ESTIMATION OF CARTESIAN VELOCITIES OF EXTENDED RADAR OBJECTS USING A RADAR SENSOR

Description

[0001] The invention relates to a method for a radar sensor, in particular for a radar sensor for motor vehicles, and to a radar sensor, in particular for motor vehicles, wherein the radar sensor has an antenna arrangement with a plurality of antennas arranged in different positions in one direction. State of the art

[0002] Conventional radar sensors detect radar targets in polar coordinates. For example, they measure a radial range, a radial relative velocity, and angles in azimuth and / or elevation. Determining a transverse (or tangential) velocity or angular velocity could only be achieved by observing an angle change over time or by using complex object models for extended radar objects.

[0003] Radar sensors are used in motor vehicles, for example, to measure the distances, relative speeds, and azimuth angles of vehicles or other radar targets detected in front of the vehicle. Several antennas are then arranged at a distance from one another on a horizontal plane, so that different azimuth angles of the detected radar targets lead to differences in the travel distances that the radar signals have to travel from the radar target to the respective antenna. These travel length differences lead to corresponding differences in the amplitude and phase of the signals received by the antennas and evaluated in the associated evaluation channels. Angle estimation takes advantage of the fact that the amplitude and phase relationships of the signals received by the various receiving antennas depend in a characteristic way on the angle of the radar target.By comparing the (complex) amplitudes received in the various channels with corresponding amplitudes in an antenna pattern, the angle of incidence of the radar signal and thus the azimuth angle of the radar target can be determined. In a similar way, the elevation angle of a radar target can also be estimated using antennas arranged vertically one above the other.

[0004] For the treatment of a transversely moving, extended radar object in object tracking, it was proposed to treat the point targets together, based on a hypothesis of an extended radar object and the corresponding assignment of point targets to the radar object, and to estimate the actual direction of the velocity (vx ,vy ) of the extended object from different observed radial velocities v 1 ,...,vn and respective measured angles α 1 ,..., α n of the n point targets assigned to the object using the overdetermined system of equations ν 1 ν 2 ⋮ ν n = cosα 1 sinα 1 cosα 2 sinα 2 ⋮ ⋮ cosα n sinα n ν x ν y , which is written as n M = M n . The solution is determined using the root mean square technique as n̂ = ( MT< M ) - 1< MT< n M .

[0005] DE 10 2012 024998 A1 describes a method in which, by means of at least one radar sensor, a transmission signal is emitted in successive measuring cycles and a signal reflected by the target object is received, wherein at least two target echoes are detected for the target object and a value of the radial velocity and a value of the target angle are determined for each of the at least two target echoes, and wherein the lateral velocity of the target object is determined as a function of the values ​​of the radial velocity and the values ​​of the target angle.

[0006] EP 2 068 173 A1 describes a method for determining a dynamic variable representative of a transverse movement of a located object in a driver assistance system for motor vehicles, which has two angle-resolving distance sensors arranged laterally offset from one another, wherein the dynamic variable is calculated on the basis of the radial components of the relative speed measured by the two distance sensors and the azimuth angles measured by the distance sensors.

[0007] DE 10 2013 019804 A1 describes a method for determining the movement of an object. The object is detected simultaneously with at least two radar sensors within an angular range, whereby movement information of the object is determined based on at least three object points arranged at different positions on the object, assuming a rigid object.

[0008] DE 10 2013 011239 A1 describes a method for determining the movement of an object. Using data acquired by a radar sensor, a speed profile of all measurement points assigned to the object is created from the radial velocities of the measurement points. The speed profile is used to determine the absolute speed and / or direction of movement of the object.

[0009] US 2015 / 0198711 A1 describes a method for detecting and tracking objects using multiple radar sensors. Radar data includes Doppler measurement data. Clusters are formed based on this data.

[0010] WO 2010 / 000252 A2 describes a radar system for detecting the surroundings of a motor vehicle, with at least two transmitting antennas. In one example, a linear phase change generated by a relative velocity is compensated for using eight antenna combinations. Disclosure of the invention

[0011] As radar sensor performance continues to improve, d,v estimates will be made with increased resolution. An increase in the usable sensor size, i.e., the size or aperture of the antenna array, will also enable increased angle estimation accuracy and improved angle separation. In an FMCW (frequency modulated continuous wave) measurement method with linear frequency ramps and evaluation of the received signals using discrete Fourier transformation, in particular an FFT (Fast Fourier Transformation), the width of a range bin of the Fourier transform corresponds to a distance difference Δr with Δr = c / (2F), where c is the speed of light and F is the frequency deviation of a linear frequency ramp of the FMCW transmitted signal. This distance difference is also referred to here as the range resolution.

[0012] The range resolution is therefore the smallest distance difference at which (given the same relative speed) two measured values ​​of the distance from the radar sensor can still be mapped to separate bins in the given operating mode of the radar sensor. When performing an FFT, the distance resolution corresponds to the separation between two distance bins in the FFT, i.e. the width of a distance bin. Here and in the following, the terms distance resolution and width of the distance bin are used synonymously. In contrast, the distance separation capability is understood to be twice the width of the distance bin. If the bandwidth of a radar sensor is increased, for example, with a frequency deviation of the transmitted signal of F = 2 GHz, a distance resolution of Δr = 7.5 cm is possible.If, at the same time, the aperture or, in the case of a MIMO (Multiple Input Multiple Output) radar sensor, the virtual aperture is increased to values ​​of a similar magnitude, then, depending on the angle of a radar target, the run length differences between received signals from individual antennas or evaluation channels can be so large that, in the Fourier spectra of the evaluation channels, information about the amplitude and / or phase of the received signals is contained not only in a frequency bin determined by the d,v estimate of a detected radar target, but also in one or more neighboring frequency bins. At a distance of 5 m from a radar target directly ahead and an offset between a central and an outer antenna position of, for example, 40 mm, an angular difference of approximately 0.5° results. If the radar target has a lateral speed of, for example,2.7 m / s (10 km / h), no relative velocity is observed at the center antenna position, but a radial relative velocity of 0.025 m / s is observed at the outer antenna position, or -0.025 m / s on the other side. With an FFT bin size of 0.1 m / s (velocity resolution of the measurement), this corresponds to a frequency shift of -1 / 4 bin, 0 bin, or +1 / 4 bin in the corresponding three evaluation channels.

[0013] The object of the invention is to provide a method and a radar sensor which enable a direct association of point targets to an extended radar object.

[0014] A further object of the invention is to provide a method and a radar sensor which enable a quick and simple estimation of a speed of an extended radar object, in particular a Cartesian speed.

[0015] To achieve at least one of the objects, a method according to the invention for a radar sensor for motor vehicles comprises claim 1. The estimated speed comprises information about a speed in the forward direction relative to the radar sensor and a tangential speed, thus comprising a two-dimensional speed. In particular, the speed can be a Cartesian speed, i.e., a speed specified in an orthogonal coordinate system. For a radar sensor oriented directly in the forward direction, the tangential speed corresponds to the lateral speed.

[0016] This makes it possible to estimate the (Cartesian) velocities of the respective radar targets based on a single measurement with a single transmitted signal modulation cycle, particularly for a transmitted signal with a frequency modulation pattern in the form of a frequency ramp or a rapid chirp sequence. In particular, the (Cartesian) velocities of the radar targets can be estimated from the respective frequency spectra based on a single measurement of the radial velocity in the respective evaluation channels.

[0017] The measurement of speeds and the assignment (association) of radar targets to an extended radar object can therefore be performed very quickly. In particular, the measurement of (Cartesian) speeds can be performed directly for individual radar targets, based on detection of the radar target within a single modulation cycle or a frequency ramp of the transmitted signal. This makes it possible to measure Cartesian speeds and perform an association with an extended object, even for objects that are only visible in one or a few cycles and are therefore difficult to handle with a complex object model, such as fast-moving objects or roadside structures.

[0018] This enables a significantly faster response from a driver assistance system or an automated driving system, especially when a radar target is at a relatively short distance. This is particularly beneficial for protecting vulnerable road users (VRUs), such as pedestrians.

[0019] A radar target is defined as a reflection center to which only a single position is assigned, in contrast to an extended radar object, which is assigned multiple reflection centers. The term "radar target" is used synonymously with "point target."

[0020] In the method, a transmission signal of the radar sensor is transmitted by at least one antenna, and a signal is received by at least one antenna. In one expedient embodiment, a transmitted signal is frequency-modulated in a ramp-like manner. In one expedient embodiment, the radar sensor is an FMCW radar sensor.

[0021] The method may further comprise the step of estimating a speed of the extended radar object based on the determined individual radial velocities of the associated radar targets. This is thus a method for estimating a speed of an extended radar object. This enables a faster and improved speed estimation of an extended object compared to conventional methods. The speed may include information about a forward speed relative to the radar sensor and a tangential speed. In particular, it may be a Cartesian speed.

[0022] Here and in the following, the quantities relating to the radar sensor as a whole that are estimated or determined are also referred to as "global" quantities, while quantities relating to the respective evaluation channels and the relevant mean antenna positions of the transmitting and receiving antennas are referred to as "individual" quantities. For example, the tangential velocity to be estimated can be referred to as the global tangential velocity.

[0023] Furthermore, the object is achieved by a radar sensor, in particular for motor vehicles, with an antenna arrangement with a plurality of antennas arranged in different positions in one direction, and with a control and evaluation device designed to carry out the method.

[0024] Advantageous embodiments and further developments of the invention are specified in the subclaims.

[0025] In an expedient embodiment, in the step of estimating a speed of the respective radar target, a Cartesian speed of the radar target is estimated based on the relationship: ν r , 1 ⋮ ν r , i ⋮ ν r , l = cosθ 1 sinθ 1 ⋮ ⋮ cosθ i sinθ i ⋮ ⋮ cosθ l sinθ l ν x ν y , where i = 1,..., l respective evaluation channels, where ν r , i the determined individual radial velocities of the radar target are θ i are individual aspect angles of the radar target assigned to the respective evaluation channels, and (ν x ,ν y ) is the Cartesian velocity of the radar target, where ν y is the tangential velocity and ν x is the velocity of a radar target in the forward direction relative to the radar sensor.

[0026] In an expedient embodiment, in the step of determining the respective individual radial velocities of the radar target for the respective evaluation channel, the respective individual radial velocity is determined based on a respective frequency position of the signal in the evaluation channel.

[0027] In a practical embodiment, a discrete frequency spectrum is calculated in the respective evaluation channels using Fourier transformation, and the respective frequency position of the signal in the respective evaluation channel is determined for the radar target with a resolution that is finer than the frequency spacing of the sampling points of the discrete frequency spectrum. The resolution of the frequency position is understood here as a quantity that indicates how finely graded the possible values ​​of the frequency position are that can be obtained during the determination. The sampling points of the discrete frequency spectrum can also be referred to as frequency points or frequency bins.

[0028] The respective frequency position can be determined, for example, by interpolating the frequency spectrum and searching for a peak maximum of the signal, or by fitting a frequency parameter of a model function to the signal in the frequency spectrum.

[0029] The following examples are explained in more detail using the drawings. They show: Fig. 1 shows a block diagram of a radar sensor for motor vehicles according to the invention; Fig. 2 shows a schematic representation of frequency bins of Fourier spectra of respective evaluation channels; Fig. 3 shows a relationship between an antenna position and a radar target; Fig. 4 shows a block diagram for explaining a first part of a method according to the invention; Fig. 5 shows a flowchart for explaining a second part of the method; and Fig. 6 shows a sketch of a traffic situation.

[0030] The Fig. 1 The radar sensor shown has several receiving antennas or antenna elements 10, 12 on a common substrate 18. The radar sensor is installed in a motor vehicle such that several of the antennas 10, 12 are located at the same height next to each other at horizontal positions yi, i=0,...,k. In Fig. 1symbolically represent radar beams received by the antennas at a respective azimuth angle θi (aspect angle).

[0031] A high-frequency section 20 for controlling a transmitting antenna 22 includes a local oscillator 24 that generates the radar signal to be transmitted. The radar echoes received by the antennas 10, 12 are each fed to a mixer 28, where they are mixed with the transmission signal supplied by the oscillator 24. In this way, a baseband signal or intermediate frequency signal Z0, Z1, ..., Zi, ..., Zk is obtained for each of the antennas 10, 12, which is fed to an electronic control and evaluation unit 30.

[0032] The control and evaluation unit 30 contains a control section 32 that controls the function of the oscillator 24. In the example shown, the radar sensor is an FMCW radar, i.e., the frequency of the transmission signal supplied by the oscillator 24 is periodically modulated in the form of a sequence of rising and / or falling frequency ramps.

[0033] Furthermore, the control and evaluation device 30 contains an evaluation section with a k-channel analog / digital converter 34, which digitizes the intermediate frequency signals Z0 - Zk received from the k antennas 10, 12 and records each of them over the duration of a single frequency ramp. The time signals thus obtained are then converted channel by channel in a transformation stage 36 using fast Fourier transformation (FFT) into corresponding frequency spectra. In these frequency spectra, each radar target is identified in the form of a peak, the frequency position of which depends on the signal propagation time from the radar sensor to the radar target and back to the radar sensor, as well as—due to the Doppler effect—on the relative speed of the radar target.

[0034] From the frequency positions of two peaks obtained for the same radar target, but on frequency ramps with different gradients, for example, an ascending ramp and a descending ramp, the distance d and the relative velocity v of the radar target in question can then be calculated in a known manner. The estimated distance d can be referred to as the global distance of the radar target, in contrast to individual distances di, which are assigned to the respective evaluation channels.

[0035] As in Fig. 1As schematically illustrated using the radar beams, the different positions of the antennas 10, 12 mean that the radar beams emitted by one and the same antenna, reflected by the radar target, and then received by the different antennas, travel different distances and therefore exhibit phase differences that depend on the azimuth angle θ of the radar target. The associated intermediate frequency signals Z0 - Zk also exhibit corresponding phase differences. The amplitudes (magnitudes) of the received signals also vary from antenna to antenna, also depending on the azimuth angle θ.

[0036] A velocity estimator 38 uses the signals in the frequency spectra of the evaluation channels to estimate a Cartesian velocity of a respective radar target with the components vx, the forward velocity relative to the radar sensor, and vy, the tangential velocity. This is explained in more detail below. Furthermore, an angle estimator 40 uses the signals to estimate an azimuth angle of the respective radar target.

[0037] With a high bandwidth, corresponding to a large frequency deviation of the FMCW modulation, and a large extension of the antenna array, depending on the azimuth angle θ of the radar target and its distance d, the complex amplitudes in the individual receiving channels are contained at different frequency positions fa(i) in the frequency spectrum of the received signal. Fig. 2schematically illustrates the frequency positions fa(i) of the signals (peaks of the spectrum) obtained for a radar target in the evaluation channels i, whereby successive frequency bins of the Fourier spectrum are shown in the direction of increasing frequency f.

[0038] Fig. 3Illustrated in plan view for an antenna position of an evaluation channel, denoted by the index i at the coordinates (0, yi), the relationship to a point target as a radar target at the coordinates (x, y) with the Cartesian velocity (vx, vy). The distance of the point target is denoted by di, and the aspect angle of the received radar signal by θi. To simplify the illustration, it is assumed that the origin (0, 0) represents the center of the antenna array and corresponds to a mean antenna position. In the example, vx=0, corresponding to a situation in which the point target is moving exactly transversely in front of the radar sensor. For simplification, the figure shows a situation in which the relative and absolute radial velocity relative to the origin are zero. At an antenna position (0, y0) located at the origin, a radial velocity vr,0 = 0 is measured. At an antenna position (0, yi), a radial velocity vr,i is measured.This corresponds to the projection of the Cartesian velocity (vx,vy) onto the radial direction of the antenna position and thus depends on the aspect angle θi of the radar target at the antenna position.

[0039] The relationship between the aspect angles θi of the evaluation channels i, the Cartesian velocity (vx,vy) of the point target, and the individual radial velocities estimated from the spectrum in the respective evaluation channels i is given by equation (1). With a corresponding definition of the matrix M, this can be rewritten as: ν _ r = M ν _ xy

[0040] From the individual radial velocities vr,i in the respective evaluation channels, the vector of the Cartesian velocities n xy can be estimated using the least squares method (LSE estimation): ν _ ^ xy = M T M − 1 M T ν _ r

[0041] The larger the angular differences and the more accurately the relative velocities can be determined, the better the corresponding estimate will be. The least-squares estimate can be calculated numerically, for example, using a pseudo-inverse, singular value decomposition (SVD), or QR decomposition.

[0042] The distance di of the evaluation channels "seen" by the radar sensor due to the run-length differences depends on the antenna configuration. In a bistatic system or a MIMO system, the effects (distance or runtime) for the path from the transmitting antenna to the target and from the target to the receiving antenna are added and averaged. The distance di is thus the average distance of the outward and return paths over the average signal runtime. Accordingly, an average antenna position of the transmitting antenna and the receiving antenna is considered.

[0043] The control and evaluation device 30 is designed to carry out a method for estimating the Cartesian speed of the radar target, which is exemplified by Fig. 4 explained and is implemented in the speed estimator 38.

[0044] By interpolating the frequency spectra, the frequency positions fa(i) of the signals (peak positions) of the channels i are determined with high resolution in step S10.

[0045] In a firstIn step S12, for each located object, i.e., each radar target (each peak in the frequency spectrum), the complex amplitudes obtained in the I reception channels are compared with the antenna pattern to estimate the azimuth angle θ of the radar target. A vector of the complex amplitudes is evaluated at a constant frequency position fref in the respective spectra of the channels. The estimated aspect angle θ can be referred to as the "global" aspect angle, in contrast to the individual aspect angles of the respective evaluation channels.

[0046] In step S14, distance-dependent bin shifts (frequency shifts Δfa(i)) of the channels are determined from the azimuth angle θ and the antenna positions yi. These frequency shifts can be referred to as distance-dependent frequency shifts. These frequency shifts can also be referred to as frequency corrections. As explained above, they are caused by a high distance resolution of the measurement. The frequency shift occurs depending on the aspect angle. Depending on the aspect angle, it can also be zero.

[0047] In step S16, the displacements Δfa(i) are subtracted from the frequency positions fa(i), and the remaining frequency positions fa(i) - Δfa(i) are evaluated in step S18 to determine the individual radial velocities vr,i of the channels i. This is done according to the FMCW equation k = 2 c dF + f 0 ν r T . Here, k is a bin position corresponding to the remaining frequency position, c is the speed of light, d is the distance, F is the frequency deviation of the ramp, f0 is the center frequency, vr is the radial velocity, and T is the duration of the ramp. The evaluated frequency position is the remaining frequency position taking into account the frequency shift.

[0048] In step S20, individual aspect angles θi are determined from the global distance d and the estimated azimuth angle θ, for example, taking into account the antenna positions yi, e.g., based on geometric relationships between these variables. The individual aspect angles can also be calculated, for example, from Cartesian coordinates of the radar target and the mean antenna positions.

[0049] In step S22, the Cartesian velocity of the radar target is estimated using equations (1) and (2). For a radar sensor output, these can be transformed, for example, into a radial and tangential velocity (or angular velocity) relative to the origin.

[0050] In an optional step S24, the angle estimator 40 estimates the azimuth angle θ in a second, improved angle estimation, wherein a vector of the complex amplitudes is evaluated at respective frequency positions in the respective spectra of the channels, taking into account the frequency shifts Δfa(i) and / or corresponding speed-dependent frequency shifts Δfb(i) resulting from the azimuth angle θ and the antenna positions yi. In particular, this is a step of second estimation of the aspect angle of the radar target, which is performed in addition to the above-described (first) step S12 of estimating the aspect angle.

[0051] The method may in particular be an iterative method in which, based on the second estimate of the aspect angle, the steps S14, S16, S18, S20, S22 dependent on the estimate of the aspect angle are executed again.

[0052] As from theAs can be seen from the diagram, steps can be performed in parallel or in a different order.

[0053] In a practical embodiment, differences in individual radial velocities are taken into account when the distance to the radar target is below a distance threshold. At very large distances, the effects are too small to distinguish between individual radial velocities. The distance threshold can be selected based on test measurements or a theoretical velocity resolution of the radar sensor.

[0054] Fig. 5illustrates method steps for estimating the speed of an extended radar object based on the method described above, which is carried out for respective located radar targets (point targets). Based on the results of the respective steps S22, in step S30, several radar targets belonging to an extended radar object are associated (assigned to one another) based on the estimated Cartesian speeds of the radar targets. For example, a decision is made as to whether the Cartesian speeds agree within a tolerance range and whether the positions, distances, and angles of the radar targets agree within a boundary range. Fig. 6 schematically shows three detected radar targets with different radial velocities vr , for which the same Cartesian velocity (vx ,vy ) was estimated. These are associated to an extended object 60.

[0055] By extending equation (1) by rows of the matrix and the left-hand vector for the relevant further radar targets belonging to the same extended object, an estimate of the Cartesian velocity of the extended object can then be made in a corresponding manner in step S32 using equation (2).

[0056] The embodiments described here use a bistatic antenna concept.

[0057] The described method can be advantageously used particularly with FMCW radar sensors that operate with so-called rapid chirp sequences. In this case, a large number of frequency ramps (chirps) are run through in rapid succession. These ramps have a steep gradient and a relatively short duration. The 2D FFT performed on the individual ramps and the sequence of ramps results in a two-dimensional frequency spectrum for each evaluation channel i. Accordingly, the frequency positions fa(i) and the frequency shifts Δfa(i) and Δfb(i) are generally two-dimensional vectors.

Claims

1. Method for a radar sensor for motor vehicles, comprising an antenna arrangement having multiple antennas (10, 12) arranged in different positions in the horizontal direction, the method comprising the following steps: - determining, for respective evaluation channels (i) which correspond to different middle antenna positions ((0,yi)) of relevant bi-statically transmitting and receiving antennas (22, 10, 12), and for respective individual radar targets, a respective individual radial velocity (vr,i) - assigned to the respective evaluation channel (i) - of the respective radar target on the basis of signals obtained in respective evaluation channels (i) in a single measurement using a single transmission signal modulation cycle, the middle antenna positions ((0,yi)) of the respective evaluation channels (i) being different in one direction; - estimating, for the respective radar targets, a respective velocity ((vx, vy)) of the respective radar target, on the basis of the individual radial velocities (vr,i) of the radar target that have been determined in a single measurement of the radial velocity in the respective evaluation channels (i), wherein the velocity ((vx, vy)) includes information about a velocity (vx) in the forward direction relative to the radar sensor and a tangential velocity (vy) relative to the radar sensor; - determining, for the respective evaluation channels (i) and the respective individual radar targets, a respective individual aspect angle (Θi) of the respective radar target, which aspect angle is assigned to the respective evaluation channel; wherein the step of estimating a respective velocity ((vx,vy)) of the respective radar target involves the velocity ((vx,vy)) of the respective radar target being estimated on the basis of the determined individual radial velocities (vr,i) of the radar target and on the basis of the determined individual aspect angles (Θi) of the radar target; and - assigning radar targets as belonging to an extended radar object, depending at least on the estimated velocities ((vx,vy)) of the radar targets.

2. Method according to Claim 1, wherein the step of estimating, for the respective radar targets, a respective velocity ((vx,vy)) of the respective radar target involves a respective Cartesian velocity ((vx,vy)) of the respective radar target being estimated that includes the velocity (vx) in the forward direction relative to the radar sensor and the tangential velocity (vy).

3. Method according to Claim 1 or 2, comprising the following step: - estimating a velocity ((vx,obj, vy,obj)) of the extended radar object on the basis of the determined individual radial velocities ((vx, vy)) of the assigned radar targets.

4. Method according to Claim 1, wherein the step of determining the respective individual aspect angles (Θi) of the respective radar target involves the respective individual aspect angles (Θi) of the radar target being determined on the basis of an estimated aspect angle (Θ) of the radar target and an estimated distance (d) of the radar target, taking into account the relevant middle antenna position (yi) of the respective evaluation channel (i).

5. Method according to any of the preceding claims, wherein the step of determining the respective individual radial velocities (vr,i) of the respective radar target for the respective evaluation channel (i) involves the respective individual radial velocity (vr,i) being determined on the basis of a respective frequency (fa(i)) of the signal in the evaluation channel (i).

6. Method according to any of the preceding claims, wherein the step of determining the respective individual radial velocities (vr,i) of the respective radar target for the respective evaluation channel (i) involves the respective individual radial velocity (vr,i) being determined from a remaining frequency after subtracting a respective frequency shift (Δfa(i)), wherein the respective frequency shifts (Δfa(i)) for the respective evaluation channels (i) correspond to differences in distance between the radar target and the respective corresponding middle antenna positions ((0,yi)).

7. Method according to Claim 6, comprising the following steps: - estimating, for the respective radar target, an aspect angle (Θ) of the respective radar target on the basis of amplitude and / or phase relationships between signals of respective evaluation channels corresponding to different middle antenna positions (yi) of the relevant transmitting and receiving antennas (22, 10, 12) in the stated direction; - determining, for the respective radar target, respective frequency shifts (Δfa(i)) of the signals in the respective evaluation channels (i), which frequency shifts (Δfa(i)) correspond to differences in distance between the radar target and the respective corresponding middle antenna positions ((0,vi)), depending on the estimated aspect angle (Θ), wherein the step of determining the respective individual radial velocities (vr,i) of the respective radar target for the respective evaluation channel (i) involves the respective individual radial velocity (vr,i) being determined from the remaining frequency of the relevant signal in the evaluation channel (i), taking into account the respective determined frequency shift (Δfa(i)).

8. Method according to any of the preceding claims, comprising the following step: - estimating, for the respective radar target, an aspect angle (Θ) of the radar target on the basis of amplitude and / or phase relationships between signals of respective evaluation channels corresponding to different middle antenna positions ((0,yi)) of the relevant transmitting and receiving antennas (22, 10, 12) in the stated direction, wherein the signals of the respective evaluation channels (i) are evaluated at respective frequencies, taking into account respective first frequency shifts (Δfa(i)) of the signals in the respective evaluation channels (i), which first frequency shifts (Δfa(i)) correspond to differences in distance between the radar target and the respective corresponding middle antenna positions ((0,yi)), and / or taking into account respective second frequency shifts (Δfb(i)) of the signals in the respective evaluation channels (i), which second frequency shifts (Δfa(i)) correspond to differences in the individual radial velocities (vr,i) of the radar target for the respective evaluation channel (i).

9. Radar sensor for motor vehicles, comprising an antenna arrangement having multiple antennas (10, 12) arranged in different positions in the horizontal direction, and comprising a control and evaluation device (30) designed to carry out the method according to any of Claims 1 to 8.