publications

2026

1. arXiv

   

   Map2Route: Benchmarking Compositional Language-Grounded Route Planning over Semantic Maps

   Muyi Bao, Hang Xu, Jingfan Tang, and 5 more authors

   arXiv preprint arXiv:2609.17910, 2026

   Abs HTML PDF Code Website

   Map2Route is a human-curated benchmark for compositional language-grounded route planning over pre-built semantic maps. It contains 1,000 episodes across 40 scenes and evaluates ordered route requirements, scoped soft preferences, and path efficiency. The accompanying Grounding2Route framework combines executable code-as-grounding, verification-guided repair, and scope-aware planning.

2. RA-L

   

   SysNav: Multi-Level Systematic Cooperation Enables Real-World, Cross-Embodiment Object Navigation

   Haokun Zhu, Zongtai Li, Zihan Liu, and 8 more authors

   IEEE Robotics and Automation Letters, 2026

   Abs HTML PDF Code Website

   SysNav is a three-level object navigation system that decouples semantic reasoning, room-based navigation planning, and embodiment-specific motion control. It generalizes across wheeled, quadruped, and humanoid robots and reports 190 real-world experiments with strong gains in navigation efficiency over prior baselines.

3. arXiv

   Scene Change Detection with Vision-Language Representation Learning

   Diwei Sheng, Vijayraj Gohil, Satyam Gaba, and 5 more authors

   arXiv preprint arXiv:2604.11402, 2026

   Abs HTML PDF Code

   LangSCD introduces a vision-language framework for scene change detection that uses language-guided semantic reasoning, cross-modal feature enhancement, and geometric-semantic matching. The work also introduces NYC-CD, a large-scale dataset with multiclass scene-change annotations for real-world urban imagery.