Reproduce “COAST: Constraints and Streams for Task and Motion Planning” 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
Brandon Vu, Toki Migimatsu, Jeannette Bohg
COAST, developed by Stanford University researchers, introduces a new Task and Motion Planning (TAMP) algorithm that integrates stream-based motion planning with constraint-based task planning. It significantly improves planning scalability by deferring the grounding of geometric stream objects until after a symbolic plan is proposed, leading to orders of magnitude faster planning times and higher success rates on complex robotic manipulation and navigation tasks.
Key insights
- The determination of specific geometric instances (stream objects) does not need to be handled by the PDDL solver or performed prior to initial symbolic task planning.
- Separating stream object resolution into a dedicated planning phase avoids the combinatorial explosion of stream objects within the core PDDL task state, keeping task planning light.
- Leveraging PDDL-compatible constraints provides an effective feedback mechanism from motion planning failures, efficiently guiding the task planner's search.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.