Reproduce “Assemble Them All: Physics-Based Planning for Generalizable Assembly by Disassembly” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yunsheng Tian, Jie Xu, Yichen Li, Jieliang Luo, Shinjiro Sueda, Hui Li, Karl D.D. Willis, Wojciech Matusik
Researchers from MIT CSAIL and Autodesk Research developed "Assemble Them All," a physics-based framework for generating assembly sequences and motions by leveraging disassembly. This method achieves high success rates on a new large-scale dataset of thousands of industrial assemblies, notably disassembling a 53-part engine in 5 minutes, and outperforms geometric baselines in handling complex, constrained motions.
Key insights
- Physics-based simulation, particularly when guided by an assembly-by-disassembly strategy, can robustly explore and navigate highly constrained motion spaces and complex geometries by leveraging approximate force directions.
- Employing an SDF-enhanced physics engine is crucial for accurately simulating contact-rich interactions involving fine details like screw threads and highly concave shapes in real-world CAD models.
- A progressive BFS strategy in multi-part disassembly planning efficiently manages computational resources by prioritizing unblocked parts and gradually increasing search depth, preventing wasted effort on blocked components.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.