Method for compensation of transmitter-end and receiver-end interference effects
Patent Information
- Application Number
- EP2023833144
- Authority / Receiving Office
- EP · EP
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2023-02-14
- Filing Date
- 2023-12-20
- Publication Date
- 2025-12-24
AI Technical Summary
Radar sensors face interference effects due to non-linearities in components like digital-to-analog converters and amplifiers, limiting signal power and measurement quality, especially with high crest factor signals, requiring more hardware effort for compensation.
The method involves reducing input back-off to 0 dB or less, allowing maximum transmitter efficiency by scaling digital signal amplitudes, and using algorithms for IQ demodulator and amplifier interference correction, with pre-distortion and equalization of radar signals to separate and compensate for interference effects on both transmission and reception sides.
This approach enables independent and adaptive calibration of radar sensors without additional components, improving signal-to-noise ratio, reducing ghost targets, and increasing range without compromising measurement quality, while optimizing amplifier power usage.
Smart Images
Figure EP2023086929_22082024_PF_FP
Abstract
Description
[0001]R. 404620 - 1 - Description Title Method for compensating transmitter-side and receiver-side interference effects The present invention relates to a method for compensating transmitter-side and receiver-side interference effects in a radar sensor. The invention further relates to a computer program, a machine-readable storage medium, an electronic control unit and a radar sensor. State of the art Nowadays, in addition to modulation types such as frequency-modulated continuous wave radar (FMCW) or pulsed radar, digital modulation types such as orthogonal frequency division multiplexing (OFDM), pseudorandom radar (PN), phase-modulated continuous wave radar (PMCW), etc. are also used in radar sensors. These enable significantly greater flexibility in operation. For precise evaluation, it is important that the radar signals are transmitted and received without distortion. Due to non-linearities in the components, such asa digital-to-analog converter, an IQ modulator, a mixer, an amplifier, etc. This is not always the case. This is especially true if the crest factor of the transmitted signals is high. In order to suppress the non-linearities of the analog components, the signal power is typically reduced on the transmit side so that the peak values of the time signal have a defined relationship to the 1 dB compression point of the transmit amplifier. The transmitted power is reduced using an input back-off (IBO) in order to operate the transmitter as linearly as possible. For example, the transmitted power is reduced by 10 dB by the input back-off. Such compensation therefore requires more hardware outlay than, for example, FMCW, in which the full power spectrum of the amplifier can be utilized. Thus, with digitally modulated R. 404620 - 2 - radar sensors, a compromise has been made between high transmitted power or a long range and good signal orMeasurement quality is affected. Furthermore, compensation methods for nonlinearities on both the transmit and receive sides are known from communications technology. The transmitted radar signals are predistorted at the transmitter, and the received radar signals are equalized at the receiver. The article by L. Anttila, M. Valkama, and M. Renfors, "Circularity-Based I / Q Imbalance Compensation in Wideband Direct-Conversion Receivers" in IEEE Transactions on Vehicular Technology, vol. 57, no. 4, pp. 2099-2113, July 2008, doi: 10.1109 / TVT.2007.909269, describes an algorithm for correcting the receive-side interference effects of the I / Q demodulator. Disclosure of the Invention: In the present method, a desired signal power of a radar signal is set. The signal power is preferably adjusted with a fixed input backoff (IBO). Preferably, the input backoff can be reduced to achieve the desired signal power.In the present method, the nonlinearity of the components is no longer compensated via the input backoff. The input backoff can, for example, be reduced to 0 dB or less so that the transmitter can be operated with maximum efficiency. In particular, the input backoff is implemented by a digital-to-analog converter for the radar sensor transmitter by scaling the amplitude of the digital signal values. Alternatively, the input backoff can be implemented by an analog and / or digital circuit in or for the radar sensor transmitter. The signal power can be freely selected and is not limited by the nonlinearity of the components as is conventional. A transmitter of a radar sensor transmits a first radar signal into a radar channel, and a receiver of the radar sensor receives this first radar signal. The first radar signal is used to calibrate an IQ demodulator of the receiver.The first radar signal can be transmitted with the above-mentioned signal power and in particular correspond to the second radar signal described below. In general, the first radar signal can be a test signal which is transmitted with any amplitude and any signal line independently of the other radar signals. The desired signal power can therefore also be set later. The received first radar signal is mixed down to a baseband or to an intermediate frequency. An IQ mixer can preferably be used for this, but other simple mixers can also be used. The received first radar signal can then be pre-processed (calibration pre-processing). The pre-processing consists in particular of time synchronization and optionally frequency synchronization. If required, filtering and / or averaging can also be carried out.When using orthogonal frequency division multiplexing (OFDM), the cyclic prefix can also be removed. This removes all known effects not caused by the I / Q mixer (I / Q demodulator). For the first radar signal, an algorithm is run to correct the receiver's interference effects, and the algorithm's coefficients are estimated. This algorithm can, in particular, correct the interference effects of the I / Q demodulator, which is referred to as I / Q compensation (IQC). For this purpose, an algorithm with blind coefficient estimation is preferably used, such as the one described in the article by L. Anttila, M. Valkama, and M. Renfors, "Circularity-Based I / Q Imbalance Compensation in Wideband Direct-Conversion Receivers," mentioned above. However, alternative algorithms can also be used.Using the calculated coefficients, equalization of the radar signal at the receiving end is possible. Additionally or alternatively, an algorithm can be used to correct the interference effects of a receiver amplifier. The transmitter then transmits a second radar signal with the desired signal power, as described above, into the radar channel, and the receiver receives the second radar signal. The coefficients of the algorithm described above are applied when processing the received radar signal. The received second radar signal is also downconverted to a baseband or an intermediate frequency, and the received second radar signal is optionally preprocessed.Here, too, effects that are not part of analog components are compensated for, such as channel effects (propagation time, frequency shift due to the Doppler effect, and the like). Using the second radar signal and the calculated coefficients for the algorithm for correcting the interference effects of the receiver, an inverse model for the pre-distortion of the transmitted radar signals is learned or trained. The pre-distortion (digital pre-distortion, DPD) of the radar signal corrects the distortion of the radar signal caused by the interference effects in the transmitter. The inverse model can be, for example, a memory polynomial model, a parallel Hammerstein model or other known non-linear models or can be implemented by a neural network. The model represents the inverse function of the transmitter (post-inverse). The inverse model is learned using the transmitted data and the received data corrected by the algorithm.The training of the inverse model for pre-distorting the next transmitted radar signal using the previously transmitted radar signal is also referred to as an “indirect learning architecture.” The above steps can be performed iteratively several times in order to determine the algorithm coefficients and the inverse model as accurately as possible. This is particularly advantageous for arbitrary transmitted data in order to estimate the coefficients as accurately as possible. In addition, the above steps can be repeated under different conditions in order to include, for example, temperature dependence, aging effects, the influence of the environment, and the like. This results in the algorithm coefficients and the inverse model being obtained for the different conditions. In addition, the above steps can be repeated for different operating points, such asDifferent bias voltages, different power levels and frequency ranges of a linear oscillator or amplifier, different frequency ranges of the radar signal, etc. can be performed. In summary, the correction of the receive-side interference effects performed by the algorithm is integrated into the feedback path of the predistortion. The correction of the transmit-side interference effects is taken into account in the inverse model R. 404620 - 5 -. The result is a separation of the receive-side interference effects and the transmit-side interference effects of the radar sensor, which is itself not calibrated. During measurement operation, a transmitted measuring radar signal with the signal power described above is emitted from the transmitter. The transmitted measuring radar signal is predistorted using the learned inverse model, thus compensating for the transmitter's interference effects.The measurement radar signal is received by the receiver and corrected using the receive-side interference correction algorithm, thus compensating for the receiver interference. As a result, the receive-side interference and the transmit-side interference are compensated separately during the measurement. This enables independent and adaptive calibration of the radar sensor without additional components. Distributed signal processing between the transmit and receive paths offers the following advantages: The amplifier power in the transmitter no longer needs to be limited. This allows for better utilization of the available amplifier power in the transmitter. Furthermore, ghost targets, which can occur due to interference, are suppressed. Furthermore, an improvement in the signal-to-noise ratio is achieved during radar measurements with the same transmit power.Alternatively, the transmission power can be increased while maintaining the same level of distortion, which also improves the signal-to-noise ratio. This increases the range of the radar sensor without impairing the measurement quality and / or dynamic range. Furthermore, out-of-band transmissions can be reduced, thus adhering to spectral masks. There is no need for a special radar channel or scene. Furthermore, a frequency shift between the local oscillators in the transmitter and receiver can be omitted. The receiver of the radar sensor can preferably be used as the receiver. In this case, the radar signals can be recorded using the radar channel after a reflection by the receiver, or alternatively, they can be obtained directly from the crosstalk between transmitter and receiver. If the receiver of the radar sensor is used, the R.404620 - 6 - components are carried out entirely within the radar sensor, and no additional external components are required. This makes it possible to perform compensation even during radar operation. The pre-distortion and equalization of the radar signal can take place within a previously planned measurement cycle or in an additional measurement. This allows the calibration to be repeated on-site during the application, and the radar sensor can be adapted to the current conditions. Alternatively, this can also take place during the initial characterization of the radar sensor. Alternatively, an additional measuring receiver can be used to record and measure the radar signals from the radar sensor. The additional measuring receiver can be an external measuring receiver and connected to the radar sensor or an external evaluation device.Such an additional measuring receiver is preferably used for the initial characterization of the radar sensor. Alternatively, the additional measuring receiver can also be an integrated circuit on a chip or on a circuit board. The measuring receiver can, for example, be supplied with an output of the transmitted signal at the transmitter output via a directional coupler, a power detector, or a mixer. Such an integrated measuring receiver can be used both for the initial characterization and during operation or between measurements. In particular, the additional measuring receiver is part of a defined measurement setup, so that consistent results are achieved. This allows the compensation of multiple radar sensors to be carried out in rapid succession using the same measuring receiver. The coefficients for the IQC depend on the radar receiver and cannot be determined directly using an additional receiver.The measurement with an additional measuring receiver is intended to determine the transmit-side interference effects and thus the inverse model. In this mode, the calibration sequence would have to be changed. First, the coefficients of the inverse model for predistortion are to be estimated using the (preprocessed) received signals from the additional receiver. Then, using the inverse model, a predistorted signal can be transmitted to determine the IQC coefficients from the received signals of the radar receiver. If the inverse model is correctly estimated, the latter received signal is free of transmit-side interference effects. The inverse model can preferably be determined using stochastic estimation methods, such as the least squares method. Alternatively, the inverse model can be trained using machine learning with a neural network.Here, the distorted signal is used as input and the ideal input signal as output (label). Such a neural network is described in the yet-to-be-published article by R. Michev, Y. Shu, D. Werbunat, J. Hasch, C. Waldschmidt, "Adaptive Compensation of Hardware Impairments in Digitally Modulated Radars Using ML-Based Behavioral Models." If the scene is known, for example, because the ground truth for a specific scenario is known, a separate channel estimation (particularly through a different calibration procedure) was performed, or the scene was detected by other sensors, such as a camera, lidar, etc., the channel information available in this way can be used to compensate for the interference effects. In particular, the channel information can improve the separation between transmitter-side interference effects and receiver-side interference effects.In particular, the channel information can be used in a channel estimation for the inverse model. Remaining signal components from targets in the channel during calibration could corrupt the calculation of the coefficients depending on the backscatter cross-section and / or signal power during interference. A more accurate channel estimation therefore leads to better compensation. Optionally, the method can be applied to multiple-input-multiple-output (MIMO) radar sensors. For this purpose, the above-mentioned steps are performed for each receiver and each transmitter to obtain the coefficients for IQC and for the inverse model, respectively. The computer program is configured to perform each step of the method, especially when performed on a computing device or control unit. It enables the method to be implemented in a conventional electronic control unit without the need for structural modifications.For this purpose, it is stored on the machine-readable storage medium. R. 404620 - 8 - By loading the computer program onto a conventional electronic control unit, the electronic control unit is obtained, which is configured to perform receiver-side equalization of radar signals. Furthermore, a radar sensor is provided, which has a transmitter and a receiver. Furthermore, the radar sensor has a preprocessing unit and a compensation unit. In particular, the radar sensor can have the aforementioned electronic control unit. In this case, the preprocessing unit and the compensation unit can be part of the electronic control unit. The radar sensor is configured to perform each step of the method in order to compensate for the interference effects occurring during the measurement.Brief description of the drawings Exemplary embodiments of the invention are illustrated in the drawings and explained in more detail in the following description. Figure 1 shows a block diagram of a radar sensor. Figure 2 shows a flow diagram of an embodiment of the method according to the invention for a first part of a calibration cycle. Figure 3 shows a flow diagram of an embodiment of the method according to the invention for a second part of the calibration cycle. Figure 4 shows a flow diagram of an embodiment of the method according to the invention in a measuring cycle. Exemplary embodiments of the invention Figure 1 shows a radar sensor according to the invention which has a transmitter 1 and a receiver 2. A transmitter 3 generates radar signals x(n) which are output by the transmitter 1.These radar signals x(n) pass through a digital-to-analog converter 4 (DAC), an IQ modulator 5, and a power amplifier 6 before being output by a transmitting antenna 7 as a transmitted signal s(n) into a radar channel H. The receiver 2 has a receiving antenna 8, with which the radar signal is received as a received signal r(n) from the radar channel H. In addition, crosstalk CT can occur between the transmitter 1 and the receiver 2. The received radar signal r(n) passes through a low-noise amplifier 9, an IQ demodulator 10, and an analog-to-digital converter 12 (analog-to-digital converter, ADC). The IQ demodulator demodulates the received radar signal r(n) to an output signal y(n). The IQ modulator 5 and the IQ demodulator 10 are connected to each other via a local oscillator 11 in order to carry out the modulation and demodulation of the radar signal in a coordinated manner.The components digital-to-analog converter 4, IQ modulator 5, power amplifier 6, low-noise amplifier 9, IQ demodulator 10 and analog-to-digital converter 12 are typically subject to non-linearities. The output signal y(n) of the IQ demodulator 10 is fed to a signal evaluation unit 13 and evaluated there. Figures 2 and 3 each show a flow diagram of the method according to the invention during a calibration cycle, divided into two parts. In the first part in Figure 2, a digital transmission signal is generated at the beginning 19. Then, in the transmitter 3, a signal power desired for the calibration for a predefined test signal x1(n) is set via an input back-off 20. An inverse model ^ is applied to this test signal x1(n). ^ ^^^of the last calibration cycle and the test signal x1(n) is thereby predistorted 21. In the first calibration cycle (k=1), no model is applied or, alternatively, a predefined model with standard coefficients is applied. This results in a predistorted signal x 1,pd (n). The goal of predistortion is to compensate in advance for the interference effects of the analog components in the transmitter (DAC 4, IQ modulator 5, power amplifier 6). The signal x 1,pd (n) is upconverted by the IQ modulator 5 22 and has the form in equivalent baseband representation: where ^ ^,^^ ( ^ ) and ^ ^,^^ ( ^ ) are the coefficients of the impulse responses of the main and mirror paths of the IQ modulator 5, ^ ^ ∗ , ^^ (^) the complex conjugate of ^ ^,^^ ( ^ )and ∗ stands for convolution. R. 404620 - 10 - After the modulated test signal z1(n) has passed through the power amplifier 6, it is transmitted as the first transmission signal s1(n) from transmitter 1 into radar channel H 23. The influence of the power amplifier 6 is described by a memory polynomial model. Thus, the first transmission signal s1(n) in equivalent baseband representation is as follows: The receiver 2 receives 24 the transmitted first radar signal from the radar channel H. The received first radar signal ^ ^ (^) results in equivalent baseband representation as a convolution of the first transmission signal ^ ^ (^) with the channel impulse response ℎ(^): ^ ^ (^) = ^ ^(^) ∗ ℎ(^) The received first radar signal r1(n) is downconverted by the IQ demodulator 10 into a baseband or to an intermediate frequency 25. This results in an output signal y1(n) of the demodulator 10 for the first radar signal, which can be represented as a sum of convolutions with impulse responses: where ^ ^,^^ (^) and ^ ^,^^ (^) are the coefficients of the impulse responses of the main and mirror paths of the IQ modulator 5, ^ ^ ∗ (^) the complex conjugate of ^ ^ (^) and ∗ represents a convolution. The reflection is corrected using a well-known algorithm for correcting the receive-side interference effects, namely the "Circularity-Based I / Q Imbalance Compensation in Wideband Direct-Conversion Receivers" with blind coefficient estimation shown above. This algorithm is based on a largely linear filtering: where ^ ^,^^^^ ( ^ )the compensated first signal is, ^ ( ^ ) the coefficients of the algorithm are, ^ ^ ∗ (^) the complex conjugate of ^ ^ (^) and ∗ stands for a convolution. In this operation, the influence of the demodulator 10 on R. 404620 - 11 - the output signal ^ ^ ( ^ ) compensated, so this operation is also called IQ compensation or IQC (IQ compensation). Now the coefficients ^ ( ^ ) of the algorithm, ie the filter coefficients, from statistical properties of the output signal ^ ^ ( ^ ) calculated 26. Ideally, the coefficients ^ ( ^ ) only from the coefficients of the (radar) receiver ^ ^,^^ ( ^ ) and ^ ^,^^ ( ^ )In the second part of the calibration cycle in Figure 3, which is carried out after the first part, a digital transmission signal is first generated 29, then a signal power desired for the measurement for a predefined second signal x2(n) is set in the transmitter 3 via the input back-off 30. The second signal x2(n) corresponds to a measurement radar signal x(n) used in the measurement and can, for example, be generated using an orthogonal frequency division multiplexing method (OFDM). However, other methods, such as phase-modulated continuous wave radar (PMCW), pseudo-noise (PN) or others, can also be used. The inverse model ^ ^ ^^^of the last calibration cycle and the second signal x2(n) is thus pre-distorted 31. Again, no model is applied in the first calibration cycle (k=1) or, alternatively, a predefined model with standard coefficients is applied. This creates an ideal second signal x 2,pd (n). The ideal second signal x 2,pd (n) is upconverted by the IQ modulator 5 32 and has the form in equivalent baseband representation: where ^ ^,^^ ( ^ ) and ^ ^,^^ ( ^ ) are the coefficients of the impulse responses of the main and mirror paths of the IQ modulator 5, the complex conjugate of ^ ^,^^ ( ^ )and ∗ stands for convolution. After the modulated second signal z2(n) has passed through the power amplifier 6, it is transmitted as the second transmission signal s2(n) from transmitter 1 into radar channel H 33. The influence of the power amplifier 6 is described by a memory polynomial model. This results in the following for the second transmission signal s2(n) in equivalent baseband representation: R. 404620 - 12 - The receiver 2 receives 34 the transmitted second radar signal from the radar channel H. The received second radar signal ^ ^ (^) results in equivalent baseband representation as a convolution of the second transmission signal ^ ^ (^) with the channel impulse response ℎ(^): ^ ^ (^) = ^ ^ (^) ∗ ℎ(^) The received second radar signal ^ ^(^) is downconverted to a baseband or an intermediate frequency by the IQ demodulator 10 and demodulated 35. In the case of a direct path, e.g., with crosstalk CT or with an integrated receiver 2, a temporal synchronization, e.g., with a correlation and a shift, as well as an amplitude normalization, is performed. This results in an output signal y2(n) from the demodulator 10 for the second radar signal, which can be represented as a sum of convolutions with impulse responses: ^ ^ (^) = ^ ^,^^ (^) ∗ ^ ^ (^) + ^ ^,^^ (^) ∗ ^ ^ ∗ (^) where ^ ^,^^ (^) and ^ ^,^^ (^) are the coefficients of the impulse responses of the main and mirror paths of the IQ modulator 5, ^ ^ ∗ (^) the complex conjugate of ^ ^ (^) and ∗ stands for a convolution. This downconverted output signal ^ ^(^) the coefficients ^(^) of the IQC calculated in the first part are now applied 36 and a compensated second signal ^ ^,^^^^ ( ^ ) which has the following form: For radar channel H, the channel influence or channel impulse response h(n) is estimated 37. In the case of a direct path, e.g., with crosstalk CT or with an integrated receiver 2, a temporal synchronization, e.g., with a correlation and a shift, as well as a normalization of the amplitude is performed. In further embodiments, predefined radar channels H can be used, or the ground truth can be determined using other R. 404620 - 13 - sensors. Subsequently, radar channel H is compensated for the channel influence 38 and thus equalized. The coefficients w(n) of the IQC, the IQC-compensated second radar signal ^ ^,^^^^ ( ^ )and the equalized channel H are now used to create the inverse model ^ ^ ^ to train in the current calibration cycle 39. The inverse model ^ ^ ^ represents the relationship between the compensated second radar signal ^ ^,^^^^ ( ^ ) and the ideal second radar signal ^ ^,^^ ( ^ ) For this purpose, a mathematical model, such as a memory polynomial model, a parallel Hammerstein model or other non-linear models can be used to calculate an estimated second signal ^ ^,^^ ( ^ ) from the compensated second radar signal ^ ^,^^^^ ( ^ ) can be determined as follows: ^ ^,^^ ( ^ ) = ^^ ^ (^ ^,^^^^ ( ^ ) ) The model ^ ^ ^ represents the inverse function of sender 1 (post-inverse). The coefficients of the inverse model ^^ ^ are calculated using the least squares method. Alternatively, a neural network can be used to train the inverse model ^ ^ ^ The compensated second radar signal ^ ^,^^^^ ( ^ ) used as input and the ideal second signal ^ ^,^^ ( ^ ) as the output variable. Such a model is described in the above-mentioned, yet-to-be-published article "Adaptive Compensation of Hardware Impairments in Digitally Modulated Radars Using ML-Based Behavioral Models." A cost function used for training the inverse model ^ ^ ^ is defined as the difference between the model ^ ^ ^ estimated second signal ^ ^,^^ ( ^ ) and the ideal second signal ^ ^,^^ ( ^ )calculated. As described below, the inverse model ^ ^ ^then used for predistortion in a subsequent measurement cycle. R. 404620 - 14 - The calibration cycle can be repeated as often as desired. Different temperature conditions, aging effects, and the like can be taken into account. Figure 4 shows a flow diagram of the method according to the invention during a measurement cycle. At the beginning, a digital transmission signal is generated 39. In the transmitter 3, a signal power desired for the measurement for a predefined measurement radar signal x(n) is set via the input back-off 40. The measurement signal x(n) corresponds to the second signal x2(n) described above and can also be generated, for example, using an orthogonal frequency division multiplexing method (OFDM) or in other ways, such as phase-modulated continuous wave radar (PMCW), pseudo-noise (PN), or others. The inverse model calculated above ^ ^ ^of the current calibration cycle and the measurement signal x(n) is thus pre-distorted 41. The ideal measurement signal x pd (n) is up-converted by the IQ modulator 5 42 and has the form in equivalent baseband representation: where ^ ^,^^ ( ^ ) and ^ ^,^^ ( ^ ) are the coefficients of the impulse responses of the main and mirror paths of the IQ modulator 5, ^ ^ ∗ ^ (^) the complex conjugate of ^ ^^ ( ^ ) and ∗ represents a convolution. After the modulated measurement signal z(n) has passed through the power amplifier 6, it is transmitted as the measurement transmission signal s(n) from transmitter 1 into radar channel H 43. The influence of the power amplifier 6 is described by a memory polynomial model. This results in the following for the measurement transmission signal s(n) in equivalent baseband representation: The receiver 2 receives 44 the transmitted measuring radar signal from the radar channel H. The received measuring radar signal r(^) results in equivalent baseband representation as a convolution of the measuring transmission signal s(^) with the channel impulse response ℎ(^): R. 404620 - 15 - ^ ( ^ ) = ^ ( ^ ) ∗ ℎ(^) The received measurement radar signal r(^) is downconverted to a baseband or an intermediate frequency by the IQ demodulator 10 and demodulated 45. In the case of a direct path, e.g., with crosstalk CT or with an integrated receiver 2, a temporal synchronization, e.g., with a correlation and a shift, as well as an amplitude normalization, is performed. This results in an output signal y(n) from the demodulator 10 for the measurement radar signal, which can be represented as the sum of convolutions with impulse responses: ^ ( ^ ) = ^^,^^ ( ^ ) ∗ ^( ^ ) + ^^,^^ ( ^ ) ∗ ^ ∗ ( ^ ) where ^ ^,^^ ( ^ ) and ^ ^,^^ ( ^ ) are the coefficients of the impulse responses of the main and mirror paths of the IQ modulator 5, ^ ∗ (^) the complex conjugate of ^ ( ^ ) and ∗ stands for a convolution. This downconverted output signal y ( ^ ) (Down-Conversion) the coefficients calculated in the first part of the calibration cycle ^ ( ^ ) the IQC is applied 46 and a compensated measurement radar signal is generated ^ ^^^^ ( ^ ) which has the following form: In the case of an optimal estimation of the coefficients of the algorithm ^ ^^^ ( ^ ) depending only on the receiving-side parameters, an optimally compensated signal is obtained ^^^^^,^^^ ( ^ ) without receiving-side mirroring (without conjugated signal component) ^ ^^^^,^^^ ( ^ ) = ^ ^^,^^ ( ^ ) The compensated measuring radar signal ^ ^^^^ ( ^ ) is then fed to the signal evaluation unit 47, and the radar image is determined. For a radar signal with orthogonal frequency division multiplexing, radar channel H in the relevant frequency range is determined by dividing the received R. 404620 - 16 - code ^ ^^ (^, µ), which is calculated from the compensated measurement radar signal ^ ^^^^ ( ^ ) is received, and the sent ideal code ^ ^^ ( ^, µ ) , which is derived from the ideal measurement signal x pd (n) is obtained. From this, the radar image can be determined using standard radar signal processing.
Claims
R. 404620 - 17 - Claims 1. Method for compensating transmitter-side and receiver-side interference effects in a radar sensor, characterized by the following steps: ^ Transmission (23) of a first radar signal (s1(n)) by a transmitter (1) and reception (24) of the first radar signal (r1(n)) by a receiver (2) of the radar sensor; ^ Estimation (26) of coefficients (w(n)) of an algorithm for correcting the receiver-side interference effects; ^ Transmission (33) of a second radar signal (s2(n)) by the transmitter (1) and reception (34) of the second radar signal (r2(n)) by the receiver (2) of the radar sensor, wherein the calculated coefficients (w(n)) are used for compensation; ^ Training (39) of an inverse model (^ ^ ^ ) for the predistortion of the transmitter (1) with the calculated coefficients (w(n)) for the receiving-side interference effects; ^ Predistortion (41) of a transmitted measuring radar signal (s(n)) using the inverse model (^ ^^ ) and correcting (46) the received measuring radar signal (r(n)) by means of the algorithm for correcting the interference effects at the receiving end.
2. Method according to claim 1, characterized in that the receiving (24) of the first radar signal (r1(n)) is carried out by an additional measuring receiver.
3. Method according to claim 1 or 2, characterized in that the order of the calculations of the inverse model (^ ^ ^ ) and the receiving-side coefficients (w(n)).
4. Method according to claim 1, characterized in that the algorithm for correcting the receiving-side interference effects is an algorithm for correcting interference effects of an IQ modulator (10) of the receiver R. 404620 - 18 - (1) with blind coefficient estimation.
5. Method according to claim 1 or 2, characterized in that the inverse model (^ ^ ^) is learned (39) using stochastic estimation methods.
6. Method according to claim 1 or 2, characterized in that the inverse model (^ ^ ^) is learned (39) by means of machine learning.
7. Method according to one of the preceding claims, characterized in that, in a known scene, available channel information is used to compensate for the interference effects.
8. Method according to one of the preceding claims, characterized in that the steps of transmitting (23) a first radar signal (s1(n)) and receiving (24) the first radar signal (r1(n)), estimating (26) coefficients (w(n)) of the algorithm for correcting the interference effects on the receiving side, transmitting (33) a second radar signal (s2(n)), receiving the second radar signal (r2(n)), and training (39) an inverse model are carried out iteratively several times.
9. Computer program which is configured to carry out each step of the method according to one of claims 1 to 8.
10. Machine-readable storage medium on which a computer program according to claim 9 is stored. 11.An electronic control unit configured to equalize radar signals using a method according to any one of claims 1 to 8.
12. A radar sensor comprising a transmitter (1) and a receiver (2) as well as a signal processing unit (13), wherein the radar sensor is configured to compensate for interference effects using a method according to any one of claims 1 to 8.