LeHome: A Simulation Environment for Deformable Object Manipulation in Household Scenarios
Zeyi Li, Yushi Yang, Shawn Xie, Kyle Xu, Tianxing Chen, Yuran Wang, Zhenhao Shen, Yan Shen, Yue Chen, Wenjun Li, Yukun Zheng, Chaorui Zhang, Siyi Lin, Fei Teng, Hongjun Yang, Ming Chen, Steve Xie, Ruihai Wu · 2026
Read LeHome: A Simulation Environment for Deformable Object Manipulation in Household Scenarios by Zeyi Li, Yushi Yang, Shawn Xie, Kyle Xu, Tianxing Chen, Yuran Wang, Zhenhao Shen, Yan Shen, Yue Chen, Wenjun Li, Yukun Zheng, Chaorui Zhang, Siyi Lin, Fei Teng, Hongjun Yang, Ming Chen, Steve Xie, Ruihai Wu (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-22363.txt.
Introduction
Li and seventeen colleagues introduce LeHome, a simulation tool for household deformable objects. Such objects, from garments to food items, are hard to handle due to varied shapes and types. Existing simulators also lack good deformable-object support. LeHome offers high-fidelity dynamics and lifelike interactions across many such objects. It supports many robot bodies, with low-cost robots as a core focus, for end-to-end task tests on constrained gear. LeHome thus provides a scalable testbed for pushing household robotics.
Catalog record
| CREATOR | Zeyi Li, Yushi Yang, Shawn Xie, Kyle Xu, Tianxing Chen, Yuran Wang, Zhenhao Shen, Yan Shen, Yue Chen, Wenjun Li, Yukun Zheng, Chaorui Zhang, Siyi Lin, Fei Teng, Hongjun Yang, Ming Chen, Steve Xie, Ruihai Wu |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 24589 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.22363
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
LeHome: A Simulation Environment for Deformable Object Manipulation in Household Scenarios Zeyi Li, Yushi Yang, Shawn Xie, Kyle Xu, Tianxing Chen, Yuran Wang, Zhenhao Shen, Yan Shen, Yue Chen, Wenjun Li, Yukun Zheng, Chaorui Zhang, Siyi Lin, Fei Teng, Hongjun Yang, Ming Chen, Steve Xie, Ruihai Wu Abstract Household environments present one of the most common, impactful yet challenging application domains for robotics. Within household scenarios, manipulating deformable objects is particularly difficult, both in simulation and real-world execution, due to varied categories and shapes, complex dynamics, and diverse material properties, as well as the lack of reliable deformable-object support in existing simulations. We introduce LeHome, a comprehensive simulation environment designed for deformable object manipulation in household scenarios. LeHome covers a wide spectrum of deformable objects, such as garments and food items, offering high-fidelity dynamics and realistic interactions that existing simulators struggle to simulate accurately. Moreover, LeHome supports multiple robotic embodiments and emphasizes low-cost robots as a core focus, enabling end-to-end evaluation of household tasks on resource-constrained hardware. By bridging the gap between realistic deformable object simulation and practical robotic platforms, LeHome provides a scalable testbed for advancing household robotics. Webpage: lehome-web.github.io/ . I INTRODUCTION For humans, the household scenario is the living space with the highest frequency of daily activities and the most critical needs. It encompasses a wide variety of essential daily tasks, such as organizing personal belongings, preparing food, and managing clothing. Unlike those in more structured scenes such as industries, these tasks feature distinct challenges: interactions with diverse, non-standardized objects and adaptation to unstructured, dynamic environments. TABLE I : Features Comparisons across different open-source robotics environment. Feature LeHome (ours) RoboTwin2.0 DexGarmentLab Behavior-1K Libero RLbench Robocasa Household Scenarios ✓ ✗ ✓ ✓ ✓ ✗ ✓ Photorealistic Rendering ✓ ✓ ✓ ✓ ✗ ✗ ✓ Food Deformation Simulation ✓ ✗ ✗ ✗ ✗ ✗ ✗ Flame and Particle Simulation ✓ ✗ ✗ ✓ ✗ ✗ ✗ Fluid Simulation ✓ ✗ ✓ ✓ ✗ ✗ ✗ Garment Manipulation ✓ ✗ ✓ ✓ ✗ ✗ ✗ Articulation and Rigid Manipulation ✓ ✓ ✗ ✓ ✓ ✓ ✓ Multi-material Manipulation ✓ ✗ ✓ ✓ ✗ ✗ ✗ Teleoperation ✓ ✗ ✗ ✓ ✗ ✗ ✓ Low-cost Robot ✓ ✗ ✗ ✗ ✗ ✗ ✗ Amid these challenges, household scenarios have attracted increasing attention, ranging from the development of simulated household simulation platforms [ 15 , 16 , 28 ] to the design of robotic policies for household tasks [ 1 , 14 , 36 ] . However, a key mismatch remains between current benchmarks and real-world needs: most existing methods target rigid and articulated objects, while many household tasks inherently involve diverse deformable objects ( e.g. , garments and food), for which current support remains limited. Such objects lack fixed shapes, deform nonlinearly under applied forces, and exhibit dynamic physical parameters that are difficult to characterize. Consequently, creating large-scale, realistic, and diverse deformable-object datasets for training robust policies remains a major challenge, due to two fundamental issues: (i) collecting real-world household data is prohibitively expensive and labor-intensive, as deformable objects’ variable states and the inherent complexity of household environments make it difficult to obtain sufficient high-quality data; and (ii) achieving accurate and authentic modeling of deformation is intrinsically hard, since it requires simultaneously capturing complex material properties, nonlinear dynamics, and realistic interactions. As shown in Fig. , we present LeHome , a household simulation environment designed for high-fidelity manipulation of diverse deformable objects. LeHome provides a household asset library spanning liquids, gaseous fluids, granular objects, linear objects, thin shells, and volumetric objects. To model these categories, LeHome combines appropriate simulation engines, including Position-Based Dynamics (PBD), Finite Element Method (FEM), and Eulerian Fluid Simulation, to improve physical realism while preserving broad task coverage. Furthermore, to enable complex manipulation mechanisms, such as a sausage being cut into multiple pieces with a knife. LeHome introduces a graphical logical modeling method called the Action Graph , representing causal mechanisms. This method enforces causal consistency between actions and outcomes, ensuring the simulation aligns with real-world dynamics. In summary, these designs allow LeHome to provide a reliable physical foundation for deformable object manipulation and establish a strong basis for generating high-quality training data. In terms of supported robot platforms, LeHome spans multiple robotic embodiments, ranging from mainstream commercial manipulators (e.g., UR, Franka) to open-source low-cost platforms (e.g., LeRobot, XLeRobot). Compared with widely-used industrial robots, low-cost robots are more compact, easier to deploy, and cheaper to maintain. These properties make them better suited for scalable deployment in household environments. Therefore, LeHome places particular emphasis on supporting a diverse set of open-source, low-cost robot arms, which are easy to set up, modify, and maintain, while still providing sufficient manipulation capability for everyday household tasks. By lowering the hardware and deployment barriers, this emphasis enables scalable, repeatable validation of core household-robot functionalities in realistic home environments, and in turn supports broader adoption and large-scale deployment. Finally, to further strengthen LeHome in both scalable data acquisition and user-friendliness, we integrate diverse teleoperation methods, offering multiple devices to fit different demands and allowing anyone to operate it intuitively. Moreover, it is compatible with both virtual and real scenarios, enabling simulation datasets to be seamlessly supplemented with real-world demonstrations under the same pipeline, which in turn supports efficient collection of manually annotated data. Meanwhile, we employ domain randomization to augment existing demonstrations by injecting randomized scene variations during trajectory replay, thereby automatically generating more diverse data with richer visual appearances and interaction conditions. Real-robot evaluations validate that LeHome supports effective sim-to-real transfer for diverse deformable object manipulation tasks, and that domain randomization mitigates distribution shift, resulting in more robust real-world deployment. II RELATED WORK Fig. 2 : The Architecture of LeHome . (Left) LeHome Assets can deliver realistic simulations of different robots, articulation/rigid objects, deformable objects, and cover multiple household scenarios. (Right) Leveraging diverse simulation methods and mechanisms, LeHome Engine enables various simulation capabilities. (Bottom) Subsequently, LeHome Benchmark utilizes these assets to construct tasks, conduct domain randomization, and facilitate teleoperation data collection. II-A Simulation and Manipulation in Household Scenarios Household scenarios are particularly valuable for robotic simulation because they closely reflect the complexities of real-world settings. Existing simulation platforms have established a solid foundation for related research: many [ 4 , 8 , 22 , 20 , 13 ] show practical value in building desktop-level or single-object scenes, effectively supporting targeted research. Some platforms [ 28 , 15 ] further extend scene coverage to more diverse scenarios, yet their visual fidelity remains limited, falling short of near-photorealistic realism. Additionally, studies [ 16 , 29 , 23 , 9 ] have actively addressed these limitations, however, high-fidelity interactions with deformable objects ( e.g. , garments, food) remain challenging. These deformable objects are ubiquitous in daily life, and accurately simulating their manipulation is critical to realistic household scenarios. To address these existing gaps, we introduce the LeHome platform, which integrates both flexibility and scalability. As Tab. I , Lehome incorporates high-fidelity assets across various room types and enables complex interactions with deformable objects ( e.g. , folding garments and cutting sausages), thereby effectively addressing a key limitation in current simulation platforms. II-B Deformable Object Simulation and Manipulation Robotic manipulation of deformable objects has recently gained attention,