Mapping & sensing
Feed MID-360 and D455 into FAST-LIVO2 for point-cloud maps and odometry.
← Back to projects
ROS · Final-year project
Multi-floor autonomous navigation
Connecting mapping, relocalization, and floor-aware mission management for indoor navigation on a Unitree robot.
- Period
- Aug 2025 — May 2026
- Role
- System development & robot deployment
- Institution & location
- Xi’an Jiaotong-Liverpool University · Suzhou
Advisor · Dr. Fanxin Wang
Background
Multi-floor navigation requires coordinated sensing, global localization, floor maps, and robot execution. This project integrates ROS components into a reproducible system that can be remapped and redeployed in new environments.
My contributions
- Integrated FAST-LIVO2, HDL Localization, and the ROS Navigation Stack, and deployed the system on a Unitree robot.
- Adapted Livox MID-360 and Intel RealSense D455, configuring sensor parameters, launch files, and point-cloud map saving.
- Connected point-cloud maps and TF transforms to enable relocalization in a prebuilt map.
- Developed floor-map switching, waypoint management, stair traversal, and mission-level coordination.
Technical approach
Localization & frames
Use HDL Localization to align global-map and odometry frames.
Floor-aware execution
Coordinate floor maps, stair connections, and waypoints, and bridge navigation commands to Go2.
Outcomes & evaluation
MID-360LiDAR
D455Vision sensor
ROSMapping-to-execution integration
- Established an end-to-end workflow from sensor bringup and map preparation to navigation execution.
- Published modular reproduction documentation covering new-environment mapping, floor configuration, and code changes.
Resources
Next project