Improve usability and documentation

[mfehr]

After getting some feedback from our first "customers" there are serveral things that need to be improved:

• Give the user a feedback of how good/bad the calibration is.
  Suggestions/Ideas:
  • Run the python evaluation step after the optimization and optionally plot the results
  • Write a summary pdf at the end of the calibration with the most important plots:
    • initial time alignment
    • initial 3D trajectory, after dual quaternion calibration, after optimization
    • RMSE plot after dual quaternion, after optimization
  • Provide a separate script to plot the 3D trajectory before and after some given calibration
  • Add a short wiki description that describes the 3 steps (time alignment, coarse initial calibration, optimization) and what the plots mean.

[HannesSommer]

What are the good solutions for plotting the trajectories? I'd love to have something nice here.

[mfehr]

We already have functions that plot the 3D trajectories, but TBH the 3D plotting tools of matplotlib are not very nice to look at. Actually I just recently started using mayavi/mlab in python and it is very easy to use and a lot nicer and faster (you can still move the 3D plot even with 1000s of elements in it). Here is sth i did there:

I've already implemented the plotting of a pose graph with coordinate frames at the nodes, so I could transfer this.

[HannesSommer]

Ah, yes, I totally forgot about these 3d plots - I think I could hardly use them on Trusty. Sorry. mayavi/mlab looks very interesting.

[ffurrer]

@mfehr if you can easily transfer them, it of course would be awesome :)