Decompose and Reorganize: Planning with Primitives and Visuomotor Policies Learned from Demonstrations
Decompose and Reorganize: Planning with Primitives and Visuomotor Policies Learned from Demonstrations
This research introduces DR-LfD, a framework that integrates visuomotor policies into a task and motion planning system to automate dexterous bimanual manipulation. By decomposing human demonstrations into atomic skills based on contact relationships, the method reduces data requirements for skill composition. Benchmarks show that DR-LfD outperforms diffusion-based imitation policies using significantly fewer demonstrations in unseen setups.
The paper proposes DR-LfD, a framework combining task and motion planning with visuomotor policies to improve data efficiency in bimanual manipulation, demonstrating superior performance on DexMimicGen benchmarks using fewer demonstrations compared to diffusion-based imitation policies.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
demonstrations
100 episodes
View original evidence
stration burden scales with the number of distinct skill types, with limited data for each skill. Real-world and simulation benchmarks demonstrate data-efficient skill composition under unseen setups and physical constraints. On two DexMimicGen tasks, DR-LfD with 100 demonstrations outperforms diffusion-based imitation policies with 1000.
Open source S5demonstrations
1,000
View original evidence
stration burden scales with the number of distinct skill types, with limited data for each skill. Real-world and simulation benchmarks demonstrate data-efficient skill composition under unseen setups and physical constraints. On two DexMimicGen tasks, DR-LfD with 100 demonstrations outperforms diffusion-based imitation policies with 1000.
Open source S5
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
stration burden scales with the number of distinct skill types, with limited data for each skill. Real-world and simulation benchmarks demonstrate data-efficient skill composition under unseen setups and physical constraints. On two DexMimicGen tasks, DR-LfD with 100 demonstrations outperforms diffusion-based imitation policies with 1000.
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org