Todos

[Jmeyer1292]

• [x] Add pictures to the rct_examples readme
• [x] Increase the robustness and visibility of the rct_examples for offline data processing. Make them roslaunch files that take well described yaml files with lots of pose info.
• [x] In the example calibrations, print out the results in URDF xyz, rpy format
• [x] Schedule a review.

[Jmeyer1292]

• [ ] Add test data sets. They are large, ~50MB, and I'd rather not add to Git. Where to host?
• [x] Add unit testing - This is an example of something that makes a lot of sense to test constantly.
• [ ] Provide verification routines
  • [x] Visualize the residuals for points on the target
  • [x] Re-draw the projected (and maybe initial points)?
  • [ ] Provide an example of computing error on a test set (different from training). Add functions necessary for this task.
  • [ ] It'd be really cool to have a "live-stream" of reprojection error and position "delta" - As you jog you get feedback on the calibration accuracy.
• [ ] Need to improve the "command line" data collector
  • [x] Documentation
  • [ ] Not use special linux headers to make directories (use boost instead?)

Further Layers While the math functions are nice, the data collection process is often the most difficult. I would like to see:

• [ ] GUI alternative to the command line data collector
  • Could make extensible - Interface for acquiring data, displaying data
  • Could be plugins? E.g. write a plugin to capture your sensors kind of data.

• [ ] A 'path planner' for calibration motions.
  • Generate sample positions for a given calibration
  • Generate robot motions to achieve this pose subject to constraints
  • Visibility
  • Collision
  • Relative joint limits (cable windup is a huge problem with these cal procedures)

Other cool calibrations

• [ ] Laser line scanner to wrist
• [ ] Robot kinematics