![[2602.14968] PhyScensis: Physics-Augmented LLM Agents for Complex Physical Scene Arrangement](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.14968] PhyScensis: Physics-Augmented LLM Agents for Complex Physical Scene Arrangement
Summary
The paper introduces PhyScensis, a framework that uses physics-augmented LLM agents to generate complex 3D physical scenes for robotic manipulation, addressing challenges in spatial arrangement and physical accuracy.
Why It Matters
As robotics increasingly relies on realistic simulations for training and data collection, PhyScensis offers a significant advancement by enabling the generation of complex physical scenes that enhance the realism and effectiveness of robotic interactions. This innovation could improve robotic applications in various fields, including logistics, manufacturing, and autonomous systems.
Key Takeaways
- PhyScensis integrates LLM agents with a physics engine for scene generation.
- The framework addresses challenges like object density and spatial relationships.
- Experimental results show improved scene complexity and physical accuracy over previous methods.
- It allows fine control over scene parameters through probabilistic programming.
- This approach has potential applications in robotic manipulation and simulation.
Computer Science > Robotics arXiv:2602.14968 (cs) [Submitted on 16 Feb 2026] Title:PhyScensis: Physics-Augmented LLM Agents for Complex Physical Scene Arrangement Authors:Yian Wang, Han Yang, Minghao Guo, Xiaowen Qiu, Tsun-Hsuan Wang, Wojciech Matusik, Joshua B. Tenenbaum, Chuang Gan View a PDF of the paper titled PhyScensis: Physics-Augmented LLM Agents for Complex Physical Scene Arrangement, by Yian Wang and 7 other authors View PDF HTML (experimental) Abstract:Automatically generating interactive 3D environments is crucial for scaling up robotic data collection in simulation. While prior work has primarily focused on 3D asset placement, it often overlooks the physical relationships between objects (e.g., contact, support, balance, and containment), which are essential for creating complex and realistic manipulation scenarios such as tabletop arrangements, shelf organization, or box packing. Compared to classical 3D layout generation, producing complex physical scenes introduces additional challenges: (a) higher object density and complexity (e.g., a small shelf may hold dozens of books), (b) richer supporting relationships and compact spatial layouts, and (c) the need to accurately model both spatial placement and physical properties. To address these challenges, we propose PhyScensis, an LLM agent-based framework powered by a physics engine, to produce physically plausible scene configurations with high complexity. Specifically, our framework consists of three main com...
