$\textttSPIN$: distilling $\textttSkill-RRT$ for long-horizon prehensile and non-prehensile manipulation

Haewon Jung, Donguk Lee, Haecheol Park, Kim Jun Hyeop, Beomjoon Kim
Proceedings of The 9th Conference on Robot Learning, PMLR 305:1311-1351, 2025.

Abstract

Current robots struggle with long-horizon manipulation tasks requiring sequences of prehensile and non-prehensile skills, contact-rich interactions, and long-term reasoning. We present $\texttt{SPIN}$ ($\textbf{S}$kill $\textbf{P}$lanning to $\textbf{IN}$ference), a framework that distills a computationally intensive planning algorithm into a policy via imitation learning. We propose $\texttt{Skill-RRT}$, an extension of RRT that incorporates skill applicability checks and intermediate object pose sampling for solving such long-horizon problems. To chain independently trained skills, we introduce $\textit{connectors}$, goal-conditioned policies trained to minimize object disturbance during transitions. High-quality demonstrations are generated with $\texttt{Skill-RRT}$ and distilled through noise-based replay in order to reduce online computation time. The resulting policy, trained entirely in simulation, transfers zero-shot to the real world and achieves over 80% success across three challenging long-horizon manipulation tasks and outperforms state-of-the-art hierarchical RL and planning methods.

Cite this Paper

BibTeX
@InProceedings{pmlr-v305-jung25a, title = {$\texttt{SPIN}$: distilling $\texttt{Skill-RRT}$ for long-horizon prehensile and non-prehensile manipulation}, author = {Jung, Haewon and Lee, Donguk and Park, Haecheol and Hyeop, Kim Jun and Kim, Beomjoon}, booktitle = {Proceedings of The 9th Conference on Robot Learning}, pages = {1311--1351}, year = {2025}, editor = {Lim, Joseph and Song, Shuran and Park, Hae-Won}, volume = {305}, series = {Proceedings of Machine Learning Research}, month = {27--30 Sep}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v305/main/assets/jung25a/jung25a.pdf}, url = {https://proceedings.mlr.press/v305/jung25a.html}, abstract = {Current robots struggle with long-horizon manipulation tasks requiring sequences of prehensile and non-prehensile skills, contact-rich interactions, and long-term reasoning. We present $\texttt{SPIN}$ ($\textbf{S}$kill $\textbf{P}$lanning to $\textbf{IN}$ference), a framework that distills a computationally intensive planning algorithm into a policy via imitation learning. We propose $\texttt{Skill-RRT}$, an extension of RRT that incorporates skill applicability checks and intermediate object pose sampling for solving such long-horizon problems. To chain independently trained skills, we introduce $\textit{connectors}$, goal-conditioned policies trained to minimize object disturbance during transitions. High-quality demonstrations are generated with $\texttt{Skill-RRT}$ and distilled through noise-based replay in order to reduce online computation time. The resulting policy, trained entirely in simulation, transfers zero-shot to the real world and achieves over 80% success across three challenging long-horizon manipulation tasks and outperforms state-of-the-art hierarchical RL and planning methods.} }
Endnote
%0 Conference Paper %T $\textttSPIN$: distilling $\textttSkill-RRT$ for long-horizon prehensile and non-prehensile manipulation %A Haewon Jung %A Donguk Lee %A Haecheol Park %A Kim Jun Hyeop %A Beomjoon Kim %B Proceedings of The 9th Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2025 %E Joseph Lim %E Shuran Song %E Hae-Won Park %F pmlr-v305-jung25a %I PMLR %P 1311--1351 %U https://proceedings.mlr.press/v305/jung25a.html %V 305 %X Current robots struggle with long-horizon manipulation tasks requiring sequences of prehensile and non-prehensile skills, contact-rich interactions, and long-term reasoning. We present $\texttt{SPIN}$ ($\textbf{S}$kill $\textbf{P}$lanning to $\textbf{IN}$ference), a framework that distills a computationally intensive planning algorithm into a policy via imitation learning. We propose $\texttt{Skill-RRT}$, an extension of RRT that incorporates skill applicability checks and intermediate object pose sampling for solving such long-horizon problems. To chain independently trained skills, we introduce $\textit{connectors}$, goal-conditioned policies trained to minimize object disturbance during transitions. High-quality demonstrations are generated with $\texttt{Skill-RRT}$ and distilled through noise-based replay in order to reduce online computation time. The resulting policy, trained entirely in simulation, transfers zero-shot to the real world and achieves over 80% success across three challenging long-horizon manipulation tasks and outperforms state-of-the-art hierarchical RL and planning methods.
APA
Jung, H., Lee, D., Park, H., Hyeop, K.J. & Kim, B.. (2025). $\textttSPIN$: distilling $\textttSkill-RRT$ for long-horizon prehensile and non-prehensile manipulation. Proceedings of The 9th Conference on Robot Learning, in Proceedings of Machine Learning Research 305:1311-1351 Available from https://proceedings.mlr.press/v305/jung25a.html.

Related Material