> ## Documentation Index
> Fetch the complete documentation index at: https://dimensionalos.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Navigation

> Goal-driven navigation: three ways to send a goal, live mapping or premap relocalization

Navigation turns a **goal** into planned, replanned motion. You can send that goal three ways - click a point in the viewer, call `set_goal` from Python, or say "go to the kitchen" through the agent - and all three drive the same stack. Start with [Using navigation](/capabilities/navigation/using-navigation).

<img src="https://mintcdn.com/dimensionalos/Zonji0eBDna-Xcoy/capabilities/navigation/assets/noros_nav.gif?s=f6a27a4fd79e590bd24f4fc5300480a3" alt="Live Go2 navigation in Rerun" width="1058" height="640" data-path="capabilities/navigation/assets/noros_nav.gif" />

Under the hood, the Go2 stack uses a **column-carving voxel map**: each new LiDAR frame replaces the corresponding region of the global map entirely, so the map always reflects the latest observations. Map live as you drive, or return to a known space using a saved premap and relocalization.

## Choose your workflow

| Workflow                    | When to use                                           | Blueprint                    | Docs                                                          |
| --------------------------- | ----------------------------------------------------- | ---------------------------- | ------------------------------------------------------------- |
| **Live mapping**            | Explore a new space where the map updates every frame | `unitree-go2`                | [Using navigation](/capabilities/navigation/using-navigation) |
| **Premap + relocalization** | Return to a known space and plan on a loop-closed map | `unitree-go2-relocalization` | [Relocalization](/capabilities/navigation/relocalization)     |

Live column-carving maps are fast and reactive, but odometry drifts over long distances. For spaces you revisit, record once, run pose-graph optimization (PGO) offline, then relocalize against the exported premap at runtime.

Internals (voxel mapper, costmap, planner data flow): [How navigation works](/capabilities/navigation/deep_dive). For hardware setup, simulation, and the full blueprint list, see the [Go2 platform guide](/platforms/quadruped/go2/index).