Solving multipath ambiguities in tdm-mimo radar

Abstract

The present application discloses a solution to multipath ambiguities in TDM-MIMO radars. A TDM MIMO FMCW radar comprises at least one row of physical receivers in a first direction with a first pitch (d r ) and further comprises a plurality of physical transmitters arranged in the first direction with a second pitch (d t ). To determine whether a peak in the angle spectrum corresponds to a direct reflection or to a first order multipath artifact, an inverse phase shift vector corresponding to the phase of the peak is applied and a constant signal with the amplitude of the peak is subtracted. For the intermediate signal (v) thus obtained, a further inverse phase shift vector - now corresponding to the shifted phase of the two front peak of the angle spectrum - is applied before a constant signal is subtracted. Then, it is detected whether the test signal (w) thus obtained has any non-noise content. If yes, the peak corresponds to a multipath artifact and, otherwise, the peak corresponds to a direct reflection.

Description

Technical Field

[0001] This disclosure relates to signal processing in time-division multiplexing (TDM) multiple-input multiple-output (MIMO) frequency-modulated continuous wave (FMCW) radar equipment. In particular, it proposes a method for resolving first-order multipath ambiguity that occurs in TDM-MIMO FMCW radar. Background Technology

[0002] A radar array can consist of a single physical transmitter and multiple physical receivers. The effective number of elements in a physical radar array equals the number of physical receivers. The number of elements determines the resolution of the radar array. For example, the angular resolution in angle-of-arrival (AoA) calculations increases with the number of elements in the radar array.

[0003] To increase the effective number of radar array elements, MIMO radar has been proposed. MIMO radar arrays have multiple physical receivers and M≥2 physical transmitters, resulting in arrays with M... r A virtual radar array of M elements, where M r It represents the number of physical receivers. Figure 1A An example setup with two physical transmitters 10 and four physical receivers 20 is shown. Figure 1B The resulting virtual array is shown, in which two subarrays 40 of eight virtual antenna elements 30 can be identified and traced back to the corresponding physical transmitters that generated them. Furthermore, within each subarray 40, as highlighted by the consistently used labels A, B, C, and D, the spatial configuration (e.g., geometry, orientation) of the virtual antenna elements 30 is identical to that of the physical receiver 20.

[0004] In MIMO radar, the physical transmitters can be synchronously fed using multicarrier signals such as orthogonal frequency division multiplexing (OFDM) signals. As an alternative, to limit the cost of the antenna structure and to ensure that all physical transmitters can be supplied from a common signal synthesizer, the concept of TDM MIMO radar, in which the physical transmitters are used in a time-alternating manner, has been proposed. Figure 5 The diagram illustrates the operation of a frequency-modulated continuous wave (FMCW) TDM MIMO radar with M=2 physical transmitters. Here, frequency is plotted relative to time, where solid and dashed lines represent chirps transmitted from the first and second physical transmitters, respectively. For the same transmitting antenna, the symbol T... c T represents the pulse signal length. r T represents the repetition time of the pulse signal, and T f =MT rThis indicates the repetition period of the pulse signal. The frequency axis does not necessarily start from the origin. In a typical millimeter-wave radar, each pulse signal scans from 77 GHz to 81 GHz, and the scan time T... c = 40μs duration.

[0005] While MIMO FMCW radars are popular due to their ability to achieve high angular resolution with a relatively small number of transmit and receive antennas, they perform poorly in certain known environments. For example, radars monitoring moving objects in scenarios with many static reflective surfaces (or scatterers such as metal fences, parked vehicles, building walls, and doors) can receive the expected echo along the line of sight of the moving target (target detection) and unwanted additional echoes from reflections from scatterers (multipath detection). Figure 6 In the diagram, the solid-line path RTR corresponds to target detection, while the dashed-line paths RTR and RSTR correspond to multipath detection. From a measurement perspective, echoes originating from multipath detection are undesirable, but they can also have negative secondary effects, for example, when a radar signal receiver mistakenly detects that a person in the scene has entered a prohibited area (when this is actually the person's echo) and triggers a costly false alarm, or when a target tracking algorithm is confused by an unlikely sudden movement (which is actually the echo of a target).

[0006] Patent applications WO2023021586A1 and WO2023021587A1 address the problem of distinguishing target detection from first-order multipath detection. Further reference is made to the research paper "Multipath Signal Mitigation for Indoor Positioning Based on MIMO FMCW Radar System" by Jeong-Ki Park, Jae-Hyun Park, and Kyung-Tae Kim, accepted for publication in the IEEE Internet of Things Journal.

[0007] US20200233076A1 discloses a method and apparatus for implementing time-frequency multiplexing for MIMO radar. A known drawback of some conventional MIMO radars is that reflected signals may leak into the cross-correlation window of the next antenna, thus appearing as phantom targets at a closer distance than the actual reflector. To suppress this phantom target, the suggestion in US20200233076A1 is to apply a slow-time phase coding scheme to scramble each pulse signal within each pulse signal period of a complete cyclic pulse signal period radar frame. Specifically, a random scrambling code is applied to each transmitted pulse signal over K pulse signal periods of the radar frame. The conjugate of the scrambling code is applied to the assumed range response of the transmitter, which allows the phase term of the signal to be recovered. Summary of the Invention

[0008] One object of this disclosure is to provide an efficient method for resolving the computational aspects of multipath ambiguity in virtual arrays of TDM MIMO radar. A further object is to propose such a method that can be executed without accessing dynamic information about the monitored scene being monitored by the radar. (It should be understood that the execution of the method may require some static information about the radar, as well as indications of the waveform and / or spectrum of the transmitted signal, such as the geometry of the radar's physical receiver and transmitter.) A further object is to propose a multipath ambiguity resolution method with good robustness, i.e., a method that can be reliably executed under a wide range of imaging conditions. A further object is to provide a signal processing apparatus and computer program for performing phase ambiguity resolution.

[0009] At least some of these objectives are achieved by the invention as defined in the independent claims. The dependent claims relate to advantageous embodiments of the invention.

[0010] In a first aspect of this disclosure, a method is provided for resolving first-order multipath ambiguity in a virtual array of a TDM MIMO FMCW radar. It should be understood that the TDM MIMO FMCW radar includes arrays with a first spacing d along a first direction. r At least one row of physical receivers and further including at a second spacing d in the first direction. t Multiple physical transmitters are deployed. The method includes: acquiring a scene-related virtual array signal x, where each element of the virtual array signal corresponds to a virtual antenna element of the virtual array; calculating the range-Doppler interval Z of the virtual array signal. (i,j) The method provides an angular spectrum; an intermediate signal v representing the first leading-edge peak in the angular spectrum (i.e., one of the leading-edge peaks, such as the largest peak or the second largest peak in the angular spectrum); a test signal w; and detection of the non-noise content of the test signal. At this point, if the test signal has non-noise content, it is inferred that the virtual array signal contains first-order multipath artifacts, and if the test signal has only noise content, it is inferred that the first leading-edge peak corresponds to a direct reflection in the scene. In the method, it is possible to select the first leading-edge peak in the angular spectrum of the virtual array signal, and then... (The sentence is incomplete and requires further context to translate accurately.) Corresponding inverse phase shift vector An intermediate signal v is provided by applying it to the Doppler range and subtracting a constant signal with an amplitude corresponding to the first leading edge peak. Furthermore, this can be achieved by adjusting the estimated offset phase... The corresponding inverse phase shift vector The intermediate signal is applied and a constant signal is subtracted to provide the test signal w, which is the estimated offset phase. It is related to the first and second leading-edge peaks in the angular spectrum of the virtual array signal (e.g., the two largest peaks in the angular spectrum).

[0011] As shown in the method, the inventors have developed a very simple test for determining whether a leading-edge peak is a first-order multipath artifact or a direct reflection. The only dynamic information required to perform the test is the phase difference of the leading-edge peak in the angular spectrum, which is relatively easy to obtain, i.e., requiring no extensive computation or non-standard input data. The inverse phase shift vector corresponding to this estimated offset phase is then calculated. When applied to intermediate signals, the result will be zero (plus noise) if the first leading-edge peak is an artifact, and non-zero if the first leading-edge peak is a direct reflection. These two results are mathematically demonstrated below. The direct technical advantage of the first approach is its ability to eliminate artifacts from radar signals. The second technical advantage is that radar-based object detection becomes more reliable (false alarms are suppressed). Furthermore, when the post-processed radar signal according to the first approach is used in control systems, control becomes more economical because less resources are wasted on false alarms and it may become more stable over time.

[0012] In some embodiments, the offset phase is calculated based on the angular spectrum of the virtual array signal x. This can include, for example, converting the angles θ1 and θ2 of the leading edge peak into two phases.

[0013]

[0014] Where λ is the wavelength, and the subtraction is performed:

[0015] Alternatively, in other embodiments, the offset phase is calculated based on the angular spectrum of the intermediate signal v. To calculate the phase difference between the two leading peaks.

[0016] More accurately, offset phase Corresponding to the estimated angle θ0 of the leading edge peak in the angular spectrum, such as:

[0017]

[0018] Assuming that the intermediate signal v is a virtual array signal, the angle θ0 will be interpreted as AoA of an object in the scene associated with the intermediate signal v. In these embodiments, the constant signal subtracted from the intermediate signal v may further have an amplitude corresponding to a leading-edge peak selected in the angular spectrum of the intermediate signal v.

[0019] In some embodiments, the detection of non-noise content includes a ratio test that compares the signal energy of a selected leading-edge peak to the total signal energy. Specifically, the ratio test may include a comparison of the signal energy (numerator) of a constant signal subtracted from the intermediate signal and the total signal energy (denominator) of the intermediate signal v. Alternatively, in other embodiments, the detection of non-noise content can be performed based on a noise floor acquired for the virtual array signal. This is an advantageous and simple way to determine the noise floor of the test signal, which is derived from the virtual array signal through a series of operations, based on the inventors' understanding, that can be expected to maintain a constant noise floor.

[0020] In a further development of the method in the first aspect, an offset phase was provided. With respect to the first spacing d of the physical receiver r AoA corresponds to a path length difference equal to an integer number of wavelengths, or to a second spacing d relative to the physical transmitter. t This results in an exception (degradation) corresponding to an AoD equal to an integer number of wavelengths of path length difference. The standard for an integer number of wavelengths can be expressed by formula d. r sinθ0=λn or d t sinθ0=λn′, where the left side represents the path length difference and n, Substituting into (1), these two conditions are considered equivalent to or If it is determined that the abnormal condition is about to occur, the execution of the method can be aborted. Alternatively, user notifications can be provided that the method cannot reliably resolve first-order multipath ambiguity.

[0021] In a second aspect of this disclosure, a method for resolving first-order or second-order multipath ambiguity is provided. The method includes: obtaining a scene-related virtual array signal x, each element of which corresponds to a virtual antenna element of a virtual array of a TDM MIMO FMCW radar as described above; calculating the angular spectrum of the virtual array signal over a range-Doppler interval; and counting the number of peaks in the angular spectrum. If the angular spectrum has a single peak, it is inferred that the angular spectrum corresponds to a second-order multipath artifact or direct reflection in the scene. Conversely, if the angular spectrum has at least two peaks, it is inferred that the angular spectrum corresponds to two first-order multipath artifacts or two direct reflections in the scene; using this result, the method of the first aspect is performed.

[0022] The second aspect allows for a more complete evaluation of virtual array signals by using the ability to distinguish direct reflections from first- and second-order multipath artifacts.

[0023] In a third aspect of this disclosure, a signal processing apparatus for a TDM MIMO FMCW radar having the above-described features is provided. The signal processing apparatus includes processing circuitry configured to resolve first-order multipath ambiguity of the virtual array of the TDM MIMO FMCW radar in a virtual array signal comprising at least one range Doppler interval by performing the method of the first aspect.

[0024] The signal processing apparatus according to the third aspect generally shares the advantages of the method of the first aspect and can be implemented with the same degree of technical change.

[0025] The present invention further relates to a computer program containing instructions for causing a computer or, in particular, a signal processing device to implement the methods described above. The computer program may be stored or distributed on a data carrier. As used herein, "data carrier" can be a non-transitory data carrier or a temporary data carrier such as modulated electromagnetic waves or light waves. Non-transitory data carriers include volatile and non-volatile memories such as permanent and non-permanent storage media of the magnetic, optical, or solid-state types. Still within the scope of "data carrier," such memory may be fixedly mounted or portable.

[0026] Generally, unless expressly defined herein, all terms used in the claims should be interpreted according to their ordinary meaning in the art. Unless otherwise expressly stated, all references to “a / the (described) element, apparatus, component, method, step, etc.” should be openly interpreted as referring to at least one instance of that element, apparatus, component, method, step, etc. Unless expressly stated otherwise, the steps of any method disclosed herein need not be performed in the exact order described. Attached Figure Description

[0027] Aspects and embodiments will now be described by way of example with reference to the accompanying drawings, in which:

[0028] Figure 1A and Figure 1B A one-dimensional array of physical transmitters and physical receivers, and the resulting one-dimensional virtual array, are shown.

[0029] Figure 2A and Figure 2B The diagram shows a two-dimensional array of physical transmitters, a one-dimensional array of physical receivers, and the resulting two-dimensional virtual array.

[0030] Figure 3A and Figure 3B The diagram shows a two-dimensional array of physical transmitters, a one-dimensional array of physical receivers, and the resulting two-dimensional virtual array.

[0031] Figure 4A and Figure 4BA two-dimensional array of physical transmitters and physical receivers, and the resulting two-dimensional virtual array, is shown.

[0032] Figure 5 It is a graph showing the relationship between the frequency and time of two physical transmitters in TDM operation;

[0033] Figure 6 The diagram illustrates object detection and first-order multipath detection.

[0034] Figure 7 This is a flowchart of a method for resolving first-order multipath ambiguity in a virtual array of a TDM MIMO FMCW radar according to embodiments herein;

[0035] Figure 8 This is a flowchart of a method for resolving first-order or second-order multipath ambiguity in a virtual array of a TDM MIMO FMCW radar according to embodiments herein;

[0036] Figure 9 It will be based on Figure 7 The method's steps are correlated with the input, output, and intermediate data in a signal processing diagram;

[0037] Figure 10 This is a graph of a 2×8 MIMO radar that has obtained virtual array signals (16 elements) after different processing steps. The left column indicates imaging of two targets by direct reflection, and the right column indicates two first-order multipath artifacts.

[0038] Figure 11 The angular spectra of a TDM MIMO FMCW radar imaging two targets are shown before processing (solid line) and after removing the leading edge peak (dashed line).

[0039] Figure 12 The image shows the angular spectra of a TDM MIMO FMCW radar before processing (solid line) and after removing the leading-edge peak (dashed line) for capturing two first-order multipath artifacts; and...

[0040] Figure 13 The diagram illustrates an exception (degeneration) that requires special measures to distinguish between object detection and first-order multipath detection. Detailed Implementation

[0041] In the following description, aspects of this disclosure will be more fully described with reference to the accompanying drawings, in which certain embodiments of the invention are illustrated. However, these aspects may be embodied in many different forms and should not be construed as limiting; rather, these embodiments are provided by way of example so that this disclosure will be comprehensive and complete, and will fully convey the scope of all aspects of the invention to those skilled in the art. Throughout the specification, the same reference numerals refer to the same elements.

[0042] Figure 1A A one-dimensional array of physical transmitters 10 configured to transmit an radio frequency (RF) beam 11 toward a scene is shown, along with a one-dimensional array of physical receivers 20 configured to receive an RF beam 21 reflected by an object (not shown) in the scene. It should be understood that, as enumerated starting from the antenna end, a complete radar device may include functional levels such as mixing, analog-to-digital conversion, RF front-end processing (based on the IF signal), and digital beamforming, in addition to the physical transmitters 10 and 20. Here, physical transmitters 10 are used according to the configured transmission schedule, corresponding to the repeating sequences TX1, TX2.

[0043] The reflecting object is observed at an angle θ relative to the main direction of transmission and reception (the main lobe, which corresponds to the vertical direction in the diagram). Angle θ corresponds to the object's angle AoA. To avoid ambiguity, the physical transmitter 10 is generally configured to emit in all directions (including the direction of angle θ) within a non-zero angular range, but is not limited to this. Furthermore, as... Figure 1A As shown, for a certain constant d, the physical transmitter 10 emits d... t = 4d units for spacing, and physical receiver 20 in d r = d units of equal spacing.

[0044] Referring to the attached patent claims, it should be noted that Figure 1A The physical transmitter 10 and the physical receiver 20 in the middle satisfy the common first direction ( Figure 1A The horizontal direction (in the middle) requires a non-zero spacing. Although Figure 1A The rows of physical transmitters 10 and physical receivers 20 are arranged parallel to each other, but this is not necessary for the applicability of method 700. In fact, the requirement is satisfied as long as the physical transmitters 10 and physical receivers 20 each have a non-zero component of spacing in a common first direction, for example, if the rows of physical receivers 20 are tilted upward relative to the rows of physical transmitters 10.

[0045] Figure 1B This illustrates the virtual array generated when the array of physical transmitters 10 and the array of physical receivers 20 are operated together. Proposed One-dimensional virtual array of array elements 30. During one pulse signal repetition period T f , for each active element 30The physical transmitter 10 generates a virtual array in the sense that it reads measurement data once from each physical receiver 20. (To recap, for example, to provide a Doppler FFT, any velocity calculation typically requires data from multiple repetitions of pulse signals. In other words, the data is collected from multiple pulse signals from each virtual array element.) The measurement data collected in this way (the virtual array signals) can be organized into a matrix with a dimension equal to the dimension of the virtual array. Figure 1B In the 1×8 case shown in the diagram, the matrix can take the following form:

[0046] X = [x TX1,A x TX1,B x TX1,C x TX1,D x TX2,A x TX2,B x Tx2,C x TX2,D ], (2)

[0047] Where x TX1,A This represents the measurement data read from the physical receiver 20, labeled A, when excited by the first physical transmitter 10 (TX1). TX2,A This refers to the measurement data read from the same physical receiver 20 when excited by the second physical transmitter 10 (TX2), and so on. It can be considered that... Figure 1B The virtual array in X is divided into two subarrays 40, each with a one-to-one relationship to the physical transmitter 10 that provides the excitation. Each measurement data entry in X can be, for example, a digital representation of an intermediate frequency (IF) signal obtained by mixing the signal fed to the physical transmitter 10 with the signal received from the physical receiver 20. The digital representation can be, for example, a row matrix of time samples. Furthermore, each entry in X can be a data structure that collects measurement data from multiple radar pulse signals; for example, the measurement data entry can be represented as a matrix, where the pulse signal corresponds to a row.

[0048] Apart from the frequency folding discussed below, the virtual array signal X is generally indistinguishable from the physical array signal collected by a 1×8 array of physical receivers excited by a single physical transmitter.

[0049] Within each subarray 40, the geometry and orientation of the array of physical receivers 20 are preserved, including their spacing d. rThis is visualized by using the same labels A, B, C, and D for the physical receiver 20 and the virtual antenna elements 30 of each subarray 40. Two virtual antenna elements 30 in different subarrays 40 generated by the same physical receiver 20 will be referred to as homogeneous in this disclosure. In the figures, two homogeneous virtual antenna elements 30 share the same label, for example, A. The spacing of the subarrays 40 is equal to the spacing of the physical transmitters 10, i.e., d in the first direction. t unit.

[0050] Reference Figure 2A and Figure 2B , Figure 3A and Figure 3B as well as Figure 4A and Figure 4B Examples are provided to briefly discuss the effects of using a two-dimensional array of physical transmitter 10 or a two-dimensional array of physical receiver 20, or both. As will be apparent from these examples, the virtual array generated by the array of physical transmitter 10 and the array of physical receiver 20 corresponds to the convolution of these two arrays. For a general introduction to the structure and operation of MIMO radar, please refer to “MIMO Radar”, Application Report SWRA554A, S. Rao, Texas Instruments, July 2018, Dallas, Texas.

[0051] exist Figure 2A In the array of physical emitters 10, there are two rows and two columns. The column spacing is d. t = 4d units (in the first horizontal direction on the attached diagram), and the line spacing is D. t Unit (in the vertical second direction in the attached diagram). The array of physical receivers 20 has a dimension of 1×4 and has d r = d units of spacing. Although Figure 2A The physical emitters 10 are arranged in two orthogonal directions, but such orthogonality is by no means necessary for the applicability of the invention. Rather, where applicable, a spacing with a non-zero component in either the first or second direction is sufficient.

[0052] exist Figure 2B The diagram shows the resulting virtual array with four subarrays 40, where the spacing between example elements is indicated. The spacing between consecutive virtual antenna elements within subarray 40 in the first direction is equal to the spacing d of the physical receivers 20. r Unit. The spacing d between the homogeneous virtual antenna elements 30 belonging to the subarray pair in the first direction. t One unit (in the TX1-TX2 pair and the TX3-TX4 pair) or zero (in the TX1-TX3 pair and the TX2-TX4 pair). For any pair of co-originating virtual antenna elements 30, the spacing relative to the second direction does not exceed d. tUnit. The spacing between the homogeneous virtual antenna elements 30 belonging to the subarray pair in the second direction is zero (in the TX1-TX2 pair and the TX3-TX4 pair) or D. t Each unit (in the TX1-TX3 pair and the TX2-TX4 pair). In either the first or second direction, the spacing between any pair of non-homogeneous virtual antenna elements can be calculated as a linear combination of these fundamental distances. For example, the distance between the second virtual antenna element B in the third subarray 40 (TX3) and the third virtual antenna element C in the fourth subarray 40 (TX4) is d. t +d r Units. Similarly, the distance between the third virtual antenna element C in the third subarray 40 (TX3) and the first virtual antenna element A in the fourth subarray 40 (TX4) is d. t -2d r Units.

[0053] use Figure 2B The virtual array signals collected by the virtual array in the image can be represented as a 2×8 matrix with the following general form:

[0054]

[0055] Alternatively, the matrix elements can be arranged in a single row. In this way, data from different pulse signals can correspond to different rows of the matrix.

[0056] exist Figure 3A In the middle, the physical transmitter 10 is in the first direction with Figure 2A The columns are arranged with different 2d unit column spacings, and the rows of the physical receiver 20 are spaced in the order of d, 3d, and d units. For example... Figure 3B As shown, despite having different subarray structures, the resulting virtual array will have the same characteristics as... Figure 2B The virtual arrays in the array have the same dimension of 2×8. The first subarray 40 (TX1) includes a first virtual antenna element 30, a second virtual antenna element 30, a fifth virtual antenna element 30, and a sixth virtual antenna element 30 in the first row of the virtual array; and the second subarray includes a third virtual antenna element, a fourth virtual antenna element, a seventh virtual antenna element, and an eighth virtual antenna element in the first row. The virtual antenna elements 30 in each row are equidistant and have a spacing d.

[0057] at last, Figure 4A This illustration shows a scenario where both the physical transmitter 10 and the physical receiver 20 are arranged in a two-dimensional pattern. More precisely, the physical transmitter 10 and the physical receiver 20 are arranged with non-zero spacing in a first direction (horizontal) and a second direction (vertical). Figure 4BThe resulting virtual array is shown in the figure. Here, to avoid unnecessary repetition, only the virtual array elements 30 in the first subarray 40 (TX1) are labeled with the letters A, B, C, ..., H, according to the physical receiver 20. It should be understood that this structure is repeated in the other seven subarrays 40, and therefore, the homology relationship is repeated in the other seven subarrays 40. Figure 4A The spacing d in the middle t d r D t D r It can be specified to any value. To make the virtual antenna elements 30 equidistant in the first direction, d can be set. t / d r =4. Similarly, if D t / D r If the value is 2, then the equidistant spacing in the second direction will be obtained.

[0058] Now refer to Figure 7 The flowchart in the document describes a method 700 for resolving first-order multipath ambiguity in the virtual array of a TDM MIMO FMCW radar. Method 700 takes the virtual array signal in the range-Doppler interval as input. This makes it possible to execute method 700 on a general-purpose processor with universal data input and output capabilities. Possible applications of method 700 include target detection or tracking, and more precisely, the identification of artifacts to avoid confusing them with actual targets in the scene. Furthermore, artifacts can be eliminated before data from the TDM MIMO FMCW radar is used for subsequent processing and / or before the data is presented to the user.

[0059] Alternatively, a signal processing device with processing circuitry configured to execute method 700 via programming or hard-coding can be used. The processing circuitry can be, for example, a general-purpose (programmable) circuit, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or a system-on-a-chip (SoC) with one or more processing cores. To recap, a radar signal processing chain can include a sequence of functional levels starting from the antenna side: mixing, analog-to-digital conversion, RF front-end processing (based on the IF signal), and digital beamforming. Different processing chains can integrate these levels to varying degrees. Thus, the signal processing device executing method 700 can be adapted to be deployed as a general-purpose radar baseband processor, a combined front-end and beamforming device, or a dedicated digital beamforming device.

[0060] In the first step 701 of method 700, a scene-related virtual array signal is obtained. The virtual array signal can represent a range-Doppler interval, which is an element of the range-Doppler spectrum. Since a two-dimensional peak search indicates the presence of moving or stationary objects, a range-Doppler interval can be selected. The virtual array signal representing the range-Doppler interval has as many elements as the elements present in the virtual array of a TDM MIMO radar; this number is denoted as M. r M. This virtual array signal can be obtained from multiple IF signals corresponding to multiple pulse signals and each virtual array element of the virtual array. In this disclosure, the term "virtual array signal" is used to refer to a signal that has a value for each virtual array element of the virtual array. Figure 1A In the example of the virtual array depicted, the IF signal (i.e., the data collected by the first physical receiver 20 (marked A) when excited by the first physical transmitter 10 (TX1)) and the six pulse signals c0, c1, ..., c5 of the leftmost virtual array element can have the following schematic structure:

[0061]

[0062] Where t0, t1, ..., t7 are intervals [0, T] c Discretization of [the signal]. (In practical implementations, discretization can be more refined, and computation can be based on data from a large number of pulse signals.) TX1,A Each row of the IF signal corresponds to one of the pulse signals, and each entry can be understood as a time sample of that pulse signal. Distance information can be obtained by applying a discrete harmonic transform, such as a DFT or FFT, to each row of the IF signal. If an FFT is used, the following distance spectrum (“Distance FFT”) will be generated:

[0063]

[0064] The row dimensions of this matrix now correspond to the distances, where r0, r1, ..., r7 can be interpreted as distance intervals, i.e., the intervals in the radial distance to the reflecting object. The column dimensions still correspond to the six pulse signals, and all the information in the matrix has been changed from... Figure 1B This is derived from the measurement data read by the leftmost virtual array element in the diagram. (The last part, "by analyzing y," appears to be incomplete and lacks context. It's unclear what "y" means in this context.) TX1,A Apply a further FFT to each column to obtain the distance-Doppler spectrum (or "Doppler FFT"):

[0065]

[0066] Matrix z TX1,AEach entry in the table is typically a complex number and can be understood as an element in a discrete representation of the distance-Doppler spectrum. For example, v i ,r j The superscript should be understood as referring to the i-th velocity (or Doppler) interval and the j-th range interval, or simply the (i,j)-th range-Doppler interval. It should be noted that velocity is a signed quantity; in this sense, the range-Doppler spectrum allows for the differentiation between radially toward the radar and radially away from it.

[0067] A range-Doppler interval of the input virtual array signal processed in step 702 of the forming method 700 can be represented as the following vector:

[0068]

[0069] Each element is the range-Doppler interval of the virtual antenna element 30 of the virtual array, i.e., a matrix term from equation (6). In the continued calculations discussed below, for simplicity, the virtual array signal in the range-Doppler interval is represented as z, where the exponent (i,j) is implicit. The phase shift between elements is the sum of the phase shift caused by velocity and the phase shift caused by AoA. Due to the path differences between the virtual array elements, the phase shift caused by AoA can be observed when AoA is non-zero in the plane of the virtual antenna array. Preprocessing aimed at eliminating the phase shift caused by velocity can be performed before calculating the angular spectrum; this is not a necessary element of the invention. In fact, since there is dedicated hardware for calculating the range-Doppler spectrum (e.g., chipsets optionally integrated in TDM MIMO radar devices), and therefore a way to obtain the virtual array signal in the range-Doppler interval that does not include performing the calculations outlined above, step 701 of method 700 should be considered complete once the data according to equation (7) is available.

[0070] This can be illustrated by considering object detection and first-order multipath detection. Figure 6 To understand the expected content of this data, the path RTR, drawn with solid lines, corresponds to object detection (i.e., direct reflection on physical objects in the scene). In this case, the departure angle (AoD) and arrival angle (AoA) are the same and equal to θ1. However, for first-order multipath reflection, the receiver receives echoes from two paths of the same length (both drawn with dashed lines as RTSR and RSTR). The first path has an AoD of θ1 and an AoA of θ2. Conversely, the second path has an AoD of θ2 and an AoA of θ1. It is observed that AoD and AoA are different for multipath reflection, which is key to the method 700 proposed in this paper.

[0071] The content of the virtual array signal z can be represented by the transmit steering vector and the receive steering vector. Therefore, the virtual antenna elements of the virtual array are equidistant in the first direction, and these steering vectors have the following form:

[0072]

[0073] Where d t d r The intervals described above are used, and each guide vector has M. r M element. For target detection, the virtual array signal is:

[0074]

[0075] in Let φ1 and φ2 represent the Kronecker product, where φ1 and φ2 are the phases calculated using (1), and It has variance σ 2 The noise term. Amplitudes s1 and s2 are real or complex numbers. Throughout this explanation, the symbol n indicates a probability distribution. Any noise term (noise vector). In other words, unless the noise terms are different (e.g., n and e). jφ n) have different probability distributions; otherwise, they would not be distinguishable by signs. For first-order multipath reflection, the sum of the virtual array signals is:

[0076]

[0077] In the second step 702 of method 700, the angle spectrum is calculated based on the phase shift between the radar array elements in (7), for example by performing an angle FFT or an AoA FFT. More fully, step 702 includes calculating the angle spectrum of those elements of the virtual array signal corresponding to consecutive virtual antenna elements generated by physical receivers belonging to the same row. The angle spectrum indicates the amplitude of each AoA, and this allows for the detection of all objects in the scene, including multipath artifacts, that have both distance and radial velocity in the (i,j)th range Doppler interval.

[0078] In the third step 703, an intermediate signal v is provided based on the information from the angular spectrum of the (i,j)th distance Doppler interval and the virtual array signal z.

[0079] This includes selecting the first leading edge peak in the angular spectrum of the virtual array signal and estimating the phase of the first leading edge peak. Sub-step 703.1. The first leading-edge peak should be one of the peaks with the largest amplitude in the angular spectrum, for example, a peak with the largest or second largest amplitude. For possible implementation, it should be noted that the ability to resolve (or distinguish) one or more peaks in this step 703.1 is related to the number of elements of the virtual array signal, and therefore, to the number of (virtual) antenna elements. If method 700 is performed on a virtual array signal with too few elements, the angular resolution may be poor, resulting in multiple peaks clustering or merging.

[0080] Step 703 further includes inverting the phase shift vector Sub-step 703.2 applied to the virtual array signal. The inverse phase shift vector is the Kronecker product of the conjugates of the transmit and receive steering vectors. The transmit and receive steering vectors are oriented towards and away from the first leading edge peak of the angular spectrum. The phase of the conjugate of the transmit and receive steering vectors is the phase corresponding to the angle of the first leading edge peak. the opposite number Applying the inverse phase shift vector (e.g., by element-wise multiplication ⊙) can be considered as corresponding to the projection of the strongly varying virtual array signal v onto the subspace corresponding to the first leading-edge peak. If the range Doppler interval contains direct reflections, such a projection returns a signal component without phase shift, i.e., a constant term, which can then be conveniently subtracted. This is attempted in substep 703.3, where a constant signal that may or may not correspond to the constant term of the signal is subtracted, depending on whether direct reflections or multipath artifacts are nearby.

[0081] In the formula, when applied to a signal with two direct reflections (8), the operation in 703.2 can be written as:

[0082]

[0083] The second equation uses the property of identity. If estimate If it is accurate, then And this expression will equal:

[0084]

[0085] Where Δφ = φ2 - φ1 will be called the offset phase. The first term s11 is all of its (M r M) 2 The elements are all equal to the vector of s1, and are therefore constant terms. To perform substep 703.3, the amplitude s1 can be obtained from the angular spectrum; alternatively, it is possible to estimate the amplitude of the constant term directly from the intermediate signal v. The intermediate signal generated after performing substep 703.3 can be represented as v.two .

[0086] Conversely, if the operation in substep 703.2 is applied to the signal (9) with two multipath reflections, then:

[0087]

[0088] None of the constant terms can be identified. Because x two (φ1,φ2) and x mp (φ1,φ2) are indistinguishable until method 700 is completed, so sub-step 703.3 is applied to (11), and the result is obtained by v mp express.

[0089] (The combination of substeps 703.2 and 703.3 can be considered as a null space projection. Matrix)

[0090]

[0091] This represents the projection onto the subspace corresponding to the φ1 peak in the angular spectrum. The associated null space projection matrix is ​​derived from the following equation:

[0092]

[0093] This allows the intermediate signal to be simply written as )

[0094] Next, in step 705, another inverse phase shift vector is applied. (Sub-step 705.2) and subtract another constant signal (sub-step 705.3) to provide the test signal w. Inverse phase shift vector Compared with the estimation of the offset phase introduced above Correspondingly, the estimation of this offset phase It is related to the first and second leading-edge peaks in the angular spectrum of the virtual array signal.

[0095] In some embodiments, the estimated offset phase is calculated based on the angular spectrum of the virtual array signal, i.e., by subtracting the phase corresponding to the angles of the two leading peaks.

[0096] In other embodiments, the estimation of the offset phase The angle spectrum is obtained from the intermediate signal v. The angle spectrum is calculated in step 704 above. The angle spectrum can be calculated, for example, by performing an angle FFT or an AoA FFT based on the phase shift in the signal value. In the angle spectrum, the leading peak whose phase is determined (sub-step 705.1) is selected. The leading peak can be the peak with the largest amplitude in the angle spectrum. It is obvious from each of equations (10) and (11) that the phase of the leading peak will correspond to the offset phase.

[0097] In this embodiment of the method 700 for calculating the angular spectrum of the intermediate signal v, a preliminary guess can be made about whether the signal is related to direct reflection or first-order multipath reflection based on the number of peaks in the angular spectrum: if there is only one peak, the signal is unlikely to contain multipath reflection. The converse inference is generally not true. Therefore, such a preliminary guess is mainly used to verify or confirm the final conclusion of the complete method 700 (step 707 or step 708).

[0098] When substep 705.2 is applied to the intermediate signal v of the two targets two At that time, we obtained:

[0099]

[0100] After eliminating the constant term s21 by subtraction in substep 705.3, step 705.3 returns the test signal w. In this case, the amplitude s2 of the constant term is usually easy to determine, as it is the only non-noise term in (12). The amplitude can be determined directly from (12), or it can be determined as equal to the amplitude of the leading peak in the angular spectrum of the intermediate signal v. Alternatively, the amplitude can be determined by estimating the DC (or non-oscillatory) component of (12), calculating the average, or calculating the average of the maximum and minimum values.

[0101] Alternatively, suppose substep 705.2 is applied to the intermediate signal v of a signal having two multipath reflections. mp This will not yield a simple expression. Direct inspection may not be a feasible way to estimate the magnitude of the constant signal to be subtracted, but the other options mentioned can be used.

[0102] Then, the execution flow of method 700 reaches a decision point in step 706 to determine whether the test signal w has any content other than noise n. If the test signal w has non-noise content, then in step 707 it is inferred that the virtual array signal z contains at least one first-order multipath artifact. (In some cases, this can even prove that the conclusion that the first leading edge peak corresponds to the first-order multipath artifact is correct.) Otherwise, if the test signal only has noise content, then in step 708 it is inferred that the first leading edge peak corresponds to a direct reflection in the scene.

[0103] In some embodiments, the detection in step 706 is performed based on the noise floor, in the sense that signal content of the test signal w below the noise floor (example unit: 1 dB) is considered noise. Only content with higher signal power will result in a positive result for non-noise content detection. According to these embodiments, for the virtual array signal z, the noise floor used in step 706 is obtained. The noise floor of the virtual array signal z can be obtained based on measurements of the virtual array signal z, or the noise floor can be obtained from the specifications (datasheets) of a TDM MIMO radar device that provides the virtual array signal or its underlying data. It should be apparent from the above mathematical derivation that the test signal w has the same noise content as the virtual array signal z; in particular, the noise does not undergo any non-unitary rescaling.

[0104] In other embodiments, step 706 can be performed as a ratio test (sub-step 706.1) that compares the signal energy of a constant signal with the total signal energy of an intermediate signal v, which is subtracted from the intermediate signal (sub-step 705.3).

[0105] Figure 9 The above calculations are summarized in the form of a typical left-to-right information flow. The virtual array signal z is processed through sub-steps 703.2 and 703.3, each of which is parametrically dependent on the first leading-edge peak in the angular spectrum calculated in step 702 (its estimated phase). (and amplitude). Furthermore, the operations applied to sub-steps 705.2 and 705.3 of the intermediate signal v are parametrically dependent on the estimated offset phase. In at least some embodiments of method 700, the leading peak in the angular spectrum of the intermediate signal v is used to provide an estimate of the offset phase and, optionally, the amplitude of the constant signal to be subtracted in substep 705.3.

[0106] From another perspective, the calculations in steps 702 to 705 can be described as two iterative applications of an algorithm for removing the strongest peak. In pseudo-instructions, this algorithm can be expressed as:

[0107] Algorithm 1

[0108] Input: Virtual array signal x

[0109] Output: Virtual array signal y

[0110] S←Angle spectrum(x)

[0111]

[0112] In this regard, the intermediate signal and the test signal are given by the following formula:

[0113] v = ALGORITHM1(z),

[0114] w = ALGORITHM1(v).

[0115] Figure 10 , Figure 11 and Figure 12 This is a graph of simulated data representing a typical execution of method 700. In detail, Figure 10 The outputs for each step of Method 700 for two direct-reflection targets (left column) and two first-order multipath artifacts (right column) for a 2×8 MIMO radar are shown. Each horizontal axis indicates the number of sixteen elements for each signal, with each of the sixteen elements having a one-to-one relationship with a virtual antenna element. The simulation is based on the following parameter values:

[0116]

[0117] For the wavelength under consideration, the angles and phases φ1 = 0.500 and φ2 = 0.766 correspond, resulting in a phase offset of Δφ = 0.266. Figure 10 In the top row, the vertical axis represents the real part of the received signal. The middle row represents the corresponding intermediate signal v. two v mp The chart shows the results after attempting to remove the two strongest components corresponding to the phase φ1 and offset phase Δφ1 of the leading edge peak. As expected, the results for the signal with two directly reflecting targets are close to zero, while the variable components are preserved in the multipath case. A systematic approach to determining whether the test signal with two directly reflecting targets has non-noise content would be to determine the relevant noise floor based on measurement or sensor performance, and then assess whether the deviation from zero is lower in power than the noise floor.

[0118] Accordingly, the simulation indirectly shows that in the case of direct reflection, all non-noise content can be eliminated by two iterations of null-space projection, but not when the signal contains first-order multipath artifacts.

[0119] Based on the same simulated dataset, Figure 11 and Figure 12 This provides an understanding of the impact of the first null space projection on the angular spectrum of the TDM MIMOFMCW radar. Figure 11 The angular spectrum is shown before any treatment (solid line) and after the leading edge peak (dashed line) is removed, in the case of two direct reflections. Figure 12 The same data is shown for the angle spectrum of a similar radar that is capturing two first-order multipath artifacts as well as other objects or artifacts that may be present in the scene.

[0120] Further development of method 700 enables the handling of anomalous cases where the phase is offset. With respect to the first spacing d of the physical receiver r This corresponds to AoA, which generates a path length difference equal to an integer number of wavelengths, or to a second spacing d relative to the physical transmitter. t This corresponds to an AoD (Aspect of Distance) that produces a path length difference equal to an integer number of wavelengths. In other words, when the angle of an object (or artifact) in one aspect of the scene is different from the distance d between the emitters on the other side... t An exception occurs when there is a specific geometric relationship between the receiver spacing d. Due to the degradation of the steering vector, this will result in intermediate signal v of first-order multipath reflection. mp It contains one or two constant terms. More precisely, the fact is that for some integers... d t Δφ=2πn, then

[0121]

[0122] As a result, expression (11) equals

[0123]

[0124] And therefore, it is indistinguishable from expression (10). Similarly, if for some d r Δφ=2πn′, then a r (Δφ)=1 and expression (11) becomes

[0125]

[0126] Degradation conditions

[0127] d r Δφ = 2πn or d t Δφ=2πn′

[0128] By applying (1), it is equivalent to

[0129] d r (sinθ2-sinθ1)=λn or d t (sinθ2-sinθ1)=λn′.

[0130] Without loss of generality, by rotating the coordinate axes to make sinθ1 = 0, this is equivalent to...

[0131] d r sinθ0=λn or d t sinθ0=λn′,

[0132] Wherein, the left side represents the path length difference at the transmitter and receiver arrays, respectively. Angle θ0 is the hypothetical AoA corresponding to the offset phase Δφ, which corresponds to the leading edge peak of the angular spectrum of the intermediate signal v. In other words, when the offset phase Δφ is relative to the first spacing d relative to the physical receiver... r Or the second spacing d relative to the physical transmitter t An exception occurs when the angle corresponds to a path length difference equal to an integer number of wavelengths.

[0133] In a further development of method 700, if an anomaly (degradation) is determined to occur, execution of method 700 can be aborted for that distance Doppler interval. Optionally, upon aborting execution, a message can be generated notifying the user that the method cannot reliably resolve first-order multipath ambiguity.

[0134] Figure 13 This is a simulation of a 2×8 MIMO radar. In this simulation, a first target is placed at θ1 = 0 and a second target is scanned, such that sinθ2 varies in the range (-1, 1). The variance (signal power) of the test signal after removing the two leading edge components of the virtual array signal through steps 702 to 705 is... Figure 13 The value is plotted as a function of the offset phase Δφ. The vertical dashed lines represent those values ​​for which multipath ambiguity cannot be resolved by directly applying method 700.

[0135] As by Figure 8 The flowchart in the diagram indirectly shows that method 700 for resolving first-order multipath ambiguity can be embedded in method 800 for resolving first-order or second-order multipath ambiguity. It is envisioned that method 800 can be used in technical applications similar to those discussed above for method 700.

[0136] The aspects of this disclosure have been described above with reference to several embodiments. However, as will be readily understood by those skilled in the art, other embodiments besides those disclosed above are also possible within the scope of the invention as defined by the appended claims.

Claims

1. A method for resolving first-order multipath ambiguity in the virtual array of a time-division multiplexed multiple-input multiple-output frequency-modulated continuous wave radar. in, The time-division multiplexing multiple-input multiple-output frequency-modulated continuous wave radar includes a first direction with a first spacing. At least one row of physical receivers and further including at a second spacing in the first direction. Multiple physical transmitters were deployed. The method includes: Obtain the virtual array signal in the range-Doppler interval relevant to the scene. Each element of the virtual array signal corresponds to a virtual antenna element of the virtual array; Calculate the angular spectrum of the distance-Doppler interval; Provide intermediate signals through the following steps. Select a first leading-edge peak from the angular spectrum of the virtual array signal; The estimated phase with the first leading edge peak Corresponding inverse phase shift vector Applied to the virtual array signal; and Subtract the constant signal with an amplitude corresponding to the first leading edge peak; Provide test signals through the following steps. Will be compared with the estimated offset phase Corresponding inverse phase shift vector The estimated offset phase is applied to the intermediate signal. The first and second leading-edge peaks in the angular spectrum of the virtual array signal are related to the first and second leading-edge peaks, which are the two largest peaks in the angular spectrum; and Subtract the constant signal; Detect the non-noise content of the test signal; If the test signal contains non-noise content, it is inferred that the virtual array signal contains first-order multipath artifacts; and If the test signal contains only noise, then it is inferred that the first leading edge peak corresponds to a direct reflection in the scene.

2. The method according to claim 1, wherein, The offset phase is calculated based on the angle spectrum of the virtual array signal. .

3. The method according to claim 1, further comprising: Calculate the intermediate signal Angular spectrum; as well as Select the leading peak in the angular spectrum of the intermediate signal, wherein the leading peak has the largest amplitude in the angular spectrum. Wherein, the offset phase Corresponding to the estimated angle of the selected leading edge peak.

4. The method according to claim 3, wherein, The constant signal subtracted from the intermediate signal has an amplitude corresponding to the selected leading edge peak.

5. The method according to claim 1, wherein, The detection of non-noise content includes subtracting the signal energy of the constant signal from the intermediate signal and the intermediate signal. The ratio test compares the total signal energy.

6. The method of claim 1, further comprising: Acquire the virtual array signal The noise floor, The test signal is executed based on the noise floor. The detection of the non-noise content.

7. The method according to claim 1, wherein, The first spacing and the second spacing The ratio of the two values ​​makes the virtual antenna elements of the virtual array equidistant in the first direction.

8. The method according to claim 1, wherein, Calculating the spectrum for each angle involves performing a Fast Fourier Transform (FFT) relative to a series of virtual antenna elements.

9. The method of claim 1, further comprising: Evaluate the offset phase Whether relative to the first spacing of the physical receiver or the second spacing of the physical transmitter The angles corresponding to the path length differences that produce an integer number of wavelengths.

10. A method for resolving first-order or second-order multipath ambiguity, comprising: Acquire scene-related virtual array signals Each element of the virtual array signal corresponds to a virtual antenna element of the virtual array of the time-division multiplexed multiple-input multiple-output frequency-modulated continuous wave radar as described in claim 1; Calculate the angular spectrum of the virtual array signal over a distance-Doppler interval; Count the number of peaks in the angular spectrum; If the angular spectrum has a single peak, it is inferred that the angular spectrum corresponds to a second-order multipath artifact or direct reflection in the scene; as well as If the angular spectrum has at least two peaks, it is inferred that the angular spectrum corresponds to two first-order multipath artifacts or two direct reflections in the scene, and the method of claim 1 is further performed.

11. A signal processing device for time-division multiplexed multiple-input multiple-output frequency-modulated continuous wave radar, in, The time-division multiplexing multiple-input multiple-output frequency-modulated continuous wave radar includes a first direction with a first spacing. At least one row of physical receivers and further including at a second spacing in the first direction. Multiple physical transmitters were deployed. The signal processing device includes processing circuitry configured to resolve first-order multipath ambiguity of the virtual array of the time-division multiplexed multiple-input multiple-output frequency-modulated continuous wave radar in a virtual array signal comprising at least one range Doppler interval by performing the method of claim 1.

12. A computer program comprising instructions that, when executed by the signal processing apparatus of claim 11, cause the signal processing apparatus to perform the method of claim 1.

Images