← Back ICRA 2026

Degradation-Aware LiDAR-Thermal-Inertial SLAM

Yu Wang, Yufeng Liu, Lingxu Chen, Haoyao Chen, Shiwu Zhang

PDF

AI summary

Key figure (auto-extracted from paper)

DaLiTI dynamically weights LiDAR and thermal sensor data in real time, enabling robust state estimation in smoke-filled and GPS-denied disaster environments.

Adaptive fusion LiDAR-thermal-inertial SLAM Perceptual degradation Disaster relief robotics State estimation Multi-modal sensor fusion

Problem

Single-modality SLAM systems struggle in perceptually degraded disaster zones like smoke or darkness, while existing multi-modal approaches lack mechanisms to prevent corrupted sensor data from degrading overall state estimation.

Approach

The system loosely couples LiDAR, thermal infrared, and IMU measurements within an iterated error state Kalman filter, using a real-time degradation quantizer to adaptively weight each sensor's contribution based on perceived environmental quality.

Key results

Novel loosely coupled LiDAR-thermal-inertial SLAM framework
Real-time degradation quantizer for adaptive multi-modal fusion
Superior robustness and accuracy in smoke and gas leak scenarios
Public release of datasets and open-source code

Why it matters

Provides disaster response robots with reliable navigation and mapping capabilities when conventional sensors fail due to smoke, darkness, or lack of GPS.

Abstract

During robotic disaster relief missions, state estima- Q1 3 tion still faces significant challenges, especially when GNSS is 4 denied or sensor perception undergoes degradation. In this let- 5 ter, we introduce a degradation-aware LiDAR-Thermal-Inertial 6 SLAM, DaLiTI, that leverages the complementary nature of multi- 7 modal information to achieve robust and precise state estimation 8 in perceptually challenging environments. The system utilizes an 9 iterated error state Kalman filter (IESKF) to loosely integrate 10 LiDAR, thermal infrared camera, and IMU measurements. We 11 propose an adaptive fusion mechanism that dynamically weights 12 and fuses LiDAR and thermal measurements based on real-time 13 modal quality to prevent failure information from propagating 14 throughout the system. Experimental results demonstrate that, 15 compared with state-of-the-art methods, DaLiTI maintains com- 16 petitive performance in conventional environments and exhibits 17 superior robustness and accuracy in degraded scenarios such as 18 fire scenes or chemical plants with gas leaks. 19

Index terms

SLAM Search and Rescue Robots