# US 2024/0004055 A1 · Radar Device and Method of Operating a Radar Device

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2024/0004055 A1
- Document ID: US20240004055A1
- Status: Published application
- Kind: A1
- Application number: 18/334,013
- Filed: 2023-06-13
- Published: 2024-01-04
- Earliest claimed: 2022-07-01
- Assignees: Infineon Technologies AG
- Applicants: Infineon Technologies AG
- Inventors: Julius Ott; Avik Santra; Lorenzo Servadei
- Web page: https://patentgrep.com/patent/US20240004055A1

## CPC

- G01S 13/581: {using transmission of interrupted pulse modulated waves and based upon the Doppler effect resulting from movement of targets}
- G01S 7/412: {based on a comparison between measured values and known or stored values}
- G01S 7/417: {involving the use of neural networks}
- G01S 13/26: wherein the transmitted pulses use a frequency- or phase-modulated carrier wave
- G01S 13/34: using transmission of continuous, frequency-modulated waves while heterodyning the received signal, or a signal derived therefrom, with a locally-generated signal related to the contemporaneously transmitted signal
- G01S: RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES

## Abstract

A radar device includes a radar front end configured to send radar signals and to receive reflected radar signals, processing circuitry configured to provide digital radar data based on the received reflected radar signals, and a digital filter configured to process the digital radar data to obtain information about objects which reflected the radar signals. The device further comprises machine learning logic with a policy network configured to set the parameters of the digital filter based on the digital radar data, and a reward value generating network including a plurality of heads, each head configured to provide a respective expected reward value for a setting of parameters by the policy network. The radar device is further configured to detect that a scene captured by the radar device is not reliably processable based on a distribution of the expected reward values generated by the plurality of heads.

## Claims

### Claim 1

A radar device, comprising:
a radar front end configured to send radar signals and to receive reflected radar signals;
processing circuitry configured to provide digital radar data based on the received reflected radar signals;
a digital filter configured to process the digital radar data to obtain information about objects which reflected the radar signals; and
a machine learning logic, comprising:
a policy network configured to set parameters of the digital filter based on the digital radar data,
a reward value generating network including a plurality of heads, each head configured to provide a respective expected reward value for a setting of parameters by the policy network, and
a detector configured to detect that a scene captured by the radar device is not reliably processable based on a distribution of expected reward values generated by the plurality of heads.

### Claim 2 (depends on 1)

The radar device of claim 1, wherein the radar device is configured to determine that the scene captured by the radar device is not reliably processable based on a comparison of the distribution of the expected reward values with predefined distribution parameters.

### Claim 3 (depends on 1)

The radar device of claim 1, wherein the radar device is configured to determine that the scene captured by the radar device is not reliably processable by determining if one of the expected reward values lies outside a predefined range about a mean value of the expected reward values.

### Claim 4 (depends on 1)

The radar device of claim 1, wherein the plurality of heads are configured to generate the respective reward values based on noise of the digital radar data.

### Claim 5 (depends on 1)

The radar device of claim 1, wherein the machine learning logic further comprises a context network configured to generate a context variable based on the digital radar data, the context variable characterizing a scene captured by the radar device, wherein the reward value generating network is configured to generate the respective expected reward value based on the context variable.

### Claim 6 (depends on 1)

The radar device of claim 1, wherein the information about the objected includes tracking information.

### Claim 7 (depends on 1)

The radar device of claim 1, wherein the digital filter comprises a Kalman filter.

### Claim 8 (depends on 1)

The radar device of claim 1, wherein the reward value generating network includes a critic network.

### Claim 9

A method of operating a radar device, comprising:
transmitting radar signals and receiving reflected radar signals;
providing digital radar data based on the received reflected radar signals;
digitally filtering the digital radar data in order to obtain information about objects that reflected the radar signals;
by a policy network of a machine learning logic, setting parameters of the digitally filtering based on the digital radar data;
by a reward value generating network of the machine learning logic including a plurality of heads, providing a respective expected reward value for each of the plurality of heads for a setting of parameters by the policy network; and
detecting that a scene captured is not reliably processable based on a distribution of the expected reward values generated by the plurality of heads.

### Claim 10 (depends on 9)

The method of claim 9, wherein determining that the scene captured is not reliably processable comprises comparing the distribution of the expected reward values with predefined distribution parameters.

### Claim 11 (depends on 9)

The method of claim 9, wherein determining that the scene captured is not reliably processable comprises determining if one of expected reward values lies outside a predefined range about a mean value of the expected reward values.

### Claim 12 (depends on 9)

The method of claim 9, wherein providing the respective expected reward value for each of the plurality of heads comprises generating the respective reward values based on noise of the digital radar data.

### Claim 13 (depends on 9)

The method of claim 9, further comprising generating, by a context network of the machine learning logic, a context variable based on the digital radar data, the context variable characterizing the captured scene, wherein providing the respective expected reward value for each of the plurality of heads comprises generating the respective reward values based on the context variable.

### Claim 14 (depends on 9)

The method of claim 9, wherein the information about the objects includes tracking information.

### Claim 15 (depends on 9)

The method of claim 9, wherein the digitally filtering comprises Kalman filtering.

### Claim 16

A radar system, comprising:
a Kalman filter configured to process digital radar data generated by a radar device to obtain information about objects represented by the digital radar data;
a policy neural network configured to set parameters of the digital filter based on the digital radar data; and
a critic neural network including a plurality of heads, each head configured to provide a respective expected reward value for a setting of parameters by the policy neural network; and
an out of distribution (OOD) detector configured to determine whether a scene captured by the radar device is not reliably processable based on a distribution of expected reward values generated by the plurality of heads.

### Claim 17 (depends on 16)

The radar system of claim 16, wherein the OOD detector is configured to determine that the scene captured by the radar device is not reliably processable based on a comparison of the distribution of the expected reward values with predefined distribution parameters.

### Claim 18 (depends on 16)

The radar system of claim 16, wherein the OOD detector is configured to determine that the scene captured by the radar device is not reliably processable by determining if one of the expected reward values lies outside a predefined range about a mean value of the expected reward values.

### Claim 19 (depends on 16)

The radar system of claim 16, wherein the plurality of heads are configured to generate the respective reward values based on noise of the digital radar data.

### Claim 20 (depends on 16)

The radar system of claim 16, wherein the policy neural network and the critic neural network each comprise a convolutional neural network.
