Kinematic-Aware Robust LIO-W: A Tightly-Coupled Framework with Trajectory-Based Asynchronous Compensation for Degraded Environments

LiDAR–Inertial Odometry (LIO) can become unreliable in geometrically degenerate environments (e.g., long corridors and tunnels), where point-cloud registration provides weak constraints and certain degrees of freedom become poorly observable. Although wheel odometry offers high-rate planar motion cues, tightly fusing wheel, IMU, and LiDAR measurements in practice remains challenging due to heterogeneous noise characteristics and asynchronous sampling. This paper proposesKinematic-Aware Robust LIO-W, a hierarchical tightly-coupled wheel–inertial–LiDAR fusion framework for robust localization in feature-poor scenes. In a local high-frequency layer, wheel encoders and IMU pre-integration are jointly optimized to build a smooth analytical motion baseline with higher-order dynamics. Exploiting its differentiability, we perform trajectory-based nonlinear extrapolation to compensate for timestamp misalignment in low-rate exteroceptive measurements, mitigating errors introduced by conventional linear interpolation under dynamic maneuvers. Moreover, a kinematic-consistency module evaluates residuals on the Lie algebra and adaptively down-weights inconsistent exteroceptive factors, enabling graceful degradation when LiDAR constraints are corrupted. Extensive experiments on public datasets and real-world degenerate corridors demonstrate that the proposed method remains stable in cases where representative baselines become ill-conditioned. On the challenging M3DGR corridor sequences, our method achieves sub-meter end-to-end translation drift of 0.43–0.64 m, compared with 10.37–12.22 m for LIW-OAM; pure LIO baselines (LIO-SAM and FAST-LIO2) exhibit severe drift and may become unstable in these corridor runs. These results indicate a practical solution for reliable localization of autonomous wheeled robots in geometrically degraded environments.