Radar system and method thereof

Abstract

A radar system includes a plurality of Tx antennas and a plurality of Rx antennas electrically coupled to a transmit (Tx) circuit and a receive (Rx) circuit, and measurement processing circuitry electrically coupled to the Tx circuit and the Rx circuit configured to produce multiple-input multiple-output (MIMO) virtual array measurements of reflections to a target. Direction processing circuitry is configured to arrange the array measurements into a MIMO matrix, configured to compute a singular value decomposition (SVD) vector of the MIMO matrix. Direction processing circuitry is configured to calculate a direction of arrival (DOA) spectrum and a direction of departure (DOD) spectrum. The detection processing circuitry is configured to detect a target in the calculated DOA spectrum and DOD spectrum, and the data interface is configured to output estimated target angle information.

Description

Technical Field

[0001] Embodiments of the subject matter described herein generally relate to radar systems. Background Art

[0002] High-resolution direction-of-arrival and direction-of-departure estimation is increasingly used in radar systems.

[0003] High-resolution automotive radars typically require super-resolution angle estimation capabilities to image inseparable targets using the natural resolution provided by their multiple-input multiple-output (MIMO) virtual array aperture. Conventional super-resolution algorithms used in radar systems require uniform arrays in order to perform spatial smoothing on the autocorrelation matrix computed in a conventional radar system as a preprocessing step for subsequent super-resolution processing. Due to hardware resource constraints, the aperture formed by a uniform array is naturally smaller than that of a sparse array, resulting in poorer natural resolution. Therefore, there is a need for improved super-resolution processing of radar systems. Summary of the Invention

[0004] In one aspect, an embodiment of a radar system may include a plurality of transmit (Tx) antennas and a plurality of receive (Rx) antennas, a Tx circuit, and an Rx circuit, where the Tx circuit and the Rx circuit are electrically coupled to the Tx antennas and the Rx antennas. In an embodiment, a measurement processing circuitry may be electrically coupled to the Tx circuit and the Rx circuit. According to an embodiment, the Tx circuit and the Rx circuit may be configured to perform multiple-input multiple-output (MIMO) virtual array measurements on target reflections generated by the radar system. In an embodiment, a direction processing circuitry may be configured to arrange the virtual array measurement results into a MIMO matrix. In an embodiment, the direction processing circuitry may be configured to compute singular value decomposition (SVD) vectors of the MIMO matrix. According to an embodiment, the radar system may further include a direction processing circuitry, where the direction processing circuitry is configured to compute a direction-of-arrival (DOA) spectrum and a direction-of-departure (DOD) spectrum.

[0005] In an embodiment, the direction processing circuitry may be configured to compute a direction-of-arrival (DOA) spectrum and a direction-of-departure (DOD) spectrum.

[0006] In an embodiment, the plurality of Tx antennas may form a sparse MIMO virtual array geometry, and the plurality of Rx antennas may form a sparse MIMO virtual array geometry.

[0007] In an embodiment, the direction processing circuitry may be configured to compute a combined DOA and DOD angle spectrum by searching for steering vectors or calibration array angular response vectors that are orthogonal to the noise subspace of the DOA and DOD singular vectors.

[0008] According to an embodiment, a plurality of Tx antennas may form a non-uniform array, and a plurality of Rx antennas may form a non-uniform array.

[0009] In an embodiment, the MIMO matrix may be modeled as a sum of a noise matrix and a matrix product, the matrix product including a Tx steering vector matrix, a target amplitude matrix, and an Rx steering vector matrix.

[0010] According to an embodiment, left singular vectors from a first side of the SVD of the MIMO matrix may indicate Rx or direction-of-arrival singular vectors, and right singular vectors from a second side of the SVD of the MIMO matrix may indicate Tx or direction-of-departure singular vectors.

[0011] In an embodiment, the least significant left singular vector may be orthogonal to the Rx steering vector, and the least significant right singular vector may be orthogonal to the Tx steering vector.

[0012] In an embodiment, the least significant left singular vector may be orthogonal to the Tx steering vector, and the least significant right singular vector may be orthogonal to the Rx steering vector.

[0013] In an embodiment, the minimum of the number of the plurality of Tx antennas and the number of the plurality of Rx antennas exceeds a target number.

[0014] In an embodiment, the detection processing circuitry may be configured to detect targets in the calculated DOA spectrum and DOD spectrum.

[0015] According to an embodiment, the data interface may be configured to output the estimated target angle information.

[0016] In another aspect, an embodiment may include a method of processing multi-input multi-output (MIMO) virtual array measurements in a radar system. The method may include: obtaining, by the radar system, MIMO virtual array measurements regarding target reflections; arranging, by the array processing circuitry, the array measurements into a MIMO matrix; calculating, by the direction processing circuitry, a singular value decomposition (SVD) of the MIMO matrix; calculating, by the direction processing circuitry, a direction-of-arrival (DOA) spectrum and a direction-of-departure (DOD) spectrum; detecting, by the detection processing circuitry, targets in the calculated DOA and DOD spectra; and outputting the estimated target angles.

[0017] In an embodiment of the method, the MIMO matrix may include a sum of a matrix product of a Tx steering vector matrix, a target amplitude matrix, and an Rx steering vector matrix and a noise matrix.

[0018] In an embodiment of the method, calculating the DOA and DOD angle spectra may include searching for Rx and Tx steering vectors orthogonal to the noise subspace of the DOA and DOD singular vectors.

[0019] In an embodiment of the method, calculating the DOA spectrum and the DOD spectrum by the direction processing circuitry may include using the orthogonality between the target DOA steering vector and the DOA noise subspace, and the orthogonality between the target DOD steering vector and the DOD noise subspace.

[0020] In an embodiment of the method, calculating the DOA spectrum and the DOD spectrum may include evaluating the expression where is the Rx steering vector evaluated at elevation angle θ and azimuth U is the matrix of the lowest left singular vectors of the MIMO matrix, Σ w is the submatrix with signal components zeroed out and noise components retained, V H is the conjugate transpose matrix of the lowest right singular vectors of the MIMO matrix, and is the Tx steering vector evaluated at elevation angle θ and azimuth .

[0021] In an embodiment of the method, detecting a target in the calculated DOA and DOD spectra by the detection processing circuitry may include using apodization, where multiple algorithms are used to detect the target.

[0022] In an embodiment of the method, calculating the combined DOA and DOD spectrum may include calculating the left singular vectors indicating the Rx or direction-of-arrival singular vectors from the first side of the SVD of the MIMO matrix, and calculating the right singular vectors indicating the Tx or direction-of-departure singular vectors from the second side of the SVD of the MIMO matrix.

[0023] In an embodiment of the method, the lowest significant left singular vector may be orthogonal to the Rx steering vector, and the lowest significant right singular vector may be orthogonal to the Tx steering vector.

[0024] In an embodiment, the method may include finding the left singular vectors indicating the Rx or direction-of-arrival singular vectors from the first side of the SVD of the MIMO matrix, and finding the right singular vectors indicating the Tx or direction-of-departure singular vectors from the second side of the SVD of the MIMO matrix. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] A more complete understanding of the subject matter may be obtained by considering the following drawings, taken in conjunction with the detailed description and the claims, in which like reference numerals refer to like elements throughout the drawings.

[0026] Figure 1 is a block diagram of a multiple-input multiple-output (MIMO) radar system according to an embodiment;

[0027] Figure 2 is according to an embodiment of Figure 1Schematic diagram of the signal model of the MIMO matrix of the MIMO radar system;

[0028] Figure 3 It is a flowchart depicting the method according to the embodiment;

[0029] Figure 4 It is according to the embodiment Figure 1 Schematic diagram of the signal model and solution principle of the multiple signal classification (M 3 ) of the MIMO matrix of the MIMO radar system;

[0030] Figure 5A 、 5B 、5C, 5D are the simulation results of the 2-target MIMO sparse array scenario according to the embodiment;

[0031] Figure 6A and 6B are the simulation results of the 4-target MIMO sparse array scenario according to the embodiment;

[0032] Figure 7A 、 7B 、7C and 7D are Figure 6A and 6B Tx antenna, Rx antenna and MIMO virtual antenna X-Z positions of the simulation example of;

[0033] Figure 8A 、 8B and 8C are schematic diagrams of the sparse array scenario according to the embodiment;

[0034] Figure 9A 、 9B 、9C are the simulation results of the 4-target MIMO sparse array scenario according to the embodiment;

[0035] Figure 10A 、 10B 、10C and 10D are the simulation results of the 4-target MIMO sparse array scenario according to the embodiment;

[0036] Figure 11A 、 11B 、11C and 11D are the simulation results of the 4-target MIMO sparse array scenario according to the embodiment;

[0037] Figure 12A 、 12B 、12C and 12D are the simulation results of the 2-target MIMO sparse array scenario according to the embodiment; and

[0038] Figure 13A 、 13B 、13C and 13D are the simulation results of the 2-target MIMO sparse array scenario according to the embodiment. Detailed implementation mode

[0039] Super-resolution imaging radar technology is the basis for the safe and efficient operation of autonomous driving (AD) and more advanced advanced driver assistance systems (ADAS). In these systems, computational and algorithmic enhancements are employed to achieve the required angular resolution that is superior to the natural resolution provided by the physical aperture of the antenna array of the radar system or the virtual array aperture of multiple-input multiple-output (MIMO).

[0040] When applying conventional super-resolution processing to radar, spatial smoothing or Hankel / Toeplitz data matrix construction is necessary for decorrelating coherent signals so that targets are separable in the eigen-space of the data matrix generated by the radar system. However, these operations require a uniform array geometry, so conventional super-resolution processing and associated benefits are not applicable to radars employing MIMO sparse array designs. There is a need for super-resolution radars and methods that are applicable to MIMO sparse array designs and can manage coherent signals and single-snapshot conditions.

[0041] The radar system and method embodiments provided herein can overcome some or all of the foregoing problems for use in radars performing super-resolution processing using the orthogonality between the noise subspace and the signal subspace. A radar system employing a new M 3 or MIMO matrix multiple signal classification super-resolution angle estimation processor and method is disclosed. Specifically, in some embodiments, the radar system and method embodiments described herein allow for MIMO sparse array designs and can handle coherent signals and single-snapshot conditions. According to an embodiment, the M 3 super-resolution angle estimation processor allows for super-resolution processing for any MIMO sparse array design. In an embodiment, a MIMO matrix (MM) is formed directly from the MIMO virtual array output. According to an embodiment, the formation of the MM allows for direct separation of targets in the decomposed left and right singular vectors, which in turn allows for super-resolution resolution of targets by employing the orthogonality between the noise subspace and the signal subspace of the left / right singular vectors, where the signal subspace is spanned by the Rx / Tx steering vectors towards the targets.

[0042] The following detailed description is merely illustrative in nature and is not intended to limit the embodiments of the subject matter or the application and use of such embodiments. As used herein, the words "exemplary" and "example" mean "serving as an example, instance, or illustration". Any embodiment described herein as exemplary or an example is not necessarily to be construed as preferred or advantageous over other embodiments. Furthermore, there is no intention to be bound by any theory expressed or implied in the foregoing technical field, background, or the following detailed description.

[0043] Figure 1 is M3 Block diagram of radar system 100 (also referred to herein as "radar system"). In an embodiment, M 3 Radar system 100 may include a plurality of transmit (Tx) antennas 110 and a plurality of receive (Rx) antennas 120, Tx circuitry 130 and Rx circuitry 140, where Tx circuitry 130 and Rx circuitry 140 are electrically coupled to Tx antennas 110 and Rx antennas 120. Processing circuitry 150 (also referred to herein as "M 3 super-resolution processor") may be coupled to Tx antennas 110 and Rx antennas 120. Processing circuitry 150 includes a radar controller 155, measurement circuitry 160, an array processor 170 (also referred to herein as "array processing circuitry") and a direction processor 180 (also referred to herein as "direction processing circuitry").

[0044] According to an embodiment, Tx circuitry 130 and Rx circuitry 140 may be configured to perform multiple-input multiple-output (MIMO) virtual array measurements on target reflections 122 generated by radar system 100. In an embodiment, direction processor 180 may be configured to arrange the array measurement results into a MIMO matrix.

[0045] Referring again to Figure 1 , in an embodiment, the plurality of Rx antennas 120 and Tx antennas 110 may be formed into an array with non-uniform and / or random spacing. In an embodiment, the spacing between some or all elements of the plurality of Rx antennas 120 and Tx antennas 110 may exceed one-half wavelength of the center frequency of the transmitted signal 111 of radar system 100. Additionally, according to an embodiment, the plurality of Tx antennas 110 and the plurality of Rx antennas 120 may form a sparse MIMO virtual array geometry. As used herein, "sparse array" means an antenna array having randomly placed individual antenna elements with any number of wavelengths or fractions of a wavelength between each antenna element.

[0046] According to an embodiment, the Tx circuit 130 and the Rx circuit 140 can be electrically coupled to a plurality of Tx antennas 110 and Rx antennas 120. In an embodiment, the chirp generator 132 can generate a series of frequency chirp pulses, where the center frequency of each pulse varies over time. The chirp generator 132 can be electrically coupled to the RF conditioning circuitry 134 within the Tx circuit 130. According to an embodiment, the RF conditioning circuitry 134 performs filtering (e.g., band-limiting, removing parasitic frequency components), applies phase rotation, and adjusts the input amplitude of the series of frequency chirp pulses to the power amplifier 136. Additionally, the local reference signal path 135 can be electrically coupled to the RF conditioning circuitry 134. The RF conditioning circuitry 134 can be electrically coupled to the power amplifier 136 in each Tx circuit 130. The power amplifier 136 boosts the power of the signal output by the RF conditioning circuitry 134 to provide sufficient power to transmit the transmitted signal 111 to the target 115 such that the signal 122 reflected by the target has sufficient amplitude to be received by the Rx antenna 120 and amplified and detected by the Rx circuit 140.

[0047] Each of the Rx circuits 140 can include a low-noise amplifier 143, a mixer device 144, a first filter network 145, an intermediate frequency (IF) amplifier 147, a second filter network 148, and an analog-to-digital converter (ADC) 149. According to an embodiment, the low-noise amplifier 143 can be electrically coupled to the Rx antenna 120. In an embodiment, the low-noise amplifier 143 can include a low-noise first amplifier stage and can also include additional gain stages to minimize the noise figure and generate sufficient power to provide a sufficiently high signal-to-noise ratio (SNR) for the Rx circuit 140. According to an embodiment, the output of the low-noise amplifier 143 can be electrically coupled to the mixer device 144. In an embodiment, the mixer device 144 can be electrically coupled to the local reference signal path 135 and electrically coupled to the RF conditioning circuit 134. The output of the mixer device 144 can be electrically coupled to the filter network 145. According to an embodiment, the first filter network 145 can be configured as a high-pass filter configured to remove an unwanted portion of the signal spectrum generated by the mixer device 144. According to an embodiment, the output of the filter network 145 can be electrically coupled to the IF amplifier 147. In an embodiment, the gain of the IF amplifier 147 can be adjusted to generate a signal level that the ADC 149 can effectively acquire. According to an embodiment, the second filter network 148 can be electrically coupled to the output of the IF amplifier 147. In an embodiment, the second filter network 148 can be configured as a low-pass filter. In an embodiment, the output of the second filter network 148 can be electrically coupled to the ADC 149. The output of the ADC 149 can be electrically coupled to the input 151 of the processing circuitry 150.

[0048] In an embodiment, and as previously stated, the processing circuitry 150 may include a radar controller 155, a measurement processor 160, an array processor 170, and an azimuth processor. It should be understood that the processing circuitry 150 may be implemented by components including, but not limited to, a microprocessor, a microcontroller, a field programmable reconfigurable gate array (FPGA), and an application specific integrated circuit (ASIC). Additionally, it should be understood that in an embodiment, each functional block of the processing circuitry 150 (e.g., the radar controller 155, the measurement processor 160, the array processor 170, and the azimuth processor 180) may represent different functions of the same overall processor unit. In other embodiments, some or all of these functional blocks of the processing circuitry 150 may include physically separate processing units.

[0049] The radar controller 155 may be electrically coupled to the chirp generator 132 and the Rx circuitry 140, and may receive control signals from the radar controller 155. According to an embodiment, the radar controller 155 controls the functions of the Tx circuit 130, the Rx circuit 140, and the measurement processor 160.

[0050] The measurement processor 160 may be electrically coupled to an input 151 of the processing circuitry 150, which receives digitized signals from the ADC 149. The measurement processor 160 may include interference mitigation circuitry 162, fast time (range) and slow time (Doppler) spectral processing blocks 164, 166, and a detection block 168. The detection block 168 may include constant false alarm rate (CFAR) processing.

[0051] According to an embodiment, the array processor 170 may be electrically coupled to the measurement processor 160. The array processor 170 receives data from the measurement processor 160 and may be used to perform MIMO demultiplexing processing, construct a MIMO virtual array measurement vector from range-Doppler data, and perform array calibration.

[0052] According to an embodiment, the azimuth processor 180 may be electrically coupled to the array processor 170. In an embodiment, the azimuth processor 180 may be configured to implement MIMO matrix multiple signal classification (M 3) An algorithm to detect targets in the calculated DOD spectrum and DOA spectrum. In an embodiment, the direction processor 180 may receive an array measurement vector from the array processor 170, and may be used to construct a MIMO matrix from the array measurement vector, and is configured to calculate the singular value decomposition (SVD) of the MIMO matrix, and may be configured to construct a DOD noise subspace and a DOA noise subspace based on the lowest left and right singular vectors, and calculate the combined DOD and DOA angle spectra by searching for steering vectors or calibrated array angular response vectors orthogonal to the noise subspaces of the DOD and DOA singular vectors. In an embodiment, the detection processor 182 may be configured to detect targets in the calculated DOD spectrum and DOA spectrum.

[0053] According to an embodiment, the data interface 190 may be configured to output the estimated target angle information. The data interface 190 may include various interfaces configured to convert the data received from the direction processor 180 into a format usable at other locations in the radar system 100 or in an external display, computing device, or processing unit (not shown).

[0054] Figure 2 is according to an embodiment Figure 1 Schematic of the signal model 200 of the MIMO matrix of the MIMO radar system according to an embodiment. According to an embodiment, the signal model 200 of the MIMO matrix of an exemplary 3 transmitter 4 receiver (3T4R) MIMO radar system in a 2-target scenario is depicted. In some embodiments, the radar system 100 may be configured such that the minimum of the number of multiple Tx antennas and the number of multiple Rx antennas may exceed the number of targets. In an embodiment, the phase delays of the radar signals 112 transmitted across 3 transmitters 230 to two targets (i.e., departure paths) are [ρ α,1 , ρ α,2 , ρ α,3 and [ρ β,1 , ρ β,2 , ρ β,3 . The phase delays of the reflected radar signals 222 received across 4 receivers 240 from two targets (i.e., arrival paths) are [p α,1 , p α,2 , p α,3 , p α,4 and [p β,1 , p β,2 , p β,3 , p β,4 .

[0055] In an embodiment, at the receiver 240, before MIMO demultiplexing, the signals originating from the transmitter 230 may not have been separated. The Rx array measurement vector It can be modeled using the following equation, where α and β are the complex reflection coefficients of the target and n1…n4 represent measurement noise:

[0056]

[0057] After MIMO demultiplexing by the array processor 170, the different signals from each transmitter can be separated and the virtual antenna array measurement vector has x i,j elements, where i = 1…4 indicating the Rx antenna index and j = 1…3 indicating the Tx antenna index can be modeled using the following equation, where the noise contained in the measurement result is represented by n i,j and where i and j follow the same rule:

[0058]

[0059] In an embodiment, the MIMO matrix can be modeled as the sum of a noise matrix and a matrix product that includes a Tx steering vector matrix, a target amplitude matrix, and an Rx steering vector matrix. Specifically, in an embodiment, X can be modeled using the following equation:

[0060] X = BΨA T + W.

[0061] Here, and in an embodiment, B is the Rx antenna array target steering vector whose columns correspond to the Rx array response to the target, A is the Tx antenna array target steering vector whose columns correspond to the Tx array response to the target, Ψ is the target amplitude matrix whose main diagonal indicates the reflection coefficient of the target, and W is the noise matrix whose elements indicate the MIMO virtual antenna measurement noise. In some embodiments, the roles of B as Rx and A as Tx can be swapped without loss of generality. According to an embodiment, the {i,j} element of the MIMO matrix X is x i,j . For the above example of a 2-target 3T4R scenario, and in an embodiment, the X equation can be explicitly expressed as follows:

[0062]

[0063] In an embodiment, the lowest effective left singular vector can be orthogonal to the Rx steering vector, and where the lowest effective right singular vector is orthogonal to the Tx steering vector. Specifically, and in an exemplary embodiment, there is a part contributing to the target (signal) of rank 2 (i.e., BΨA T ) and a part contributing to the noise of rank 3 (W). According to an embodiment, the resulting X is of rank 3 (full rank) and can be decomposed using singular value decomposition (SVD). According to an embodiment, when Figure 1When the power of the reflected signal 122 is significantly higher than the noise power, the decomposed singular values and vectors can include the two strongest target (signal) components and the weakest noise component.

[0064] According to an embodiment, the left singular vectors from the first side (i.e., the left side) of the SVD of the MIMO matrix can indicate the Rx or direction of arrival (DOA) singular vectors, and the right singular vectors from the second side (i.e., the right side) of the SVD of the MIMO matrix can indicate the Tx or direction of departure (DOD) singular vectors. Specifically, as shown in this 2-target 3T4R example and in the embodiment, the two strongest left singular vectors can be spanned by the Rx array steering vectors of the targets, and similarly, the two strongest right singular vectors can be spanned by the Tx array steering vectors of the targets. Thus, according to an embodiment, the weakest left singular vector can span the DOA noise subspace or null space of the Rx array steering vectors of the targets, and similarly, the weakest right singular vector can span the DOD noise subspace or null space of the Tx array steering vectors of the targets. Therefore, and in the embodiment, the Rx steering vector of any target can be orthogonal to the noise subspace of the left singular vectors, and the Tx steering vector of any target can be orthogonal to the noise subspace of the right singular vectors. According to an embodiment, the target angles can be estimated by employing this orthogonality. In the embodiment, when the DOA of a target is found, the inner product of the Rx steering vector of the target and the noise subspace of the left singular vectors can be approximately zero. Similarly, according to an embodiment, when the DOD of a target is found, the inner product of the Tx array steering vector of the target and the noise subspace of the right singular vectors can be approximately zero.

[0065] For illustrative purposes, the embodiments described above include linear arrays and relate to 1D angle estimation problems. The 1D formulas can be extended to any 2D MIMO sparse array for solving 2D angle (i.e., azimuth and elevation) problems, but are not limited thereto.

[0066] Figure 3 is a flowchart depicting a method 300 for processing multi-input multi-output (MIMO) virtual array measurements in a radar system. By looking at Figures 1 - 3 The method can be best understood. According to an embodiment, the formulas described in connection with the method are generalized for any 2D sparse array and for estimating the azimuth and elevation angle (θ).

[0067] Referring to Figure 1 and Figure 3 , at block 310, the method can include obtaining MIMO virtual array measurements of the reflected target signal 122 (i.e., "target reflection") by the radar system 100.

[0068] Referring to Figures 1 - 2and Figure 3 , frame 320, the method may include arranging the array measurement results into a MIMO matrix by the direction processor 180. In an embodiment of the method, the MIMO matrix may be modeled as the sum of the matrix product of a Tx steering vector matrix, a target amplitude matrix, and an Rx steering vector matrix and a noise matrix. Specifically, in an embodiment, arranging the array measurement results into a MIMO matrix using the direction processor 180 may include modeling the MIMO virtual array measurement results in the form X = BΨA T +W, where B is the target angle RX steering vector matrix, A is the target angle TX steering vector matrix, Ψ is the diagonal target amplitude matrix, and W is the measurement noise matrix. In an embodiment, where N r ×N t matrix X measures q targets, preferably such that q < N to allow separable signal and noise subspaces, where N = min{N r , N t}}.

[0069] Reference Figures 1 - 2 and Figure 3 , frame 330, an embodiment of the method may additionally include calculating the singular value decomposition (SVD) of the MIMO matrix X by the direction processor 180, as given by X = U∑V H , where U is the first orthogonal matrix containing the left singular vectors of X (i.e., "the first side of the SVD"), V H is the conjugate transpose of the second orthogonal matrix containing the right singular vectors of X (i.e., "the second side of the SVD"), and Σ is the diagonal matrix of the singular values of X.

[0070] Reference Figure 2 and Figure 3 , frame 340, an embodiment of the method may additionally include calculating or finding the left singular vectors indicating the Rx or direction of arrival singular vectors from the first side of the SVD of the MIMO matrix, and calculating or finding the right singular vectors indicating the Tx or direction of departure singular vectors from the second side of the SVD of the MIMO matrix. In an embodiment, the least significant left singular vectors may be orthogonal to the Rx steering vectors, and the least significant right singular vectors may be orthogonal to the Tx steering vectors. According to an embodiment, calculating the DOA and DOD angle spectra may include searching for the Rx and Tx steering vectors orthogonal to the noise subspaces of the DOA and DOD singular vectors. Specifically, and in an embodiment, the least significant N r -q left singular vectors u (u i ∈{u q+1 , u q+2 , …, u Nr}) and u i are orthogonal to B, i.e., where to correspond to the RX steering vector. In an embodiment, the method may further include finding the lowest significant N t -q right singular vectors v (v i ∈ {v q+1 , v q+2 , …, v Nt}) and being orthogonal to A, i.e., where is the TX steering vector corresponding to .

[0071] Referring to Figure 2 and Figure 3 , block 350, an embodiment of the method may further include determining the DOA spectrum and the DOD spectrum by the direction processor 180 by searching for vectors in a group consisting of steering vectors and calibration array angular response vectors that are orthogonal to the noise subspaces of the direction-of-arrival (DOA) singular vectors and the direction-of-departure (DOD) singular vectors in the SVD of the MIMO matrix. In an embodiment, a consistent DOA-DOD solution may include:

[0072] a.

[0073] b.

[0074] c. where ∑ w is the ∑ whose signal components are zeroed out (and the noise components are retained)

[0075] d. where ∑ n is the submatrix corresponding to the noise components of ∑ (removing columns and rows of non-noise main diagonal elements), and U n and V n are the columns corresponding to the noise components of U and V (removing non-noise columns).

[0076] Referring to Figure 1 , Figure 2 and Figure 3 , block 360, an embodiment of the method may further include detecting a target by the detection processor 180 from the calculated DOA and DOD spectra.

[0077] Referring to Figure 1 , Figure 2 and Figure 3, frame 370, embodiments of the method may include outputting an estimated target angle. In an embodiment, the detection processor 180 may send target information, such as a target angle, to the data interface 190.

[0078] In some embodiments, the main computational burden is caused by the singular value decomposition (SVD) of the MIMO matrix processing step and the grid-based angle spectrum search processing step. In an embodiment of an N transmitter N receiver (NTNR) MIMO system, the size of the matrix for which SVD decomposition can be used is N×N, so the complexity is approximately N^3 complex multiply-accumulate operations (MAC). For each evaluation of the angle search, N element-by-element multiplications may be required, followed by a summation. Assuming there are K azimuth points and L elevation points, considering the Tx and Rx arrays, the complexity is approximately 2*N*K*L complex MACs. Therefore, the complexity for generating the M 3 angle spectrum is controlled by 2*N*K*L + N^3. In some embodiments, the computational complexity can be reduced to 4*N*K + N^3 = N*(4K + N^2). Typical values of K are in the low hundreds, and N values are between 4 and 16. Overall, according to the embodiments, compared with other sparse array estimation methods, the M 3 computational burden of super-resolution processing may be lower.

[0079] Figure 4 is a schematic diagram of the signal model 400 and the solution principle 410 of the MIMO matrix multiple signal classification (M 3 ) of a 4T4R MIMO radar system according to an embodiment, which shows the usage of the radar system 100 and the method 300 described in combination Figures 1 to 3 . In an embodiment, SVD is performed on the MIMO matrix 412, and then the right and left singular vector noise subspaces are used to correlate with the DOA and DOD steering vectors, and the angles of the targets are found by calculating the angle spectrum 420 and observing the peaks 422, 424 in the angle spectrum 420.

[0080] Figure 5A and 5B and the diagrams 500, 520 graphically depict the simulation results of the spectral amplitudes 502, 522 and the normalized spatial frequencies 504, 524 for three target MIMO sparse array scenarios according to an embodiment, and compare the results with Figure 5C and 5D the simulation results of the 16-element array in

[0081] Reference Figure 5A and 5B, in an embodiment, a 4T4R sparse MIMO array is simulated to resolve three randomly generated targets in the following two scenarios. In each of the two scenarios, according to the embodiment, the criterion is used to generate the angle spectra 508, 528 based on the orthogonality between the DOA steering vectors and the decomposed DOA singular vectors spanning the DOA noise subspace, and the orthogonality between the DOD steering vectors and the decomposed DOD singular vectors spanning the DOD noise subspace. In an embodiment, the traces 506, 526 show the Tx spectra, and the traces 507, 527 show the Rx angle spectra for each of the two scenarios for comparison. The points 509, 529 show the true target angles and amplitudes for reference.

[0082] Reference Figure 5C , 5D , FIGS. 540, 560 show the amplitude spectra 542, 562 and the spatial frequencies 544, 564. The traces 546 and 566 show the results of a 16-element uniform linear array (ULA), and the traces 547, 567 show the beamforming results of a 4T4R sparse array. The actual target truth points 549, 569 are shown for reference. The spectra plotted using the conventional multiple signal classification (MUSIC) algorithm for a 16-element ULA (with forward and backward spatial smoothing) are also plotted in the traces 548, 568 for reference. It is noted that, according to the embodiment, Figure 5A , 5B the M 3 angle spectra of the traces 508, 528 remove the prominent ambiguities in the Tx and Rx angle spectra and provide higher resolution than the ULA beamforming in the traces 546, 566. In an embodiment, Figure 5A , 5B the M 3 angle spectra of the traces 508, 528 have performance comparable to that of the traces 548, 568 using ULA MUSIC. The advantages of the embodiments described herein are manifested in the flexibility of the array design, as conventional MUSIC cannot be applied to sparse arrays. In some embodiments, for the case of a two-dimensional (2D) array, especially for a 2D quasi-random sparse array using a limited number of Tx and Rx elements, the ability to use a sparse array becomes prominent.

[0083] Figure 6A and 6B depict the simulation results and the estimated target angle errors for a 4-target MIMO sparse array scenario according to an embodiment. For this purpose, according to the embodiment, a 6T6R MIMO sparse linear array is simulated to resolve four targets with varying spatial frequencies (angles) as well as different amplitudes and random phases. Figure 6A, Diagram 600 depicts the simulated intensity and angle 604 at incremental time samples 602 of a sparse array using only beamforming. Figure 6B , Diagram 610 depicts the simulated intensity and angle 614 at incremental time samples 612 of an M 3 super-resolution radar system. Figure 6A and 6B each horizontal row in depicts an angle scenario, as the angle of the target varies continuously between instances, starting from the true initial positions 606, 616. As Figure 6A shown, using the beamforming method produces only two traces 608 as the target angle changes. As Figure 6B shown, according to an embodiment, traces 618 corresponding to all four targets are generated by the change in the target angle. The results clearly demonstrate the effectiveness of M 3 in reducing target angle ambiguity.

[0084] In Figure 7A , 7B , 7C and 7D, the estimated target angle errors (compared to the true value) from all simulated instances represented in Figure 6A and 6B are combined and classified to form a cumulative probability distribution function (CDF). Figure 7A and 7B , according to an embodiment, diagrams 700 and 710 correspond to Figure 6A as well as Figure 7C and 7D beamforming results, diagrams 720 and 730 correspond to Figure 6B the M 3 super-resolution radar results. Figure 7A , in diagram 700, trace 703 shows that the 90th percentile error of the beamformer at point 704 is approximately 70 degrees. In contrast, Figure 7D , in diagram 730, trace 733 shows that the 90th percentile error of the M 3 super-resolution processor at point 734 is approximately 0.5 degrees. When observed over an axis range of 0 to 1 degree as Figure 7B shown, trace 713 in diagram 710 shows that the 50th percentile error of the beamformer at point 714 is approximately 0.4 degrees. Similarly, Figure 7D , when magnifying Figure 7D diagram 730 of, the axis range showing trace 733 is 0 to 1 degree, and the M 3 super-resolution processor shows a 50th percentile error of 0.1 degree at point 734. In Figure 7A , the 95th percentile error of the beamformer is approximately 95 degrees (point 706), while in diagram 720, at point 724 on trace 722, the M 3is 20 degrees. One way to measure performance is to observe the first inflection point, at which the error sharply turns from usually very low to very high. The error can only decrease when all targets are resolved and close to the true position. When any one target is missed and / or a wrong target is reported, the error increases sharply. Therefore, the inflection point is a reasonable metric for the percentage of instances in which all targets are correctly found. The inflection point of the beamformer-based radar occurs at Figure 7A at approximately the 54th percentile 709 in, and M 3 The inflection point of the super-resolution radar occurs at Figure 7C at the 90th percentile 729 in. Intuitively, it can be understood that the beamformer-based radar has a 50% chance of capturing all four targets, so the average number of targets that can be correctly resolved is 2 (i.e., 0.5 * 2). For M 3 super-resolution radar, the average number of targets that can be correctly resolved is 3.6 (i.e., 0.9 * 4). Therefore, the difference indicates better performance of M 3 super-resolution radar. In the case of correctly resolving targets, in an embodiment, M 3 super-resolution radar also produces a lower angle estimation error, as shown by the vertical position of the inflection point.

[0085] Figure 8A 、 8B and 8C depict the Tx antenna, Rx antenna, and MIMO virtual antenna X-Z positions of the simulation example discussed in conjunction with Figures 9A to 9C 、10A to 10D, 11A to 11D, 12A to 12D, and 13A to 13D. Figure 8A 、 8B and 8C show the X-Z positions of these antennas, the Tx position 815, the Rx position 825, and the MIMO virtual antenna position 835 in the corresponding diagrams 810, 820, 830. The Tx and Rx antenna positions are provided below:

[0086] Tx antenna position [λ]:

[0087]

[0088]

[0089] Rx Antenna Position [λ]:

[0090]

[0091] Figure 9A 、 9B9C shows the antenna gain patterns of the Rx array in illustration 900, the Tx array in illustration 910, and the full virtual array in illustration 920 according to an embodiment, all expressed in elevation and azimuth (in degrees). In all three illustrations 900, 910, 920, the sidelobe maxima 905, 915, 925 are minimized such that they are as low as possible, and the other sidelobes 907, 917, 927 are substantially uniform for a given desired aperture. In an embodiment, it is preferred to jointly optimize the Tx, Rx, and virtual arrays to produce low sidelobes in all three gain patterns.

[0092] Figure 10A 、 10B 、10C, 10D, 11A, 11B, 11C, 11D, 12A, 12B, 12C, 12D, 13A, 13B, 13C, and 13D are the simulation results of the 4-target MIMO sparse array scenario of Figure 8A 、 8B and 8C. An 8T8R 2D MIMO sparse array is simulated to resolve targets in the azimuth and elevation domains. According to an embodiment, several target examples are simulated, and the results are shown in these figures as intensity curves in decibels versus azimuth and elevation in degrees.

[0093] Figures 10A to 10D The simulation results are generated assuming four targets, a 20 dB minimum signal-to-noise ratio (SNR), and a 10 dB target dynamic range (DR). Under these SNR and DR conditions, Figure 10A , illustration 1000 depicts an intensity map 1001 of the elevation 1002 versus azimuth 1004 angle spectrum of a beamforming-based radar. The true target position points 1005 show the true positions of the four targets. Figure 10B , illustration 1020 depicts an intensity map 1021 of the elevation 1022 versus azimuth 1024 angle spectrum of an M 3 ultra-resolution radar. The true target position points 1025 (shown as crosses) show the true positions of the four targets. Figure 10C , illustration 1040 depicts an intensity map 1041 of elevation 1042 versus azimuth 1044. The true target position 1045 shows the true position, and the resolved position 1047 shows, in circles, the four targets resolved by the M 3 ultra-resolution radar. Figure 10D , illustration 1060 depicts the M 3 ultra-resolution radar target output plotted against elevation 1062 and azimuth 1064. According to an embodiment, the true target position 1065 shows the true position and the resolved position of the M 3 ultra-resolution radar with the four strongest peaks 1067 and the two next-highest peaks 1069.

[0094] Figures 11A to 11D The simulation results are generated assuming four targets, a minimum SNR of 15 dB, and a target DR of 10 dB. Under these SNR and DR conditions, Figure 11A , illustration 1100 depicts the intensity map 1101 of the elevation angle 1102 and azimuth angle 1104 angular spectrum of a beamforming-based radar. The true target location points 1105 show the true locations of the four targets. Figure 11B , illustration 1120 depicts the intensity map 1121 of the elevation angle 1122 and azimuth angle 1124 angular spectrum of an M 3 super-resolution radar. The true target location points 1125 (shown as crosses) show the true locations of the four targets. Figure 11C , illustration 1140 depicts the intensity map 1141 of the elevation angle 1142 and azimuth angle 1144. The true target location 1145 shows the true location, and the resolved location 1147 is shown as a circle for the four targets resolved by the M 3 super-resolution radar. Figure 11D , illustration 1160 depicts the M 3 super-resolution radar target output plotted against elevation angle 1162 and azimuth angle 1164. According to an embodiment, the true target location 1165 shows the true location and the resolved location using the M 3 super-resolution radar with the four strongest peaks 1167 and the two next-highest peaks 1169.

[0095] Figures 12A to 12D The simulation results are generated assuming two targets, a minimum SNR of 15 dB, and a target DR of 10 dB. Under these SNR and DR conditions, Figure 12A , illustration 1200 depicts the intensity map 1201 of the elevation angle 1202 and azimuth angle 1204 angular spectrum of a beamforming-based radar. The true target location points 1205 show the true locations of the four targets. Figure 12B , illustration 1220 depicts the intensity map 1221 of the elevation angle 1222 and azimuth angle 1224 angular spectrum of an M 3 super-resolution radar. The true target location points 1225 (shown as crosses) show the true locations of the four targets. Figure 12C , illustration 1240 depicts the intensity map 1241 of the elevation angle 1242 and azimuth angle 1244. The true target location 1245 shows the true location, and the resolved location 1247 is shown as a circle for the four targets resolved by the M 3 super-resolution radar. Figure 12D , illustration 1260 depicts the M 3 super-resolution radar target output plotted against elevation angle 1262 and azimuth angle 1264. According to an embodiment, the true target location 1265 shows the true location using the M3 The true and resolved positions of the four strongest peaks 1267 and the two next strongest peaks 1269 of the super-resolution radar.

[0096] Figures 13A to 13D The simulation results were generated assuming two targets, a minimum SNR of 15 dB, and a target DR of 10 dB. Under these SNR and DR conditions, Figure 13A , Diagram 1300 depicts the intensity map 1301 of the elevation angle 1302 and azimuth angle 1304 angle spectrum of a beamforming-based radar. The true target position points 1305 show the true positions of the four targets. Figure 13B , Diagram 1320 depicts the intensity map 1321 of the elevation angle 1322 and azimuth angle 1324 angle spectrum of the M 3 super-resolution radar. The true target position points 1325 (shown as crosses) show the true positions of the four targets. Figure 13C , Diagram 1340 depicts the intensity map 1341 of the elevation angle 1342 and azimuth angle 1344. The true target position 1345 shows the true position, and the resolved position 1347 is shown as a circle for the four targets resolved by the M 3 super-resolution radar. Figure 13D , Diagram 1360 depicts the M plotted at the elevation angle 1362 and azimuth angle 1364 3 super-resolution radar target output. According to an embodiment, the true target position 1365 shows the true and resolved positions of the four strongest peaks 1367 and the two next strongest peaks 1369 using the M 3 super-resolution radar.

[0097] The results show that for 4 targets, a higher SNR (20 dB) may be required to clearly separate the 4 targets. For 2 targets, a much lower SNR (10 dB) is required. The results clearly demonstrate the 2D super-resolution angle estimation ability using a true 2D MIMO sparse array with a quasi-random pattern in the X (right) and Z (up) dimensions. In an embodiment, as in these examples, an omnidirectional azimuth and elevation search can be performed within the assumed antenna field of view. According to an embodiment, an intelligent search strategy can be further applied to reduce the search space to accelerate the process and further reduce ambiguity. In an embodiment, the search space can be reduced by checking the observed radial velocity of the target against the self-velocity of the (radar) projected in the spatial position direction indicated by the searched azimuth and elevation steering angles and the range of the target. If the observed self-velocity is inconsistent with the projected self-velocity by more than a threshold, the search angles can be excluded for stationary objects. For moving objects, search elevation angles below the ground or above the maximum car height can be excluded.

[0098] Without departing from the scope of the inventive subject matter disclosed herein, M 3 The spectrum and method of the super-resolution radar can also be used in combination with other angular spectrum estimation algorithms. For example, according to an embodiment, using multi-angle spectra, a tapering method or a minimum pooling operation can be performed, where the magnitude of each support is determined based on the minimum magnitude supported by all algorithms. In some embodiments, the tapering method may be effective for spurious targets because they tend to be inconsistent between the outputs of different algorithms. In an embodiment, a weighted average between two or more spectra can be used to achieve better performance and may be effective for spurious targets.

[0099] For the sake of brevity, certain terms may also be used herein for reference purposes only, and thus these terms are not intended to be restrictive, and unless the context clearly indicates otherwise, the terms "first", "second" and other such numerical terms referring to structures do not imply order or sequence.

[0100] As used herein, a "node" means any internal or external reference point, connection point, junction point, signal line, conductive element, etc., at which a given signal, logic level, voltage, data pattern, current or quantity exists. In addition, two or more than two nodes can be implemented by one physical element (and although received or output at a common node, two or more signals can still be multiplexed, modulated or distinguished).

[0101] The foregoing description refers to elements or nodes or features being "connected" or "coupled" together. As used herein, unless otherwise explicitly stated, "connected" means that one element is directly joined to another element (or directly in communication with another element) and not necessarily in a mechanical manner. Similarly, unless otherwise explicitly stated, "coupled" means that one element is directly or indirectly joined to another element (or directly or indirectly in communication with another element) and not necessarily joined in a mechanical manner. Thus, while the schematic illustrations shown in the figures depict an exemplary arrangement of elements, additional intervening elements, devices, features or components may be present in embodiments of the depicted subject matter.

[0102] Although at least one exemplary embodiment has been presented in the foregoing detailed description, it should be understood that there are a large number of variations. It should also be understood that the exemplary embodiments described herein are not intended to limit in any way the scope, applicability or configuration of the claimed subject matter. Indeed, the foregoing detailed description will provide those skilled in the art with a convenient guide for implementing the described embodiments. It should be understood that various changes can be made to the functions and arrangements of the elements without departing from the scope defined by the claims, which scope includes known equivalents and foreseeable equivalents at the time of filing this patent application.

Claims

1. A radar system, characterized in that: include: multiple transmit (Tx) antennas and multiple receive (Rx) antennas; Tx circuitry and Rx circuitry, wherein the Tx circuitry and the Rx circuitry are electrically coupled to the Tx antenna and the Rx antenna; measurement processing circuitry electrically coupled to the Tx circuitry and the Rx circuitry, wherein the measurement processing circuitry is configured to generate a multiple-input multiple-output (MIMO) virtual array measurement of target reflectance to a target; as well as Direction processing circuitry, wherein the direction processing circuitry is configured to arrange the virtual array measurements into a MIMO matrix, and wherein the direction processing circuitry is further configured to compute a singular value decomposition (SVD) vector of the MIMO matrix.

2. The radar system according to claim 1, characterized in that The direction processing circuitry is configured to compute a direction of arrival (DOA) spectrum and a direction of departure (DOD) spectrum.

3. The radar system according to claim 2, characterized in that Additionally included is detection processing circuitry configured to detect targets in the calculated DOA and DOD spectra.

4. The radar system according to claim 2, characterized in that The direction processing circuitry computes the combined DOA and DOD angular spectrum by searching for a steering vector or calibration array angular response vector that is orthogonal to the noise subspace of the DOA and DOD singular vectors.

5. The radar system according to claim 1, characterized in that The plurality of Tx antennas form a sparse MIMO virtual array geometry, and the plurality of Rx antennas form a sparse MIMO virtual array geometry.

6. The radar system according to claim 1, characterized in that The plurality of Tx antennas form a non-uniform array, and the plurality of Rx antennas form a non-uniform array.

7. The radar system according to claim 1, characterized in that The MIMO matrix is ​​modeled as the sum of: Matrix product of the Tx steering vector matrix, the target amplitude matrix, and the Rx steering vector matrix; and Noise matrix.

8. The radar system according to claim 1, characterized in that The left singular vectors from a first side of the SVD of the MIMO matrix indicate Rx or arrival direction singular vectors, and the right singular vectors from a second side of the SVD of the MIMO matrix indicate Tx or departure direction singular vectors.

9. The radar system according to claim 1, characterized in that Also included is a data interface configured to output the estimated target angle information.

10. A method for processing multiple-input multiple-output (MIMO) virtual array measurement results in a radar system, characterized in that: The method comprises the following steps: Acquiring, by the radar system, a MIMO virtual array measurement result regarding target reflection; Arranging the array measurement results into a MIMO matrix by a directional processing circuit system; calculating, by the directional processing circuitry, a singular value decomposition (SVD) of the MIMO matrix; calculating a direction of destination (DOD) spectrum and a direction of arrival (DOA) spectrum by a direction processing circuit system; detecting, by the detection processing circuit system, a target in the calculated DOD and DOA spectrum; and Outputs the estimated target angle.

Images