Reproduce “Object-Centric Task and Motion Planning in Dynamic Environments” 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
Toki Migimatsu, Jeannette Bohg
Researchers at Stanford University developed an object-centric Task and Motion Planning framework that optimizes relative Cartesian frames, allowing robots to robustly execute complex manipulation sequences in dynamic environments. This approach enables real-time adaptation through reactive controllers, even with noisy perception and imprecise control, eliminating the need for computationally expensive replanning during execution.
Key insights
- Optimizing relative Cartesian frames yields plans robust to dynamic environments, perception inaccuracies, and control imprecision, as local controllers can adapt in real-time.
- Defining manipulation actions in object-centric Cartesian space simplifies constraint definitions and decouples geometric constraints across timesteps in complex sequences.
- Closed-form Jacobians for the end-effector's absolute world pose, derived from a kinematic tree of relative frames, enable efficient gradient-based nonlinear optimization.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.