Methods, apparatus, systems and articles of manufacture are disclosed for distributed, multi-node, low frequency radar systems for degraded visual environments. An example system includes a transmitter to transmit a radar signal. The example system includes a distributed network of radar receivers to receive the radar signal at each receiver. The example system includes a processor to determine a first range and a first angular position of a background point based on return time, wherein the first range and the first angular position are included in first data; determine a second range and a second angular position of the background point based on doppler shift, wherein the second range and the second angular position are included in second data; determine a refined range and a refined angular position, wherein the refined range and refined angular position are included in third data, and generate a radar map based on third data.

A device includes: a millimeter-wave radar sensor circuit configured to generate N virtual channels of sensed data, where N is an integer number greater than one; and a processor configured to: generate a 2D radar image of a surface in a field of view of the millimeter-wave radar sensor circuit based on sensed data from the N virtual channels of sensed data, where the 2D radar image includes azimuth and range information, generate a multi-dimensional data structure based on the 2D radar image using a transform function, compare the multi-dimensional data structure with a reference multi-dimensional data structure, and determine whether liquid is present in the field of view of the millimeter-wave radar sensor circuit based on comparing the multi-dimensional data structure with the reference multi-dimensional data structure.

A method and an apparatus for removing a motion artifact of a radar are provided. The method includes: obtaining a radar signal for a target to be measured by the radar; measuring posture of the radar; estimating a motion artifact caused by movement of the radar based on a vertical angle, a horizontal angle based on the posture of the radar, and displacement; and correcting the radar signal according to the motion artifact. The posture of the radar includes the vertical angle at which the radar signal is radiated in a vertical direction about a central axis, the horizontal angle at which the radar signal is radiated in a horizontal direction about the central axis, and the displacement of the radar according to the movement of the radar.

A method of motion compensation for a Doppler radar system includes receiving, for each transmitted pulse of a set of transmitted pulses, a respective set of echo signals returned from a plurality of distance ranges, performing Doppler Fourier transforms on the sets of echo signals for the set of transmitted pulses to generate outputs that include detected signals in a plurality of velocity bins, and applying a respective pre-determined compensation phase vector to the detected signals in each velocity bin of the plurality of velocity bins. The respective pre-determined compensation phase vector applied to the detected signals in each velocity bin includes at least one of a first component proportional to a velocity of the velocity bin or a second component for compensating a phase compensation error associated with Doppler velocity aliasing.

A method (100; 200) for digital signal processing (S(t)) of a pulse and scanning radar during an observation of a coastal zone in land/sea detection mode, the signal being sampled according to a two-dimensional temporal map, a distance dimension (d) and a recurrence dimension (rec), comprising: selecting a digital terrain model file (MNT) of the observed coastal zone; transforming (110; 210) the temporal map and/or the digital terrain model file to obtain a transformed temporal map and/or a transformed digital terrain model file the data of which are expressed in a common reference frame; constructing (120) a mask (MT; MF) from the transformed digital terrain model file; and applying (130) the mask to the samples (E(d, rec); E(d, f)) of the map associated with the transformed temporal map, in such a way as to obtain filtered samples (Ef(d, rec); Ef(d, f)).

A radar device is mounted on a vehicle, which is a moving object, and includes a doppler correction phase-rotation controller and a phase rotator. Based on the speed of the vehicle, the doppler correction phase-rotation controller calculates a Doppler correction phase-rotation amount for correcting a Doppler frequency due to movement of the vehicle. By using the Doppler correction phase-rotation amount, the phase rotator pre-corrects Doppler frequency components with respect to a radar transmission signal in each transmission interval of the radar transmission signal.

The present application discloses a new form of -STAP, referred to herein as post -STAP or P-STAP, which overcomes the drawbacks associated with existing -STAP techniques. The P-STAP techniques described herein facilitate the generation of additional training data and homogenization after pulse compression. For example, P-STAP techniques may apply a plurality of homogenization filters to a pulse compressed datacube generated from an input radar waveform, which produces a plurality of new pulse compressed datacubes with improved characteristics. Unlike existing -STAP techniques described above, which require pre-pulse compressed data to operate, the P-STAP techniques disclosed in the present application are designed to utilize pulse compressed data, and therefore may be readily applied to legacy radar systems.

A method for determining direction information for at least one target object in a radar system for a vehicle. The first detection information is provided by at least two receive antennas of the radar system, wherein the first detection information is specific for a first radar signal transmitted by a first transmit antenna of the radar system. The second detection information is provided by the at least two receive antennas of the radar system, wherein the second detection information is specific for a second radar signal transmitted by a second transmit antenna of the radar system. A first angle determination and a second angle determination are performed. At least one comparison of the first angle information with the second angle information is performed in order to detect an ambiguity in the first angle determination for the determination of the direction information.

A method and apparatus for processing a transceiver signal (115) detected by a transceiver (110). The method includes obtaining (S1) a processed signal from the transceiver signal (115), the processed signal having frames (200, 300) corresponding to respective time intervals (t1, t2, t3, t4), wherein the frames define bins (210, 310) configured according to a quantized resolution (dr) of the transceiver signal (115). The method further includes obtaining (S2) data related to a relative motion of the transceiver (110) during a time interval (t1, t2, t3, t4) and initializing (S3) a residual distance to zero. For each frame (200, 300) and each respective time interval (t1, t2, t3, t4) the method further includes determining (S4) a shift distance (ds1, ds3) corresponding to a sum of the residual distance and a distance value (d1, d2) corresponding to a relative motion of the transceiver (110) in the respective time interval (t1, t2, t3, t4) and rounding (S5) the determined shift distance (ds1, ds3) with respect to the distance resolution (dr) to a rounded shift distance. The method then further includes updating (S6) the residual distance based on a difference between the determined shift distance (ds1, ds3) and the rounded shift distance, and generating (S7) an adjusted frame (304) by shifting the bins (310) of the frame by the rounded shift distance to account for relative transceiver motion with respect to the object (150) in the respective time interval. The method finally includes processing (S8) the signal by integrating bin values (210, 310) over the adjusted frames (300).

In various examples, a deep neural network(s) (e.g., a convolutional neural network) may be trained to detect moving and stationary obstacles from RADAR data of a three dimensional (3D) space, in both highway and urban scenarios. RADAR detections may be accumulated, ego-motion-compensated, orthographically projected, and fed into a neural network(s). The neural network(s) may include a common trunk with a feature extractor and several heads that predict different outputs such as a class confidence head that predicts a confidence map and an instance regression head that predicts object instance data for detected objects. The outputs may be decoded, filtered, and/or clustered to form bounding shapes identifying the location, size, and/or orientation of detected object instances. The detected object instances may be provided to an autonomous vehicle drive stack to enable safe planning and control of the autonomous vehicle.