System and method for locating the position of a road object by unsupervised machine learning

1. A vehicle system for locating the position of at least one road object associated with a portion of a road network on which the vehicle is traveling, the vehicle system comprising:

at least one vehicle sensor for sensing a road object; and

a vehicle processor configured to:

receive a plurality of geolocated road information items which relate to the road object, the plurality of road information items being obtained from the at least one sensor of the vehicle, and received from a plurality of road vehicles,

identify the type of the road object from the plurality of road information items and calculate a plurality of geographic coordinates associated with the road object, each geographic coordinate being calculated from the plurality of road information items,

create, in a first creation step, for each type of road object, a first characteristic vector comprising the type of the road object and the associated plurality of geographic coordinates,

partition, in a first partitioning step, on the basis of an unsupervised classification algorithm based on distribution density of the plurality of geographic coordinates in the first characteristic vector for each type of the road object, at least a portion of the set of first characteristic vectors into a plurality of classes, and

for each class determined in the first partitioning step:

create, in a second creation step, a second characteristic vector comprising at least one component for each geographic coordinate of the class, each component being representative of an azimuth angle of a direction of observation of the road object associated with the corresponding geographic coordinate, partition, in a second partitioning step, on the basis of the unsupervised classification algorithm based on distribution density of the plurality of geographic coordinates in the second characteristic vector for each type of the road object, at least a portion of the set of second characteristic vectors into a plurality of subclasses, and

calculate, in a calculating step, for each subclass, at least one measure of central tendency from the geographic coordinates associated with the subclass, the measure of central tendency being representative of the position of the road object in a direction of observation of the road object as indicated by the azimuth angle of the direction of observation of the road object; and

control the vehicle based on the measure of central tendency of the road object.

2. The system as claimed in claim 1 , wherein the vehicle processor is further configured to separate, in a first separation step of the first creation step, each geographic coordinate into a position component in terms of latitude and a position component in terms of longitude.

3. The system as claimed in claim 1 , wherein the vehicle processor is further configured to separate, in a second separation step of the second creation step, each azimuth angle into an angular component in terms of sine and an angular component in terms of cosine.

4. A vehicle method for locating the position of at least one road object associated with a portion of a road network on which the vehicle is traveling, the type of road object of which is identified from a plurality of geolocated road information items which relate to the road object, the plurality of road information items being obtained from at least one vehicle sensor associated with each one of a plurality of road vehicles, the road object further being associated with a plurality of geographic coordinates, each geographic coordinate being calculated, by a vehicle processor, from the plurality of road information items associated with one of the plurality of road vehicles, the method comprising:

receiving, by the processor, a plurality of geolocated road information items which relate to the road object, the plurality of road information items being obtained from the at least one sensor of the vehicle, and received from a plurality of road vehicles, identifying, by the processor, the type of the road object from the plurality of road information items and calculating a plurality of geographic coordinates associated with the road object, each geographic coordinate being calculated from the plurality of road information items,

executing, by the processor, a first creation step in which is created, for each type of road object, a first characteristic vector comprising the type of the road object and the associated plurality of geographic coordinates,

executing, by the processor, a first partitioning step in which is partitioned, on the basis of an unsupervised classification algorithm based on distribution density of the plurality of geographic coordinates in the first characteristic vector for each type of the road object, at least a portion of the set of first characteristic vectors into a plurality of classes,

for each class determined in the partitioning step:

executing, by the processor, a second creation step in which is created a second characteristic vector comprising at least one component for each geographic coordinate of the class, each component being representative of an azimuth angle of a direction of observation of the road object associated with the corresponding geographic coordinate,

executing, by the processor, a second partitioning step in which is partitioned, on the basis of the unsupervised classification algorithm based on distribution density of the plurality of geographic coordinates in the second characteristic vector for each type of the road object, at least a portion of the set of second characteristic vectors into a plurality of subclasses,

executing, by the processor, in a calculating step, for each subclass, at least one measure of central tendency from the geographic coordinates associated with the subclass, the measure of central tendency being representative of the position of the road object in a direction of observation of the road object as indicated by the azimuth angle of the direction of observation of the road object, and

control the vehicle based on the measure of central tendency of the road object.

5. The method as claimed in claim 4 , wherein the first creation step further comprises a first separation step in which each geographic coordinate is separated into a position component in terms of latitude and a position component in terms of longitude.

6. The method as claimed in claim 4 , wherein the second creation step further comprises a second separation step in which each azimuth angle is separated into an angular component in terms of sine and an angular component in terms of cosine.

7. The method as claimed in claim 4 , wherein the unsupervised classification algorithm based on distribution density is of a type chosen from: DBSCAN, OPTICS, CLARANS, DENCLUE and CLIQUE.

8. The method as claimed in claim 4 , wherein the measure of central tendency of a class or of a subclass is chosen from: a centroid and a medoid.

9. The system as claimed in claim 1 , wherein the unsupervised classification algorithm based on distribution density is of a type chosen from: DBSCAN, OPTICS, CLARANS, DENCLUE and CLIQUE.

10. The system as claimed in claim 1 , wherein the measure of central tendency of a class or of a subclass is chosen from: a centroid and a medoid.

11. The system as claimed in claim 2 , wherein the vehicle processor is further configured to separate, in a second separation step of the second creation step, each azimuth angle into an angular component in terms of sine and an angular component in terms of cosine.

12. The method as claimed in claim 5 , wherein the second creation step further comprises a second separation step in which each azimuth angle is separated into an angular component in terms of sine and an angular component in terms of cosine.