NM-LIO: Multiple LiDAR-Inertial Odometry Addressing LiDAR Measurement Noise Discrepancy
This study addresses the accuracy degradation in multi-LiDAR odometry caused by heterogeneous sensor measurement noise. To this end, we propose a noise-aware multi-LiDAR-inertial odometry framework. This work is the first to explicitly model inter-sensor measurement noise inconsistencies across multiple LiDARs. By integrating a dedicated noise model, the proposed approach quantifies the noise level of each individual LiDAR and adaptively captures dynamic discrepancies based on residual uncertainty, thereby enabling robust sensor fusion. Extensive experiments conducted on public datasets demonstrate that, compared with existing state-of-the-art methods, the proposed system significantly improves both estimation accuracy and robustness across diverse complex environments.