Single tf noise

[brunofavs]

Hey, this PR comes as a request from #1006.

It is a sub-PR of #989, containing a smaller set of features.

Features :

• Add noise to a specific transformation.
• Overhaul of noise adding mechanism ( Refactor only, functionality remains the same).
• Rotation noise now uses uniform distribution instead of choice.

  atom/atom_core/src/atom_core/dataset_io.py

  Lines 923 to 924 in c571011

  	v = np.random.uniform(-1.0, 1.0, 3)
  	v = v / np.linalg.norm(v)

This was the least amount of features I could add to implement specific transformation noise. There were implemented only the strictly necessary dependencies.

I tested with Softbot and it worked well. I used the `-ctgt' flag to ensure the noise added was correct.

A MWE example to test the feature would be something like this. Any softbot dataset shall work as long as the most recent config.yml is used. The one in atom examples works.

rosrun atom_calibration calibrate \                                                                                                         
-json ${Dataset_of_Choice} \                                                                                                                   
-v  -ctgt \                                                                                                                                    
-nig 0.1 0.1 \                                                                                                                           
-ntfv 0.1 0.1 \                                                                                                                                
-ntfl "world:base_footprint"                                                                                                                  

[miguelriemoliveira]

Hey @brunofavs ,

thanks for the PR.

I inspected it and it seems fine. Even though you tried to compartmentalize, still it affects 9 files!

I cannot make sure it has no side effects, but let's accept it and then be on the lookout for those.

[brunofavs]

Many of those 9 files are minor things though. 2 of them are just due to the renaming of a function, 2 are xacros that I actually uploaded on accident. If something were to break, I would bet it would have to do with dataset_io.py.

But yeah, lets be on the lookout for unexpected things, better safe than sorry.