Reconstruction of elevation information from radar data

Patent metadata

Field	Value
Publication number	US-12392871-B2
Application number	US-202117147079-A
Country	US
Kind code	B2
Filing date	Jan 12, 2021
Priority date	Jan 22, 2020
Publication date	Aug 19, 2025
Grant date	Aug 19, 2025

Abstract

A method for reconstructing elevation information from measured data that were recorded with the aid of at least one radar device and include a two-dimensional spatial distribution of at least one physical measured variable. The measured data are fed as input variables to at least one generator module that is designed as a neural network. At least one output variable is retrieved from the generator module that represents a measure of the elevation angles from which radar radiation was reflected to the radar device from at least one object. A method for training a generator module, and a method including a complete active chain up to activating a vehicle, are also described.

First claim

What is claimed is:

1. A method for a vehicle, the method comprising the following steps:
during a drive of the vehicle:
recording with at least one radar device of the vehicle, radar data that form a two-dimensional spatial distribution of at least one physical measured variable;
feeding the radar data as input values to at least one generator module of the vehicle that is configured as a neural network and generating, from the input values, surroundings output values that represent elevation angles from which radar radiation was reflected to the radar device from at least one object;
generating, by the vehicle, a representation of surroundings of the vehicle based on the surroundings output values; and
performing, by the vehicle, an automated drive control of the vehicle according to the representation of surroundings;

wherein the method includes at least one of the following two features (I)-(II):
(I) (i) the feeding of the radar data includes dividing the input values into a plurality of segments, each of the segments formed of a respective contiguous portion of the input values, (ii) the generating of the surroundings output values includes separately processing each of the segments to obtain a respective set of segment output values and forming from the respective sets of segment output values, the surroundings output values as a single contiguous three-dimensional distribution, and (iii) with respect to each segment of at least two of the plurality of segments, representations of the segment output values of the respective segment are interspersed amongst the representations of the segment output values of one or more other ones of the segments, and are therefore non-contiguous, within the single contiguous three-dimensional distribution; and
(II) the generating of the surroundings output values includes (i) generating a plurality of two-dimensional spatial distributions, with respect to a predefined coordinate space, of the at least one physical measured variable, each of the plurality of two-dimensional spatial distributions corresponding to a respective elevation region, and (ii) stacking the plurality of two-dimensional spatial distributions to form a contiguous three-dimensional distribution spanning the respective elevation regions of the plurality of two-dimensional spatial distributions.

2. The method as recited in claim 1 , wherein the surroundings output values indicate, with regard to locations in the two-dimensional spatial distribution, from which elevation angles radar radiation that contributed to a value of a physical measured variable at each location was reflected to the radar device.

3. The method as recited in claim 1 , wherein the generator module includes:
an encoder configured to translate the input values into latent variables in a space, whose dimensionality is smaller than a dimensionality of a space of the input values and smaller than a dimensionality of a space of the surroundings output values; and
a decoder configured to translate the latent variables into the surroundings output values.

4. The method as recited in claim 1 , wherein:
the antenna array of the at least one radar device is arranged within a single elevation range such that, without use of the generator module, the antenna array is capable of performing measurements of only the single elevation range; and
information of one or more elevations outside of the single elevation range is obtainable using the generator module.

5. The method as recited in 1 , wherein the at least one radar device is a moved synthetic aperture radar device.

6. The method as recited in claim 1 , wherein the generating of the representation of the surroundings includes evaluating:
at least one category of a predefined classification of traffic signs, or other road users, or roadway markings, or signaling systems, or other traffic-relevant objects, in surroundings of a vehicle, and/or
at least one position, and/or spatial measurement and/or speed of an object including a traffic sign, or other road user, or roadway marking, or signaling system, or another traffic-relevant object, in the surroundings of a vehicle.

7. The method as recited in claim 1 , further comprising training the generator module by:
providing training measured radar data;
ascertaining a two-dimensional spatial distribution of the at least one physical measured variable from the training measured radar data;
ascertaining setpoint output values from the training measured radar data that represent a measure of elevation angles from which radar radiation was reflected back to the radar device from at least one object;
processing the training measured radar data to form a generator module version of output values using the generator module; and
optimizing generator parameters that characterize a behavior of the generator module to maximize a degree to which the generator module version of the output values reproduce the setpoint output values in accordance with a predefined generator cost function.

8. The method as recited in claim 7 , wherein the training of the generator module further includes:
feeding the generator module version of the output values and the setpoint output values to a discriminator module that is configured as a further neural network; and
optimizing discriminator parameters that characterize a behavior of the discriminator module alternatingly with the generator parameters for optimizing a differentiation performed by the discriminator module to differentiate the generator module version of the output values from the setpoint output values in accordance with a discriminator cost function.

9. The method as recited in claim 7 , wherein the radar device includes multiple channels outside of a plane in which its azimuth angle is defined.

10. The method as recited in claim 9 , wherein a distance relative speed spectrum of the training measured radar data is ascertained for each of the multiple channels and, during the training, input for the generator module and the setpoint output values are ascertained from the distance relative speed spectra.

11. The method as recited in claim 1 , wherein:
(i) the feeding of the radar data includes the dividing the input values into the plurality of segments, each of the segments formed of the respective contiguous portion of the input values;
(ii) the generating of the surroundings output values includes the separately processing of each of the segments to obtain the respective set of segment output values and the forming, from the respective sets of segment output values, the surroundings output values as the single contiguous three-dimensional distribution; and
(iii) with respect to each segment of the at least two of the plurality of segments, the representations of the segment output values of the respective segment are interspersed amongst the representations of the segment output values of the one or more other ones of the segments, and are therefore non-contiguous, within the single contiguous three-dimensional distribution.

12. The method as recited in claim 11 , wherein the at least one generator module includes a plurality of generator modules executing on different processors, by which execution the plurality of generator modules process the plurality of segments in parallel.

13. The method as recited in claim 1 , wherein the generating of the surroundings output values includes:
(i) the generating of the plurality of two-dimensional spatial distributions, with respect to the predefined coordinate space, of the at least one physical measured variable, each of the plurality of two-dimensional spatial distributions corresponding

Assignees

• Bosch Gmbh Robert

Inventors

• Hakobyan Gor
• Rambach Kilian

Classifications

• G01S13/90 Primary

  using synthetic aperture techniques {, e.g. synthetic aperture radar [SAR] techniques} · CPC title

• G01S13/06

  Systems determining position data of a target · CPC title

• G01S13/931

  of land vehicles · CPC title

• G06N3/08

  Learning methods · CPC title

• G06N3/04

  Architecture, e.g. interconnection topology · CPC title