Method for analyzing radar signals in a radar system with multiple sensor units - Patents.com

Abstract

To enable more reliable object detection with a small amount of communication data between different components.SOLUTION: A method for analyzing radar signals in a radar system 10 including multiple sensor units 12, which are each capable of detecting objects in the surroundings of the radar system, the detection areas of the sensor units at least overlapping one another. The method includes a step in which each sensor unit calculates a short message 16 from the radar signals received by it, whose data volume is smaller than the complete detection result, but which at least contains data (distance index, relative velocity index, quality scale) which allow for the identification of potential objects and for the determination of a probability that the potential object is a real object, and a step in which, on the basis of all of the short messages and selected detection results of the sensor units, at least one analysis unit 14 calculates a consolidated detection result.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a method for analyzing radar signals in a radar system comprising a plurality of sensor units each capable of detecting objects in the vicinity of the radar system, the detection areas of the sensor units at least overlapping each other.

[0002] In particular, the present invention relates to radar systems for vehicles, which serve to detect objects in the vehicle's surroundings and provide measurement data about the objects to various driver assistance or safety functions, such as automatic distance control or automatic emergency braking functions.

[0003] Each sensor unit has an analog radio frequency section with one or more antennas configured to transmit radar signals and receive radar echoes reflected from objects. [Background technology]

[0004] Vehicles commonly use FMCW (Frequency Modulated Continuous Wave) radar systems, in which the frequency of the transmitted signal is ramp modulated. The received signal is mixed with a portion of the simultaneously transmitted signal, resulting in a beat signal whose frequency corresponds to the frequency difference between the transmitted and received signals. This frequency difference depends on the propagation time of the signal from the sensor unit to the object and back, and the relative velocity of the object. The beat signal is recorded and digitized over the duration of each measurement cycle.

[0005] The low-frequency part of the sensor unit performs a pre-analysis of the digital signal. In particular, a fast Fourier transform is used to calculate the frequency spectrum of the beat signal. In this spectrum, each object to be located stands out as a peak at a specific frequency. The frequency position of this peak provides information about the object's distance and relative velocity. By analyzing peaks from the same object at different frequency ramps, distance and velocity information can be separated.

[0006] A more common implementation uses a time sequence of multiple frequency ramps with identical ramp slopes (also called a chirp sequence). A two-dimensional spectrum is obtained using spectral analysis of each individual ramp (fast time) and all ramps in the sequence (slow time).

[0007] Furthermore, instead of the classic FMCW-based modulation scheme, digitally modulated waveforms are also considered. While the high-frequency structure of such digitally modulated radar systems is naturally different, most of the signal processing remains the same. For example, after appropriate demodulation of the signal, a similar two-dimensional frequency spectrum can be calculated. Therefore, subsequent signal analysis can be performed in the same way as for analog chirp sequences.

[0008] Each sensor unit often has multiple receive channels that analyze signals from multiple spatially displaced receive antennas, and the positioning angle of an object in azimuth and / or elevation can be determined from the amplitude and phase relationships of the signals received at the different antennas.

[0009] Since the detection areas of the different sensor units overlap or are identical, an individual object is typically detected by multiple sensor units, ideally by all of the sensor units.

[0010] In conventional radar systems, the high-frequency or analog and digital sections of each sensor unit are implemented on separate integrated circuits (ICs). However, recent developments aim to create radar chips that integrate the analog and digital sections (J. Singh, B. Ginsburg, S. Rao, and K. Ramasubramanian: "AWR1642 mmWave sensor: 76-81-GHz radar-on-chip for short-range radar applications," White Paper SPYY006, Texas Instruments, Inc. (May 2017); "AWR1642 Single-Chip 77- and 79-GHz FMCW Radar Sensor," Datasheet SWRS203A - A Revision, Texas Instruments, Inc. (April 2018)).

[0011] Such chips are also called SoCs (System-on-Chip). For example, RFCMOS (radio-frequency complementary metal-oxide-semiconductor) technology with structures as small as 22 nm makes it possible to develop complex SoCs for radar systems. This integration allows for significantly lower power consumption and lower costs.

[0012] Ideally, a chip would integrate all necessary circuits, such as transmit / receive paths, transmission control, analog / digital conversion, and digital signal processing. However, in practice, the number of transmit / receive channels per chip is limited due to issues such as chip heat dissipation, crosstalk between channels, and limitations on pin count and package size. For this reason, a single chip typically implements a maximum of four transmit antennas and four receive antennas. Therefore, systems requiring a larger number of antennas require multiple sensor units, each fabricated on an SoC, to be networked together.

[0013] Signal processing typically uses the spectra of all received channels. These raw data spectra are integrated either coherently or noncoherently, and threshold detection is performed on the resulting spectrum (M.A. Richards, "Noncoherent integration gain, and its approximation," Georgia Institute of Technology, Tech. Rep. (June 2010)).

[0014] For optimal detection results, information from all available channels in the system must always be used, even if the information is distributed across multiple SoCs. However, this creates a problem: the data from all SoCs must be transmitted to a central processing unit. With current radar sensors, the amount of data that must be transmitted can be on the order of hundreds of megabytes to several gigabytes per second. This data transmission has several drawbacks, including increased power dissipation, adverse effects on analog performance, and increased costs due to the additional circuitry, pins, and data lines required.

[0015] To reduce the amount of data transmitted, U.S. Patent Application Publication No. 2016018511 proposes threshold detection based only on the channels available to each SoC. This method integrates the detection spectrum over only a portion of the channels, resulting in targets in this spectrum having a worse signal-to-noise ratio than if they were analyzed over all channels.

[0016] A further drawback is that the different strength amplitudes of the different channels can only be poorly resolved: in the worst case, a target may be barely visible in the first observed SoC and therefore not detected, while in other SoCs the target has a very strong signal amplitude and is therefore very well detected.

[0017] Therefore, the above-mentioned US document proposes a two-stage detection process. In the first stage, detection is performed by each SoC using a very low threshold, which reduces the probability of missing a target on the one hand, and increases the false alarm rate (i.e., the number of targets incorrectly detected) as the threshold decreases. Each SoC sends the detected targets to a central processing unit, which then performs a second detection based on all channels. For all targets detected as targets by all SoCs, a second detection is performed with a more restrictive threshold, thereby obtaining the final reflex list. [Prior art documents] [Patent documents]

[0018] [Patent Document 1] U.S. Patent Application Publication No. 2016018511 [Non-patent literature]

[0019] [Non-Patent Document 1] J. Singh, B. Ginsburg, S. Rao, and K. Ramasubramanian: "AWR1642 mmWave sensor: 76-81-GHz radar-on-chip for short-range radar applications", White Paper SPYY006, Texas Instruments, Inc. (May 2017) [Non-patent document 2] "AWR1642 Single-Chip 77- and 79-GHz FMCW Radar Sensor", Datasheet SWRS203A - A Revision, Texas Instruments, Inc. (April 2018) [Non-patent document 3] MA Richards, "Noncoherent integration gain, and its approximation," Georgia Institute of Technology, Tech. Rep. (June 2010) Summary of the Invention [Problem to be solved by the invention]

[0020] An object of the present invention is to enable more reliable object detection with a reduced amount of communication data between different components of the system. [Means for solving the problem]

[0021] This problem is solved according to the present invention as follows. - each sensor unit calculates a short message from the radar signal received by each sensor unit, the amount of data in the short message being smaller than the complete detection result, the short message containing at least data that makes it possible to identify a potential object and to determine the probability that the potential object is a real object. - at least one analysis unit (analysis instance) receives all the short messages of all the sensor units and calculates an integrated detection result based on the short messages of all the sensor units and selected detection results of the sensor units;

[0022] The essential advantage of this method is that the first detection step is based not only on data from one sensor unit but on data from all sensor units, significantly improving detection security. Data volume reduction is achieved by transmitting only a short outline of the detection results, rather than the complete spectra of all sensor units, to the analysis unit. These detection results indicate potential objects and implicitly or explicitly represent the probability of their presence. These probabilities, which essentially depend only on the data from each sensor unit, are integrated in the analysis unit. This allows for more accurate denial of only potential objects whose presence is unanimously assessed as unlikely by all sensor units. The calculation of the integrated detection result requires only the portions of the spectra generated by each sensor unit that represent objects with a high probability of presence. This method limits communication to data containing actual useful information, while spectral portions containing only noise signals are never transmitted to the analysis unit.

[0023] Advantageous embodiments and refinements of the invention are set forth in the dependent claims.

[0024] In one embodiment, the analysis part is a central analysis unit that receives the short messages of all sensor units and requests further data from the sensor units in order to calculate an integrated detection result.

[0025] In another embodiment, one of the sensor units also constitutes the analyzer, for example, by having a sensor unit with fewer receive channels than the other sensor units take over the function of the analyzer, thereby achieving equal utilization of the sensor units (SoCs).

[0026] In yet another embodiment, the functionality of the analysis unit can be distributed across multiple different SoCs that make up the sensor unit.

[0027] The communications network interconnecting the different components of a radar system need not be a star-shaped master / slave configuration connecting each sensor unit to a central analysis unit, but can instead consist of point-to-point connections between individual components. In an advantageous embodiment, the communications network has a chain or ring structure that transfers data from sensor unit to sensor unit. This not only reduces the amount of wiring but also makes it easier to expand the radar system by adding additional sensor units.

[0028] In addition to object identification data (e.g., distance and relative velocity), the short messages calculated by the sensor units can also contain intermediate results resulting from the signal analysis at the sensor unit. Examples of such intermediate results include the integrated amplitude for all channels of the sensor unit, detection flags, detection counters, local signal-to-noise ratios, etc. To further reduce the amount of data, the intermediate results and possibly the complete short messages can be compressed.

[0029] Hereinafter, the embodiments will be described in detail with reference to the drawings. [Brief explanation of the drawings]

[0030] [Figure 1] 1 is a block diagram of a radar system to which the method according to the invention can be applied; [Figure 2] 2 is a diagram of the radar system according to FIG. 1 in a state at a later step of the method; [Figure 3] FIG. 10 is a block diagram of a radar system according to another embodiment. [Figure 4] FIG. 10 is a block diagram of a radar system according to another embodiment. [Figure 5] 1 is a table showing an example of a short message and a detection result when the method according to the present invention is implemented; FIG. DETAILED DESCRIPTION OF THE INVENTION

[0031] 1 shows a block diagram of a radar system 10 with three sensor units 12 and one central analysis unit 14. Each sensor unit 12 is configured as a system-on-chip (SoC) that integrates the functions of the high-frequency part of a radar sensor with multiple receiving channels and the function of digitally pre-analyzing the received signals. The central analysis unit 14 can be configured as a processor that undertakes further analysis of the signals pre-analyzed by the sensor units.

[0032] The communication network interconnecting the components of the radar system 10 has a star-shaped master / slave configuration with the analysis unit 14 as the master.

[0033] As an example, the sensor unit 12 is assumed to be an FMCW sensor unit, but the method proposed here is equally feasible when using chirp sequences or digital modulation schemes.

[0034] Within each measurement cycle, each sensor unit 12 calculates for each of its receiving channels a two-dimensional spectrum, in which one dimension represents the distance of the measured object and the other dimension represents the relative velocity. Each measured object appears more or less clearly in this spectrum above the noise background and stands out as a peak whose position in the spectrum indicates the object's distance and relative velocity.

[0035] In a radar system for a vehicle, multiple sensor units 12 may be installed at different locations on the vehicle or may be co-located on a common substrate, but preferably the antenna elements of all sensor units together form a one- or two-dimensional array with a large aperture, which enables object detection with high angular resolution in azimuth and / or elevation. The detection region, i.e., the region around the vehicle in which objects can be detected, may be, and preferably is, the same for all three sensor units 12 in this example; therefore, objects within this detection region should theoretically be "seen" by each sensor unit 12. The analysis unit 14 can then calculate the angular position of the object with high resolution based on the complex amplitudes of the signals simultaneously received for a given object on all receive channels of all three sensor units 12.

[0036] For objects that produce relatively weak radar echoes, the peaks associated with these objects in the spectrum often only slightly or not at all distinguishable from the noise background, making them undetectable by several different sensor units 12, and in some cases, only by one or two of the sensor units. Therefore, if a signal distortion is detected at a given position in the spectrum by a single sensor unit 12, it is impossible to determine with certainty whether this distortion represents a noise signal or a real object. Even if all receiving channels of the sensor unit are analyzed, a certain degree of uncertainty remains. Only when the detection results of all sensor units 12 are considered together in the analysis unit 14 can real objects be identified from the noise background with greater certainty.

[0037] However, in the proposed method, the sensor units 12 do not transmit their complete detection results, i.e., the complete two-dimensional spectrum for each receiving channel, to the analysis unit 14. Instead, in a first method step, each sensor unit 12 transmits a short message 16 to the analysis unit 14, which represents only a very simplified (compressed) outline of its detection results. In particular, even if the sensor units 12 have multiple receiving channels, it may be sufficient to enter each object once in the short message 16. For example, the short message 16 may contain a distance index and a velocity index for each object measured or supposedly measured by the sensor unit, which together indicate the position of this peak in the spectrum, as well as a scalar quality measure indicating the probability that the detected peak is a real object. The calculation of this quality measure may take into account information from multiple receiving channels. Methods for calculating the quality measure are known. For example, the quality measure may be calculated based on the peak height of the peak above the noise background (preferably averaged over all receiving channels) and / or based on the integrated power of the peak relative to the noise power and / or based on the quality (width) of the peak. A putative or real object identified by its distance and velocity index is included in the short message 16 as a detected object only if the quality measure exceeds a certain threshold.

[0038] The analysis unit 14 calculates the probability of presence of each object detected by at least one of the sensor units 12 based on the short messages 16 from all three sensor units. For example, this probability of presence may be proportional to the sum of the quality measures reported by the three sensor units.

[0039] In a further step, analysis unit 14 compares the probability of presence of each real or putative object with a threshold that is greater than the sum of the thresholds used by sensor units 12 to decide whether the object should not be reported at all. For example, an object that is located only just above the thresholds in all three sensor units 12 will be discarded by analysis unit 14 as not present.

[0040] This strategy makes it possible to use very low detection thresholds in the individual sensor units 12 while still ensuring that relevant objects are not missed, and by using higher thresholds in the analysis unit 14, the number of objects considered to be real can be reduced to a practical extent.

[0041] Next, via the return channel 18, the analysis unit 14 sends a request to each of the plurality of sensor units 12, for each object determined to be real, to transmit a portion of the two-dimensional spectrum containing the peaks belonging to this object. Using these portions of the spectrum, the analysis unit 14 can then make a more accurate angle estimate for each object and, optionally, improve the accuracy of the measured object's distance and relative velocity by suppressing statistical fluctuations by averaging the measurements of all three sensor units. Portions of the spectra recorded by the individual sensor units that do not contain real objects are not transmitted to the analysis unit 14, thereby reducing the amount of data and thus the load on the communication network, without compromising the accuracy or reliability of the detection results.

[0042] FIG. 2 shows the radar system 10 during the second step, in which the sensor unit 12 transmits a portion 20 of the spectrum to the analysis unit 14 for each object that has been located and determined to be real, in response to a request from the analysis unit 14, and the analysis unit calculates and outputs an integrated detection result 22 based on this data, which includes distance data, relative velocity data, and angle data for the object that has been determined to be real.

[0043] In a modified embodiment, the method may be supplemented with at least one of the following steps: the analysis unit 14 instructs the sensor units 12 that do not see a particular object to repeat the analysis of the spectrum with a lower threshold and to first transmit the result in the form of a modified short message, on the basis of which the probability of the presence of this object is then calculated.

[0044] FIG. 3 shows another example of a radar system 24 with three sensor units 12 communicating with each other via a ring bus 26. Here, the sensor units are additionally labeled S1, S2, and S3 for better differentiation. In the radar system 24, the processors of the sensor units 12 also perform at least some of the functions performed by the analysis unit 14 in FIG. 1. FIG. 3 illustrates the first three steps of the analysis method. In the first step, sensor unit S1 sends a short message 16 to sensor unit S2. This short message has the content K1, which includes distance and relative velocity indexes of all objects detected by sensor unit S1, as well as a quality measure. Sensor unit S2 can compare this data with its own detection results. During the comparison, sensor unit S2 supplements all its own detection values ​​not previously included in the short message K1.

[0045] Furthermore, the quality measures of all detections detected in both S1 and S2 are integrated.

[0046] In the following, a comparison of multiple short messages is represented by the symbol "&". Sensor unit S2 then sends to sensor unit S3 a short message with the content K1&K2, which contains the distance and angular velocity indices of all objects detected by sensor unit S1, S2 or at least one of the sensor units, as well as the cumulative quality measures associated with these objects by the sensor units. In the following description, the quality measures are combined by unweighted addition, but any mathematical operation is also possible, such as a weighted sum, a product, or a sum of logarithmic values.

[0047] Sensor unit S3 then compares the contents K1&K2 with its own detection results and sends a short message with the contents K1&K2&K3 back to sensor unit S1. This short message K1&K2&K3 contains the distance index and relative velocity index of all objects detected by at least one of the three sensor units, as well as the sum of all three quality measures assigned to these objects by the sensor units. As far as determining the probability of object presence is concerned, the short message with the contents K1&K2&K3 represents the already combined detection result. Sensor unit S1 uses this result to compare the sum of the quality measures with a higher threshold and discards objects whose sum is below the threshold as non-existent.

[0048] 4 shows two further steps of the method, in which the detection results K1, K2, and K3 are transferred from sensor unit S1 to sensor unit S2 and finally to sensor unit S3, so that all three sensor units have the same level of information regarding the probability of an object's presence and can perform further signal analysis based on their own spectra. Optionally, the ring bus 24 can also be used to transmit parts of the spectra from one sensor unit to the next, allowing at least one of the sensor units to perform the full analysis, which in FIG. 1 is performed by analysis unit 14. However, these analysis functions can also be optionally distributed among the processors of the different sensor units S1, S2, and S3, for example, so that the processing load is evenly distributed among the three sensor units regarding the objects that are considered to be real and that require further analysis.

[0049] It should be noted that in other embodiments not shown, certain analysis functions, such as angle estimation, may be delegated to a central analysis unit.

[0050] 5, the possible contents of short messages exchanged by the radar system 24 are represented in the form of a table. Each table contains the distance index of the measured object in the first column D, the relative velocity index in the second column V, and the quality measure in the third column Q. In this example, the sensor unit S1 detected six objects A to F, so the short message K1 consists of six rows.

[0051] In contrast, sensor unit S2 only detected four objects, and therefore short message K2 consists of only four rows. The distance indexes and relative speed indexes in two rows, namely rows 2 and 4, are the same as those of objects C and E in short message K1. Therefore, these rows, or the related objects, can be identified as objects C and E. In contrast, the distance indexes and relative speed indexes in the remaining two rows have no equivalents in short message K1, and therefore these are "new" objects that were only seen by sensor unit S2.

[0052] The short message K1&K2 is the union of the short messages K1 and K2. For two new objects G and H, two rows are added to the short message K1. Also, for objects C and E seen by the two sensor units, quality measures are added to the column Q.

[0053] Sensor unit S3 detects the four objects A, B, E, and F that were detected by sensor unit S1, and three more objects I, J, and K that were not detected by the other two sensor units.

[0054] The short message K1&K2&K3 is the integration of the contents of the short messages K1, K2, and K3. As a result, three more rows are added to the short messages K1 and K2 for objects I, J, and K, and the quality measures of K3 and K1&K2 for each of the objects A, B, E, and F are added to column Q.

[0055] To form the integrated detection result, in this example the threshold of the quality measure is set to a value of 10 in the sensor unit S3. Therefore, objects D, G, H, I, K that do not reach this threshold of 10 in K1&K2&K3 are discarded as spurious objects. In this way, an integrated short message K0 is obtained that contains only data of objects that are considered to be real and forms the basis for calculating the integrated detection result.

Claims

1. 1. A method for analyzing radar signals in a radar system (10; 24) comprising a plurality of sensor units (12) each capable of detecting an object (A-K) in the vicinity of the radar system, comprising: The detection areas of the plurality of sensor units (12) at least overlap with each other, The method comprises: detecting, by each sensor unit (12) of the plurality of sensor units, one or more potential objects from the received radar signal of each sensor unit (12); each of said sensor units (12) assigning to each potential object of said one or more potential objects a quality measure Q as a measure of the probability that each potential object is a real object; calculating, by each of said sensor units (12), a short message (16) containing data enabling the sensor unit (12) to identify only those detected objects among said one or more potential objects having a quality measure Q equal to or greater than a predetermined detection threshold, and the quality measure Q assigned to said detected objects; and a step in which at least one analysis unit receives all short messages (16) of the plurality of sensor units (12) and calculates an integrated detection result (22) based on the short messages and selected detection results (20) of the sensor units (12).

2. The method described in claim 1, wherein the integrated detection result (22) includes only data of objects whose cumulative quality measure generated from the quality measures Q of all sensor units (12) is greater than or equal to a predetermined threshold.

3. 3. The method according to claim 2, wherein the analysis unit instructs a sensor unit (12) that has not detected an object detected by at least one other sensor unit to lower its detection threshold and to send a new short message based on the lowered detection threshold.

4. A method for analyzing radar signals in a radar system (10; 24) comprising a plurality of sensor units (12), each capable of detecting an object (A-K) in the vicinity of the radar system, comprising: The detection areas of the plurality of sensor units (12) at least overlap with each other, The method comprises: - calculating, by each sensor unit (12) of said plurality of sensor units, a short message (16) from the radar signal received by said each sensor unit (12), said short message (16) including at least data (D, V, Q) that allows to identify a potential object and to determine the probability that said potential object is a real object; and a step of at least one analysis unit receiving all short messages (16) of the plurality of sensor units (12) and calculating an integrated detection result (22) based on the short messages and selected detection results (20) of the sensor units (12), The method, wherein the at least one analysis unit functions as a central analysis unit and calculates, based on the short messages, a probability of presence for each potential object detected by at least one of the plurality of sensor units (12), determines whether each potential object is an actual object, and, for each actual object, requests data (20) from the sensor unit (12) that characterizes the object in more detail.

5. 1. A radar system (10; 24) comprising a plurality of sensor units (12) each configured to transmit and receive radar signals and to perform a digital pre-analysis of the received signals, and an analysis unit in communication with the sensor units (12), characterized in that the sensor units (12) and the analysis unit are configured to perform the method according to any one of claims 1 to 4.

6. The radar system (10) according to claim 5, wherein the analysis section comprises a central analysis unit (14).

7. 7. The radar system (24) of claim 5 or 6, comprising a communication network (26) through which the sensor units (12) communicate with each other.

8. The radar system (24) of claim 7, wherein the communication network comprises a ring bus (26).

9. 9. A radar system (10; 24) according to any one of claims 5 to 8, wherein each of the sensor units (12) has multiple receiving channels and / or multiple transmitting channels.

10. 10. The radar system (10; 24) according to any one of claims 5 to 9, wherein the sensor unit (12) is a SoC chip in which the functions of transmitting and receiving radar signals and the functions of digital pre-analysis are integrated into a common component.

Images