0029] Figure 1 This illustrates frequency modulation consisting of a series of frequency ramps.

[0030] Figure 2 Showing N TX =4 transmitting antennas (TX1-4) and N RX =An antenna arrangement scheme with 4 receiving antennas (RX1-4).

[0031] Figure 3a The received signal is shown as a sinusoidal useful signal and superimposed interference. After suppressing the interference by setting the values ​​outside the boundaries marked by the dashed lines to zero, the result is... Figure 3b The signal shown has a spectrum in Figure 3c It is shown as a dashed curve in the middle; for contrast, Figure 3c The spectrum without interference suppression (solid line curve) and the spectrum obtained without superimposed interference (dotted line curve) are also shown.

[0032] Figure 4 The processing procedure according to the present invention is shown.

[0033] Figure 5a A two-dimensional FFT of the received signal is shown for example for four objects at distances of 10, 50, 100, and 200 m and with RCS values ​​of 10 dBsm, -5 dBsm, 25 dBsm, and 10 dBsm, respectively. Strong interference from another radar system is present, generating increased noise in the spectrum that overshadows the power peaks of both objects. The high-power peaks extracted from this spectrum (four power peaks for each of the two objects) are shown in... Figure 5bAs shown in the image.

[0034] Figure 6a The above example shows the sampled values ​​after the inverse transform and before the first interference suppression; Figure 6b The sampled values ​​are shown after the first interference suppression.

[0035] Figure 7 The example above shows a two-dimensional FFT after interference suppression and summation of the extracted power peaks; compared with... Figure 5a Compared to the original spectrum, the noise was significantly reduced, so the power peaks of all four objects became visible.

[0036] Figure 8 Another flow of the method according to the invention is shown, wherein additional digital beamforming is performed on the signal at the receiving antenna / across the signal at each receiving antenna. Detailed Implementation

[0037] In radar systems used in motor vehicles, a method based on... Figure 1 The transmitted signal is modulated by a sequence of K linear frequency ramps, which are distributed at least approximately equidistant in time. A detailed description of this modulation and related signal evaluation can be found in EP 2 629 113 B1; therefore, only a brief description is given here. The low-frequency received signal, generated after mixing / mixing the high-frequency received signal with the current high-frequency signal (part of which is transmitted and the other part is used for this mixing), contains components with different frequencies—the transmitted signal reflected back from the object forms a signal of constant frequency after mixing, where this frequency is substantially proportional to the distance to the object (the frequency component generated by the Doppler effect, i.e., the radial relative motion of the object with respect to the sensor, is very small relative to this distance-related frequency component).

[0038] During each frequency ramp, the low-frequency signal is sampled I times and digitized. Based on these I sampled values, a first Discrete Fourier Transform (FFT) is performed; in this FFT, the received signal reflected by the object generates power peaks at frequency support points corresponding to the respective distances. At each potentially functionally significant frequency support point of the first FFT, a second FFT is calculated based on the values ​​generated in the K frequency ramps; the frequencies of the power peaks generated in the second FFT are proportional to the radial relative velocities of the corresponding objects. After performing a two-dimensional FFT based on the I sampled values ​​from the K frequency ramps, a two-dimensional field is obtained in which each object generates a power peak whose position corresponds proportionally to both distance and relative velocity (except that the relative velocity contributes very little to the distance dimension) – therefore, the unit of the field obtained after the two-dimensional FFT is also called the range-relative-velocity gate or the range-Doppler gate.

[0039] As in Figure 2 As exemplarily illustrated, N exists in a radar system. RX One receiving antenna and N TX One transmitting antenna (in) Figure 2 N RX =4 and N TX =4, the receiving antenna is labeled RX1-4 and the transmitting antenna is labeled TX1-4). For each N RX Calculate the above two-dimensional FFT using the received signal from the receiving antenna. In N TX Simultaneous transmission occurs on multiple transmit antennas, where the transmitted signals are modulated to separate the components that contribute to the received signal. One possible modulation involves the phase of the transmitted signal changing linearly on a frequency ramp for each transmit antenna, where the rates of these phase changes differ between the transmit antennas. This results in a different shift of the power peak in the velocity dimension (i.e., the Doppler dimension) after a two-dimensional FFT. Thus, components from different transmit antennas can be separated, and the N-axis of the transmit and receive antennas can be used to determine the components. TX ·N RX The combination yields N TX ·N RX One signal, for Figure 2 Antenna examples, these signals correspond to having one transmitting antenna and N TX ·N RX =An antenna (array) with 16 receiving antennas spaced λ / 2 (λ is the wavelength of the center frequency of the transmitted signal, and the center frequency is approximately 76.5 GHz). For digital beamforming, in N... TX ·N RX On a signal (or across N) TX ·N RXFor each distance Doppler gate, a third FFT is calculated for each signal; the frequency of the power peak generated in the third FFT is proportional to the electrical angle of the corresponding object (the geometric angle and electrical angle are related by a simple nonlinear trigonometric relationship). This is achieved on I sampling values ​​across K frequency ramps and at N points between the transmitting and receiving antennas. TX ·N RX After performing a 3D FFT on the combination, a 3D field is obtained in which each object generates a power peak, the position of which corresponds substantially proportionally to its range, relative velocity, and electrical angle (except for displacement due to phase modulation of the transmitted signal in the relative velocity dimension, i.e., the Doppler dimension) – therefore, this cell of the field obtained after the 3D FFT is also called the range-Doppler angle gate. Thus, the range, relative velocity, and angle of an object can be determined from the position of the power peaks that appear after the 3D FFT; these power peaks are distinct from system noise (especially inherent noise generated by thermal noise); the radar cross section (RCS) of the object is derived from the height of the corresponding power peak.

[0040] Radar systems can receive signals not only emitted by themselves and reflected off objects, but also signals from other radar systems operating in the same frequency domain. Since almost all radar systems used in motor vehicles operate in the 76-77 GHz band, and use a large portion or even the entire band for frequency modulation, mutual illumination (also known as interference) occurs very frequently with the increasing prevalence of radar systems in motor vehicles. This interference signal can be particularly strong in cases of direct mutual illumination. Figure 3aThe diagram illustrates a received signal with respect to a frequency ramp, consisting of interference from illumination by another radar system and a strong intrinsic signal (i.e., a signal emitted by itself and reflected at a large and nearby object). (It should be noted that a real-valued mixer is considered here first; therefore, the received signal is real-valued.) As mentioned above, the intrinsic signal to be detected has a constant frequency (corresponding to the distance to the object). The interference signal (visible approximately in the time index range 170–255 and dominant in the range 180–240) has a rapidly changing frequency. This should now be explained in more detail: the interference radar system under consideration also has linear frequency modulation, but with a different slope than this system. In the received signal, the interference is visible in a short time domain around the intersection of the two frequency ramps (i.e., the frequency modulation of the jamming radar system and its own); the width of this time domain is defined by the width of the low-pass filter implemented after the mixer (such a low-pass filter is particularly used to suppress frequency components of signals and noise above half the sampling frequency) - therefore, in the digitized received signal, the interference is only visible when the frequency difference between the jamming frequency and its own radar frequency is numerically less than half the sampling frequency.

[0041] A common method to suppress this type of interference is to zero out signal values ​​that exhibit upward outliers in their amplitude (i.e., significantly higher than the average signal amplitude). This is also illustrated in Figure 3; a numerical threshold is determined from the value of the received signal (by multiplying the average signal amplitude by a factor of approximately 2.5; shown by dashed lines), and all values ​​above this threshold are zeroed out, thus obtaining... Figure 3b The curve shown has its FFT, i.e., spectrum, in Figure 3c The spectrum is shown as dashed curves representing power and dB—since a real-valued mixer is considered here, resulting in a real-valued sampled signal, only the lower half of the spectrum is shown; the upper half is its complex conjugate, carrying no further information and therefore functionally irrelevant. For comparison, Figure 3c The solid line curve shows the spectrum that may be generated without setting the abnormal signal values ​​caused by interference to zero (i.e., corresponding to...). Figure 3a The spectrum is shown as a dotted line curve representing the inherent signal, i.e., the spectrum generated without interference. All three spectra show the peak power of the inherent signal of interest (i.e., the useful signal), but the spectrum without interference suppression (solid line curve) and the spectrum with the signal anomalies caused by interference set to zero (dashed line curve) show, in addition to the peak power, components exceeding the noise amplitude of the interference-free spectrum (dotted line curve). Although these components are reduced in the spectrum with the signal anomalies caused by interference set to zero (dashed line curve), it can only be described as partial interference suppression because they are not completely eliminated.

[0042] These remaining components have two reasons: on the one hand, from Figure 3a and Figure 3b It can be seen that the interference signal is not zeroed out at all time points; that is, there are still some interference components in the signal. On the other hand, at those time points where the overall signal is zeroed out, the useful signal component itself is also zeroed out. Therefore, the sinusoidal useful signal is effectively / actually multiplied by a window function before the FFT, in which the values ​​are zeroed out in the inner region—the spectrum of such a window and the spectrum of the signal obtained by multiplying the window by the sinusoidal signal have obvious sidelobes.

[0043] These residual components beyond the power peaks have two negative impacts: First, they cause the noise level after the 3D FFT to rise above the normal range, i.e., to a level caused by its own noise, resulting in reduced detection sensitivity—smaller or more distant objects may no longer be detected in certain situations. Second, the spectral sidelobes generated by zeroing out the useful signal components in the first FFT may be cross-frequency ramp correlated in their phase, which could ultimately lead to additional power peaks after the 3D FFT, resulting in so-called ghost detection—that is, generating detections that do not actually exist in reality, and these detections could cause driver assistance systems implemented with this radar system to react completely incorrectly.

[0044] In particular, this situation, involving ghost detection, may occur when the jamming radar system emits at a constant frequency. Then, according to... Figure 1The radar system modulated as shown is interfered with at the same time position in each frequency ramp, so that the useful signal (the self-signal reflected from the object) is always zeroed in the same time domain. Therefore, the spectral sidelobes resulting from zeroing the useful signal are at least approximately the same in amplitude, while their phases vary linearly, similar to the phase of the useful signal, resulting in power peaks at the same relative velocity as the object after the second FFT; these peaks appear at multiple range positions because the sidelobes extend over multiple frequency support points of the first FFT. Regarding the signal on each receiving antenna (or across each receiving antenna), the sidelobes also have the same phase transition as the useful signal—and thus correspond to the angle of the object. However, for signal components assigned to different transmitting antennas (which originate from different Doppler gates under different linear phase modulations as described above), the phases of these sidelobes resulting from zeroing the useful signal are generally random. Therefore, after the three-dimensional FFT, ghost detection may appear at angles different from the object angle (corresponding to the useful signal). The ghost detection range also differs from the object distance, while the ghost detection relative velocity is consistent with the relative velocity of the real object. However, the latter only applies to the specific case where the illumination radar system has a constant frequency; if its frequency changes very slowly (e.g., according to...). Figure 1 During the entire frequency modulation process (that is, the overall linear change is about 50MHz within about 18ms), the relative velocity of the ghost detection will shift towards the relative velocity of the real object, and the greater the distance difference between it and the real object, the stronger the shift.

[0045] Therefore, the method of setting the outlier values ​​of the signal generated by interference to zero is limited in its ability to suppress interference; especially in the presence of relatively strong useful signals, not all interference is eliminated, and ghost detection may occur due to real objects. This constitutes an unacceptable limitation, especially given the large number of fairly strong useful signals that often appear in real-world scenarios (not only from the vehicle but also from the environment itself), particularly in autonomous driving where very high availability and detection quality are required.

[0046] These adverse effects can be largely eliminated by having interference suppression performed, at least in part, only after these strong useful signals have been eliminated. However, it is difficult, if not impossible, to determine these useful signals directly in the input signal (i.e., the sampled values ​​of each frequency ramp) (i.e., before the multidimensional FFT) in a simple manner: on the one hand, it is often difficult to distinguish between useful and interference signals. On the other hand, even relatively strong useful signals may not be higher than or significantly higher than the system noise within the sampled value range—because the integral gain of the three-dimensional FFT is in the range of 60 dB, meaning that useful signals as strong as the system noise within the sampled value range are 60 dB higher after the three-dimensional FFT, an amplitude far greater than that of weaker but still detectable useful signals, which are only 10 dB higher than the noise after the three-dimensional FFT. Due to the problem of not being able to extract strong useful signals simply and efficiently within the sampled value range, according to the present invention, a two-dimensional Fourier transform (first positive transform) with an integral gain close to 50 dB is first performed, and then the power peaks significantly higher than the noise are extracted (buffered and zeroed in the two-dimensional spectrum). Next, a two-dimensional inverse Fourier transform (i.e., inverse transform) is performed, the result of which represents the sampled values ​​without a strong useful signal. Based on these sampled values, known interference suppression methods are applied, and a second two-dimensional FFT (second forward transform) is performed; then, the power peak of the strong useful signal extracted after the first forward transform is added to the result.

[0047] Below, according to Figure 4 The method of the present invention will be described in more detail below. Once the sampled values ​​of the frequency ramp are available, then for N... RX For each of the receiving antennas, the sampled value is compared with the window function w. R1 The samples are multiplied and then transformed using an FFT (for the distance dimension). The window function assigns the highest weight to the intermediate sample values ​​and reduces the weight to the outer sample values ​​(i.e., the outer sample values ​​are less involved). Thus, although the spectral power peaks generated by the sinusoidal useful signal (after the FFT) are slightly wider, only a small number of sidelobes (i.e., spectral components located next to these power peaks) are produced. The FFT values ​​are compressed (e.g., using the method according to EP 3152587 B1) to occupy as little space as possible in memory (RAM = Random Access Memory). Here, in the case of the real-valued mixer considered first, only the first half of the spectrum consisting of I / 2 = 256 values ​​is stored (the second half is conjugate to the first half and does not carry information).

[0048] After processing and storing the sampled values ​​of all frequency ramps in this way (i.e., completing the processing loop on (all) frequency ramps (or across frequency ramps), its... Figure 4This is referred to as "Loop 1". The second processing loop (Loop 2) is executed at the frequency support points of the first FFT, i.e., at the distance gates (or across frequency support points or across distance gates). First, for the corresponding distance gate, the values ​​of the corresponding K frequency ramps are read from memory and decompressed; then, for N... RX For each of the receiving antennas, the values ​​from the K frequency ramps are multiplied by the window function w. D Then, an FFT (for dimensional Doppler, i.e., relative velocity) is used for transformation. For the resulting N... RX Statistical calculations are performed using the modulus K=1024; for example, these modulus values ​​are sorted, and then the value ranked 70th, starting from the minimum, is used (i.e., 70% of the values ​​are less than this value, and 30% of the values ​​are greater than this value); this value is usually called the OS70% value (OS = Ordered Statistics). The so-called extraction threshold is set to be 12dB above this OS70%; the distance gate N... RX All values ​​with moduli higher than the threshold in the Doppler FFT will be extracted. Figure 4 The extracted values ​​are at least mostly the power peaks of the useful signal, because the inherent noise and the noise generated by interference (which is significantly higher than the inherent noise in the example considered below) are statistically distributed such that they are below the determined threshold with almost 100% probability. It should also be noted that the interference signal caused by interference appears as noise after the two-dimensional FFT because its phase in each frequency ramp is random, which can be caused in particular by the pseudo-random part of the modulation in this radar system; for example, the distance between frequency ramps and the sampling time interval within the frequency ramp are slightly different (see also EP 2057480 B1), and the phase modulation of the transmitted signal (i.e., its phase changes on the frequency ramp sequence) has the same random component for all transmitting antennas.

[0049] The values ​​extracted from the corresponding Doppler FFT are stored (in...). Figure 4 In the "memory" module), and the Doppler FFT is set to zero at these locations. Then, the inverse FFT (needling the Doppler dimension) is performed. Figure 4 (The block "IFFT Doppler" in the middle), then divide by the Doppler window w DThis yields values ​​equivalent to those before the initial Doppler FFT, excluding the component generated by the extracted power peak. Finally, compression is performed; the compressed values ​​are stored in RAM in the same location as they were before this second processing loop (i.e., the original values ​​for the corresponding distance gates are overwritten by the new values)—meaning no additional memory is needed, and this also applies to other processing loops, where the same memory is always reused. If the power peak in the distance gate does not exceed the extraction threshold (i.e., no power peak is extracted), the subsequent processing steps of the second loop can be omitted, because the data after this loop is the same as the data before, and the original values ​​are still stored in memory.

[0050] Now, as an example, consider four objects at distances of 10, 50, 100, and 200 from the gate, with RCS values ​​of 10 dBsm, -5 dBsm, 25 dBsm, and 10 dBsm, respectively, where there is strong interference from another radar system. Figure 5a The table shows the values ​​of all range gates after the initial Doppler FFT (where, due to sequential loop processing, these values ​​actually exist and are processed once for each range gate in turn). Figure 5b The power peak extracted from this spectrum is shown (i.e., Figure 5b All values ​​greater than zero are stored and set to zero in the 2D FFT for subsequent processing; power values ​​are shown in dB. It should be noted that N... TX The different phase modulations of the signals transmitted by the four TX antennas result in four power peaks for a single object—each power peak corresponding to one TX antenna. For example... Figure 5b As shown, power peaks are extracted from only two objects (objects located at distances of 10 and 100 from the gate); the other two objects are overwhelmed by high noise generated by interference (see [reference]). Figure 5a (Due to its smaller RCS or greater distance, the received signal it produces is smaller than that of the other two objects).

[0051] Next, a third processing loop is performed, which is similar to the first processing loop, and further processes the data of each frequency ramp in turn. Figure 4(See "Loop 3" in the original text). After reading from and decompressing from memory, an inverse FFT (for the distance dimension) is performed. As mentioned above, in the case of the real-valued mixer under consideration, each frequency ramp and each receiving antenna stores only I / 2 = 256 values ​​in memory, that is, only half of the actual spectrum of the first FFT, since the other half is its conjugate complex number and does not carry additional information. Therefore, the first approach is to reconstruct the original spectrum of length I from the stored I / 2 values ​​accordingly (more precisely, except for the intermediate values, unless they are also not in memory, i.e., not storing I / 2+1 values ​​therein). The inverse FFT yields I real-valued sampled values. However, alternatively, the inverse FFT can also be performed on only the stored I / 2 spectral values ​​(i.e., an inverse FFT with half the length I / 2); then, after the inverse transform, I / 2 complex-valued sampled values ​​are obtained—thus equivalent to implementing the Hilbert transform and halving the sampling rate. The advantage of this alternative method is that it reduces the computational workload in the third processing loop considered here; therefore, the method of using I / 2 complex samples will be considered below.

[0052] After the inverse FFT, divide by the window function w R2 Now the original sampling range is reached again because the two-dimensional forward transform is compensated for by the two-dimensional inverse transform, i.e., it is canceled. However, to be precise, this only holds true if the window function w used for division after the inverse FFT of the distance is... R2 The window function w of the FFT with the initial distance R1 The same applies, and it only applies to cases where the extracted power peak is not considered, the transition to the complex sampling domain is not considered, and there may be different numerical scaling.

[0053] However, especially when using strong compression, if the window function w R2 A weaker effect (i.e., a gentler tilt from the center outwards) or complete omission of the window function w (i.e., essentially a constant value of 1) can be advantageous, because otherwise the following problem may arise: in the presence of strong interference or multiple strong useful signals from objects at different distances, compression can introduce noise that is significantly higher than the system noise (i.e., noise generated by the analog receiver). Therefore, for this compressed noise, the window function w of the initial range FFT... R1 It had no effect; after the inverse transform, i.e., after the inverse FFT of the distance, the compressed noise approximates white noise, meaning it exhibits an approximately constant amplitude across the sampled values ​​(except for statistical fluctuations in the noise). If this approximate white noise is divided by the window function, it may have a much higher amplitude in the outer regions than in the middle regions (especially in the case of strong windows), which may be detrimental to subsequent interference suppression (affecting both the calculation of the threshold and potentially causing these outer values ​​to be suppressed even when there is no interference at these locations).

[0054] In the third processing loop under consideration, after the inverse FFT of the distance and possibly subsequently divided by the window function, interference suppression is performed within the sampled value range (in... Figure 4 (This is referred to as "IFU1 = First Interference Suppression" in Chinese). It targets the corresponding frequency ramp and N. RX N receiving antennas RX • Statistical calculations are performed on 1 / 2 complex sampled values; for example, these moduli are sorted, and then the value ranked 50th, starting from the minimum, is used, i.e., the OS50% value. The interference suppression threshold is set 14 dB higher than this OS50% value. For the frequency ramp and N considered... RX Each receiving antenna sets all sampled values ​​above the threshold to zero. Because a strong useful signal has already been extracted, the remaining useful signal within the sampling range is now typically below the noise level. As a result, the interference suppression threshold is no longer (at least not significantly) increased by the useful signal, thus enabling better identification and suppression of interference.

[0055] After that, for N RX For each of the receiving antennas, execute the window function w R2 The multiplication operation is performed (if the window function was divided after the inverse FFT of the distance), and an FFT (for the distance dimension) is performed, where the FFT now has only a length of I / 2; except for the suppressed interference components, the equivalent value before the inverse FFT of the distance is obtained again. Finally, compression is performed; the compressed value is stored in memory (RAM) in the same location as it was before the third processing loop (i.e., the original value of the corresponding frequency ramp is overwritten by the new value). If there are no values ​​in the frequency ramp above the interference suppression threshold (i.e., no value is set to zero), the subsequent processing steps of this third loop can be omitted, because the data after this loop will be the same as the data before, and the original value is still stored in memory.

[0056] For the receiving antenna, Figure 6a The figure shows the magnitude of the sample values ​​before the first interference suppression. Figure 6b The table shows the magnitudes of the corresponding sample values ​​for each frequency ramp (where these sample values ​​are actually processed sequentially for only one frequency ramp through successive loop processing). It can be seen that the interference (upward signal outliers) is effectively suppressed, i.e., zeroed out, while the remaining signal appears as noise—where the level of useful signal still contained is lower than the noise level, and only becomes visible afterward through a multidimensional FFT (with high integral gain). It should also be noted that the above relationship causes the noise level to vary with the time index and is not entirely constant (especially since a window function w is not used here). R2 ).

[0057] Finally, the fourth processing loop, similar to the second, further processes the data for each distance gate. First, the data for the corresponding distance gate is read from memory and decompressed. Then, further interference suppression is performed (in... Figure 2 This is referred to as "IFU2 = Second Interference Suppression" in Chinese, meaning that in the corresponding range gate, at K frequency ramps and N... RX N receiving antennas RX • Eliminate the interference effects that occur there at K values. Interference typically occurs only at certain frequency ramps, or at least only at certain frequency ramps relative to the range gate (i.e., the frequency in the down-mixed received signal)—this is especially true when the frequency ramp slopes of this radar system and the interfering radar system are similar or even the same. Therefore, in the second interference suppression, the upward-deviational outliers in the signal level are again zeroed out. For this purpose, for all N values ​​in the considered range gate… RX N receiving antennas RX • Statistical calculations are performed on the moduli of K values; for example, these moduli are reordered, and then the value ranked 80th, starting from the minimum, is used, i.e., the OS80% value. The interference suppression threshold is set 11 dB higher than this OS80% value. In N RX • Of the K values, all values ​​above this threshold are set to zero.

[0058] Similar to the second processing loop, here for each N RX The receiving antenna combines the values ​​from K frequency ramps with a window function w. D The signals are multiplied and then transformed using an FFT (for the Doppler dimension). Within the currently reached signal range (i.e., after the 2D FFT), the high-power peaks extracted and stored in memory in the second processing loop are now read from there and re-added—thus obtaining a 2D FFT for the corresponding range gate, with interference largely suppressed and containing all useful signals. Finally, in the fourth processing loop, for the corresponding range gate, an FFT for angle formation (which performs digital beamforming for the equidistant antenna channels present here) is performed, and power peaks exceeding a detection threshold (determined adaptively by the data itself) are extracted; these detected power peaks are referred to as peaks. These peaks are further processed to derive an environment description interface for implementing driver assistance functions. Figure 4 These modules are no longer shown in the document.

[0059] The effectiveness of interference suppression will now be illustrated using four examples. Therefore, in Figure 7The diagram shows a two-dimensional FFT for the fourth processing loop, i.e., for all range gates, after performing the first and second interference suppressions and summing the power peaks extracted in the second loop (due to the successive loop processing, this two-dimensional FFT actually exists only for each range gate sequentially and is then further processed; that is, the processing does not require storing the entire two-dimensional FFT shown in memory at any given time). This is achieved by comparing... Figure 5a A comparison of the original two-dimensional FFT (i.e., before interference suppression) shows that, on the one hand, noise is now reduced (i.e., strong noise caused by interference is essentially eliminated), and on the other hand, all four objects are now visible, including the two smaller or farther objects located at distances of 50 and 150 from the gates (due to N). TX The transmitted signals from each antenna are modulated in different phases, with four power peaks corresponding to each object. No additional power peaks are generated, i.e., no ghost detection is produced. Finally, it should be noted that the power peaks of the two smaller or more distant objects that are now visible have slightly lower levels compared to the interference-free case; this is because some values ​​are zeroed out during interference suppression, thus the contribution of these locations to the power peaks is lost. In principle, this percentage loss can be determined (e.g., based on the number of values ​​zeroed out, or more precisely, by combining the values ​​at these locations with the corresponding window function); for example, it can be corrected only during peak processing (to obtain the correct RCS value) or it can be corrected before or during the summation of the extracted power peaks in the fourth processing loop (by applying appropriate weights to the Doppler FFT).

[0060] As described above, the power peaks extracted in the second loop are added again in the fourth processing loop. An alternative approach is to generate peaks directly from the extracted power peaks (after the FFT used for angle formation), and peak detection in the fourth processing loop is performed only based on the result without the extracted power peaks; however, this has two drawbacks: firstly, local maxima are often generated at the transition to the extraction region (because these values ​​are approximately zero in the extraction region), and these local maxima must be eliminated in peak detection. Secondly, the extracted high power peaks also contain noise components caused by interference, which, although significantly smaller than the corresponding power peaks themselves, still cause some measurement errors in distance, relative velocity, and angle (from a functional perspective, angle measurement error is the most critical). In the method shown here, noise components are largely avoided by adding the extracted power peaks because the Doppler FFT in the fourth processing loop (i.e., before the power peaks are added) approximately has a negative value for the noise component at the location of the extracted power peaks, thus eliminating the noise component after adding the extracted power peaks. The background to this correlation is that extraction in the second loop can be mathematically understood as subtraction—that is, adding the negative value of the interference noise at the extraction location. Therefore, for subsequent signal processing, what is obtained is the sum of the original interference noise distributed across all distance Doppler gates and this negative interference noise at a few locations (i.e., the extraction locations). After the inverse transform, this negative interference noise from the few locations results in a widely distributed noise that cannot be eliminated by interference suppression; in contrast, the original interference noise component is largely suppressed because, after the inverse transform, it only appears in isolated regions (i.e., locations where interference exists in the sampling area). Therefore, after another forward transform, this negative interference noise is retained at the extraction locations, and the positive interference noise contained therein is compensated after summing the extracted power peaks.

[0061] The following text explains the basis. Figure 4 A variation of the previously considered method. Regarding the phase relationship across the receiving antenna, the interference behaves like a normal reflection (i.e., like a received signal formed by reflecting a signal from its own transmitted source onto the object); according to... Figure 2 When the receiving antennas are equidistantly arranged, the phase changes linearly with the angle. If digital beamforming (hereinafter referred to as RX beamforming) is performed on the signal of the receiving antennas, both normal reflections and interference will be focused onto the corresponding receiving direction—this has a positive impact on the extraction of strong useful signals and interference suppression, especially because, through RX beamforming, useful signals and interference can be better distinguished from other signal components, thus making them easier to identify; therefore, a threshold for extraction or interference suppression is determined separately for each digital beam direction (while in the above method, all N...RX (All receiving antennas use the same threshold). Because... Figure 2 N in RX The equidistant arrangement of the receiving antennas allows for digital RX beamforming via FFT; its length is at least N. RX —The length can be increased accordingly by adding zero values ​​(so-called zero-padding). The FFT used for RX beamforming must be inversely operated on again at the end of the interference processing, that is, in the FFT-based, for all N... TX ·N RX Digital beamforming of each antenna channel is performed beforehand. It should be considered that compression efficiency is lower in the RX beamforming region (due to the different levels in the RX beam direction); to avoid this, such as... Figure 8 As shown, in each processing loop with interference suppression and / or high-power peak extraction / summing (i.e., loops 2, 3, and 4), RX beamforming FFT can be performed at the beginning. Figure 8 The FFT RX module in the FFT module performs inverse RX beamforming FFT (IFFT RX) at the end. It should also be mentioned that, in principle, RX beamforming can also be used only for extracting strong useful signals or only for interference suppression (i.e., it does not have to be used for both at the same time).

[0062] It should also be mentioned that, for the sake of simplicity, Figure 4 and Figure 8 Processing steps that are not important to the overall concept are not shown—for example, compensation for pseudo-random coding parameters in modulation (which can be used in conjunction with a Doppler window), multiplication operations for antenna channel calibration, and demodulation for components from different transmitted signals.

[0063] The following will further explain some modifications and alternatives to the previously described methods for improving interference suppression: - Extraction of strong useful signals can also be performed for all N TX ·N RX antenna channels (i.e., from N) RX One receiving antenna and N TX The process is performed after additional digital beamforming is applied to all combinations of signals from each transmitting antenna.

[0064] As described above, at the end of the first processing loop, the distance FFT is stored for all distance gates that do not carry redundant information (i.e., only half the spectrum is stored since the input signal is real). In principle, a smaller portion of the distance FFT can also be stored and used later (e.g., 0.45I values ​​instead of I / 2 values). This is particularly meaningful when the receive low-pass filter has already achieved significant suppression before reaching half the sampling frequency due to the ever-present filter transition band, thus allowing, for example, the use of only 0.4I distance gates for peak detection. The slightly higher number of values ​​(0.45I) stored after the first processing loop and used at least before the first interference suppression is because using a reduced number of values ​​to perform the inverse distance FFT in the third processing loop results in poorer interference suppression at the highest distance gates; that is, significant interference noise may still remain at these locations.

[0065] So far, we have considered real-valued mixers. If a complex-valued mixer (also known as an IQ mixer) is used, the above processing remains essentially unchanged. However, the distance FFT must be stored in large or all of it, i.e., not just the lower half, because the spectrum carries information across its entire range. Therefore, the inverse distance FFT and all other steps in the second processing loop must be computed on I values ​​(instead of just I / 2 values). It should also be noted that an inherent advantage of complex-valued receivers is that interference typically only lasts for half the duration, as it only passes through the receiver filter when the frequency difference between the local and interfering systems has a value of one sign (i.e., it is only effective in the region before or after the frequency ramps of the local and interfering systems intersect, not both regions simultaneously).

[0066] - In the above description, the transmitting and receiving antennas are used in parallel. The described method can also be used in cases of partial or complete serial operation.

[0067] For a sinusoidal useful signal, not all power is located within the power peak in the frequency domain; due to the windowing effect, a portion is located in the sidelobes. The stronger the window, the smaller this power component becomes (however, this widens the power peak, which is detrimental to object separation and reduces the integration gain). To extract this power component from the secondary sidelobes in the second processing loop (so that it does not negatively impact subsequent interference suppression), all spectral values ​​below the maximum spectral value by a certain amount (e.g., 35 dB) are zeroed; however, unlike the extracted high-power peaks, these values ​​are not stored (to save storage space) and therefore are not added back in the fourth processing loop.

[0068] - Due to the aforementioned situation (i.e., a portion of the power lies in the sidelobes within the spectral range), it may be advantageous to use a Doppler window with small sidelobes (i.e., a strong window) in the second processing loop to extract the strong useful signal. This window will then be used in the Doppler FFT in the fourth processing loop (otherwise, inconsistencies will occur during subsequent extractions). If a weaker Doppler window is required in subsequent processing, an inverse Doppler FFT should be performed after summing the extracted power peaks in the fourth processing loop. The result should then be multiplied by the quotient of the weaker Doppler window and the previously used Doppler window, and finally, the Doppler FFT should be performed again.

[0069] - Based on the previously considered basis Figure 1 In the modulation, all frequency ramps have the same frequency position, i.e., the starting frequency (and consequently the center frequency). To improve distance resolution, as described in DE 10 2020 210 079 B3, the starting frequency can vary linearly along the frequency ramps, wherein the distance between the frequency ramps preferably also varies linearly. This has no effect on the aforementioned interference suppression method and therefore remains unchanged.

[0070] So far, as a sequentially repeating transmitted signal, a linear frequency ramp has been considered. However, other signal forms besides frequency ramps can also be used, such as signals with pseudo-random binary phase modulation (i.e., the symbols within the signal change very rapidly in a pseudo-random manner) or OFDM signals (OFDM = Orthogonal Frequency Division Multiplexing). Then, as input values ​​for the above method, the sampled values ​​cannot be used directly (after analog-to-digital conversion). Instead, the Fourier transform of each transmitted signal and each receiving antenna is first calculated and divided by the spectrum of the transmitted signal.

[0071] The narrower the interference signal appears in the sampled values ​​over time, the better the first interference suppression (in the third processing loop) is. To reduce the width of the interference signal, pulse compression can be performed. If the slope difference between the frequency ramps of this system and the jamming radar system is known, the spectrum of the interference signal (ignoring the linear phase component corresponding to the time of interference occurrence) can preferably be determined using the low-pass filter function implemented in the receiving path; the slope difference between the frequency ramps of this system and the jamming radar system can be estimated, for example, based on the extent of the interference spread in the received signal. At the beginning of the third processing loop, before performing the inverse range FFT, the spectrum is divided by this spectrum that does not contain the linear phase component, resulting in a very sharp and high interference spike after the inverse range FFT, i.e., within the sampled value range—the interference energy is effectively compressed to a single moment. This spectral division must be compensated for by a corresponding multiplication at the end of the third processing loop, i.e., after the range FFT, so that the overall useful signal remains unchanged.

[0072] - In cases of very strong interference, the high interference noise generated may also obscure relatively strong useful signals in a 2D FFT, preventing their extraction. These relatively strong useful signals remain even during interference suppression and are zeroed out during the suppression process. Subsequent multidimensional FFTs may cause these unextracted useful signals to generate sidelobes, which may exceed the detection threshold; i.e., they then generate one or more ghost detections. Therefore, to determine confidence levels, detections can be assigned to different categories: Category 1 detections, which belong to the power peaks extracted in the second processing cycle (i.e., before interference suppression); Category 2 detections, which are only visible after interference suppression (i.e., belong to the extracted power peaks in the second processing cycle), but cannot be artifacts of interference suppression; and Category 3 detections, which become visible only after interference suppression and are likely interference suppression artifacts because their levels are significantly lower than those of other detections that are only visible after interference suppression. The level ratio / level difference used to distinguish between Categories 2 and 3 is based on an estimate of the useful signal energy loss caused by interference suppression.

[0073] - In the method described above, at the beginning of the fourth processing loop (i.e., before the Doppler FFT), a second interference suppression is performed. This second interference suppression sets outliers in the numerical sequence across the K frequency ramps to zero at each range gate. This is particularly effective when the frequency ramp slopes of the radar system and the jamming radar system are similar or even identical. This interference suppression can also be applied at the beginning of the second processing loop because, especially when the frequency ramps are the same, a single very large value may appear in the value sequence before the Doppler FFT. These values ​​would still produce such a high noise level after the Doppler FFT that the extraction of strong useful signals would only be effective within a limited range.

[0074] So far, in the case of interference suppression, values ​​above the suppression threshold have been set to zero. Alternatively, they can simply be attenuated / damped, i.e., multiplied by a coefficient between zero and one; this results in less error in the remaining useful signal. This is particularly meaningful for values ​​whose levels are not significantly above the suppression threshold, since in this case the ratio between interference and the remaining useful signal is lower, and therefore the trade-offs of interference suppression are less favorable. The value can also be adaptive relative to the level; thus, for example, all values ​​are weakened only to the level of the suppression threshold.

[0075] - So far, the distance dimension and the Doppler dimension have always been used in both the forward and inverse transforms; however, it is also possible to use only one of these two dimensions and replace the other with the second dimension, especially the dimension generated by RX beamforming.

[0076] - The steps used in the above method, such as forward transformation, extraction of identifiable useful signals, inverse transformation, and suppression of interference, can also be performed multiple times, especially in an iterative manner.

Claims

1. A method for a radar system that robustly performs environmental detection even in the presence of other potential interferences, the radar system comprising: - A transmitting device having one or more transmitting antennas that operate serially or in parallel for transmitting a transmitted signal, the transmitted signal comprising one or more sequences of K single signals, the single signals preferably being identical or similar; - A receiving device having one or more receiving antennas operating serially or in parallel for receiving transmitted signals reflected from an object, wherein, for each of K individual transmitted signals, at least I digital received values ​​are formed for each receiving antenna in the receiving device, and the received values ​​can include the following components: a) The received signal from the self-emitted signal reflected off the object, hereinafter referred to as the useful signal. b) Received signals from transmitted signals of other radar systems, hereinafter referred to as interference. c) Noise generated within the sensor, hereinafter referred to as inherent noise. - A signal processing device for processing received signals. The method is characterized in that, in a digital signal processing apparatus, in addition to additional processing steps that can further transform received signals from multiple transmitting and / or receiving antennas, if necessary, it also includes... - Using K individual transmitted signals and N RX At least I×K×N obtained from each receiving antenna RX A received value, or a value derived therefrom, is used to perform a multidimensional signal transformation, hereinafter referred to as the first forward transformation, and the multidimensional signal transformation can suitably be implemented in a multi-stage manner by performing one-dimensional transformations in a corresponding sequence. - After the first positive transformation in the multidimensional dimension, at least some of the identifiable useful signal components are extracted. - Subsequently, a complete or partial one-dimensional or multi-dimensional transformation is performed, which is at least partially the opposite of the forward transformation, and is referred to below as the inverse transformation. - After the inverse transformation, at least some of the identifiable interference is suppressed. - Subsequently, a full or partial, one-dimensional or multi-dimensional second positive transformation is performed, wherein preferably, any interference that can be identified during the multi-dimensional positive transformation is also at least partially suppressed. - Use the results of the first and second positive transforms to detect useful signals, and then detect objects in the environment.

2. The method according to claim 1, characterized in that, After the first forward transform, the extraction of identifiable useful signals is performed as follows: in the multidimensional numerical field obtained by the forward transform, values ​​that have or may have significant useful signal components are extracted, i.e., they are cached and / or set to zero for use in the subsequent inverse transform, wherein, if necessary, more values ​​are set to zero than are cached, especially in order to save storage space.

3. The method according to claim 2, characterized in that, Values ​​containing, or at least potentially containing, significant useful signal artifacts caused by the signal window used in the positive transform are simply set to zero but are not stored.

4. The method according to any one of the preceding claims, characterized in that, After a multidimensional positive transformation, the interference is mapped to noise in at least one dimension, and the values ​​distinguished from the noise are identified as useful signals, wherein statistical methods are preferably used.

5. The method according to any one of the preceding claims, characterized in that, The inverse transform is not exactly the inverse of the forward transform, especially because different signal windows are used, and / or different numerical scaling is used, and / or all values ​​that are not part of the forward transform are used in the inverse transform, and / or different number domains are used, especially the real and complex domains.

6. The method according to any one of the preceding claims, characterized in that, The first positive transform and / or the second positive transform are implemented in a multi-level manner through multiple one-dimensional transforms, in which identifiable interference is also suppressed at least partially.

7. The method according to any one of the preceding claims, characterized in that, Interference can be identified by distinguishing the values ​​affected by interference from the useful signal and / or inherent noise, preferably using statistical methods.

8. The method according to any one of the preceding claims, characterized in that, The interference can be suppressed by reducing or setting the amplitude of the affected value to zero.

9. The method according to any one of the preceding claims, characterized in that, For subsequent detection of useful signals, the results of the first forward transform and the second forward transform are combined in the following manner: the value extracted after the first forward transform, which is cached and set to zero for the subsequent inverse transform, is added to the result of the second forward transform. Preferably, when interference suppression changes the levels of other signal components, the corresponding weighting is considered during the addition.

10. The method according to any one of the preceding claims, characterized in that, When detecting useful signals, especially when different categories are distinguished to determine confidence levels, these categories can include: a) The useful signal that has been identified after the first positive transform. b) Useful signals that are only identified after the second positive transform, but are unlikely to be interference suppression artifacts. c) Useful signals that are only identified after the second positive transform and may be interference suppression artifacts are preferably identified by using an estimate of the energy of the useful signals removed due to interference suppression and a comparison with the level of other useful signals contained after the second positive transform.

11. The method according to any one of the preceding claims, characterized in that, - A central buffer is provided, which preferably contains, in compressed form, two-dimensional data of a transformation region formed after performing a one-dimensional transformation on I received values ​​for each of K individual transmitted signals. The computational steps are executed sequentially, i.e., within a loop of one-dimensional data, and require at most one-dimensional, and thus significantly less, additional memory. - The data in the central cache changes continuously throughout the computation process.

12. The method according to any one of the preceding claims, characterized in that, Multiple receiving antennas are provided, and digital beamforming is performed on the signals from the multiple receiving antennas, wherein, - The extraction of the useful signal and / or at least one interference suppression is performed in the region obtained by digital beamforming after the first positive transform. - When necessary, compressed data stored in the central buffer is not digitally beamformed. - If necessary, digital beamforming can be canceled through inverse operations, and the beamforming and inverse operations process can be performed multiple times. - Preferably, digital beamforming and its inverse operation are performed by means of a discrete Fourier transform in the form of a fast Fourier transform, especially when the receiving antennas are arranged in an equidistant grid.

13. The method according to any one of the preceding claims, characterized in that, For data whose content has not changed due to the extraction of useful signals or the suppression of interference, no inverse transform and / or second forward transform are performed.

14. The method according to any one of the preceding claims, characterized in that, The process, which includes forward transformation, extraction of identifiable useful signals, inverse transformation, and suppression of interference, is performed multiple times, especially iteratively.

15. The method according to any one of the preceding claims, characterized in that, The parameters of the transmitted signal—such as time interval, frequency position, and / or phase position—are varied in a random or pseudo-random manner to ensure that the decorrelation of this radar system relative to other radar systems reaches at least the following extent: After the multidimensional positive transformation, the disturbance is mapped to noise in at least one dimension.

16. The method according to any one of the preceding claims, characterized in that, Within each of the K individual transmitted signals, the transmission frequency varies linearly with at least approximately the same slope, wherein the frequency position of the transmitted signal—characterized in particular by its starting frequency—is at least approximately constant or varies linearly. - Where necessary, the spacing between transmitted signals should be at least approximately constant, except for a small portion of random or pseudo-random and / or linear variations. - In the receiving device, the received signal is down-converted to the current transmission frequency, and filtered, in particular, using a low-pass filter. - The forward and inverse transformations are performed using the Discrete Fourier Transform, preferably calculated using the Fast Fourier Transform.

17. The method according to claim 16, characterized in that, The interference is caused by or is assumed to exist by at least one radar system that is also at least locally linearly frequency modulated, particularly for each transmitted signal, pulse compression is performed on the received value in the time or frequency domain, wherein the pulse compression is based on a filtered signal with a linear frequency variation, the slope of which is preferably estimated by the identified interference.

18. The method according to any one of claims 1 to 15, characterized in that, - Within a single transmitted signal, the amplitude, frequency, and / or phase change, for example, in the form of an OFDM signal or a signal with pseudo-random binary phase modulation. - The spacing between transmitted signals is at least approximately constant. - In the receiving device, the received signal is down-converted using a constant frequency. - I received values ​​for each transmitted signal and receiving antenna are generated by Fourier transform and then divided by the spectrum of the transmitted signal. - The forward and inverse transformations are performed using the Discrete Fourier Transform, preferably calculated using the Fast Fourier Transform.

19. A radar system that can robustly perform environmental detection even in the presence of other potentially interfering radar systems, said radar system comprising: - A transmitting device having one or more transmitting antennas that operate serially or in parallel for transmitting a transmitted signal, the transmitted signal comprising one or more sequences of K single signals, the single signals preferably being identical or similar. - A receiving device having one or more receiving antennas that operate serially or in parallel, the receiving antennas being used to receive transmitted signals reflected off an object, wherein, For each of the K individual transmitted signals, at least I digital received values ​​are formed in the receiving device for each receiving antenna, and these received values ​​can contain the following components: a) the received signal from the self-transmitted signal reflected on the object, hereinafter referred to as the useful signal; b) the received signal from the transmitted signals of other radar systems, hereinafter referred to as interference; and c) the noise generated in the sensor itself, hereinafter referred to as inherent noise. - A signal processing device for processing received signals. The radar system is preferably configured to implement the method according to any one of the preceding claims, characterized in that, in the digital signal processing apparatus, in addition to additional processing steps already included that can further transform the received signals from multiple transmitting and / or receiving antennas, it also includes... - Using K individual transmitted signals and N RX At least I×K×N obtained from each receiving antenna RX A received value, or a value derived therefrom, is used to perform a multidimensional signal transformation, hereinafter referred to as the first forward transformation, and the multidimensional signal transformation can suitably be implemented in a multi-stage manner by performing one-dimensional transformations in a corresponding sequence. - After the first positive transformation in the multidimensional dimension, at least some of the identifiable useful signal components are extracted. - Subsequently, a complete or partial one-dimensional or multi-dimensional transformation is performed, which is at least partially the opposite of the forward transformation, and is referred to below as the inverse transformation. - After the inverse transformation, at least some of the identifiable interference is suppressed. - Subsequently, a full or partial, one-dimensional or multi-dimensional second positive transformation is performed, wherein preferably, any interference that can be identified during the multi-dimensional positive transformation is also at least partially suppressed. - Use the results of the first and second positive transforms to detect useful signals, and then detect objects in the environment.

Images