Robotics Engineer · Visual SLAM & Vision-and-Language Navigation
Teaching robots to see where they are — and to follow instructions to where they should go.
👁️ Visual SLAM (VSLAM) — my primary focus
- Feature-based & learning-based front-ends, illumination-resilient deep features
- Visual-Inertial Odometry (VIO), stereo & voxel-map based estimation
- Multi-sensor fusion: LiDAR-Visual-Inertial (LVIO) tightly-coupled systems
- Loop closure, pose-graph & factor-graph back-end optimization
🧭 Vision-and-Language Navigation (VLN) — where I'm heading
- Instruction-guided navigation and embodied agents
- Foundation models for perception, mapping & decision-making
- Bridging classical geometric SLAM with language-conditioned policies
🛰️ Foundations I build on
- LiDAR-Inertial Odometry (LIO), real-time mapping, coverage path planning
- Sensor calibration (LiDAR-camera), 3D navigation stacks, legged/mobile platforms
| Project | What it does |
|---|---|
| fast-lio-loop | FAST-LIO extended with loop closure for drift-free LiDAR-inertial SLAM |
| cartographer_detailed_comments_ws | Google Cartographer workspace annotated line-by-line for study |
| complete_coverage | Online complete-coverage path planning for unmanned surface vehicles |
| small_car | ROS-based autonomous ground vehicle (lingao_ros) |
| pupper | Experiments on an open-source quadruped robot |
Curated from my stars — the work shaping how I think about perception and navigation:
Visual / Visual-Inertial SLAM
sair-lab/AirSLAM— efficient hybrid visual SLAM (TRO 2025)ZikangYuan/voxel_svio— stereo VIO on a voxel mapkoide3/glim_ros2— versatile 3D range-based SLAMdatawhalechina/smoothly-vslam— open VSLAM tutorial & exercises
LiDAR-Visual-Inertial Fusion
IF-A-CAT/LIR-LIVO— illumination-resilient LVIO with deep featuressjtuyinjie/Ultra-Fusion— tightly-coupled multi-sensor fusion SLAMxuankuzcr/FAST-LIVO2-RTK— FAST-LIVO2 with GNSS/RTK
Vision-and-Language Navigation & Embodied AI
qiyuanlanniao/OmniVLN— omnidirectional vision-language navigationNVlabs/GR00T-WholeBodyControl— humanoid whole-body controlRobbyant/lingbot-map— feed-forward 3D foundation model for scene reconstruction
🌴 Currently on vacation — feel free to open an issue, I'll get back to you.