Woowa Bros. Robot Delivery Challenge
GPS-denied localization with 3D LiDAR + IMU and collision-risk-aware DWA planning using pedestrian motion estimation
Overview
For the Woowa Bros. Robot Delivery Challenge, I focused on reliable navigation in complex environments by stabilizing GPS-denied localization and improving local planning safety with collision-risk-aware DWA. :contentReference[oaicite:29]{index=29}
What I Did
- GPS-denied localization stabilization
- Stabilized localization by fusing:
- 3D LiDAR-based pose (against a prebuilt 3D map)
- IMU measurements :contentReference[oaicite:30]{index=30}
- Stabilized localization by fusing:
- Pedestrian motion estimation and safety-aware planning
- Estimated pedestrian speed and direction from YOLOv5 + LiDAR data.
- Integrated collision risk into DWA to achieve safer local path planning. :contentReference[oaicite:31]{index=31}
My Role
- Implemented the localization fusion approach for GPS-denied operation
- Built the pedestrian motion feature extraction and risk integration into DWA planning :contentReference[oaicite:32]{index=32}
Tech Stack
- Localization: 3D LiDAR-based pose on prebuilt 3D map + IMU fusion
- Perception: YOLOv5 + LiDAR-based pedestrian motion estimation
- Local Planning: DWA with collision-risk augmentation :contentReference[oaicite:33]{index=33}
Media
Replace placeholders with:
- localization plot (drift before/after)
- pedestrian motion vectors overlay
- DWA trajectory samples with risk scores
Suggested visuals: localization robustness, motion estimation overlays, and safer trajectory selection via risk-aware DWA.
Takeaway
This project demonstrates practical autonomy for real deployment conditions:
- robust localization when GPS is unreliable
- safe navigation among pedestrians via risk-aware planning :contentReference[oaicite:34]{index=34}