A map begins as measurements. A quadruped's navigation stack typically combines LiDAR, which returns precise distances in every direction, with depth cameras that add dense shape information at shorter range, and an inertial unit that tracks the robot's own motion between sensor frames. Fused together, these streams describe the world as geometry: walls, columns, kerbs, stair edges, parked equipment. The fusion matters because each sensor fails differently - LiDAR sees through darkness but not glass, depth cameras read surfaces LiDAR glances off - and a navigation system that depends on one sensor inherits that sensor's blind spots.
Start With Sensing: The Raw Material of a Map
Booster's quadruped platforms carry this sensing as part of the flagship industrial quadruped fit, and the same perception approach scales across the inspection line. The point for buyers is not the sensor list itself but what the list produces: a continuous, metrically accurate picture of the operating space, refreshed as the robot moves.
From Points to Places: Building the 3D Map
Raw sensor returns are point clouds - millions of disconnected measurements. Mapping software stitches successive clouds into a single consistent 3D model by matching overlapping geometry, correcting drift with loop closure when the robot revisits a place, and separating the stable structure of the site from things that move. The result is not a picture but a working model: floor levels, obstacles, passages and no-go volumes, stored at a resolution that lets a route planner reason about clearances for a machine of a specific size.
Three-dimensional mapping earns its cost wherever sites stop being flat. Stairs, loading ramps, cable trays and mezzanines are invisible or ambiguous in a 2D floor plan; in a 3D map they are simply terrain. For a legged robot whose whole value is going where wheels cannot, the map has to represent the vertical world as faithfully as the horizontal one - otherwise the robot's mobility advantage is planned away.
Localisation: Knowing Where You Are in Your Own Map
A map is only useful if the robot knows where it stands inside it. Localisation continuously matches live sensor data against the stored map to fix the robot's position and heading, while odometry from legs and inertial sensing bridges the gaps between matches. Good localisation is quiet infrastructure: patrols repeat to the same inspection points, readings are taken from the same angles, and routes stay consistent shift after shift - the repeatability that makes robot-collected data comparable over time.
Degraded localisation announces itself honestly in the data: viewpoints drift, readings wander, routes vary. When evaluating an embodied navigation robot, ask how the system behaves in feature-poor spaces - long uniform corridors, open yards - and what it does when confidence drops. A professional system slows, re-acquires, or stops and calls for help; it does not guess its way down a staircase.
Planning Routes and Handling the Unexpected
On top of the map and the fix sits the planner: given a destination or a patrol route, it threads a path through free space, then a local layer adjusts that path around whatever the world has added since - a parked trolley, a closed gate, a person crossing. For inspection work the planner also serves viewpoints: the robot does not merely pass an asset, it positions itself at the marked point, at the marked distance, facing the marked gauge, because the reading is the product.
Booster's industrial inspection and public safety solution is built around this discipline - routes, points and readings defined once, executed repeatably. Buyers comparing systems should watch a live re-plan, not a clean demo run: block the route, move an obstacle, and see whether the robot recovers gracefully or waits for rescue.
What to Ask Before You Buy the Map
Four questions separate real navigation from choreography. Was the map built by the robot itself, on a site like yours, or by engineers with survey equipment? How are map updates handled when the site changes - a new partition, a relocated cabinet? What positioning repeatability does the supplier claim at inspection points, and how was it measured? And what happens, exactly, when localisation is lost mid-route? The answers, demonstrated live, are the difference between an autonomous patrol asset and an expensive remote-controlled one.
Mapping your site for a quadruped patrol or inspection pilot? Compare the relevant product range and deployment scenarios to take the next step.
