Method and assistance device for assisting driving operation of a motor vehicle, and motor vehicle

1 . A method for assisting at least partially automated driving operation of a motor vehicle, comprising:

recording, via a surroundings sensor system of the motor vehicle, surroundings raw data that map particular current surroundings of the motor vehicle;

generating, via an assistance device of the motor vehicle, semantic surroundings data from the surroundings raw data by carrying out semantic object recognition on the surroundings raw data;

acquiring predefined semantically annotated map data of the surroundings of the motor vehicle, in which static objects in the surroundings are indicated;

comparing the map data against the semantic surroundings data, wherein the static objects in the map data are identified in the semantic surroundings data; and

detecting discrepancies between the map data and the semantic surroundings data during to recognize perception errors of the assistance device;

wherein a recognized perception error prompts a predefined safety measure to be carried out, and

wherein a scene complexity for the respective surroundings is automatically assessed on the basis of the recognized discrepancies and/or perception errors and a scene complexity rating is generated based on the recognized discrepancies and/or perception errors, wherein a higher scene complexity rating is generated as a result of a greater number, a greater extent, and/or a greater severity of the recognized discrepancies and/or perception errors.

2 . The method according to claim 1 , wherein

the comparison involves at least some of the map data being projected into the respective semantic surroundings data or a surroundings model of the respective surroundings that is generated therefrom and a resultant object-by-object overlap being determined.

3 . The method according to claim 2 , wherein

a discrepancy and/or a perception error is recognized whenever the overlap between an object in the map data and, when present, an object in the semantic surroundings data or in the model generated therefrom that has been localized at an appropriate place is below a predefined threshold.

4 . The method according to claim 1 , wherein

the comparison and/or the recognized discrepancies is/are taken as a basis for assigning confidence values to the semantic surroundings data on an object-by-object basis.

5 . The method according to claim 1 , wherein

the map data comprise further details relating to the static objects beyond a semantic identification and a position of the static objects, and said further details are taken into consideration for the comparison and/or the at least partially automated driving operation of the motor vehicle.

6 . The method according to claim 5 , wherein

the further details include a size of the static objects.

7 . The method according to claim 1 , wherein

a perception error is recognized only when the applicable discrepancy has been recognized repeatedly, at least over a predefined period of time and/or at least over a predefined distance covered by the motor vehicle, on the basis of surroundings raw data recorded at different times and/or from different positions of the motor vehicle.

8 . The method according to claim 1 , wherein

the map data that are relevant to a particular current or impending journey section are automatically retrieved from a vehicle-external server device by degrees during driving operation of the motor vehicle, and

the particular relevant map data describes the surroundings of the motor vehicle in each case to a predefined extent beyond a range of the surroundings sensor system of said motor vehicle that is configured to record the surroundings raw data.

9 . The method according to claim 1 , comprising:

determining that the scene complexity rating exceeds a threshold; and

in response to determining that the scene complexity rating exceeds the threshold, storing the surroundings raw data and/or the semantic surroundings data for subsequent error analysis, improvement of the assistance device, and/or generation of training data for a machine learning device.

10 . An assistance device for a motor vehicle, comprising:

an input interface that acquires surroundings raw data and semantically annotated map data;

a data processor that processes the surroundings raw data and map data; and

an output interface that outputs a resultant result signal,

wherein the data processor is configured to:

record surroundings raw data that map particular current surroundings of the motor vehicle;

generate semantic surroundings data from the surroundings raw data by carrying out semantic object recognition on the surroundings raw data;

acquire predefined semantically annotated map data of the surroundings of the motor vehicle, in which static objects in the surroundings are indicated;

compare the map data against the semantic surroundings data, wherein the static objects in the map data are identified in the semantic surroundings data;

detect discrepancies between the map data and the semantic surroundings data to recognize perception errors of the assistance device, wherein a recognized perception error prompts a predefined safety measure to be carried out; and

automatically assess a scene complexity for the respective surroundings on a basis of the recognized discrepancies and/or perception errors, and generating a scene complexity rating based on the recognized discrepancies and/or perception errors, wherein a higher scene complexity rating results from a greater number, a greater extent, and/or a greater severity of the recognized discrepancies and/or perception errors.

11 . A motor vehicle, comprising:

a surroundings sensor system that records surrounding raw data; and

an assistance device according to claim 10 .

12 . The assistance device according to claim 10 ,

wherein the data processor is configured to:

determine whether the scene complexity rating exceeds a threshold; and

in response to determining that the scene complexity rating exceeds the threshold, store the surroundings raw data and/or the semantic surroundings data for subsequent error analysis, improvement of the assistance device, and/or generation of training data for a machine learning device.

13 . The assistance device according to claim 12 ,

wherein the data processor is configured to:

in response to determining that the scene complexity rating does not exceed the threshold, abstain from storing the surroundings raw data and/or the semantic surroundings data for subsequent use.

14 . The assistance device according to claim 10 ,

wherein the data processor is configured to:

determine that the scene complexity rating does not exceed a threshold; and

in response to determining that the scene complexity rating does not exceed the threshold, abstain from storing the surroundings raw data and/or the semantic surroundings data for subsequent use.