Learning to Correct Mistakes: Backjumping in Long-Horizon Task and Motion Planning

Yoonchang Sung, Zizhao Wang, Peter Stone
Proceedings of The 6th Conference on Robot Learning, PMLR 205:2115-2124, 2023.

Abstract

As robots become increasingly capable of manipulation and long-term autonomy, long-horizon task and motion planning problems are becoming increasingly important. A key challenge in such problems is that early actions in the plan may make future actions infeasible. When reaching a dead-end in the search, most existing planners use backtracking, which exhaustively reevaluates motion-level actions, often resulting in inefficient planning, especially when the search depth is large. In this paper, we propose to learn backjumping heuristics which identify the culprit action directly using supervised learning models to guide the task-level search. Based on evaluations of two different tasks, we find that our method significantly improves planning efficiency compared to backtracking and also generalizes to problems with novel numbers of objects.

Cite this Paper

BibTeX
@InProceedings{pmlr-v205-sung23a, title = {Learning to Correct Mistakes: Backjumping in Long-Horizon Task and Motion Planning}, author = {Sung, Yoonchang and Wang, Zizhao and Stone, Peter}, booktitle = {Proceedings of The 6th Conference on Robot Learning}, pages = {2115--2124}, year = {2023}, editor = {Liu, Karen and Kulic, Dana and Ichnowski, Jeff}, volume = {205}, series = {Proceedings of Machine Learning Research}, month = {14--18 Dec}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v205/sung23a/sung23a.pdf}, url = {https://proceedings.mlr.press/v205/sung23a.html}, abstract = {As robots become increasingly capable of manipulation and long-term autonomy, long-horizon task and motion planning problems are becoming increasingly important. A key challenge in such problems is that early actions in the plan may make future actions infeasible. When reaching a dead-end in the search, most existing planners use backtracking, which exhaustively reevaluates motion-level actions, often resulting in inefficient planning, especially when the search depth is large. In this paper, we propose to learn backjumping heuristics which identify the culprit action directly using supervised learning models to guide the task-level search. Based on evaluations of two different tasks, we find that our method significantly improves planning efficiency compared to backtracking and also generalizes to problems with novel numbers of objects.} }
Endnote
%0 Conference Paper %T Learning to Correct Mistakes: Backjumping in Long-Horizon Task and Motion Planning %A Yoonchang Sung %A Zizhao Wang %A Peter Stone %B Proceedings of The 6th Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2023 %E Karen Liu %E Dana Kulic %E Jeff Ichnowski %F pmlr-v205-sung23a %I PMLR %P 2115--2124 %U https://proceedings.mlr.press/v205/sung23a.html %V 205 %X As robots become increasingly capable of manipulation and long-term autonomy, long-horizon task and motion planning problems are becoming increasingly important. A key challenge in such problems is that early actions in the plan may make future actions infeasible. When reaching a dead-end in the search, most existing planners use backtracking, which exhaustively reevaluates motion-level actions, often resulting in inefficient planning, especially when the search depth is large. In this paper, we propose to learn backjumping heuristics which identify the culprit action directly using supervised learning models to guide the task-level search. Based on evaluations of two different tasks, we find that our method significantly improves planning efficiency compared to backtracking and also generalizes to problems with novel numbers of objects.
APA
Sung, Y., Wang, Z. & Stone, P.. (2023). Learning to Correct Mistakes: Backjumping in Long-Horizon Task and Motion Planning. Proceedings of The 6th Conference on Robot Learning, in Proceedings of Machine Learning Research 205:2115-2124 Available from https://proceedings.mlr.press/v205/sung23a.html.

Related Material