TravExplorer: Cross-Floor Embodied Exploration via Traversability-Aware 3D Planning

Han Zheng¹, Zhe Chen¹, Yudong Huang¹, Haoran Liu¹, Jinghao Wang¹, Ming Yang¹, Tong Qing¹,

¹Shanghai Jiao Tong University,

Key Contributions

3D Traversable Frontier

We pioneer a traversable frontier-based exploration paradigm for ground robots, which unifies 2D ray casting for efficiency and 3D semantic traversability for multi-floor awareness, while leveraging active perception to mitigate FOV-limited occlusions.

Spatial Semantic Guidance

We design spatial semantic guidance that couples a probabilistic instance map from open-vocabulary semantic segmentation with a spatial value map from fast image-to-text matching, enabling zero-shot search to focus on semantically promising regions for efficient exploration.

Hierarchical Cross-Floor Planning

We develop a hierarchical cross-floor exploration planner that integrates TSP-based global tours, foothold-guided 3D path search, an execute-review mechanism, and vertically constrained local planning, enhancing both search efficiency and traversal robustness.

Large-Scale Validation

We conduct extensive evaluations, including 50 real-world experiments and 4,195 simulated episodes on HM3D and MP3D benchmarks, achieving state-of-the-art performance. To the best of our knowledge, this is the first system that efficiently handles both single-floor and cross-floor object navigation through 3D traversability mapping.