Thank you for opening this issue.
Currently the dynamic_obstacle_stop module does not use the predicted paths but uses a simple "constant heading + constant velocity" model to predict the future positions of the dynamic objects (see the documentation page).

I agree we may want to consider predicted paths of dynamic objects. How about adding two parameters to the dynamic_obstacle_stop module ? One parameter ignore_objects_behind_ego, and one parameter use_predicted_paths.

By the way, I am planning to move the dynamic_obstacle_module to the new motion_velocity_planner (still a work in progress, see #7064) to allow more accurate collision detection.

Thanks for your comment @maxime-clem
Yes, I am aware that dynamic_obstacle_stop_module does not use predicted_path.

I am considering proceeding as follows:

• First of all, I am thinking of adding a lambda function to filter_predicted_objects to take into account the objects coming from behind by adding a parameter named ignore_objects_behind_ego. So the objects behind the vehicle will also be included in the filtered_object, which is the output of this function.

• At this stage, I plan to add a parameter to use_predicted_path of objects and check the intersection of the EGO's trajectory with the predicted paths of the filtered objects.

What do you think?

Additionally, I agree with you. I think it would be more logical for the dynamic_obstacle_stop_module to be inside the motion_velocity_planner.

Your plan sounds good thank you 👍 If you need any help with developing these new features let me know. I will also be happy to review the PR once it is ready.