SNAIL-Radar-Indoor:A large-scale handheld LiDAR–4D radar–IMU dataset for robust SLAM in degraded environments

This is the authors' abstract. We don't add key points for this paper.

Multimodal sensing, particularly 4D imaging radar, is promising for emergency search and rescue because it remains informative under smoke and highly reflective structures. However, existing public radar datasets predominantly record outdoor vehicle motion or clean indoor spaces. We present SNAIL-Radar-Indoor, a large-scale handheld LiDAR–4D radar–IMU dataset for SLAM in smoke and other degraded environments. Extending our previous outdoor SNAIL-Radar benchmark, the new release contains 217 sequences totaling approximately 850 min, collected at 13 sites across six scene types, including offices, abandoned buildings, halls, factory workshops, underground parking areas, and connecting outdoor spaces. The sequences cover corridors, rooms, staircases, narrow corners, indoor–outdoor transitions, reflective structures, dynamic crowds, and real smoke. Data were recorded using a compact self-developed integrated platform comprising a 3D LiDAR, a low-cost single-chip 4D imaging radar, a consumer-grade IMU, and an optional barometer. Repeated traversals and closed paths are included where feasible to support place recognition and loop-closure evaluation. The released trajectories are method-generated references and do not constitute independent absolute ground truth. Smoke-free sequences mainly use loop-consistent LIO references; fully smoke-dominated sequences use an iRIOM-based radar–IMU reference with AHRS and barometer constraints; and partially degraded sequences use a smoke-aware piecewise RLIO reference. Closed sequences additionally include start–end LiDAR registration constraints, while non-closed sequences are accompanied by map-based end-to-end displacement checks. The dataset supports research on radar and multimodal registration, smoke-robust odometry and mapping, denoising, environmental perception, and navigation in degraded environments.