Online Radar-LiDAR-Camera Calibration for Autonomous Vehicles

LG Innotek 2025 – 2027

Confidential project. Figures and results shown are from the published paper.

LiDAR points in the camera image: reference extrinsics (left) and the online estimate converging (right).

Overview

Platform
Passenger vehicle
Sensors
4D radar, LiDAR, camera
Environment
Urban driving
The online calibration framework (from the paper). I built all of it.
The online calibration framework (from the paper). I built all of it.

My Part

Lead developer, planning to evaluation

Targets, every module, vehicle data and ground truth, and the evaluation.

Rotation and translation errors vs. CMRNext (from the paper).
Rotation and translation errors vs. CMRNext (from the paper).

Joint optimization over all three pairs

All three extrinsics optimized together, so the strong camera-LiDAR pair stabilizes the radar pairs.

Lower error on all pairs; largest gain with radar

Making radar usable

An adaptive radar noise filter and accumulation of sparse radar matches over frames.