FIRE-LIVWO: Robust LiDAR-Inertial-Visual-Wheel Odometry via Failure-Immune mmWave Radar Enhancement

· Editorial Team estimated
SLAM odometry mmWave-radar multi-modal-fusion mining-robotics

FIRE-LIVWO is a failure-immune multi-modal odometry framework for large-scale underground coal mines. Built on an iterated error-state Kalman filter (IESKF), it tightly couples 4D mmWave radar, LiDAR, visual features, and wheel odometry inside a unified VoxelMap, jointly constructing LiDAR-radar point-to-plane residuals and sparse visual photometric residuals. In smoke-filled environments it exploits mmWave radar penetration with pointwise Doppler velocity constraints to preserve observability; in geometrically degenerate corridors it couples wheel odometry via non-holonomic constraints with online lever-arm compensation. Its central contribution is a degeneration-detection and adaptive fusion-model switching strategy grounded in geometric and visual observability analysis that quantifies observability online and dynamically adjusts modality weights. Real-world underground coal-mine experiments show reliable failure-boundary identification and modality switching with a 5.677 m average localization error, outperforming baselines; code is open-sourced.

Paper · arXiv:2609.05325

Large-scale underground coal mines are hostile to SLAM: dense smoke and dust strip away visual information and degrade LiDAR point-cloud features, while long, self-similar corridors induce geometric degeneration and pronounced odometry drift. Most multi-modal systems fuse sensors with fixed weights and have no way to know when a modality has become untrustworthy, so they diverge precisely in the conditions that matter most. FIRE-LIVWO treats “which sensor to trust right now” as a first-class problem, estimating system observability online and switching fusion strategies at failure boundaries.

Core Innovation

FIRE-LIVWO is a tightly coupled multi-modal odometry framework built on an iterated error-state Kalman filter (IESKF). It fuses 4D mmWave radar, LiDAR, and visual features within a unified VoxelMap, jointly constructing LiDAR-radar point-to-plane residuals and sparse visual photometric residuals. In smoke-filled environments, it exploits the strong penetration of 4D mmWave radar and adds pointwise Doppler velocity constraints to preserve state observability. In geometrically degenerate corridors, it tightly couples wheel odometry using non-holonomic constraints (NHC) with online lever-arm compensation to reduce drift. The central contribution is a degeneration-detection and adaptive fusion-model switching strategy grounded in geometric and visual observability analysis: the system quantifies observability online and dynamically adjusts modality weights, making the fusion itself failure-immune.

Results

  • Real-world experiments in underground coal mines show the system accurately identifies failure boundaries and performs reliable modality switching under extreme conditions.
  • Superior accuracy and robustness over baselines, with an average localization error of 5.677 m.
  • Source code is open-sourced on GitHub for community reproduction.

Limitations

The abstract reports a single average localization error (5.677 m) without a per-condition breakdown across smoke density or corridor geometry, so the boundary of the claimed gains is hard to assess independently. The system requires a rich sensor suite (radar, LiDAR, camera, and wheel encoders), raising the hardware barrier. Details of the observability quantification, switching thresholds, and comparisons against recent radar-inertial systems need the full paper.

Industry Implications

Mining automation is one of the highest-value proving grounds for mmWave-radar-enhanced fusion: the autonomy of expensive underground equipment (longwall, roadheader, and auxiliary vehicles) has long been blocked by unreliable localization in dust and degenerate geometry. FIRE-LIVWO’s observability-aware, on-demand switching philosophy transfers directly to tunnel construction, firefighting and search, and underground logistics. For system integrators, it offers a concrete template for making multi-sensor estimation degrade gracefully, and the open-source release lowers the cost of entry for the whole field.