map a space

⏱ ~5 min record

Walk a room, save a kiss-icp map, reload + localize next session.

Prereqs: lidar spinning, ~5 min clear floor, operator walks alongside (0.2 m/s).

1 · record

> start SLAM, I'll guide you around. Save it as 'living-room' when done.

[ g1_lidar_switch(on=True), g1_slam_start() ]   # wait ~3s
g1_slam_reset()                                  # fresh map

Then walk it through the space (walk forward 2m, turn 90 right, …), then:

> save the map as 'living-room'      →  g1_slam_save(name="living-room")

2 · localize later

> load the 'living-room' map and tell me where you are

[ g1_lidar_switch(on=True), g1_slam_start(), g1_slam_load(name="living-room") ]
pose = g1_slam_pose()   # after ~5s ICP converge

Localized in 'living-room' · x=1.23 y=0.47 yaw=12° · near the door.

list maps

g1_slam_list_maps()   # → [{name, size_kb, points, saved}, ...]

tips

• Rooms need visual diversity for ICP (bare walls drift — add a few obstacles).
• ICP uses ~15-20% of one Jetson core. Maps are pointclouds, not meshes.