search

awpollon / cs685_locobot_blocks

0 stars 0 forks source link

"Kidnapped Robot Problem" Demonstration #6

Open awpollon opened 1 year ago

awpollon commented 1 year ago

See how well our localization handles a kidnapped robot.

(Proof I didn't make this up: https://en.wikipedia.org/wiki/Kidnapped_robot_problem)

Proposed procedure:

  1. Have robot move around in sight of landmarks to establish landmark data.
  2. While stationary, pick the robot up and move it to the other side of the landmarks, still in sight.
  3. Move around a little establish the new locations relative to landmarks (maybe try to capture estimated pose and confidence as it moves)
  4. Direct back home and see how well it does.

Could experiment with more landmarks, as well as changing the noise models (does a higher variance in the odometry noise do better?)

This is not necessarily expected to be successful, but a fun experiment to see how well landmark localization does.