Methods and systems for determining a direction of arrival of a radar detection

Patent metadata

Field	Value
Publication number	US-12044787-B2
Application number	US-202117513797-A
Country	US
Kind code	B2
Filing date	Oct 28, 2021
Priority date	Nov 2, 2020
Publication date	Jul 23, 2024
Grant date	Jul 23, 2024

Abstract

A computer implemented method for determining a direction of arrival of a radar detection comprises the following steps carried out by computer hardware components: acquiring a complex-valued beamvector of the radar detection; processing the complex-valued beamvector by a machine learning module in the complex domain; and obtaining the direction of arrival as an output of the machine learning module.

First claim

What is claimed is:

1. A method of operating a radar system, the method comprising:
acquiring, from a radar sensor, a complex-valued beamvector of a radar detection;
processing the complex-valued beamvector by a machine learning module in a complex domain, the processing including,
determining, based on the complex-valued beamvector, a plurality of phase differences for antennas spaced apart a pre-determined distance and for antennas spaced apart at least substantially fifty percent more than the pre-determined distance,
expanding a plurality of angles based on the plurality of phase differences such that outputs of a layer included in the machine learning module have a same dimension as the plurality of angles, and
combining the output of the layer and the plurality of angles; and

obtaining a direction of arrival as an output of the machine learning module.

2. The method of claim 1 , wherein the machine learning module comprises a plurality of dense layers.

3. The method of claim 2 , wherein each of the plurality of dense layers is followed by a respective complex activation function.

4. The method of claim 3 , wherein the machine learning module further comprises a further dense layer.

5. The method of claim 4 , wherein the further dense layer is followed by at least one of a square layer, a convolution layer, or a real activation function.

6. The method of claim 5 , wherein
the further dense layer is followed by the square layer, and
the combining includes combining the output of the square layer and the plurality of angles.

7. The method of claim 1 , wherein the machine learning module further comprises at least one of a convolution layer, a real activation function, or a dense combination layer.

8. The method of claim 1 , wherein the machine learning module further comprises a probability-like activation function.

9. The method of claim 1 , wherein the plurality of angles are expanded by applying a multilayer perceptron (MLP) that shares weights between the plurality of angles.

10. A vehicle comprising:
a radar sensor; and
a computer system configured to:
acquire, from the radar sensor, a complex-valued beamvector of a radar detection;
process the complex-valued beamvector by a machine learning module in a complex domain including,
determining, based on the complex-valued beamvector, a plurality of phase differences for antennas spaced apart a pre-determined distance and for antennas spaced apart at least substantially fifty percent more than the pre-determined distance,
expanding a plurality of angles based on the plurality of phase differences such that outputs of a layer included in the machine learning module have a same dimension as the plurality of angles, and
combining the output of the layer and the plurality of angles; and

obtain a direction of arrival as an output of the machine learning module.

11. The vehicle of claim 10 , wherein the machine learning module comprises a plurality of dense layers.

12. The vehicle of claim 11 , wherein each of the plurality of dense layers is followed by a respective complex activation function.

13. The vehicle of claim 12 , wherein the machine learning module further comprises a further dense layer.

14. The vehicle of claim 13 , wherein the further dense layer is followed by at least one of a square layer, a convolution layer, or a real activation function.

15. The vehicle of claim 10 , wherein the plurality of angles are expanded by applying a multilayer perceptron (MLP) that shares weights between the plurality of angles.

16. A non-transitory computer readable media comprising computer-executable instructions that, when executed, cause a computer system to:
acquire a complex-valued beamvector of a radar detection;
process the complex-valued beamvector by a machine learning module in a complex domain including,
determining, based on the complex-valued beamvector, a plurality of phase differences for antennas spaced apart a pre-determined distance and for antennas spaced apart at least substantially fifty percent more than the pre-determined distance,
expanding a plurality of angles based on the plurality of phase differences such that outputs of a layer included in the machine learning module have a same dimension as the plurality of angles, and

combining the output of the layer and the plurality of angles; and
obtain a direction of arrival as an output of the machine learning module.

17. The non-transitory computer readable media of claim 16 , wherein the plurality of angles are expanded by applying a multilayer perceptron (MLP) that shares weights between the plurality of angles.

Assignees

• Aptiv Technologies AG

Inventors

• Becker Adrian
• Feldmann Anton
• Grumpe Arne
• Stefer Markus

Classifications

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/048

  Activation functions · CPC title

• G06N3/047

  Probabilistic or stochastic networks · CPC title

• G06N20/00

  Machine learning · CPC title

• G01S2013/93185

  Controlling the brakes · CPC title