SC-LIO: A Semi-Continuous-Time LiDAR-Inertial Odometry - Xingyu Chen - IEEE ICIEA 2026

LiDAR-inertial odometry is widely used for robot navigation and autonomous systems, but its accuracy is still sensitive to motion distortion caused by non-instantaneous scan acquisition. Most existing methods compensate this distortion in a separate preprocessing stage and then perform scan matching on the corrected point cloud, which ignores the fact that the deskewed points are still coupled with the underlying system state and inertial measurement uncertainty.

This paper presents SC LIO, a semi-continuous-time LiDAR-inertial odometry framework that maintains discrete-time anchor states while modeling intra-scan point motion continuously through local IMU-driven propagation. Based on this representation, motion undistortion is integrated directly into the state estimation process instead of being treated as an isolated front-end operation. We further construct an undistortion-aware point-to-plane measurement model for iterated error-state Kalman filtering so that the influence of scan-time motion is explicitly reflected in the residual linearization. This design preserves the compactness of discrete-time LIO while improving the consistency of scan registration under aggressive motion. 

The proposed framework is particularly suitable for resource-constrained odometry systems that require both efficiency and robustness. Experimental evaluation on representative handheld or mobile platform sequences will demonstrate the effectiveness of the proposed method in terms of accuracy and stability.