Robust Thermal-Radar-Inertial SLAM for Fireground Environments
A fireground-oriented localization system for perception-degraded environments where smoke, water spray, fire, darkness, and weak geometry can destabilize conventional odometry. The system uses thermal, radar, and inertial sensing to maintain robot pose estimates across NTU, HTTC, CDA, and warehouse validation sequences.
Overview
A fireground-oriented localization system for perception-degraded environments where smoke, water spray, fire, darkness, and weak geometry can destabilize conventional odometry. The system uses thermal, radar, and inertial sensing to maintain robot pose estimates across NTU, HTTC, CDA, and warehouse validation sequences.
Details
Fireground Localization Context
Fireground autonomy first depends on reliable pose. Dense maps, operator interfaces, and higher-level navigation all become fragile if the robot cannot maintain localization while moving through smoke, water spray, fire, darkness, or visually textureless interiors. The same sensing background as the dense mapping project applies here, but the emphasis shifts from scene reconstruction to real-time state estimation.
In these environments, conventional RGB- or LiDAR-dominant odometry can lose the assumptions it relies on. Smoke suppresses visible texture, water droplets scatter geometric measurements, and fireground layouts often contain reflective, low-texture, or partially occluded structures. The project therefore treats robust localization as the core capability required before mapping and autonomy can be trusted.
Thermal-Radar-Inertial Backbone
The system is built around thermal, 4D radar, and inertial sensing because the modalities fail differently. Thermal imagery keeps heat-based scene cues available when visible appearance degrades. Radar contributes geometric and motion-related evidence under smoke, dust, rain, and poor illumination. Inertial sensing keeps motion continuous through short intervals where external observations become unreliable.
The important point is the effect: the robot keeps an aligned pose estimate in conditions where a single sensing modality can become unreliable.
Validation Across Degraded Sites
The evaluation covers indoor-outdoor NTU sequences, HTTC fireground-style trials, CDA smoke/fire scenarios, and warehouse tests. These settings include heavy rain, semi-outdoor fire, mist, smoke, water spray, dynamic pedestrians, stairs, outdoor-to-indoor transitions, and warehouse smoke.
Across the evaluated sequences with ground truth, the system tracks trajectories from short indoor loops to several-hundred-meter runs. Representative results include:
| Setting | Representative Result |
|---|---|
| NTU-B6 | 57.9 m trajectory, 0.115 m APE RMSE |
| HTTC-seq1 / seq2 / seq3 | 336-628 m trajectories, 0.564-0.688 m APE RMSE |
| CDA-seq1 / seq2 / seq3 | 182-190 m trajectories, 0.595-1.375 m APE RMSE |
| Warehouse time trials | 31.8-58.5 m trajectories, 0.147-0.284 m APE RMSE |
| Warehouse smoke trial | 164.8 m trajectory logged under smoke without ground-truth reference |
Fireground-Oriented Outcome
The project is positioned as the localization foundation for the broader fireground robotics payload. It supports robot pose feedback, map alignment, and operator situational awareness in scenes where teleoperation alone is limiting and single-modality perception is too brittle.
Together, these results position robust localization as the real-time backbone for later mapping, navigation, and human-robot coordination in degraded fireground environments.