Motion dataset: How to find if two objects are on the same lane?

[18627242758]

Hello!

Currently I am using the Waymo motion dataset to extract car-following behaviors. In the dataset, there are some columns about the coordinates of the objects. But I wonder how to tell whether two vehicles are driving on the same lane.

[s-ettinger]

Hi. There is not a direct way to determine this. One way you could approach this is to find the nearest lane points to the centers of the vehicles and compare the feature IDs of those closest points to see if they are from the same lane feature.

[ghost]

Thanks for your reply!

Following your advice, if two vehicles are both closed to the same lane center, is that safe to say that they are in the same lane? The premise is that vehicles are keeping centered in their lanes.

[s-ettinger]

I think it will depend on how you define being in a lane, but proximity to the lane center is I think the best indicator. If two objects are very close to a point on the same map feature, the feature is continuous and has a unique ID so they will be close to the same lane center. Keep in mind that sometimes lanes may consist of multiple features (after branches, etc.) so depending on your application you may need to take this into account.

[ghost]

Thanks again.

I just have one more question. Currently I am using Scenario Protos. Is the direction of lane centers the same as the forwarding direction of vehicles?

[s-ettinger]

The direction of the lanes is not necessarily the same as the direction of vehicles. The lane direction can be determined by the vector between sequential points in the lane polylines.

[ghost]

But the lane center should be unidirectional, correct?

[Draupadin]

@18627242758 Did you get any closer to a solution to this task? I'm trying to solve the same problem and would be glad if you could share your progress.

[EnnaSachdeva]

Is there any update on this task?