Focusing method for radar detection of relative motion

Abstract

The invention relates to a method according to the invention for a radar system for detecting the surroundings of a motor vehicle, comprising:-a transmitting device for directionally radiating a transmission signal; a receiving device for directionally receiving the transmission signal reflected back on the object; and-a signal processing device for processing the received signal, the modulation method of the frequency of the radiated transmission signal being made such that the frequency contains a sequence of K linear ramps having at least substantially the same slope and duration, hereinafter referred to as frequency ramps,-in the signal processing device, the frequency of the radiated transmission signal is modulated by the sequence of K linear ramps having at least substantially the same slope and duration, hereinafter referred to as frequency ramps. A signal having a transient transmission frequency or having a constant offset from the transient transmission frequency is mixed with a transmission signal received by the receiving device and reflected back from the object,-after appropriate pre-processing is performed if necessary in the signal processing device, the signal is processed by the signal processing device. The mixed output signal is sampled I times during each of the K frequency ramps,-a two-dimensional discrete time-frequency transform is completely or partially determined for these I * K sampled values after pre-processing in the signal processing device, characterized in that, preferably depending on the vehicle motion, the two-dimensional discrete time-frequency transform is determined for the I * K sampled values. The pre-processing of the I * K sampling values comprises: frequency shifting the signal composed of the I sampling values of the respective frequency ramps such that the frequency of the signal composed of the respective I sampling values remains unchanged over the K frequency ramps for the object having the defined radial relative movement, in particular, the blur / trailing of power spikes generated by such objects in a two-dimensional time-frequency transform is thereby counteracted.

Description

Technical Field

[0001] The invention relates to a radar method and a radar system for a driver assistance system in a motor vehicle. The radar system comprises a focusing method according to the invention for radar detection of relative motion. Background Art

[0002] More and more motor vehicles are equipped with driver assistance systems, which detect the surroundings or the respective traffic situation with the aid of sensor systems and initiate an automatic reaction of the vehicle and / or provide instructions to the driver, in particular warnings, based on the traffic situation thus detected. In this context, the system functions are divided into comfort functions and safety functions.

[0003] Currently under development, FSRA (Full Speed ​​Range Adaptive Cruise Control) plays an important role as a comfort function. As long as the traffic situation allows, the vehicle will adjust its speed to the desired speed set by the driver in advance, otherwise its speed will automatically adjust to the actual traffic situation.

[0004] Today, safety functions exist in various forms. One group consists of functions that shorten the braking or stopping path in an emergency situation up to autonomous emergency braking. Another group is lane change functions: if the driver wants to make a dangerous lane change, i.e. if a vehicle in the adjacent lane is in the blind spot (known as BSD – “Blind Spot Detection”) or is approaching quickly from behind (LCA – “Lane Change Assist”), the lane change function warns the driver or intervenes with the steering system.

[0005] Currently, not only is the driver assisted, but his tasks are increasingly performed autonomously by the vehicle, i.e. the driver's actions are increasingly replaced; this is known as autonomous driving.

[0006] Radar sensors are often combined with sensors from other technologies, such as camera sensors, for use in systems of the type described above. The advantages of radar sensors include reliable operation even in adverse weather conditions and the ability to measure not only the distance to an object but also its radial relative velocity directly via the Doppler effect. Frequencies of 24 GHz, 77 GHz and 79 GHz are often used as transmission frequencies.

[0007] In addition to a large sensor detection range, the above functions also require high measurement accuracy, resolution and separation performance for distance and relative speed. High resolution and high separation performance for distance and relative speed are important because they can at least partially compensate for the poor angular resolution and angular separation performance of motor vehicle radar sensors (due to their small size). The problem in current radar systems is that high resolution for distance and relative speed can only be fully achieved at the same time if the relative movement between the object to be measured and the radar system is relatively small. In the case of high relative speeds, "blurring" or "smearing" phenomena occur in the radar detection. The terms "blurring" or "smearing" here refer in particular to the fact that, for example, point-like objects in the radar image are extended to multiple detection units and thus look like extended objects. Summary of the invention

[0008] The object of the present invention is to achieve high distance resolution and high relative speed resolution for relatively moving objects using a motor vehicle radar sensor. Here, stationary objects are particularly important for sensors that observe the direction of travel.

[0009] This object is achieved in principle by means of a radar method and a radar system according to claims 1 to 16. Here, according to the invention, it is shown how to focus the radar detection of an object with a defined relative motion.

[0010] The advantages of the present invention derive from improvements in radar detection.

[0011] In the method of the radar system for detecting the surroundings of a motor vehicle according to the present invention, the radar system comprises a transmitting device for radiating a transmission signal, a receiving device for receiving the transmission signal reflected by an object, and a signal processing device for processing the received signal.

[0012] The frequency of the radiated transmission signal is modulated in such a way that the frequency comprises a series of K linear ramps, hereinafter referred to as frequency ramps, of at least approximately the same slope and duration. In the signal processing device, the transmission signal received by the receiving device and reflected on the object is mixed with a substantially transient transmission frequency or a signal having a constant offset therefrom. In addition, if necessary, the signal processing device samples the mixed output signal I times during each frequency ramp of the K frequency ramps after appropriate preprocessing (such as amplification, bandpass filtering, etc.), and after the preprocessing by the signal processing device, the two-dimensional discrete time-frequency transformation is completely or only partially determined for these I·K sampling values, wherein, preferably according to the vehicle movement, the preprocessing of the I·K sampling values ​​includes: frequency shifting the signal composed of the I sampling values ​​of each corresponding frequency ramp so that the frequency of the signal formed by each corresponding I sampling value remains unchanged on the K frequency ramps for objects with a defined radial relative motion, thereby offsetting the blurring or tailing of the signal peaks generated by these objects in the two-dimensional time-frequency transformation.

[0013] Depending on the intended use, the method can be used for a radar system whose detection range includes the direction of travel, wherein the defined radial relative motion is the negative of the vehicle's own motion, so that for stationary objects in the direction of travel, the frequency of the signal consisting of each corresponding I sampling values ​​remains constant on K frequency ramps.

[0014] Furthermore, in the method, the frequency shift may be achieved by multiplying with a rotated complex unit vector.

[0015] The frequency shift is preferably achieved by multiplication with a rotated complex unit vector, wherein the respective I sample values ​​of the frequency ramp are equidistant in time and the rotation speed of the complex unit vector is respectively constant during a frequency ramp but varies over the individual frequency ramps.

[0016] Here, according to a preferred embodiment, the change of the rotational speed of the complex element vector on each frequency ramp may be proportional to the integral of the defined radial relative motion speed, ie, in particular, proportional to the integral of the vehicle's own speed.

[0017] The center frequency and the time interval of the frequency ramp are preferably at least approximately constant, and a linear variation of the complex unit vector rotation speed is used on the individual frequency ramps, which corresponds to the possible simplifying assumption that the speed of the radial relative motion defined during the acquisition of I·K sampling values ​​is constant, i.e., in particular the vehicle's own speed is constant.

[0018] As an alternative to this, the center frequency and the time interval of the frequency ramp can also vary at least approximately linearly, wherein the relative change of the time interval is at least approximately twice the relative change of the center frequency in terms of amplitude and the signs of these changes are opposite, using a linear change of the complex unit vector rotation speed on the individual frequency ramps, which corresponds to the necessary simplifying assumption that the radial relative motion speed defined during the acquisition of I·K sampling values ​​is constant, that is, in particular the vehicle's own speed is constant.

[0019] The phase of the complex unit vector is preferably point-symmetric at I sampling values ​​and K frequency ramps, ie the center point is zero in each corresponding case, so that the position of the power peak generated by the object in the two-dimensional discrete time-frequency transform does not change.

[0020] Depending on the purpose of use, the first stage of the two-dimensional discrete time-frequency transform can be performed through the corresponding I sampling values ​​of each frequency ramp. It is preferred to use a fast Fourier transform to efficiently implement the discrete Fourier transform, and the frequency shift can be implemented in combination with the window function used for the transformation.

[0021] According to a special design of the present invention, the window function can be iterated from one frequency ramp to another frequency ramp by multiplying each of them with the same constant rotation complex unit vector.

[0022] In addition, the frequency shifts corresponding to different radial relative motions may be calculated using at least part of the same sampling values ​​of a plurality of two-dimensional discrete time-frequency transforms.

[0023] Depending on the purpose of use, preferably based on the movement of the vehicle, the preprocessing of the I·K sampling values ​​may include: phase shifting the signal composed of the I sampling values ​​of each corresponding frequency ramp, so that the phase of the signal composed of the corresponding I sampling values ​​has a purely linear change on the K frequency ramps for an object with a defined radial relative motion, thereby preventing the power spikes generated by such objects in the two-dimensional time-frequency transform from being smeared or blurred in the dimension generated by the K frequency ramps.

[0024] The method is preferably used in conjunction with a radar system whose detection range includes the direction of travel, wherein the defined radial relative movement is the negative of the vehicle's own movement, so that for stationary objects in the direction of travel, the signal phase formed by each corresponding I sampling values ​​has a purely linear variation over K frequency ramps.

[0025] Additionally, phase shifting can be achieved by multiplying with a rotated complex unit vector.

[0026] Depending on the intended use, the phase shift can also be achieved in combination with a window function used in the dimensional transformation generated by the K frequency ramps.

[0027] Another subject of the invention is also a radar system for detecting the surroundings of a motor vehicle, which radar system is particularly focused on the radar detection of relative motion according to the method of the invention. The radar system has a transmitting device for directional radiating a transmission signal, a receiving device for directional receiving of the transmission signal reflected on an object, and a signal processing device for processing the received signal, wherein the frequency of the radiated transmission signal is modulated to include a series of K linear ramps (frequency ramps) with at least approximately the same slope and duration, in the signal processing device, a signal having approximately the instantaneous transmission frequency or having a constant offset from the instantaneous transmission frequency is mixed with the transmission signal received by the receiving device and reflected on the object, in the signal processing device, the mixed output signal is sampled I times during each of the K frequency ramps after appropriate preprocessing, and in the signal processing device, after preprocessing, a two-dimensional discrete time-frequency transformation is determined in whole or in part from these I·K sampling values. In addition, the radar system is characterized in that, preferably depending on the vehicle movement, the preprocessing of I·K sampling values ​​includes: frequency shifting the signal composed of I sampling values ​​of each corresponding frequency ramp, so that the signal frequency composed of each corresponding I sampling value remains unchanged on the K frequency ramps for an object with a defined radial relative motion, thereby offsetting the tailing / blurring of the power peaks generated by such objects in the two-dimensional time-frequency transform, that is, a broadening phenomenon. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 An example implementation of a radar system is described.

[0029] Figure 2 The transmitted signal frequency is shown to illustrate a so-called frequency ramp, the frequency position of which is constant.

[0030] Figure 3 The absolute value spectra of four objects after two-dimensional discrete Fourier transformation without using the method according to the present invention are shown, wherein the objects have no relative acceleration with respect to the radar system.

[0031] Figure 4 The absolute value spectra of four objects after two-dimensional discrete Fourier transformation using the method according to the present invention are shown, wherein the objects have no relative acceleration with respect to the radar system.

[0032] Figure 5 The absolute value spectra of four objects having relative accelerations with respect to the radar system after two-dimensional discrete Fourier transformation are shown, wherein only the first step of the method according to the present invention is applied, namely only frequency shifting the received signal.

[0033] Figure 6The absolute value spectra of four objects with relative acceleration relative to the radar system after two-dimensional discrete Fourier transformation are shown, wherein two steps of the method according to the present invention are applied, namely, frequency shift and phase shift of the received signal.

[0034] Figure 7 Explain the frequency of the transmitted signal with a linear change in frequency position. DETAILED DESCRIPTION

[0035] We considered roughly Figure 1 An example of a radar system design is shown. The radar system has a transmitting antenna TX0 for radiating a transmit signal and M=4 receiving antennas RX0 to RX3 for receiving the transmit signal reflected on an object; the antenna is located on a planar circuit board 1.1 using planar technology and is arranged in the form of a patch antenna, wherein the circuit board is oriented in the horizontal and vertical directions in the vehicle as shown in the figure and faces the direction of travel. All antennas (transmitting antennas and receiving antennas) should have the same beam characteristics in elevation and azimuth. The four receiving antennas (and the phase centers of the receiving antennas, i.e., the radiation centers) should have the same lateral spacing from each other, i.e., the horizontal spacing d=λ / 2=1.96mm, wherein λ=c / 76.5GHz=3.92mm is the average wavelength of the transmit signal in the 76GHz to 77GHz frequency band used, and c=3*10 8 Meters per second is the speed of light.

[0036] The transmission signal radiated from the transmitting antenna is generated by a high-frequency oscillator 1.2 in the range of 76 GHz to 77 GHz, the frequency of which can be adjusted by controlling the voltage v 控制装置 The control voltage is generated in the control device 1.7, wherein these control devices include, for example, a phase-locked loop or a digital-to-analog converter, which controls the oscillator in such a way that the frequency response corresponds to the desired frequency modulation.

[0037] The signals received by the four receiving antennas are mixed in parallel with the signals of the oscillator 1.2 into the low frequency range in the real-valued mixer 1.3. The received signals then pass through a bandpass filter 1.4 with the transfer function shown, an amplifier 1.5 and a digital-to-analog converter 1.6. They are then further processed in the digital signal processing unit 1.8.

[0038] like Figure 2 As shown, the frequency f of the high-frequency oscillator can measure the distance of the object. TX The frequency of the transmitted signal will change rapidly and linearly (around B ch =600MHz at T ch =51.2 microseconds, where the center frequency is f c=76.5GHz); here, this is called a frequency ramp (also commonly referred to as a "linear frequency modulation signal"). The frequency ramp is on a fixed grid T D = 70 microseconds; there are a total of K = 512 frequency ramps, which all have the same frequency curve, that is, the same frequency gradient, the same frequency position (especially the same starting frequency and center frequency) and the same duration. In recent years, this modulation method has become increasingly popular when using radar to detect the surrounding environment of motor vehicles. This modulation method can achieve a higher sensor detection range and speed resolution (due to the longer data acquisition time) and a higher distance resolution (due to the use of a higher modulation bandwidth).

[0039] During each frequency ramp k=0, ..., K-1, the received signal of each of the M=4 digital-to-analog converters m=0, ..., M-1 is respectively T s = 25 nanoseconds (ie at a frequency of 40 MHz), each correspondingly being scanned I = 2048 times, wherein the sampling always starts at the same time relative to the start of the ramp (see Figure 2 ); the resulting digital sample values ​​with index i=0,…,I-1 are calibrated as s(i,k,m). It is only meaningful to sample the signal within the time range when the received signal of the object reaches the distance range of interest, that is, after the ramp starts, it is necessary to wait at least until the propagation time corresponding to the maximum distance of interest (when the maximum distance of interest is 200 meters, this time is equivalent to 1.33 microseconds) ends; it should be noted that the distance here and below always refers to the radial distance, and the relative speed always refers to its radial component.

[0040] It is known from the prior art that it is also easy to deduce that, in the case of a single point object at a distance r(k), the sampled signal s(i,k,m) describes a sinusoidal oscillation at index i, as described below with very good approximation:

[0041]

[0042] That is, the oscillation frequency is proportional to the object distance r(k), which varies slightly over the K=512 frequency ramps k=0,...,K-1 as the object moves radially relative to the sensor. The relative motion also affects the phase of the sinusoidal oscillation. For a motion with a constant radial velocity component v, the result is:

[0043] That is, the phase changes linearly on the frequency ramp k, wherein the phase change speed is proportional to the radial relative velocity v of the object. Due to the linearity of the receiver, in the case of multiple and / or extended objects, the sampled signal s(i, k, m) results in a linear superposition of sinusoidal functions of the above shape.

[0044] If the variation of the object distance r(k) is neglected in relation (1), i.e., the frequency of all frequency ramps is assumed to be constant, then the optimal filtering of the signal shape (1) for each reception channel m corresponds to a two-dimensional discrete Fourier transform (DFT=Discrete Fourier Transform), which can be implemented very efficiently in two stages by means of two one-dimensional fast Fourier transforms (FFT=Fast Fourier Transform). Including appropriate signal windows (in two dimensions), this has become Figure 2 Standard evaluation method for the modulation forms shown.

[0045] According to this two-dimensional discrete Fourier transform (DFT), power peaks appear in the resulting spectrum S(j,l,m), whose respective positions correspond to the average distance r and relative velocity v of the objects involved, see Figure 3 , where the absolute value spectrum / amplitude spectrum |S(j,l,m) / A(m)| of four point-like objects, independent of the receiving channel m, is shown in dB, with the same radar cross section and an azimuth of about 0°, and the following distances and relative velocities: [r1=50 m, v1=-50 m / s], [r2=100 m, v2=0 m / s], [r3=149.25 m, v3=-50 m / s] and [r4=150 m, v4=-50 m / s]; the receiver noise is also superimposed on the object signal, which is significantly lower than the power peak of the object marked with the object number in the spectrum. Negative relative velocity indicates that the object is approaching the vehicle; the vehicle's own velocity is assumed to be 50 m / s, so objects 1, 3 and 4 are stationary. Moving object 2 has no relative velocity and travels at an absolute velocity of 50 m / s. The dimension j=0,...,J-1 formed by the dimension i (sampling value index) is represented by the range gate, and the dimension l=-L / 2,...,L / 2-1 formed by the dimension k (frequency ramp) is represented by the Doppler gate. This is because the power peak position in the j dimension is mainly determined by the object distance, while the power peak position in the l dimension is mainly determined by the radial relative velocity (mapped by the Doppler effect), so it can be ignored here. The power peak position also shows very little dependence on the other of the two physical quantities, distance and relative velocity. It should be noted that the speed cannot be explicitly calculated by the Doppler gate of the power peak. This is because in this design, only a clear range of 28 m / s can be achieved through K=L=512 Doppler gates. For example, the distance T of the frequency ramp can be changed in different radar cycles. D Implements fuzzy computing (see also below). Figure 3As described above, the number of range gates is only J=801, which is significantly less than the number of sample values ​​I=2048; the background is that, on the one hand, the sample values ​​are real values, so their spectrum is symmetrical, that is, the upper half of their discrete Fourier transform (DFT) does not contain additional information, and on the other hand, according to Figure 1 As shown, the frequency bandwidth of the upper transition range of the analog bandpass filter 1.4 is 8.75 MHz (corresponding to the 448 frequency support points of the discrete Fourier transform (DFT)). ch =600MHz, the distance gate width is 150MHz / B ch 1 meter is equal to 25 centimeters, so the maximum range allowed by J=801 range gates is 200 meters.

[0046] from Figure 3 As can be seen in the figure, the two stationary objects 3 and 4 at [r3 = 149.25 m, v3 = -50 m / s] and [r4 = 150 m, v4 = -50 m / s] cannot be separated but merged into one power spike, even though their distance is three times the range gate width and is expected to be separable (typically a difference of about two range gates is required to separate two point targets of equal intensity). The reason can be seen by comparing stationary object 1 ([r1 = 50 m, v1 = -50 m / s]) and moving object 2 ([r2 = 100 m, v2 = 0 m / s]): the power spike of the moving object with zero relative velocity has the expected sharp shape, while the power spike of the stationary object with a higher relative velocity is greatly broadened - this broadening of the power spike causes the power spike to merge in stationary objects 3 and 4. During the entire data acquisition time of 512·70 μs = 35.84 ms, the object moved about 1.79 m at a relative speed of -50 m / s, i.e., it passed through more than 7 range gates, which can explain this broadening. The broadening occurs not only in the range dimension, but also in the Doppler dimension, because the object is not in the range gate all the time, but only in the range gate for a shorter time, which can be imagined as a Doppler window with reduced width - the spectrum of such a window is broadened, and therefore the shape of the power peak corresponding to the spectrum is also broadened. In addition to reducing the object separation performance, the broadening of the power peak has two disadvantages: on the one hand, it reduces the possible detection range (because the energy dissipation leads to a reduction in amplitude); on the other hand, the measurement accuracy of the distance and relative speed is reduced (because the noise superimposed on the blurred power peak will lead to greater errors). Figure 3 The example in shows that achieving high resolution (and thus high separation performance) simultaneously in terms of distance and relative velocity is limited when there is relative motion of the objects to be detected.

[0047] The two-dimensional discrete Fourier transform (DFT) used ignores the fact that the frequency of the received signal varies slightly with the object distance r(k) on the frequency ramp in relation (1); thus, the higher the relative speed of the object, the greater the deviation from the optimal filtering. If the relative speed v is assumed to remain constant (this is usually a simplification allowed when the data acquisition time is short), the variation of the object distance between the frequency ramps is v·T D (T D = 70 microseconds is the spacing of the frequency ramp), therefore, according to the relationship (1), the change of the time discrete frequency (related to the sampling value index i, that is, independent of the continuous time t) is

[0048] Δf=1 / I·(v·T D ) / (m)·B ch / 150MHz. (3)

[0049] This change Δf of the time-discrete frequency of the received signal from one frequency ramp to another can be compensated by a corresponding inverse frequency shift, i.e. the received signal frequency of the k-th frequency ramp k=0, ..., K-1 is shifted by -k·Δf relative to the received signal frequency of the first frequency ramp k=0. According to the frequency shift theorem of the Fourier transform, this is achieved in the time domain (i.e. by multiplying the I sample values ​​indexed i=0, ..., I-1 in each corresponding frequency ramp k) with a rotated complex unit vector:

[0050] (where "exp" represents an exponential function, and the symbol represents the imaginary unit - not to be confused with the running variable i of the sampled values); the rotation speed of the complex unit vector is constant within a frequency ramp and increases linearly on the frequency ramp. According to relation (4), after multiplying the sampled values ​​s(i,k,m) by the complex unit vector p1(i,k), and then according to relation (3), using the frequency change Δf, the received frequency of an object with radial relative velocity v is the same on all frequency ramps, so the power peak of the object in the two-dimensional spectrum is clear / sharp, that is, it is focused.

[0051] The definition of the ramp index k is asymmetrical (the ramp index starts at zero, so the ramp index of the "intermediate" frequency ramp is not equal to zero); therefore, when frequency shifting with the rotated complex unit vector p1(i,k) according to relation (4), the reception frequency of each frequency ramp is shifted to the reception frequency of the first frequency ramp, so the distance measurement result will be based on the distance of the first frequency ramp, i.e. the distance at the beginning of the data acquisition. Typically, it is necessary to determine the distance at the middle of the data acquisition, which can be achieved by replacing k with the (point) symmetric k-(K-1) / 2 (the "intermediate" frequency ramp (K-1) / 2) is zero). The definition of the index i of the sample value is also asymmetrical; therefore, according to relation (4), the complex unit vector p1(i,k) has a linearly varying phase at the "average" index (I-1) / 2 on the frequency ramp k, which can lead to a slight Doppler shift, i.e. the relative velocity measurement value will be slightly biased; to avoid this, i can be replaced by the (point) symmetric i-(I-1) / 2. This gives the frequency-shifted complex unit vector p1(i,k), that is,

[0052]

[0053] The frequency change Δf is calculated according to equation (3).

[0054] A relative speed v must be defined for this Δf, that is, the broadening of the power peak in the two-dimensional discrete Fourier transform (DFT) can only be completely avoided at the relative speed v of the object; for objects with different relative speeds, the greater the difference between the relative speed of the object and the relative speed defined when designing Δf, the more obvious the broadening of the peak.

[0055] The accurate detection of stationary objects is crucial for sensors that observe the direction of travel, especially in the context of autonomous braking functions. False braking must not be permitted, for example if two stationary objects (stationary vehicles, crash barriers, buildings, etc.) merge on the left and right of the vehicle's own lane and are thus mistakenly identified as being in the vehicle's own lane, leading to false braking. In this context, it is important to understand that the purely angular separation performance of radar systems is very poor; in a radar system with only one transmitting antenna and four receiving antennas, as shown here, angular separation is only possible if the angular difference between the two objects is very large. Even today's radar systems, which have significantly more transmitting and receiving antennas, can only separate the two objects in most cases if the angular difference between the two objects is very large and the backscatter cross sections are very similar. Therefore, a good separation performance must be achieved by means of range measurement and / or Doppler (Doppler separation performance also helps for stationary targets, since the radial relative velocity depends on the cosine of the azimuth angle); if there is only one target in a range-Doppler cell, the angle determination is usually accurate enough. In addition to the above-mentioned erroneous braking, it is of course dangerous to ignore real stationary obstacles such as stationary vehicles next to a crash barrier or under a bridge; in this case, good separation performance by ranging and / or Doppler is also important in order to determine the azimuth angle and thus the position relative to the own lane with sufficient accuracy. The above-mentioned smearing / blurring effects of the power peaks of stationary objects are therefore very disadvantageous in terms of the ranging and Doppler dimensions. It is therefore very advantageous for the sensor considering here that observes in the direction of travel to select the relative speed v for setting Δf and thus the rotation speed of the rotated complex unit vector p1(i,k) as the relative speed of the stationary object at an azimuth angle of 0°, that is, as the vehicle's own speed v ego According to equation (3), we can get:

[0056] Δf=1 / I·(-v ego ·T D ) / (m)·B ch / 150MHz. (6)

[0057] If the frequency shift defined in this way is applied to the example including the above four objects, that is, when the frequency change Δf is calculated according to the relation (6), the sample value s(i,k,m) is multiplied by the complex unit vector p1(i,k) according to the relation (5), and then the two-dimensional discrete Fourier transform (DFT) is performed, the following is obtained: Figure 4The absolute value spectrum |S(j,l,m) / A(m)| in dB is shown. Clear power peaks appear for all three stationary objects 1, 3 and 4, so that the two objects 3 and 4 can also be clearly distinguished at a distance of only 0.75 meters. The power peak amplitude of these objects is also increased by nearly 6 dB (because the power is now focused on a clear peak).

[0058] The power spike for moving object 2 is now broader and has also been reduced in amplitude by nearly 6 dB (which results in a reduced detection range). Figure 3 The initial spectrum shown has no frequency shift, i.e., it is effectively focused at the position where the relative velocity is zero, and the relative velocity of object 2 is exactly zero; Figure 4 Focusing on relative velocity -v in the spectrum ego , the relative velocity of object 2 now deviates from v ego , which leads to the broadening of the spectrum of moving objects and Figure 3 The spectral broadening for stationary objects is the same in the original spectrum. However, this is not critical for a forward-oriented sensor from a functional point of view, since on the one hand, the best separation of ranging and Doppler is not so important for moving objects (seldom do multiple moving objects with very similar distances and relative velocities occur), and on the other hand, a large detection range is not required for objects traveling at the same speed (which is the case with object 2). However, if it is still desired to focus on the power peak of moving object 2, a second 2D discrete Fourier transform (DFT) must be calculated - in this case, the multiplication for the frequency shift can be omitted (since the frequency shift is zero at zero relative velocity). In general, if multiple relative velocities are to be focused, the multiplication for the frequency shift and the 2D discrete Fourier transform (DFT) must be calculated several times; then, for each object, the spectrum whose relative velocity is closest to the one being focused on is used.

[0059] The situation considered so far is that the relative speed v used to focus the power peak is constant during the data acquisition or can be assumed to be constant. In particular, at high relative accelerations and long data acquisition times, this assumption can lead to blurring of the power peak. Therefore, the time-discrete frequency change Δf of the received signal is no longer constant from frequency ramp to frequency ramp (as shown in relation (3)), but varies over the frequency ramp k=0,...,K-1 with the relative speed v(k):

[0060] Δf(k)=1 / I·(v(k)·T D ) / (m)·B ch / 150MHz. (7)

[0061] The rotation of the complex unit vector p1(i,k) for frequency shifting requires summing Δf(k); to do this, the distance change v(k)·T in the above equation must be calculated. D Sum. Discrete time value v(k)·T D (T D The sum of the discrete time values) when converted to the time continuous domain corresponds to the integration int[v(t)](k) of the time continuous v(t) at the time point t(k) belonging to the frequency ramp k; for example, the center of the data acquisition time can be used as a reference point for the integration. This results in the rotation of the complex unit vector p1(i,k):

[0062]

[0063] Note that this relationship also assumes that the relative speed during the ramp is constant, i.e. the rotation speed of p1(i,k) is constant during each corresponding frequency ramp (since the frequency ramp is very short, this assumption also applies to high relative acceleration).

[0064] After multiplying the sampled values ​​s(i,k,m) by the unit vector p1(i,k), all frequency ramps are received at the same frequency, so the power peaks in the spectrum will not be broadened in the distance dimension. However, it must also be considered that the relative velocity changes with the relative acceleration, which means that multiple Doppler gates can be scanned during the data acquisition time, which still leads to a broadening of the power peaks in the Doppler dimension; Figure 5 Show that the above four objects are subjected to an additional relative acceleration of 10 m / s 2 This effect is caused by the vehicle braking completely. The change in relative speed means that the phase of the received signal no longer changes linearly on the frequency ramp, so that relation (2) no longer applies; the change is now proportional to the non-constant relative speed, so that k·T in relation (2) D The term v must be replaced by the integral int[v(t)](k) (similar to the derivation of p1(i,k) above from relation (8)):

[0065]

[0066] Phase The nonlinear component in is given by the difference between relation (8) and the original relation (2), which can be compensated by multiplying the sampled value s(i,k,m) by the corresponding second complex unit vector p2(k), which changes from frequency ramp to frequency ramp but remains constant on each frequency ramp:

[0067]

[0068] Among them, v av is the average speed during the data collection period. Figure 6 Demonstrates a relative acceleration of 10 m / s 2 The absolute value spectrum |S(j,l,m) / A(m)| obtained using this second complex unit vector p2(k) is shown in Figure 2; now, the stationary object shows a clear power spike again (note that in Figure 5 and Figure 6 In the figure, the two objects 3 and 4 are separated in the distance dimension, which is not obvious due to the angular illustration).

[0069] After multiplying the sampled values ​​s(i,k,m) by the two complex unit vectors p1(i,k) and p2(k), a clear power peak appears in the spectrum in both dimensions (i.e., range and Doppler) as long as the object has relative motion, and the settings of p1(i,k) and p2(k) are based on this relative motion. As mentioned above, for a sensor oriented forward, stationary objects near azimuth angle 0° are the most important, so the relative motion that needs to be considered is the generally non-constant own motion v ego (t) is the opposite number, that is, the negative number; thus, two complex unit vectors p1(i,k) and p2(k) are obtained, namely:

[0070]

[0071] It should be noted that in relation (11b), k·T can also be omitted in principle. D ·v ego,av subtraction of ; this would imply a Doppler shift such that a stationary object with an azimuth of 0° would have a Doppler (frequency shift) of 0.

[0072] As Figure 1 As an alternative to the modulation shown, the Figure 7 The form shown, in which the bandwidth of a single frequency ramp is reduced to 25%, i.e., B ch = 150 MHz, but the frequency positions of these frequency ramps (e.g. characterized by their center frequencies) are within the bandwidth B s =600MHz. This linear variation of the center frequency results in a distance-dependent component in the Doppler dimension of the two-dimensional discrete Fourier transform (DFT), thus achieving the same effect as the modulation bandwidth B. s = 600MHz corresponding to a high distance resolution. As shown in DE 10 2020 210 079 B3, for a relatively moving object, the time distance T of the frequency ramp k = 0, ..., K-1 DThe linear variation of (k) can offset the large broadening of the power peak in the Doppler dimension; relatively speaking, the magnitude of this variation is twice that of the center frequency variation and has the opposite sign, i.e. Figure 2 In the example of , it is equivalent to -2·600MHz / 76.5GHz=-1.57% (a reduction of 1.57% on K frequency ramps). However, in this modulation method, there is still the effect of slight changes in the received frequency on the frequency ramp due to relative motion; but due to the bandwidth B of a single frequency ramp ch =150MHz is significantly smaller than the first modulation method, so the blurring / smearing effect of the power peak in the distance dimension and Doppler dimension is much smaller here (reduced to about one-fourth). However, this smaller effect can also be completely eliminated by frequency shifting, that is, multiplying the sampled value by the complex unit vector p1(i,k) according to the above formula. For non-constant relative speed, that is, the effect of relative acceleration, it can also be compensated by multiplying with the second vector p2(k) according to the above formula. The above relationship is at least partially derived under the assumption that the time intervals of the frequency ramps are equal. Although the situation here does not completely meet the assumption, the deviation from this assumption is very small and can be ignored.

[0073] The determination of the two-dimensional discrete Fourier transform (DFT) is performed in most cases by first performing the fast Fourier transform (FFT) of the distance dimension, and then performing the fast Fourier transform (FFT) of the Doppler dimension. The reason for this is that the data of the frequency ramps appear successively, and once the data of the new frequency ramps are available, the first fast Fourier transform (FFT) can be determined by the sampled values ​​of each corresponding frequency ramp. In addition, after the first fast Fourier transform (FFT) of the distance dimension, the obtained data can be highly compressed without losing a large amount of information (see EP 3 152 587 B1). Only after the data of all frequency ramps are obtained (i.e., after the entire data collection is completed), the fast Fourier transform (FFT) of the Doppler dimension can be determined; if starting from the fast Fourier transform (FFT) of the Doppler dimension, the determination of the two-dimensional discrete Fourier transform (DFT) can only begin after the entire data collection is completed. If the fast Fourier transform (FFT) of the distance dimension is first determined, that is, the fast Fourier transform is performed on dimension i of the sampled values ​​s(i,k,m), then only the complex unit vector components that depend on dimension i need to be multiplied, that is, p1(i,k)-p2(k) does not depend on dimension i, so that the multiplication operation can also be performed after the first fast Fourier transform (FFT). It is very useful to first form the product of the window function (distance dimension) of the first fast Fourier transform (FFT) and the vector p1(i,k) at one time and then apply the modified window function to the received signals of all M=4 receiving antennas. Under the assumption that the relative speed is constant, that is, p1(i,k) is calculated according to equation (5), if it is calculated by comparing it with the vector that does not depend on k (which is also a rotated complex unit vector) and iteratively generating the modified window function on the frequency ramp k is particularly advantageous when using hardware accelerated computing units. It should be noted that the multiplication of the real-valued sample values ​​s(i,k,m) with the complex unit vector p1(i,k) produces a complex-valued signal, so the potential advantage of the real-valued input signal can no longer be used in the Fast Fourier Transform (FFT). As mentioned above, the complex unit vector p2(k) depends only on the frequency ramp running variable k and can only be multiplied with the complex unit vector p2(k) after the first Fast Fourier Transform (FFT). Preferably, this can be achieved by multiplying the vector p2(k) once with the second Fast Fourier Transform (FFT) window function (for the Doppler dimension); the modified window function can be used for the signals of all M=4 receiving antennas and all J=801 range gates.

[0074] It should be emphasized that the method according to the invention is not limited to the above-described order of implementing the two-dimensional discrete Fourier transform (DFT). It is also possible to start with the fast Fourier transform (FFT) in the Doppler dimension, wherein a multiplication operation with two complex unit vectors p1(i, k) and p2(k) must first be implemented, since both depend on the frequency ramp operation variable k.

[0075] So far, we have considered sensors oriented forward. Of course, the method according to the invention can also be used for sensors oriented in other directions. However, the relative movement, on which the complex unit vectors p1(i, k) and p2(k) are set, may need to be defined differently; for sensors oriented to look backward, the most important thing is not stationary objects, but fast approaching vehicles.

[0076] In order to make the radar system robust to interference from other radar systems, it is preferred and especially preferred to adopt a method similar to that proposed in articles such as WO 2008 / 040341 ​​A1, DE 10 2009 016 480 A1 and EP 2 629 113 B1, for example, to change the following modulation parameters:

[0077] - average spacing of the frequency ramp between cycles (as mentioned above, this also allows for a simple solution to the velocity ambiguity problem);

[0078] - Modulation bandwidth B from cycle to cycle ch (magnitude and / or sign);

[0079] - The time interval T of the frequency ramp D (k) by superimposing a random or pseudo-random meanless component that varies over k, typically within a few microseconds at most; for objects in relative motion, the received phase has a component that varies slightly over the frequency ramp, but this component is so small that the effect according to the discrete Fourier transform (DFT) (reducing the amplitude of noise and power peaks) is negligible;

[0080] - the frequency position of the frequency ramp (i.e. its center frequency) is superimposed with a random or pseudo-random meanless component that varies over k; this variation of the frequency position can also be achieved by always using the same frequency ramp, but changing the points in time at which the received signal samples are obtained; the resulting phase variation of the received signal, which is proportional to the range gate, can be compensated by a corresponding general phase correction after the first one-dimensional discrete Fourier transform (DFT) for the range dimension;

[0081] - The phase of the single transmission signal is varied by an additional phase modulator in the transmission device, wherein the phase position varies randomly or pseudo-randomly on the frequency ramp and is preferably compensated again on the receiving end in the digital signal processing device.

[0082] In accordance with Figure 1 In the radar system under consideration, there are M=4 receiving antennas and associated receiving channels m=0, ..., M-1. After the two-dimensional discrete Fourier transformation (DFT), a digital beamforming is preferably calculated in each range-Doppler gate (j, l), for example also in the form of a discrete Fourier transformation (DFT) or a fast Fourier transformation (FFT); a three-dimensional Fourier transformation is thus performed. The power peaks are then determined in the three-dimensional spectrum. The azimuth angle of the object is derived from the position of its power peak in the third dimension resulting from the receiving channel dimension m; the distance and relative speed are derived from the other two dimensions according to the above relationship. In order to have more channels available for the angle formation, it is preferred to use not only multiple receiving antennas, but also multiple transmitting antennas, and to evaluate the signals of all transmitting antennas and receiving antenna combinations to realize multiple virtual receiving channels. If all or some of the transmitting and / or receiving antennas are not operated simultaneously, a plurality of frequency ramp sequences of the above-mentioned type, which are preferably similar, are nested in each other.

[0083] Please note that the ideas and designs of the present invention shown in the above application examples can be applied to general detection and parameter interpretation in a simple manner, that is, also for other values. Therefore, general parameters are also given in the relationship and the drawings.

Claims

1. A method for a radar system for detecting surroundings of a motor vehicle, the method comprising: - a transmitting device for radiating a transmission signal, - receiving means for receiving the transmitted signal reflected from the object, and - signal processing means for processing received signals, in, - modulating the frequency of the radiated transmission signal so that it comprises a sequence of K linear ramps having at least approximately the same slope and duration, hereinafter referred to as frequency ramp, - in the signal processing device, a signal having a substantially transient transmission frequency or having a constant offset from the transient transmission frequency is mixed with the transmission signal reflected from the object and received by the receiving device, - in the signal processing device, sampling the mixed output signal I times during each of the K frequency ramps, if necessary after suitable preprocessing, - in the signal processing device, after preprocessing, a two-dimensional discrete time-frequency transform is completely or partially determined for these I·K sample values, It is characterized in that Preferably, based on the vehicle movement, the preprocessing of the I·K sampling values ​​includes frequency shifting the signal formed by the I sampling values ​​of each corresponding frequency ramp, so that for an object with a defined radial relative motion, the frequency of the signal formed by the corresponding I sampling values ​​remains unchanged on the K frequency ramps, and in particular, the blurring / tailing of the power peaks generated by such objects in the two-dimensional time-frequency transform is offset.

2. The method according to claim 1, characterized in that The detection range of the radar system includes the direction of travel. The radial relative motion is defined as the negative value of the vehicle's own motion. Therefore, for stationary objects in the direction of travel, the frequency of the signal composed of each corresponding I sampling value remains unchanged on the K frequency ramps.

3. The method according to any one of the preceding claims, characterized in that Frequency shifting is achieved by multiplying with a rotated complex unit vector.

4. The method according to claim 3, characterized in that The corresponding I sample values ​​of the frequency ramp are equidistant in time, and the rotation speed of the complex unit vector remains unchanged during each corresponding frequency ramp, but changes on each frequency ramp.

5. The method according to claim 4, characterized in that The change in the rotational speed of the complex element vector over the frequency ramp is proportional to the integral of the speed of the defined radial relative movement, ie in particular to the integral of the vehicle's own speed.

6. The method according to claim 5, characterized in that The time intervals and the center frequencies of the frequency ramps are at least approximately constant, and a linear variation of the rotational speed of the complex unit vector on the individual frequency ramps is employed, which corresponds to the possibly simplified assumption that the speed of the radial relative motion defined during the acquisition of the I·K sampling values ​​is constant, i.e. in particular the vehicle's own speed is constant.

7. The method according to claim 5, characterized in that The center frequency and the time interval of the frequency ramp vary at least approximately linearly, wherein the value of the relative change in the time interval is at least approximately twice the value of the relative change in the center frequency and the signs of these changes are opposite, using a linear variation of the rotational speed of the complex unit vector on the individual frequency ramps, which corresponds to the possibly simplified assumption that the speed of the radial relative movement defined during the acquisition of I·K sampling values ​​is constant, that is, in particular the vehicle's own speed is constant.

8. The method according to claim 6 or 7, characterized in that: The phase of the complex unit vector is point-symmetric on I sampling values ​​and K frequency ramps, that is, it is zero at the center point, so the position of the power peak generated by the object will not change in the two-dimensional discrete time-frequency transform.

9. The method according to any one of the preceding claims, characterized in that The first stage of a two-dimensional discrete time-frequency transform may be performed for each corresponding I sample values ​​of each frequency ramp, preferably using a fast Fourier transform to efficiently implement a discrete Fourier transform, with the frequency shift being implemented in conjunction with a window function for the transform.

10. The method according to claim 6 or 7 and 9, characterized in that The window function is iteratively modified from one frequency ramp to another by multiplying each response by the same rotated complex unit vector.

11. The method according to any one of the preceding claims, characterized in that A plurality of two-dimensional discrete time-frequency transforms are calculated for at least part of the same sampling values ​​by means of frequency shifts corresponding to different radial relative motions.

12. The method according to any one of the preceding claims, characterized in that Preferably, based on the vehicle movement, the preprocessing of the I·K sampling values ​​includes: phase shifting the signal formed by the I sampling values ​​of each corresponding frequency ramp so that the phase of the signal formed by the corresponding I sampling values ​​at K frequencies exhibits a purely linear change for an object with a defined radial relative motion, thereby preventing the power spikes generated by such objects in the two-dimensional time-frequency transform from being smeared / blurred in the dimension generated by the K frequency ramps.

13. The method according to claim 12, characterized in that The detection range of the radar system includes the direction of travel, and the defined radial relative motion is the negative value of the vehicle's own motion. Therefore, for stationary objects in the direction of travel, the phase of the signal formed by each corresponding I sampling value changes purely linearly on K frequency ramps.

14. The method according to claim 12 or 13, characterized in that Phase shifting is achieved by multiplying with a rotated complex unit vector.

15. The method according to claim 12, 13 or 14, characterized in that The phase shift is implemented in conjunction with a window function in the transform for the dimensions generated by the K frequency ramps.

16. A radar system for detecting the surrounding environment of a motor vehicle, the radar system comprising: - a transmitting device for transmitting signals by directional radiation, - receiving means for directionally receiving the transmitted signal reflected by the object, and - signal processing means for processing the received signal, in, the frequency of the radiated transmission signal is modulated in such a way that it comprises a sequence of K linear ramps having at least approximately the same slope and duration, hereinafter referred to as frequency ramps, - in the signal processing device, a signal having a substantially transient transmission frequency or having a constant offset from the transient transmission frequency is initially mixed with the transmission signal reflected from the object and received by the receiving device, - in the signal processing device, sampling the mixed output signal I times during each of the K frequency ramps, if necessary after appropriate preprocessing, - in the signal processing device, after preprocessing, a two-dimensional discrete time-frequency transform is completely or partially determined for these I·K sample values, It is characterized in that Preferably, based on the vehicle movement, the preprocessing of I·K sampling values ​​includes: frequency shifting the signal composed of I sampling values ​​of each corresponding frequency ramp, so that for an object with a defined radial relative motion, the frequency of the signal composed of each corresponding I sampling value remains unchanged on the K frequency ramps for the object with a defined radial relative motion, and in particular, thereby offsetting the blurring / tailing of power peaks generated by such objects in the two-dimensional time-frequency transform.

Images