EP3311189A1

EP3311189A1 - Method for operating a radar device

Info

Publication number: EP3311189A1
Application number: EP16719095.8A
Authority: EP
Inventors: Christopher Brown; Michael Schoor
Original assignee: Robert Bosch GmbH
Current assignee: Robert Bosch GmbH
Priority date: 2015-06-22
Filing date: 2016-04-29
Publication date: 2018-04-25
Also published as: US10684364B2; DE102015211490A1; CN107787460B; US20180164421A1; CN107787460A; WO2016206841A1

Abstract

The invention relates to a method for operating a radar device (100), comprising the steps of: - determining a matrix with time signals of reflected radar radiation; - determining elements of a distance-velocity-power matrix of a radar target from said time signals; - carrying out a first one-dimensional discrete Fourier transform for the elements of said distance-velocity-power matrix in a first dimension; and - carrying out a second one-dimensional discrete Fourier transform for the elements of the distance-velocity-power matrix in a second dimension such that the second one-dimensional discrete Fourier transform is implemented for every second element of the distance-velocity-power matrix, in a mathematically-defined offset manner.

Description

description

title

Method for operating a radar device The invention relates to a method for operating a radar device. The

The invention further relates to a radar device.

PRIOR ART In the motor vehicle sector, radar systems are increasingly being used to support advanced driver assistance systems of motor vehicles. One of the main tasks of said radar systems is to determine a distance to objects and a speed of the objects (for example, vehicles, pedestrians, stationary obstacles, etc.) in the vicinity of the motor vehicles. This is important for the driver assistance system "Adaptive Cruise Control", for example.

(ACC), where an accurate estimate of a distance and a relative speed of the vehicle is used to determine appropriate actions of the motor vehicle. In motor vehicles, said radar systems can also be used to implement safety functions, such as warning the driver in critical situations or causing an emergency stop when a collision can no longer be avoided. Various different signal modulation methods are known, with chirp sequence modulation (CS modulation) being a type of modulation that is particularly frequently used in automotive radar systems. In this type of modulation, a sequence of so-called chirp signals (linear frequency-modulated electromagnetic signals) is emitted, in which the current frequency of the signals is linearly time-variable. One of the advantages of chirp Sequence modulation is that it allows a simultaneous estimation of distance and relative velocity of objects.

After receiving reflected chirp signals and after appropriate pre-processing, the reflected time signals are typically stored in a two-dimensional array, each column of the matrix containing values of received signals of a chirp, with a number of columns of the matrix being a number of chirps. Signals of a transmission sequence corresponds. A discrete Fourier transform (DFT) for data elements of a chirp along a column of the data matrix allows an estimate of a distance (or range) of targets in a footprint of the radar. The data elements of the columns of the matrix represent distances of target objects. Performing a second discrete Fourier transformation along the lines of the resulting matrix allows an estimate of the relative velocity of the target objects.

Performing a two-dimensional discrete Fourier transform produces a distance-speed-power (d-v) matrix, wherein amplitude values of the elements of the d-v matrix in a third dimension represent estimates of the reflected signal energy for the corresponding distance and velocity of the target object. In practice, a fast Fourier transform is typically performed for this purpose that realizes an efficient implementation of the discrete Fourier transform.

Disclosure of the invention

It is an object of the present invention to provide an improved method of operating a radar apparatus.

The object is achieved according to a first aspect with a method for operating a radar device, comprising the steps:

- determining a matrix with time signals of reflected radar radiation;

Determining elements of a distance-speed-power matrix of a radar target from the time signals; Performing a first discrete one-dimensional Fourier transform for the elements of the distance-speed power matrix in a first dimension; and

Performing a second discrete one-dimensional Fourier transform for the elements of the distance-speed power matrix in a second dimension such that the second discrete one-dimensional Fourier transform is performed mathematically defined offset for every other element of the distance-dimension-power-matrix of the second dimension ,

According to a second aspect, the object is achieved with a radar device, comprising:

a generator for generating elements of a distance-speed-power matrix of a radar target; and an evaluation device for evaluating the elements of the distance-speed-power-matrix, wherein by means of the evaluation device elements of the distance-speed-power-matrix are equally evaluable in a first dimension, and wherein by means of the evaluation device elements of the distance-speed- Performance matrix in a second dimension mathematically defined offset evaluable.

In this way, peak performance of elements of the distance-speed-power matrix can be better determined. As a result, this is made possible by realizing a kind of hexagonal "sampling grid", which advantageously requires comparison of elements of the distance-speed-power matrix with only six other values for an evaluation. Speed space) is a continuous two-dimensional space in which values at each point of the dv space correspond to an amplitude of a received radar signal generated by an object of defined distance and speed relative to the radar. Since only a limited number of input data is available, the "real" dv space can thus only be "scanned" at certain points. As a result, an increase of a "granularity" of the matrix can advantageously be provided, which advantageously supports an improved estimation of distance and relative speed of targets. Preferred embodiments of the method according to the invention are the subject of dependent claims.

An advantageous development of the method provides that the elements of the distance-speed-power-matrix by means of a chirp sequence

Modulation can be generated. In this way, a favorable in the automotive field modulation is used, which is particularly suitable for high real-time requirements. An advantageous development of the method provides that first of all the discrete Fourier transformation is carried out for all elements of columns of the distance-speed-power matrix, wherein the discrete

Fourier transform is then performed for all elements of lines of the distance-speed-power-matrix. In this way, a column-by-row generation of matrix elements can be carried out, wherein advantageously data elements of the preceding chirp signal can already be processed by means of discrete Fourier transformation already when subsequent chirp signals are received and next-following matrix elements are generated. In this way, a high real-time capability of the method is supported.

A further advantageous development of the method provides that first of all the discrete Fourier transformation is carried out for all elements of lines of the distance-speed-power matrix, wherein the discrete

Fourier transformation afterwards for all elements of columns of the distance

Speed performance matrix is performed. Thereby, the order of performing the discrete Fourier transform for the entire matrix may be reversed, thereby providing a mathematically equivalent effect which may be particularly useful for less time critical applications (e.g., in astronomy, geodesy, etc.).

A further advantageous development of the method provides that each element of every second row of the distance-speed-power matrix is multiplied by the factor e ^{jM / N} , with the parameters: n column number of the distance-speed-power matrix N .... Total number of columns of the distance-speed-power matrix In this way, a mathematically defined technical realization of the method is provided.

The invention will be described in detail below with further features and advantages with reference to several figures. All features described or illustrated alone or in any combination form the subject matter of the invention, regardless of their combination in the claims or their dependency, and regardless of their representation in the description and in the figures.

In the figures shows:

Fig. 1 shows a conventional sampling scheme for a d-v matrix with a

test cell;

Fig. 2 shows a known arrangement scheme of telecommunication devices;

FIG. 3 shows a sampling scheme according to the invention for a d-v matrix with a test cell; FIG.

4 shows a worst-case scenario of a conventional sampling scheme for a d-v matrix;

FIG. 5 shows a worst-case scenario of an inventive sampling scheme for a d-v matrix; FIG.

FIG. 6 shows a conventional scanning scheme for a d-v matrix with an elongate target object; FIG.

7 shows a scanning scheme according to the invention for a d-v matrix with an elongate target object;

Fig. 8 is a principle flowchart of an embodiment

Method for operating a radar device; and 9 shows a basic block diagram of an embodiment of the radar device according to the invention.

Embodiments of the invention

1 shows a schematic representation of a conventional sampling scheme for a distance-speed-power-matrix (d-v-matrix) in the d-v-plane. The d-v plane scales a speed v in the x direction and a distance d of a radar device of a motor vehicle one to a target object (not shown) in the y direction. It is necessary for a reliable determination of the target object to obtain peak values of elements of the distance-speed-power matrix, each peak value representing a potential target object, provided that its power amplitude, which is plotted in a third dimension, is above Noise is lying. In the illustrated sampling scheme with rectangularly arranged test cells 1 and sampling points respectively, local peaks can be found by comparing each matrix element with eight adjacent matrix elements. 1 shows, with a square framing, that a test cell 1 (English, cell under test, CUT) is surrounded in the d-v plane by eight potential neighboring elements or test cells 1.

In typical applications, a d-v spectrum of a target object has a two-dimensional pattern due to effects of limited window sizes, which may include a main lobe and several side lobes. This has the effect of spreading electromagnetic signal energy across several adjacent cells. If only the main lobe is considered, the closer the test cell 1 is to the peak value of the d-v spectrum, the higher the signal energy. Although the four closest neighboring test cells 1 (in the "north", "south",

"East" and "west") to a specific test cell 1 have a minimum distance to the test cell 1, distances to diagonally arranged neighboring test cells (in the northeast, northwest, southeast and southwest) are considerably larger. As a result, it can be seen that distances between sampling points of the distance-speed-power-matrix and the test cell 1 can be very different. This is an inherent property of rectangular sampling schemes and means that the more distant cells (Northeast, Northwest, Southeast, and Southwest) are more sensitive to influences from other targets. An alternative to rectangular sampling schemes is a hexagonal sampling scheme where every other row or column value is shifted by one half-value. The advantages of hexagonal sampling schemes relative to spatially scanned data are known, for example, for geophysical applications, thus allowing for efficient masking of a surface. This allows a maximum or worst case distance between each

Point and one sample point to be reduced. For this reason, cell towers or other wireless telecommunications devices are often located locally in a hexagonal pattern, as shown in principle in the diagram of FIG. 2 with the x and y axes.

In the aforementioned geophysical applications, the two-dimensional area to be scanned is a physical area of a ground surface in which both dimensions have the same physical unit length. It is proposed to use a conventional two-dimensional matrix with chirp

Sequence data of a discrete two-dimensional digital or discrete Fourier transform, wherein the Fourier transform is applied according to a hexagonal sampling scheme. In this case, for example, a two-dimensional dV matrix is initially generated, in which each data element corresponds to a signal energy of a temporal chirp sequence signal. For example, with 32 chirp signals of 512 values each, a d-v matrix of 512 rows and 32 columns can be generated.

Thereafter, a one-dimensional digital Fourier transform (preferably using Fast Fourier Transform, FFT) is performed along each column of the dv matrix. Thereafter, for every other line (eg, every even-numbered line) of the dv matrix, the one-dimensional digital Fourier transform is performed for all elements of the entire line.

Finally, all elements of every other row (eg, every odd-numbered row) are multiplied by the factor e ^{jM / N} , where: n ... column number of the dv matrix, 0 to N-1

N ... number of columns of the d-v matrix

Thereafter, the one-dimensional digital Fourier transform over the mathematically manipulated elements of every other line (e.g., each

odd-numbered line).

The effect of the latter step is to perform the digital Fourier transform on so-called half-frequency bins or the digital one

Fourier transform at frequencies between the standard standard frequency bins rather than evaluating, as in conventional digital Fourier transform, at k / N cycles per sample, where: k ... integer with values between 0 and N -1

As a result, the digital Fourier transform is evaluated at lobes of (k + 1/2) / N cycles per sample.

Alternatively, the method can also be realized by first performing the digital one-dimensional Fourier transform over all rows of the dv matrix until all elements of the dv matrix have been transformed. Then the multiplication of the elements of each second column is performed with the factor e ^Λ -j * π * m / M with the parameters:

... line number

.. total number of lines This makes it possible to first process the lines and then apply the shifted digital Fourier transform to alternative columns (equivalent to transposing the dv data matrix). A MATLAB code with which the method can be implemented is listed below: function XX = fft2hex (x)

% x - time domain data matrix

% XX - calculated d-v matrix

X1 = fft (x, [], 1);

N = size (x, 2); for ii = 2: 2: size (X1, 1),

X1 (ii,:) = X1 (ii,:). * Exp (-1j * pi * (0: (N-1)) / N);

end

XX = fft (X1, [], 2);

end

From Fig. 3 it can be seen that by such scanning a test cell 1 each has only six neighboring test cells 1. FIG. 3 indicates this with a hexagonal border of the test cell 1. Table 1 below shows that a number of neighboring elements to the test cell 1 has decreased in the following way:

Table 1

In comparison, a conventional scanning scheme has the following distances from neighboring cells to test cell 1, as shown in the following Table 2: Number of post-speed domain Distance domain barzellen

2 1 0

4 1 1

2 0 1

Table 2

FIG. 5 shows an advantage of such a sampling of the distance-speed-power-matrix compared to FIG. 4.

4 indicates with four arrows that in a conventional rectangular sampling scheme in a worst-case scenario, one sampling point may have four adjacent, substantially equally spaced sampling points. In this way, a reliable detection of a target object can be difficult.

In contrast, it can be seen from FIG. 5 with the modified sampling scheme that in a worst-case scenario, a sampling point has only three adjacent sampling points, each having the same spacing, the distances to the adjacent sampling points being reduced in comparison to FIG. Thereby, a process of evaluating target objects can be significantly simplified and improved because a distance of a peak in the d-v space to the nearest sampling point is reduced. From FIG. 3 it can be seen that in this case only six comparisons with adjacent sampling points are required, instead of eight, as can be seen from FIG.

FIG. 6 shows a conventional rectangular scanning scheme of a d-v matrix, wherein an elongate target object 200 is to be detected. It can be seen that, due to the regular structure of the sampling scheme, a distance of the target object 200 from all sampling points or test cells 1 is substantially the same size.

In contrast, FIG. 7 shows that with a hexagonal sampling scheme, individual sampling points or test cells 1 of the dv matrix are closer to the target object 200 than others. As a result, detection of the elongate target 200 may be easier and more accurate. This can be particularly useful if the Target object 200 is weak (eg, a low-reflectivity pedestrian), since a contribution of the signal peak value of the target object 200 to the amplitude of the nearest dv matrix element is stronger, the closer the sampling point is to the peak value of the target object 200. As a result, using the sampling scheme of FIG. 7, a distance of a "true" target in the dv space to the next sample point is only about half as large as when using the conventional sampling scheme of FIG. 6.

Fig. 8 shows in principle a sequence of an embodiment of the method according to the invention.

In a step 300, a determination of a matrix with time signals of reflected radar radiation and determination of elements of a distance-speed-power matrix of a radar target are performed.

In a step 310, a first discrete one-dimensional Fourier transform is performed for the elements of the distance-speed power matrix in a first dimension.

In a step 320, a second discrete one-dimensional Fourier transform is performed for the elements of the distance-speed-power matrix in a second dimension such that the second discrete one-dimensional Fourier transform defines mathematically for every other element of the distance-speed-power matrix offset is performed.

Fig. 9 greatly illustrates a radar apparatus 100 having a generating means 10 for generating elements of a distance-speed-power matrix of a radar target 200. The radar apparatus 100 further comprises an evaluator 20 for evaluating the elements of the pitch-speed-power matrix in which elements of the distance-speed-power matrix can be uniformly evaluated in a first dimension by means of the evaluation device 20, and elements of the distance-speed-power-matrix in a second dimension can be mathematically defined and evaluated in a defined manner by means of the evaluation device 20. Advantageously, the radar device 100 may be formed as a conventional electronic hardware, wherein by means of software, a configuration of the hardware can be performed, for example, an implementation of the Fast Fourier transform and other radar signal processing steps. Alternatively, the method may also be fully implemented in software. Advantageously, the inventive method can also be used for other applications, such as an ultrasonic device, if a chirp sequence modulation is used. The person skilled in the art can suitably modify and combine the described features of the invention without departing from the essence of the invention.

Claims

A method of operating a radar apparatus (100), comprising the steps of:

Determining a matrix with time signals of reflected radar radiation;

Determining elements of a distance-speed-power matrix of a radar target from the time signals;

Performing a first discrete one-dimensional Fourier transform for the elements of the distance-speed-power matrix in a first dimension; and

Performing a second discrete one-dimensional Fourier transform for the elements of the distance-speed-power matrix in a second dimension such that the second discrete one-dimensional Fourier transform is performed mathematically defined offset for every other element of the distance-speed-power matrix.

The method of claim 1, wherein the elements of the distance-speed power matrix are generated by chirp-sequence modulation.

The method of claim 1 or 2, wherein first the discrete

Fourier transform is performed for all elements of columns of the distance-speed power matrix, wherein the discrete Fourier transform is performed thereafter for all elements of lines of the distance-speed-power-matrix.

The method of claim 1 or 2, wherein first the discrete

Fourier transform is performed for all elements of lines of the distance-speed-power-matrix, wherein the discrete Fourier transform is then performed for all elements of columns of the distance-speed power matrix.

A method according to any one of the preceding claims, wherein each element of each second row of the distance-speed-power matrix is multiplied by the factor e ^{jM / N} , with the parameters: n column-number of the distance-speed-power-matrix

N .... The total number of columns of the distance-speed-power matrix

6. Radar apparatus (100), comprising:

generating means (10) for generating elements of a distance-speed-power matrix of a radar target (200); and

an evaluation device (20) for evaluating the elements of the distance-speed-power matrix, wherein by means of the evaluation device (20) elements of the distance-speed-power-matrix are equally evaluable in a first dimension, and wherein by means of the evaluation device ( 20) elements of the distance-speed-power-matrix in a second dimension mathematically defined evaluated are evaluable.

7. Radar device (100) according to claim 6, wherein by means of the generating means (10) is a chirp sequence modulation feasible.

A radar apparatus (100) according to claim 6 or 7, wherein, by means of said evaluating means (20), each element of every second row of the distance-speed-power matrix is multiply by the factor e ^{jM / N} , with the parameters: n column number of the distance -Geschwindigkeits-performance matrix

N .... Total number of columns of distance-speed-power matrix

9. Computer program product with program code means for carrying out the method according to one of claims 1 to 5, when it is stored on an electronic radar device (100) runs out or is stored on a computer readable medium.