Skip to content
Skild Robotics Operations

Mapping

Scope

Building and saving a 2D LiDAR floor map of a deployment area for the GO2W robot.

Purpose

A more accurate map produces more accurate autonomous navigation. A compromised map produces recurring localization loss. This SOP defines how to capture, clean, and finalize a map suitable for safe autonomy. [SOURCE: GO2_Standard_Operating_Procedure_V3 §III "The 5 Rules" Rule 3]

Required inputs / tools

  • Hardware: GO2W with clean LiDAR lens; battery at 100% before mapping begins. [SOURCE: SOP V3 §XI "Start Mapping on a Full Battery"]
  • Software: Skild tablet (controller cradled) OR browser at https://control.skild.ai/robots.
  • Permissions: Skild operator credentials.
  • Documents / channels: site map / floor plan for the area being mapped.

Procedure

Before

  1. Confirm the physical environment is in low-traffic state. Map only during low-traffic periods. [SOURCE: SOP V3 §XI]
  2. Confirm the LiDAR lens is clean (wipe with microfiber cloth).
  3. Confirm battery is at 100%. If the robot dies mid-map, all progress is lost — including glass marking. [SOURCE: SOP V3 §XI "Start Mapping on a Full Battery"]
  4. Complete the pre-power-on safety checklist (see /robots/go2/setup/pre-flight-checklist/).
  5. Power on robot and verify connection from tablet or browser.

During

  1. Initiate mapping.
    • On tablet: select the robot icon (top-right) → “Start” → “Start Mapping”.
    • On browser: click the control icon (top-right toolbar, square with joystick) to open Remote Controls → in “Start Mapping” section, click Create New Map.
  2. Name the map using convention: AreaName_Date (e.g. ConcoursA_04092026). [SOURCE: SOP V3 §XI]
  3. Teleop the robot through the space. The LiDAR builds a 2D occupancy grid in real time (each cell ≈ 5 cm).
  4. Move methodically. Avoid redundant passes. Ensure all static obstacles and boundaries are captured. [SOURCE: SOP V3 §XI "Executing the Map"]
  5. Do not capture dynamic objects: moving people, rolling carts, luggage, vehicles. Pause and wait for them to clear. [SOURCE: SOP V3 §XI]
  6. Watch for adequate coverage. White areas = navigable. Black shapes = obstacles. As scan density increases, obstacles should solidify (darker, more defined). Thin/patchy sections need another pass.
  7. Correct phantom obstacles. If the LiDAR registers a black shape that does not exist in real life, teleop the robot through that black area — the LiDAR will rescan and remove the false obstacle. This is the primary cleanup tool during mapping. [SOURCE: SOP V3 §XI "Correcting Map Artifacts"]
  8. Mode awareness (tablet only): the map view has Scroll Mode (green toggle, taps pan only) and Place Mode (red toggle, taps commit to the map). Confirm the toggle color before every interaction. On browser, the modifier is spacebar (hold spacebar + trackpad scroll = pan). [SOURCE: SOP V3 §XI "Switching Scroll Mode and Place Mode"]

After

  1. Review the complete map for accuracy and coverage. Look for patchy/thin sections — scan those again before saving.
  2. Save the floor map.
    • Tablet: “Save Floor Map” → select no portals or docks (unless specifically applicable) → “Finish Mapping”. [SOURCE: SOP V3 §XI "Saving the Map"]
  3. Glass marking (mandatory final phase).
    1. Select “Start Glass Marking”.
    2. Mark all transparent barriers the LiDAR could not detect (glass walls, windows, transparent partitions).
    3. Also use glass marking to geofence restricted areas where the robot must not navigate.
    4. When all marks are placed, select “Stop Glass Marking”.
    5. Select “Terminate Mapping”. Map is now saved. [SOURCE: SOP V3 §XI "Glass Marking"]

Quality checks

  • Completed map reads as a clean, detailed bird’s-eye blueprint of the space. [SOURCE: SOP V3 §XI "Quality Assessment"]
  • Obstacle outlines are solid and well-defined.
  • Open areas are continuous and gap-free.
  • No phantom obstacles remain.
  • All glass barriers / geofenced exclusion zones are marked.

Common failure modes

IssueLikely causeAction
Map has black blobs in empty spacePhantom obstacles from people / carts during scanTeleop through the blob; LiDAR rescans and clears it. [SOURCE: SOP V3 §XI]
Sections appear thin / patchyCoverage was inadequate from one directionRe-scan from another angle.
Robot dies mid-mappingBattery was not at 100% at startRestart mapping from scratch with a full battery. [SOURCE: SOP V3 §XI]
Recurring localization failures after deploymentMap captured dynamic objects, or environment has materially changedRe-map. Do not edit. Updating a map means creating a new one. [SOURCE: SOP V3 §XI "When to Remap" + "Update the Map Means Create a New One"]

When to re-map

  • Significant physical-environment changes (furniture, barriers, signage, equipment moved). [SOURCE: SOP V3 §XI]
  • Persistent localization failures that do not resolve through pose resets.
  • The map no longer accurately represents the static environment.

Escalation

  • Persistent mapping failures: post in the designated Skild Slack channel for the deployment.
  • Site-rule conflict (e.g., area you need to map is off-limits per site briefing): consult /sites/ before proceeding.

Source notes

  • [SOURCE: GO2_Standard_Operating_Procedure_V3.docx] — primary; sections III (5 Rules), XI (Mapping).
  • [SOURCE: GO2_Standard_Operating_Procedure_ABM.docx] (April 2026, Tyler Kugima) §XI Mapping. Authoritative for ABM-LGA.
  • [SOURCE: GO2_General_Operating_Instructions_SOP_MAPPING_EXPANDED.docx] — supplements; confirms “remap, don’t edit” and “clean phantom obstacles before saving”.