Methods for analyzing and processing radar signals in radar systems with multiple sensor units, and radar systems thereof.

Abstract

A method for analyzing and processing radar signals in a radar system (10) having multiple sensor units (12), each capable of detecting objects in the environment surrounding the radar system, wherein the detection areas of the sensor units (12) at least overlap, characterized in that each sensor unit (12) calculates a short message (16) from the radar signals it receives, the short message having a smaller data volume than a complete detection result, but the short message containing at least the following data: the data allows for the identification of potential objects and the determination of the probability that the potential object involves a real object; at least one analysis and processing unit (14) receives all the short messages (16) from all sensor units (12) and calculates a merged detection result based on the short messages and selected detection results of the sensor units (12). The invention also relates to a corresponding radar system.

Description

Technical Field

[0001] The present invention relates to a method for analyzing and processing radar signals in a radar system having multiple sensor units, each capable of detecting objects in the surrounding environment of the radar system, wherein the detection areas of the sensor units at least overlap with each other.

[0002] This invention relates particularly to radar systems for motor vehicles. Such radar systems are used to detect objects in the vehicle's surrounding environment and provide location data about said objects for various driver assistance or safety functions (such as automatic distance adjustment or automatic emergency braking).

[0003] Each sensor unit has an analog high-frequency section with one or more antennas, the analog high-frequency section being configured to transmit radar signals and receive radar echoes reflected at an object. Background Technology

[0004] In motor vehicles, FMCW (Frequency Modulated Continuous Wave) radar systems are commonly used, where the transmitted signal is modulated at a ramp-like frequency. The received signal is then mixed with a portion of the transmitted signal transmitted at the same time, resulting in a beat signal (Schwebungssignal), the frequency of which 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 to the sensor unit, as well as the relative velocity of the object. The beat signal is recorded and digitized separately over the duration of the measurement cycle.

[0005] In the low-frequency range of the sensor unit, the digital signal undergoes pre-analysis processing. Specifically, the spectrum of the beat signal is calculated using a Fast Fourier Transform. In this spectrum, each located object appears as a peak at a specific frequency. The frequency position of this peak indicates information about the object's distance and relative velocity. By analyzing the peaks obtained from the same object at different frequency ramp slopes, distance and velocity information can be separated from each other.

[0006] Another commonly used implementation uses multiple frequency ramps with the same ramp slope, which are time-oriented as a sequence (also known as a linear frequency modulation sequence). A two-dimensional spectrum is derived by spectral analysis of each individual ramp (“fast-time”) and all ramps of the sequence (“slow-time”).

[0007] Furthermore, digital modulation waveforms can also be considered as an alternative to the classic FMCW-based modulation method. This digital modulation radar system naturally has advantages in terms of high-frequency structure. While there are some differences, most of the signal processing can be preserved. For example, the two-dimensional spectrum can still be calculated after the corresponding demodulation of the signal. Subsequent signal analysis and processing can therefore be performed in the same way as in the case of analog linear frequency modulation sequences.

[0008] In most cases, each sensor unit has multiple receiving channels that analyze and process signals from multiple spatially offset receiving antennas. Based on the amplitude and phase relationships between the signals received by the different antennas, the positioning angle of the object in azimuth and / or elevation can then be determined.

[0009] Because the detection areas of different sensor units overlap or are the same, a single object is usually detected by multiple sensor units, and ideally even by all sensor units.

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

[0011] This type of chip is also known as a SoC (System-on-Chip). RFCMOS technology, with its small structure size (e.g., 22nm), enables the development of increasingly complex SoCs for radar systems. This integration leads to significantly lower power consumption and lower cost.

[0012] Ideally, this chip integrates all necessary circuitry (such as transmit and receive paths, transmit control, analog-to-digital conversion, and digital signal processing) into a single unit. However, in practice, the number of transmit / receive channels per chip is limited. This is primarily due to chip heat dissipation issues, crosstalk between channels, and limitations in pin count and package size. For this reason, typically no more than four transmit and receive antennas are implemented within a single chip. For systems requiring a larger number of antennas, multiple sensor units, each constructed via a SoC, must be networked together.

[0013] In signal processing, the spectrum of all received channels is typically considered. These raw data spectra are integrated either coherently or incoherently, and threshold probing is performed on the resulting spectrum (MARichards, "Noncoherent integration gain, and its approximation," Georgia Institute of Technology, Technical Report, June 2010).

[0014] For optimal detection results, information from all available channels of the system should always be considered, even if this information is distributed across multiple SoCs. However, this presents a problem: data from all SoCs must be sent to the central processing unit. For modern radar sensors, the amount of data that must be transmitted can easily be in the range of hundreds to thousands of gigabits per second. This data transmission is problematic due to increased power loss, negative impact on analog performance, and increased cost for additional circuitry, connectors, and data lines.

[0015] To reduce the amount of data to be transmitted, a method proposed in US2016018511A1 involves performing threshold probing only based on the channels available in each SoC. Since this method integrates the probe spectrum only on a portion of the channels, the targets contained in this spectrum have a poorer signal-to-noise ratio compared to analysis and processing on all channels.

[0016] Another drawback is that the amplitude of different intensities in different channels can only be analyzed in a flawed manner. Therefore, in the worst case, the target may be almost invisible and therefore undetectable in the first SoC under consideration, while in another SoC, the target has a very strong signal amplitude and can therefore be detected very well.

[0017] The cited US literature therefore proposes performing the detection in two phases. In the first phase, detection is performed on each SoC with a very small threshold, thereby reducing the probability of missing targets while increasing the false alarm rate (i.e., the number of targets falsely detected) as the threshold decreases. Each SoC sends its detected targets to the central processing unit, which then performs a second detection based on all channels. Now, for all targets detected as targets on all SoCs, the second detection can be performed with a more stringent threshold to obtain the final reflection list. Summary of the Invention

[0018] The objective of this invention is to achieve more reliable object detection with a low data volume of communication between different components of the system.

[0019] According to the present invention, this task is solved by a method for analyzing and processing radar signals in a radar system having multiple sensor units, each capable of detecting objects in the surrounding environment of the radar system, wherein the detection areas of the sensor units at least overlap with each other, wherein:

[0020] Each sensor unit calculates a short message (Kurznachricht) from the radar signals it receives. This short message contains less data than the complete detection result, but it includes at least the following data: data that allows identification of potential objects and allows determination of the probability that the potential object relates to a real object, and

[0021] At least one analysis and processing unit receives all short messages from all sensor units and calculates merged detection results based on the short messages and the selected detection results from the sensor units.

[0022] The main advantage of this method is that the first detection step is no longer based solely on data from a single sensor unit, but rather on data from all sensor units, thereby significantly improving detection reliability. This reduction in data volume is achieved by transmitting only a brief summary of the detection results (indicating potential objects and implicitly or explicitly stating the probability of each potential object's presence) to the analysis and processing unit, instead of the complete spectrum from all sensor units. These presence probabilities (which can naturally only be based on data from the respective sensor units) are then fused within the analysis and processing unit. This allows for the removal of only those potential objects whose presence is consistently assessed as impossible across all sensor units with improved accuracy. For the calculation of the merged detection results, only the portions of the spectrum formed across different sensor units that represent objects with a high probability of presence need to be considered. In this way, communication is limited to the transmission of data that actually contains useful information, while the portion of the spectrum containing only noise is not transmitted to the analysis and processing unit at all.

[0023] According to the present invention, this task is also solved by a radar system having a plurality of sensor units configured to transmit and receive radar signals and perform digital pre-analysis processing on the received signals, and the radar system having an analysis and processing mechanism communicating with the sensor units, wherein the sensor units and the analysis and processing mechanism are configured to implement the above-described method.

[0024] Advantageous configurations and extensions of the invention are described below.

[0025] In one embodiment, the analysis and processing mechanism is a central analysis and processing unit that receives short messages from all sensor units and requests additional data from the sensor units in order to calculate the merged detection results.

[0026] In another embodiment, one of the sensor units also constitutes the analysis and processing mechanism. For example, the function of this analysis and processing mechanism can be taken over by a sensor unit with fewer receiving channels than the other sensor units, thus enabling a more even load distribution across the sensor unit (SoC).

[0027] In another embodiment, the functions of the analysis and processing mechanism can also be distributed across different SoCs that constitute the sensor unit.

[0028] The communication network connecting the different components of a radar system does not necessarily require a star-shaped master / slave architecture where each sensor unit is connected to a central analysis and processing unit. Instead, it can be constructed through point-to-point connections between the components. In an advantageous implementation, the communication network has a chain or ring architecture, in which data is further transmitted from sensor unit to sensor unit. This allows for not only a reduction in wiring overhead but also the scaling of the radar system by adding other sensor units.

[0029] In addition to data used for object identification (such as distance and relative velocity), the short information calculated by the sensor unit can also include intermediate results accumulated in the signal analysis and processing within the associated sensor unit. Examples of such intermediate results include amplitude, detection flags, detection counters, local signal-to-noise ratio, etc., integrated across all channels of the associated sensor unit. To further reduce the amount of data, the intermediate results or, if necessary, the complete short information can be compressed. Attached Figure Description

[0030] The embodiments are further described below with reference to the accompanying drawings.

[0031] The attached diagram shows:

[0032] Figure 1 A block diagram of a radar system is shown, in which the method according to the present invention can be applied;

[0033] Figure 2 The basis for showing the state during the later step of the method. Figure 1 Radar system;

[0034] Figure 3 and Figure 4 A block diagram of a radar system according to another embodiment is shown; and

[0035] Figure 5 A tabular diagram showing examples of short messages and detection results when implementing the method according to the invention. Detailed Implementation

[0036] exist Figure 1 In the diagram, radar system 10 is shown as a block diagram. This radar system has three sensor units 12 and a central analysis and processing unit 14. Each sensor unit 12 is composed of a SoC (System-on-Chip), which integrates the high-frequency portion of the radar sensor with multiple receiving channels and the digital pre-analysis processing function of the received signals. The central analysis and processing unit 14 may be composed of a processor, which takes over the further analysis and processing of the signals pre-analyzed in the sensor units.

[0037] The communication network that connects the components of the radar system 10 to each other has a star-shaped master / slave architecture, with the analysis and processing unit 14 as the master.

[0038] As an example, it should be assumed that sensor unit 12 relates to an FMCW sensor unit. However, the method presented herein can also be implemented using linear frequency modulation sequences or digital modulation forms.

[0039] During each measurement cycle, each sensor unit 12 calculates a two-dimensional spectrum for each of its receiving channels. In this two-dimensional spectrum, one dimension represents the distance to the located object, and the other dimension represents the relative velocity of the located object. Each located object appears as a peak in the spectrum, which more or less protrudes significantly beyond the noise background, and the position of the peak in the spectrum indicates the distance and relative velocity of the relevant object.

[0040] In a radar system for motor vehicles, sensor units 12 can be mounted at different locations within the vehicle or arranged together on a common circuit board. Preferably, this mounting or arrangement ensures that the antenna elements of all sensor units together form a one-dimensional or two-dimensional array with a large aperture, enabling object detection with high angular resolution in azimuth and / or elevation. The detection area (i.e., the area in the vehicle's surrounding environment where objects can be detected) should be identical for all three sensor units 12 in this example, such that objects within this detection area must theoretically be "seen" by each sensor unit 12. In the analysis and processing unit 14, the angular position of the object can then be calculated at high resolution based on the complex amplitude of the signals received in all receiving channels of all three sensor units 12 for a given object.

[0041] In the case of objects that produce only relatively weak radar echoes, the peaks assigned to the object in the spectrum often stand out only slightly or not at all from the noise background. This makes the object undetectable by different sensor units 12 with equal clarity, and may only be detectable by one or two sensor units at all. If a certain signal deflection is determined at a specific location in the spectrum within a single sensor unit 12, it is therefore impossible to reliably determine whether the deflection relates to noise or a real object. Even when analyzing all the receiving channels of the relevant sensor units, some uncertainty remains. Only when the detection results of all sensor units 12 are considered in a correlated manner in the analysis and processing unit 14 can the real object be distinguished from the noise background with greater certainty.

[0042] However, in the method proposed herein, sensor unit 12 does not transmit its complete detection results (i.e., the complete two-dimensional spectrum of each receiving channel) to analysis and processing unit 14. Instead, in a first method step, each sensor unit 12 sends only a short message 16 to analysis and processing unit 14, which represents only a very shortened (compressed) summary of the detection results. In particular, even if sensor unit 12 has multiple receiving channels, it may be sufficient to register a single entry for each object in the short message 16. For example, the short message 16 includes a distance index and a velocity index, as well as a scalar quality metric, for each object located or assumed to be located by the sensor unit. The distance index and velocity index together indicate the location of a relevant peak in the spectrum, and the quality metric indicates the probability that the detected peak relates to a real object. The calculation of the quality metric may include information from multiple receiving channels. Methods for calculating the quality metric are known. For example, the quality metric may be calculated based on the peak height above a noise background (preferably averaged across all receiving channels) and / or based on the power integrated on the peak compared to the noise power and / or based on the peak quality (width). Only when the quality metric exceeds a certain threshold will the hypothetical or real object identified by its distance index and velocity index be recorded as the detected object in short message 16.

[0043] The analysis and processing unit 14 calculates the presence probability for 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 presence probability can be proportional to the sum of the quality metrics reported by the three sensor units.

[0044] In another step, the analysis processing unit 14 compares the probability of the presence of each real or hypothetical object with a threshold that is greater than the sum of the thresholds used in the sensor units 12 to determine whether the object should be reported or not. Thus, for example, only objects in all three sensor units 12 that just exceed the threshold are removed by the analysis processing unit 14 as non-existent.

[0045] In this strategy, a very small detection threshold can be used in a single sensor unit 12 to ensure that no important and relevant objects are missed. Then, by using a higher threshold in the analysis processing unit 14, the number of objects considered real is reduced to the actual metric.

[0046] Then, via feedback channel 18, the analysis and processing unit 14 sends a request to each of the sensor units 12 to transmit a segment of the two-dimensional spectrum containing the peaks belonging to that object, for each object assessed as real. Based on this segment of the spectrum, the analysis and processing unit 14 can then make a more accurate angle estimate for each object, and optionally also improve the accuracy of the measured object distance and relative velocity by suppressing statistical fluctuations through averaging the measurements from all three sensor units. Those portions of the spectrum recorded in each sensor unit that do not contain any real objects are not transmitted to the analysis and processing unit 14, thus reducing the amount of data and therefore the load on the communication network, without affecting the accuracy and reliability of the detection results.

[0047] Figure 2 The radar system 10 is shown during the implementation of the second step in such a manner that, at the request of the analysis and processing unit 14, the sensor unit 12 transmits a segment 20 of the spectrum to the analysis and processing unit 14 for each located and assessed real object, and the analysis and processing unit then calculates and outputs a merged detection result 22 based on the data, the merged detection result including distance data, relative velocity data, and angle data of those objects that have been assessed as real.

[0048] In a modified implementation, the method may be supplemented by at least one step in which the analysis processing unit 14 instructs the sensor unit 12, which has not yet seen the identified object, to repeat the spectral analysis processing at that location with a smaller threshold, and first transmits the result in the form of a modified short message. Then, the probability of the object's presence is calculated based on the modified short message.

[0049] Figure 3As another example, radar system 24 has three sensor units 12 that communicate with each other via a ring bus 26. Here, for better distinction, these sensor units are additionally designated as S1, S2, and S3. In radar system 24, the processor of sensor unit 12 also takes over at least some of the functions of: Figure 1 In this context, the analysis and processing unit 14 is configured for the aforementioned function. Figure 3 The first three steps of the analysis and processing method are shown. In the first step, sensor unit S1 sends a short message 16 to sensor unit S2. This short message has content K1. This content includes distance indices, relative velocity indices, and mass measurements of all objects detected by sensor unit S1. In sensor unit S2, this data can be compared with the detection results of sensor unit S2 itself. During the comparison, sensor unit S2 supplements all its own detections that have not yet been included in short message K1.

[0050] Furthermore, the quality metrics of all probes detected not only by S1 but also by S2 will be combined.

[0051] In the following text, the symbol “&” indicates the comparison of multiple short messages. Then, sensor unit S2 sends a short message containing content K1&K2 to sensor unit S3. This content includes the distance and angular velocity indices of all objects detected by at least one of sensor units S1 and S2, as well as the cumulative mass metric assigned to the objects by the sensor unit. In the following description, the mass metrics are combined using unweighted addition; however, arbitrary mathematical operations, such as weighted sums, products, or sums of logarithmic values, can also be considered.

[0052] Then, sensor unit S3 compares the contents K1 & K2 with its own detection results again and sends a short message containing 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 the object by the sensor unit. For the determination of the object's existence probability, the short message containing the contents K1 & K2 & K3 already represents the merged detection results. Sensor unit S1 uses this result to compare the sum of the quality measures with a higher threshold and removes objects with a sum below that threshold as non-existent objects.

[0053] Figure 4The method is illustrated in two additional steps, in which the detection results K1, K2, and K3 are further passed from sensor unit S1 to sensor unit S2 and then finally to sensor unit S3, thus ensuring that all three sensor units have the same information regarding the probability of the object's presence and can then perform further signal analysis processing based on their respective spectra. Optionally, the ring bus 24 can also be used to transmit segments of the spectrum from one sensor unit to the next, thus enabling at least one of the sensor units to perform further signal analysis processing based on its own spectrum. Figure 1 The complete analysis and processing is performed by the analysis and processing unit 14. However, alternatively, these analysis and processing functions can also be distributed to the processors of the individual sensor units S1, S2, and S3, for example, by dividing the workload among the three sensor units such that the objects to be further analyzed and processed are treated as real objects, thereby achieving a uniform load.

[0054] However, in another embodiment not shown, certain analysis and processing functions (such as angle estimation) may also be delegated to a central analysis and processing unit.

[0055] exist Figure 5 The table below shows the possible contents of the short messages exchanged in radar system 24. Each table contains the distance index of the located objects in the first column D, the relative velocity index in the second column V, and the mass metric Q in the third column. In this example, sensor unit S1 has located six objects A to F, so short message K1 consists of six rows.

[0056] Conversely, sensor unit S2 detects only four objects, resulting in only four lines in short message K2. In two lines (i.e., the second and fourth lines), the distance and relative velocity indices are the same as in the case of objects C and E in short message K1. These lines, or the objects they belong to, can therefore be identified as objects C and E. Conversely, for the distance and relative velocity indices in the remaining two lines, there are no corresponding data in short message K1, thus involving "new" objects seen only by sensor unit S2.

[0057] Short messages K1 & K2 merge short messages K1 and K2. For the two new objects G and H, two rows have been appended to short message K1. In addition, in column Q, the quality measures are added at objects C and E as seen by the two sensor units.

[0058] Sensor unit S3 detects four objects A, B, E and F that were also detected by sensor unit S1, as well as three other objects I, J and K that were not detected by either of the other two sensor units.

[0059] Short messages K1&K2&K3 merge the contents of short messages K1, K2, and K3. Therefore, the additional three rows for objects I, J, and K are appended to short messages K1 and K2, and in column Q, for objects A, B, E, and F, the quality metrics from K3 and K1&K2 are added together.

[0060] To construct the merged detection result, in this example, the threshold for the quality metric is set to 10 in sensor unit S3. Therefore, objects D, G, H, I, and K in K1, K2, and K3 that do not reach the threshold of 10 are removed as pseudo-objects. The merged short message K0 is obtained in this way, containing only data considered as real objects, and forms the basis for calculating the merged detection result.

Claims

1. A method for analyzing and processing radar signals in a radar system (10; 24) having a plurality of sensor units (12), each of which is capable of detecting objects in the surrounding environment of the radar system, wherein, The detection areas of the sensor units (12) at least overlap, characterized in that, Each sensor unit (12) calculates a short message (16) from the radar signals it receives. The amount of data in the short message is less than that of a complete detection result, but the short message contains at least the following data: the data allows for the identification of potential objects and allows for the determination of the probability that the potential object involves a real object, and At least one analysis processing unit receives all short messages (16) from all sensor units (12) and calculates a merged detection result (22) based on the short messages and the selected detection results of the sensor units (12). The sensor unit (12) assigns a quality metric Q to each potential object as a measure of the probability that the potential object involves a real object, and only records objects whose quality metric Q is higher than a determined threshold in the short message (16) of the sensor unit, and the merged detection result (22) includes only data about objects for which the cumulative quality metric Q of all sensor units (12) is higher than a determined threshold.

2. The method according to claim 1, wherein the analysis processing mechanism instructs the sensor unit (12) that has not detected an object that has been detected by at least one other sensor unit to lower the threshold and send a new short message based on the lower threshold.

3. The method according to any one of the preceding claims, wherein the analysis processing mechanism calculates the probability of existence for each potential object detected by at least one of the sensor units (12) based on the short message, determines whether a real object is involved, and requests data (20) from the sensor unit (12) for each real object that characterizes the object in more detail.

4. A radar system (10; 24) having a plurality of sensor units (12), each of the plurality of sensor units being configured to transmit and receive radar signals and to perform digital pre-analysis processing on the received signals, and the radar system having an analysis and processing mechanism communicating with the sensor units (12), characterized in that, The sensor unit (12) and the analysis and processing mechanism are configured to implement the method according to any one of claims 1 to 3.

5. The radar system (10; 24) according to claim 4, wherein the analysis and processing mechanism has a central analysis and processing unit (14).

6. The radar system (10; 24) according to claim 4 or 5, wherein the radar system has a communication network through which the sensor units (12) communicate with each other.

7. The radar system (10; 24) according to claim 6, wherein the communication network has a ring bus (26) in the radar system.

8. The radar system (10; 24) according to any one of claims 4 to 7, wherein at least one of the sensor units (12) is configured to take over at least some of the functions of the analysis and processing mechanism.

9. The radar system (10; 24) according to any one of claims 4 to 8, wherein the sensor unit (12) has a plurality of receiving channels and / or a plurality of transmitting channels respectively.

10. The radar system (10; 24) according to any one of claims 4 to 9, wherein the sensor unit (12) is a SoC chip in which functions for transmitting and receiving radar signals and functions for digital pre-analysis processing are integrated into a common component.

Images