HI, I am at the exercise 3.12 where i had the husky bot to make circles for lower covariance from /amcl_pose; but after trying many times and i always land on an offset between the map(/map) and the laser(/scan) even when all 3 covariance are < 0.05
The amcl node come from the command: roslaunch husky_navigation amcl_demo.launch
Try to do an square movement (with some translation) instead of a circle, since it is better for localizing the robot. I believe this should fix this offset issue.
Increasing the number of particles to 10000 gave me the right result (that is no offset) every time. I suppose this gave the program more data to work on, compared to the mere 2000 max particles we were working with initially.