Best way to organize perceptive pipeline [closed]

edit

Hello,

I have prototyped my perceptive pipeline in PCL, and now I have to port it in ROS.

I'm trying to figure out which is the best tradeoff between efficiency and time/effort to port my pipeline.

I see that there are mainly three options:

• organize everything using pcl_ros nodelets
• exploit the Ecto framework
• The third option consists in a single (or a small number of) common ros nodes executing the pipeline.

It is to be said that I don't need the pipeline to run in "real-time": I don't have to track anything in real-time. My robot perceives the scene, performs some operations without perceptive feedback, and perceives again the scene to understand how he had modified the environment. So an asynchronous paradigm based on services could be ok for my purposes, and some extra copies of point clouds could be admissible.

What are your thoughs about my issue?

Thank you very much

Cheers