Research map / 3D vision / Gaussian splatting
Habitat-GS: A High-Fidelity Navigation Simulator with Dynamic Gaussian Splatting
Ziyuan Xia, Jingyi Xu, Chong Cui, Yuanhong Yu, Jiazhao Zhang, Qingsong Yan, Ni Tao, Junbo Chen, Xiaowei Zhou, Hujun Bao, Ruizhen Hu, Sida Peng
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| train on 100 GS scenes instead of 100 mesh scenes (ablation) | DD-PPO PointNav agent · PointNav GS test | SR (%): 61.30 -> 70.70 | Improved |
| train on 100 GS scenes instead of 100 mesh scenes (ablation) | DD-PPO PointNav agent · PointNav Mesh test | SR (%): 59.00 -> 53.00 | Hurt |
| mix 20 mesh and 80 GS scenes instead of 100 mesh (ablation) | DD-PPO PointNav agent · PointNav GS test | SR (%): 61.30 -> 79.60 | Improved |
| mix 20 mesh and 80 GS scenes instead of 100 mesh (ablation) | DD-PPO PointNav agent · PointNav Mesh test | SR (%): 59.00 -> 59.60 | No clear effect |
| mix 50 mesh and 50 GS scenes instead of 100 GS only (ablation) | DD-PPO PointNav agent · PointNav Mesh test | SR (%): 53.00 -> 61.80 | Improved |
| mix mesh and GS scenes (E) instead of mesh only (ablation) | DD-PPO PointNav agent · Real video self-recorded | APA±1 (%): 55.6 -> 61.2 | Improved |
| mix mesh and GS scenes (D) instead of mesh only (ablation) | DD-PPO PointNav agent · Real video InCrowd-VI | APA±1 (%): 20.4 -> 29.6 | Improved |
| render with 3DGS instead of mesh | Habitat-GS scenes · Human ranking study | pairwise wins (%): 38.1% -> 61.9% | Improved |
| fine-tune with gaussian avatars (ablation) | PointNav agent · GS scene+avatar test | collision rate (%): 6.713 -> 4.746 | Improved |
| fine-tune with gaussian avatars (ablation) | PointNav agent · Mesh scene+avatar test | collision rate (%): 2.521 -> 2.342 | Improved |
| fine-tune with gaussian avatars (ablation) | PointNav agent · GS scene+avatar test | SPL (%): 71.35 -> 65.72 | Hurt |
| increase scene Gaussian count from 1M to 5M (ablation) | Habitat-GS renderer · Rendering benchmark | FPS: 120.9 -> 51.46 | Hurt |
Closest papers by meaning
- Image2Sim: Scaling Embodied Navigation via Generative Neural Simulator (NeurIPS 2026 · similarity 0.90)
- Habitat Synthetic Scenes Dataset (HSSD-200): An Analysis of 3D Scene Scale and Realism Tradeoffs for ObjectGoal Navigation (CVPR 2024 · similarity 0.89)
- Habitat 3.0: A Co-Habitat for Humans, Avatars, and Robots (ICLR 2024 · similarity 0.88)
- Learning to Navigate Efficiently and Precisely in Real Environments (CVPR 2024 · similarity 0.86)
- Synthesizing Photorealistic and Dynamic Urban Environments for Multimodal Robot Navigation and Collaboration (NeurIPS 2025 · similarity 0.86)
- SimWorld Studio: Automatic Environment Generation with Evolving Coding Agent for Embodied Agent Learning (NeurIPS 2026 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
