adyansh04/grove-g1/lost-on-the-map

The robot may be lost on the map

Localization can settle far off, even 90 degrees turned, and then Nav2 "arrives" at the wrong place. Everything after that fails as if perception were broken.

1. Call tf for map to pelvis and note x, y and the heading.
2. In simulation, call topic_sample on /g1_sensor_relay/base_state: it is where the robot

truly stands (its frame is labelled odom; it is the simulator's world within about 1 cm). The map starts where the robot stood when it was made, up to half a metre from that world's origin.

1. If the two disagree by more than about 1 m or 30 degrees, the robot is lost. Do not replan

the same mission: it would fail the same way.

1. Call look to see what the camera sees now, and compare it with what should be at the place.

Tell the operator the robot is mislocalized, by how much, and that it needs its pose set again (the 2D Pose Estimate in RViz, or a restart of the stack) before any walk.