All projects
SLAM & Autonomous Navigation
Mapping, localization and navigation for legged robots in real buildings.
- SLAM
- ROS 2
- Navigation
Overview
We gave legged robots the ability to map a building, know where they are in it and walk to a goal on their own.
The stack runs on quadrupeds such as the Unitree Go2 and B2W and on humanoids with a lidar head.
The Challenge
Legged robots shake and pitch as they walk, which degrades lidar and camera data.
Real buildings have glass, stairs, people and furniture that moves.
Our Approach
Lidar-inertial SLAM tuned for legged motion, with map saving and relocalization.
ROS 2 Nav2 navigation with costmaps adapted to legged footprints and step-able terrain.
A simple mission interface: pick points on the map and the robot patrols or delivers.
Results
- Reliable autonomous patrols in multi-room indoor spaces (placeholder — add metrics)
- Portable across Unitree Go2, B2W and G1 with configuration only
Tech Stack
- ROS 2
- Nav2
- Lidar SLAM
- Unitree Go2
- Unitree B2W
- C++
