Granted Patent B2

US 12,412,363 · App. 18/023,683 · Granted Sep 9, 2025

Inventor: Umair Rasheed (Wolfsburg, DE)

Assignee: Volkswagen Aktiengesellschaft

G06V10/25G06V20/56

Abstract

Claims (36)

1. A method for sensing an environment of a vehicle configured to drive in an at least semi-automated manner, comprising:

capturing an environment of the vehicle via at least one sensor to produce environment sensor data;

evaluating the environment sensor data via an evaluation unit using a first neural network at a first lower resolution to perform an environment sensing function; and

defining one or more focus regions in the environment sensor data;

evaluating the environment sensor data in the one or more defined focus regions via the evaluation unit using a second neural network, distinct from the first neural network, at a second higher resolution to perform the environment sensing function; and

combining the evaluated environment sensor data of the first lower resolution evaluation and the second higher resolution evaluation, wherein uncertainties of the first lower resolution evaluation and the second higher resolution evaluation are estimated and weighted during the combining to form an overall evaluation result, configured to affect a function of the vehicle.

2. The method according to claim 1 , wherein defining the one or more focus regions comprises defining the one or more focus regions via a rule-based system as a function of at least one of (i) a position of the vehicle, (ii) a planned trajectory of the vehicle, and/or (iii) an environment context.

3. The method according to claim 1 , wherein defining the one or more focus regions comprises defining the one or more focus regions as a function of detected temporal changes in the captured sensor data.

4. The method according to claim 1 , wherein defining the one or more focus regions comprises defining the one or more focus regions as a function of at least one detected movement in the captured sensor data.

5. The method according to claim 1 , wherein defining the one or more focus regions comprises defining the one or more focus regions via machine learning.

6. The method according to claim 1 , further comprising estimating an uncertainty of the first lower resolution evaluation and the second higher resolution evaluation during the combining of the evaluated environment sensor data.

7. The method according to claim 1 , further comprising assigning a priority to each of the one or more focus regions, and evaluating the environment sensor data in the one or more defined focus regions based on the assigned priority.

8. The method according to claim 1 , wherein evaluating the environment sensor data in the one or more defined focus regions via the evaluation unit using a second higher resolution comprises maintaining the evaluated environment sensor data in the one or more defined focus regions for a configured duration after the one or more defined focus regions is abandoned.

9. An apparatus for sensing an environment of a vehicle configured to drive in an at least semi-automated manner, comprising:

at least one sensor configured to capture an environment of the vehicle to produce environment sensor data;

an evaluation unit, operatively coupled to the at least one sensor, wherein the evaluation unit is configured to

evaluate the environment sensor data using a first neural network at a first lower resolution to perform an environment sensing function; and

define one or more focus regions in the environment sensor data;

evaluate the environment sensor data in the one or more defined focus regions using a second neural network, distinct from the first neural network, at a second higher resolution to perform the environment sensing function; and

combine the evaluated environment sensor data of the first lower resolution evaluation and the second higher resolution evaluation, wherein uncertainties of the first lower resolution evaluation and the second higher resolution evaluation are estimated and weighted during the combining to form an overall evaluation result, configured to affect a function of the vehicle.

10. The apparatus according to claim 9 , wherein the evaluation unit is configured to define the one or more focus regions by defining the one or more focus regions via a rule-based system as a function of at least one of (i) a position of the vehicle, (ii) a planned trajectory of the vehicle, and/or (iii) an environment context.

11. The apparatus according to claim 9 , wherein the evaluation unit is configured to define the one or more focus regions by defining the one or more focus regions as a function of detected temporal changes in the captured sensor data.

12. The apparatus according to claim 9 , wherein the evaluation unit is configured to define the one or more focus regions by defining the one or more focus regions as a function of at least one detected movement in the captured sensor data.

13. The apparatus according to claim 9 , wherein the evaluation unit is configured to define the one or more focus regions by defining the one or more focus regions via machine learning.

14. The apparatus according to claim 9 , wherein the evaluation unit is configured to estimate an uncertainty of the first lower resolution evaluation and the second higher resolution evaluation during the combining of the evaluated environment sensor data.

15. The apparatus according to claim 9 , wherein the evaluation unit is configured to assign a priority to each of the one or more focus regions, and evaluate the environment sensor data in the one or more defined focus regions based on the assigned priority.

16. The apparatus according to claim 9 , wherein the evaluation unit is configured to evaluate the environment sensor data in the one or more defined focus regions using a second higher resolution by maintaining the evaluated environment sensor data in the one or more defined focus regions for a configured duration, after the one or more defined focus regions is abandoned.

17. A method for sensing an environment of a vehicle configured to drive in an at least semi-automated manner, comprising:

capturing an environment of the vehicle via at least one sensor to produce environment sensor data;

evaluating the environment sensor data via an evaluation unit using a first neural network at a first lower resolution to perform an environment sensing function; and

defining one or more focus regions in the environment sensor data via machine learning;

evaluating the environment sensor data in the one or more defined focus regions via the evaluation unit using a second neural network, distinct from the first neural network, at a second higher resolution to perform the environment sensing function; and

combining the evaluated environment sensor data of the first lower resolution evaluation and the second higher resolution evaluation, wherein uncertainties of the first lower resolution evaluation and the second higher resolution evaluation are estimated and weighted during the combining to form an overall evaluation result, configured to affect a function of the vehicle.

18. The method according to claim 17 , wherein defining the one or more focus regions comprises defining the one or more focus regions via a rule-based system as a function of at least one of (i) a position of the vehicle, (ii) a planned trajectory of the vehicle, and/or (iii) an environment context.

19. The method according to claim 17 , wherein defining the one or more focus regions comprises defining the one or more focus regions as a function of detected temporal changes in the captured sensor data.

20. The method according to claim 17 , wherein defining the one or more focus regions comprises defining the one or more focus regions as a function of at least one detected movement in the captured sensor data.

Assignments (1)

Priority Claims (1)

Continuity (1)

References Cited (19)