FIRE-LIVWO: Failure-Immune mmWave Radar-Enhanced Odometry for Degenerate Underground Environments
Original title:FIRE-LIVWO: Robust LiDAR-Inertial-Visual-Wheel Odometry via Failure-Immune mmWave Radar Enhancement
Dense dust and long, repetitive corridors frequently cause conventional visual and LiDAR SLAM to drift or fail in subterranean coal mines. FIRE-LIVWO addresses this by tightly coupling 4D millimeter-wave radar and wheel odometry with visual-inertial and LiDAR inputs within an iterated error-state Kalman filter. Leveraging Doppler constraints and online observability analysis, the framework dynamically adjusts modality weights when specific sensors degrade. Real-world mine trials yielded an average localization error of 5.677 meters.
Why it's worth reading
Underground mines present extreme degenerate conditions where sensor redundancy fails; this work provides an open-source observability-guided switching architecture for real-world robotic deployments in harsh environments.