Reducing and processing simulated and real-world radar data

Patent metadata

Field	Value
Publication number	US-12221128-B2
Application number	US-202218059155-A
Country	US
Kind code	B2
Filing date	Nov 28, 2022
Priority date	Nov 28, 2022
Publication date	Feb 11, 2025
Grant date	Feb 11, 2025

Abstract

A system for filtering RADAR data includes one or more graphical processing units (GPUs) for performing steps of a filtering process in parallel. For example, a first GPU or first portion of GPU circuitry calculates a threshold parabola for a tensor of RADAR data. In parallel, a second GPU or second portion of GPU circuitry separately reduces and indexes the tensor for comparison to the threshold parabola. The threshold parabola is compared to reduced and indexed data to filter the RADAR data. Importance sampling can also be used to reduce data, e.g., if the RADAR data includes four dimensions (range, Doppler, azimuth, and elevation).

First claim

What is claimed is:

1. A computer implemented method for filtering radio detection and ranging (RADAR) data comprising:
receiving data representing a RADAR signal, the data comprising a tensor having multiple dimensions;
calculating a threshold curve using first graphical processing unit (GPU) circuitry, wherein a first portion of the first GPU circuitry processes a first portion of the tensor, and a second portion of the first GPU circuitry processes a second portion of the tensor;
reducing the data representing the RADAR signal using second GPU circuitry;
comparing the reduced data to the threshold curve; and
outputting detections based on the threshold curve.

2. The computer implemented method of claim 1 , the tensor comprising an azimuth dimension and a Doppler dimension, wherein calculating the threshold curve using the first GPU circuitry comprises:
splitting the tensor along the azimuth dimension, the first portion of the tensor comprising a first range of azimuths, and the first portion of the tensor comprising a second range of azimuths;
reducing, by the first portion of the first GPU circuitry, the first portion of the tensor along the Doppler dimension; and
reducing, by the second portion of the first GPU circuitry, the second portion of the tensor along the Doppler dimension.

3. The computer implemented method of claim 2 , the tensor further comprising a range dimension, wherein calculating the threshold curve using the first GPU circuitry further comprises:
further reducing the first portion of the tensor along the range dimension; and
further reducing the second portion of the tensor along the range dimension.

4. The computer implemented method of claim 1 , the tensor comprising an azimuth dimension and a Doppler dimension, wherein reducing the data representing the RADAR signal using second GPU circuitry comprises:
splitting the tensor along the Doppler dimension to generate a third portion of the tensor and a fourth portion of the tensor, the third portion of the tensor comprising a first range of Dopplers, and the fourth portion of the tensor comprising a second range of Dopplers;
reducing, by a first portion of the second GPU circuitry, the third portion of the tensor along the azimuth dimension; and
reducing, by a second portion of the second GPU circuitry, the fourth portion of the tensor along the azimuth dimension.

5. The computer implemented method of claim 1 , wherein the tensor has a range dimension, a Doppler dimension, an azimuth dimension, and an elevation dimension.

6. The computer implemented method of claim 5 , further comprising:
reducing a size of the tensor by performing Monte Carlo importance sampling of the tensor in the Doppler dimension.

7. The computer implemented method of claim 6 , wherein the importance sampling identifies at least two slices of data in the tensor, each of the identified slices of data having a respective Doppler value.

8. A system for filtering radio detection and ranging (RADAR) data, the system comprising:
circuitry to receive data representing a RADAR signal, the data comprising a tensor having multiple dimensions;
first graphical processing unit (GPU) circuitry to calculate a threshold curve, the first GPU circuitry comprising:
a first portion of to process a first portion of the tensor, and
a second portion to process a second portion of the tensor; and

second GPU circuitry to:
reduce the data representing the RADAR signal using parallel processing, and
compare the reduced data to the threshold curve;

the system further to output detections based on the threshold curve.

9. The system of claim 8 , wherein the first GPU circuitry comprises a first GPU, and the second GPU circuitry comprises a second GPU.

10. The system of claim 8 , further comprising third GPU circuitry to refine the detections, the system to output the refined detections.

11. The system of claim 8 , wherein the tensor comprises an azimuth dimension and a Doppler dimension, and the first GPU circuitry is to:
split the tensor along the azimuth dimension, the first portion of the tensor comprising a first range of azimuths, and the first portion of the tensor comprising a second range of azimuths;
reduce, by the first portion of the first GPU circuitry, the first portion of the tensor along the Doppler dimension; and
reduce, by the second portion of the first GPU circuitry, the second portion of the tensor along the Doppler dimension.

12. The system of claim 11 , wherein the tensor further comprises a range dimension, and the first GPU circuitry is further to:
further reduce the first portion of the tensor along the range dimension; and
further reduce the second portion of the tensor along the range dimension.

13. The system of claim 8 , wherein the tensor comprises an azimuth dimension and a Doppler dimension, and the second GPU circuitry is to:
split the tensor along the Doppler dimension to generate a third portion of the tensor and a fourth portion of the tensor, the third portion of the tensor comprising a first range of Dopplers, and the fourth portion of the tensor comprising a second range of Dopplers;
reduce, by a first portion of the second GPU circuitry, the third portion of the tensor along the azimuth dimension; and
reduce, by a second portion of the second GPU circuitry, the fourth portion of the tensor along the azimuth dimension.

14. The system of claim 8 , wherein the tensor has a range dimension, a Doppler dimension, an azimuth dimension, and an elevation dimension, and the system is further to reduce a size of the tensor by performing Monte Carlo importance sampling of the tensor in the Doppler dimension.

15. A non-transitory computer-readable medium storing instructions for filtering radio detection and ranging (RADAR) data, the instructions, when executed by a processor, cause the processor to:
receive data representing a RADAR signal, the data comprising a tensor having multiple dimensions;
calculate a threshold curve using first graphical processing unit (GPU) circuitry, wherein a first portion of the first GPU circuitry processes a first portion of the tensor, and a second portion of the first GPU circuitry processes a second portion of the tensor;
reduce the data representing the RADAR signal using second GPU circuitry;
compare the reduced data to the threshold curve; and
output detections based on the threshold curve.

16. The computer-readable medium of claim 15 , wherein the tensor comprising an azimuth dimension and a Doppler dimension, and calculating the threshold curve using the first GPU circuitry comprises:
splitting the tensor along the azimuth dimension, the first portion of the tensor comprising a first range of azimuths, and the first portion of the tensor comprising a second range of azimuths;
reducing, by the first portion of the first GPU circuitry, the first portion of the tensor along the Doppler dimension; and
reducing, by the second portion of the first GPU circuitry, the second portion of the tensor along the Doppler dimension.

17. The computer-readable medium of claim 16 , wherein the tensor further comprises a range dimension, and calculating the threshold curve using the first GPU circuitry further comprises:
further reducing the first portion of the tensor along the range dimension; and
further reducing the second portion of the tensor along the range dimension.

18. The computer-readable medium of claim 15 , wherein the tensor comprises an azimuth dimension and a Doppler dimension, wherein reducing the data representing

Assignees

• Gm Cruise Holdings Llc

Inventors

• Wang Ke

Classifications

• B60W2420/408

  Radar; Laser, e.g. lidar · CPC title

• G01S13/505

  using Doppler effect for determining closest range to a target or corresponding time, e.g. miss-distance indicator · CPC title

• G01S13/006

  Theoretical aspects (G01S7/418, G01S13/9094, G01S13/958 take precedence) · CPC title

• G09B9/54

  Simulation of radar (G09B9/40 takes precedence) · CPC title

• B60W60/001 Primary

  Planning or execution of driving tasks · CPC title