SIMPLE-C: Multi-LiDAR Extrinsic Calibration using a Simple Moving Reflective Target in Feature-Sparse Maritime Environments

Abstract Multi-LiDAR sensor configurations enhance perception for autonomous tasks in robotics compared to single sensors, but they require precise extrinsic calibration. Existing methods often depend on moving the platform or feature-rich environments, making them impractical in extremely sparse maritime settings where water surfaces yield unreliable LiDAR returns and vessels are expensive to move. This paper presents a practical tool for extrinsic online calibration of multiple synchronized LiDARs in such environments called SIMPLE-C (alibration). Our pragmatic approach detects a highly reflective target by combining intensity threshold and spatial clustering with backtracking and filtering. Singular Value Decomposition solves the extrinsics for closed-form point cloud registration. This enables calibration with arbitrary reflective objects like plastic lids with reflective tape or even conventional warning vests within the overlapping field of view. Adaptive outlier rejection and calibration weights enhance robustness. Our approach operates in real time and is quasi-invariant to target dimension or shape, and does not require target-specific geometric calculations, manual transformation inputs or additional sensors. Experiments demonstrate a pairwise calibration accuracy of 5 cm translational and $$0.3^{\circ }$$ 0 . 3 ∘ rotational Root Mean Square Error (RMSE) averaged on sparse water scenarios. On land-based scenarios, we achieve translational accuracy of 3 cm (RMSE). A comparison with state-of-the-art methods, including ICP variants, GMMCalib, CROON, and Fast Global Registration, consistently yields superior results improving translational error by up to 47 % specifically on the water side while achieving similar rotational errors. The code and dataset are publicly available.