Method and training system for training a radar device

Patent metadata

Field	Value
Publication number	US-11047957-B2
Application number	US-201816194081-A
Country	US
Kind code	B2
Filing date	Nov 16, 2018
Priority date	Nov 16, 2018
Publication date	Jun 29, 2021
Grant date	Jun 29, 2021

Abstract

A method for training a radar device using a training system is disclosed. The radar device comprises a first machine learning module to be trained. The training system comprises at least one signal generator, a signal analyzer and at least one controller being connected to both the at least one signal generator and the at least one signal analyzer. The controller causes the at least one signal generator to generate at least one wireless test signal based on an initial set of parameters corresponding to at least one object being located in the field of view of the radar device. The wireless test signal is received by the radar device and a feedback signal is generated by the radar device based on the test signal. The feedback signal is forwarded to the analyzer unit, and the set of parameters is automatically adapted by the controller based on the feedback signal. Moreover, a training system for training a radar device is disclosed.

First claim

The invention claimed is:

1. A method for training a radar device using a training system, the radar device comprising a first machine learning module to be trained, the training system comprising at least one signal generator, a signal analyzer and at least one controller being connected to both the at least one signal generator and the at least one signal analyzer, with the following steps:
causing, by a controller, the at least one signal generator to generate at least one wireless test signal based on an initial set of parameters corresponding to at least one object being located in the field of view of the radar device;
receiving the at least one wireless test signal by the radar device;
generating a feedback signal by the radar device based on the test signal, wherein the feedback signal comprises an identification or classification of the at least one object generated by the first machine learning module;
forwarding the feedback signal to the signal analyzer; and
adapting, automatically, the initial set of parameters by the controller based on the feedback signal such that the first machine learning module is automatically trained to correctly identify or classify objects in the field of view of the radar device.

2. The method of claim 1 , wherein the set of parameters comprises information on at least one of a number of objects, a radar cross section of said at least one object, a position of said at least one object, a velocity of said at least one object, an acceleration of said at least one object, an emission characteristic of said at least one object, a communication signal, an interfering transmitter and background noise.

3. The method of claim 1 , wherein the test signal is adjusted based on the adapted parameter set, wherein the adjustment comprises changing at least one of a frequency of the test signal, an amplitude of the test signal, a waveform of the test signal, an emission mode, a power level of the test signal and a pulse width of the test signal.

4. The method of claim 1 , wherein the initial set of parameters is at least one of generated by the controller, loaded from a database stored in a memory of the training system, read from a storage medium and manually input by a user.

5. The method of claim 1 , wherein the controller comprises a second machine learning module, wherein the set of parameters is automatically adapted by the second machine learning module.

6. The method of claim 5 , wherein the first machine learning module and the second machine learning module are both trained unsupervised.

7. The method of claim 5 , wherein the first and second machine learning module form a generative adversarial network.

8. The method of claim 7 , wherein the first machine learning module and the second machine learning module each comprise an artificial neural network, wherein the two artificial neural networks form the generative adversarial network.

9. A non-transitory computer readable medium for training a radar device using a training system, the radar device comprising a first machine learning module to be trained, the training system comprising at least one signal generator, a signal analyzer and at least one controller being connected to both the at least one signal generator and the at least one signal analyzer, the non-transitory computer readable medium having a program code being adapted to cause the training system to perform the following steps when executed as the at least one controller:
generating at least one wireless test signal via the at least one signal generator based on an initial set of parameters corresponding to at least one object being located in the field of view of the radar device; and
automatically adapting the set of parameters based on a feedback signal being generated by the radar device based on the test signal and being forwarded to the signal analyzer, such that the first machine learning module is automatically trained to correctly identify or classify objects in the field of view of the radar device, wherein the feedback signal comprises an identification or classification of the at least one object generated by the first machine learning module.

10. The non-transitory computer readable medium of claim 9 , further comprising a program code being adapted to cause the training system to at least one of generate the initial set of parameters, load the initial set of parameters from a database stored in a memory of the training system and read the initial set of parameters from a storage medium when the computer program is run on the at least one controller.

11. The non-transitory computer readable medium of claim 9 , wherein the computer readable medium is at least partly established as a second machine learning module being configured to automatically adapt the set of parameters.

12. The non-transitory computer readable medium of claim 9 , wherein the computer readable medium is at least partly established as an artificial neural network, wherein the first machine learning module comprises another artificial neural network, and wherein the two artificial neural networks together form a generative adversarial network.

13. A training system for training a radar device, the radar device comprising a first machine learning module to be trained, the training system comprising:
at least one signal generator, a signal analyzer and at least one controller being connected to both the at least one signal generator and the at least one signal analyzer,
the controller being configured to control the at least one signal generator to generate at least one wireless test signal based on an initial set of parameters corresponding to a at least one object being located in the field of view of the radar device,
the radar device being configured to receive and process said test signal to generate a feedback signal, wherein the feedback signal comprises an identification or classification of the at least one object generated by the first machine learning module, the radar device being further configured to forward the feedback signal to the signal analyzer, and
the controller further being configured to automatically adapt the set of parameters based on the feedback signal such that the first machine learning module is automatically trained to correctly identify or classify objects in the field of view of the radar device.

14. The training system of claim 13 , comprising several signal generators, at least one combiner and at least one transmission antenna, wherein the at least one transmission antenna is connected to the several signal generators via the at least one combiner.

15. The training system of claim 14 , wherein the controller is configured to control the several signal generators to generate signals being coherent to each other.

16. The training system of claim 13 , wherein the controller comprises a second machine learning module, the second machine learning module being configured to automatically adapt the set of parameters.

17. The training system of claim 16 , wherein the first and second machine learning module form a generative adversarial network.

18. The training system of claim 16 , wherein the first machine learning module and the second machine learning module each comprise an artificial neural network, wherein the two artificial neural networks form a generative adversarial network.

Assignees

• Rohde & Schwarz

Inventors

• Mikhailov Yassen

Classifications

• G06N3/047

  Probabilistic or stochastic networks · CPC title

• G06N3/045

  Combinations of networks · CPC title

• G06N3/088 Primary

  Non-supervised learning, e.g. competitive learning · CPC title

• G06N3/0475

  Generative networks · CPC title

• G06N3/094

  Adversarial learning · CPC title