Test ArUco detection on the robot

[vvasco]

Given the results of this #180, we want to test the ArUco detection on the robot to understand up to which distance the detection is reliable. If such distance is long enough (> 1m between robot-line), the detection could be acceptable for the TUG, considered that:

• the line should be detected reliably just once;
• once detected, as the robot moves, it knows where the line is with respect to it and to the skeleton, even if the line is not visible.

I will test the detection for both resolutions 320x240 and 640x480 px.

[vvasco]

This is the detection we have on the robot when the line is at:

• ~2m from the robot, with a resolution of 320x240 px:

  

• ~2.5m from the robot:

  • with a resolution of 320x240 px:
  • with a resolution of 640x480 px:

The pose estimated at ~2.5m at a resolution of 640x480 px is the best one and the most stable (the estimation at same distance but lower resolution is instead rather unstable). However, the pose estimated at ~2m at a resolution of 320x240 px looks good as well and 2m distance can be sufficient for the TUG test. Furthermore, since we require a single robust detection, we can grab more poses to get a final robust pose. To sum up: line at ~2m at a resolution of 320x240 px seems the best option.

[vvasco]

This is using a board of 6x1 smaller markers:

The pose looks good. So far, this is the best board I could find (I tried smaller markers, but the line has to be too close to the robot), therefore I'll use this one as default.