Device and a method for distinguishing between traversable and nontraversable objects

1. A device ( 1 ) for distinguishing between traversable and non-traversable objects, said device ( 1 ) comprising:

an object-detection sensor unit ( 2 ) configured to monitor a space around the object-detection sensor unit ( 2 ) and to generate data about the space,

a map generation unit ( 3 ) configured to use the data generated by the object-detection sensor unit ( 2 ) in order to generate an occupancy grid map and a free space grid map for the space around the object-detection sensor unit ( 2 ), wherein each of the occupancy grid map and the free space grid map comprises an array of cells, each cell of the occupancy grid map contains an occupied space probability value representing the probability for the respective cell being occupied and each cell of the free grid map contains a free space probability value representing the probability for the respective cell being free, and

a classifier unit ( 4 ) configured to classify at least one of the cells as containing a traversable or non-traversable object by calculating a difference of two free space probability values derived from the free space probability values of at least two of the cells adjacent to the at least one cell and by using the difference as well as the occupied space probability value of the at least one cell to classify the at least one cell.

2. The device ( 1 ) as claimed in claim 1 ,

wherein a plurality of groups of cells that are adjacent to the at least one cell are pre-determined,

wherein the cells of each group are located adjacent to each other,

wherein the classifier unit ( 4 ) is further configured to calculate for each group the average value of the free space probability values of the cells of the respective group and to determine the highest average value and the lowest average value, and

wherein the difference of the two free space probability values calculated by the classifier unit ( 4 ) is the difference between the highest average value and the lowest average value.

3. The device ( 1 ) as claimed in claim 2 ,

wherein each group contains a pre-determined number of cells,

wherein each group contains different cells, and

wherein the number of the groups is equal to the number of cells located adjacent to the at least one cell.

4. The device ( 1 ) as claimed in claim 1 ,

wherein the object-detection sensor unit ( 2 ) comprises a radar unit, in particular a 2D radar unit.

5. The device ( 1 ) as claimed in claim 1 ,

wherein the classifier unit ( 4 ) is further configured to classify a plurality of the cells as containing a traversable or non-traversable object, and

wherein the classifier unit ( 4 ) is configured to classify only those cells, the occupied space probability value of which is equal to or greater than a pre-determined value.

6. The device ( 1 ) as claimed in claim 1 ,

wherein the classifier unit ( 4 ) comprises a classifier model,

wherein a feature vector is input into the classifier model for each cell and the classifier model uses the feature vector to classify the respective cell as containing a traversable or non-traversable object,

wherein the feature vector for a respective cell comprises a first feature and a second feature, and

wherein the first feature is derived from the occupied space probability value of the respective cell and the second feature is derived from the difference of the two free space probability values derived from the free space probability values of the at least two cells adjacent to the respective cell.

7. The device ( 1 ) as claimed in claim 6 ,

wherein the classifier model is established by a machine learning algorithm.

8. A vehicle comprising the device ( 1 ) as claimed in claim 1 .

9. A method for distinguishing between traversable and non-traversable objects, said method comprising:

using an object-detection sensor unit ( 2 ) to monitor a space around the object-detection sensor unit ( 2 ) and to generate data about the space;

generating an occupancy grid map and a free space grid map for the space around the object-detection sensor unit ( 2 ) by means of the data generated by the object-detection sensor unit ( 2 ), wherein each of the occupancy grid map and the free space grid map comprises an array of cells, each cell of the occupancy grid map contains an occupied space probability value representing the probability for the respective cell being occupied and each cell of the free space grid map contains a free space probability value representing the probability for the respective cell being free; and

classifying at least one of the cells as containing a traversable or non-traversable object by calculating a difference of two free space probability values derived from the free space probability values of at least two of the cells adjacent to the at least one cell and by using the difference as well as the occupied space probability value of the at least one cell to classify the at least one cell.

10. The method as claimed in claim 9 ,

wherein a plurality of groups of cells that are adjacent to the at least one cell is pre-determined,

wherein the cells of each group are located adjacent to each other,

wherein for each group the average value of the free space probability values of the cells of the respective group is calculated and the highest average value and the lowest average value are determined, and

wherein the calculated difference of the two free space probability values is the difference between the highest average value and the lowest average value.

11. The method as claimed in claim 10 ,

wherein each group contains a pre-determined number of cells,

wherein each group contains different cells, and

wherein the number of the groups is equal to the number of cells located adjacent to the at least one cell.

12. The method as claimed in claim 9 ,

wherein the object-detection sensor unit ( 2 ) comprises a 2D radar unit.

13. The method as claimed in claim 9 ,

wherein a plurality of the cells is classified as containing a traversable or non-traversable object, and

wherein in particular only those cells are classified, the occupied space probability value of which is equal to or greater than a pre-determined value.

14. The method as claimed in claim 9 ,

wherein a classifier model is provided,

wherein a feature vector is input into the classifier model for each cell and the classifier model uses the feature vector to classify the respective cell as containing a traversable or non-traversable object,

wherein the feature vector for a respective cell comprises a first feature and a second feature, and

wherein the first feature is derived from the occupied space probability value of the respective cell and the second feature is derived from the difference of the two free space probability values derived from the free space probability values of the at least two cells adjacent to the respective cell.

15. The method as claimed in claim 14 ,

wherein a machine learning algorithm is used to establish the classifier model.