Discovering State and Action Abstractions for Generalized Task and Motion Planning
<jats:p>Generalized planning accelerates classical planning by finding an algorithm-like policy that solves multiple instances of a task. A generalized plan can be learned from a few training examples and applied to an entire domain of problems. Generalized planning approaches perform well in...
| Main Authors: | , , , , |
|---|---|
| Other Authors: | |
| Format: | Article |
| Language: | English |
| Published: |
Association for the Advancement of Artificial Intelligence (AAAI)
2023
|
| Online Access: | https://hdl.handle.net/1721.1/150399 |
| _version_ | 1826191153275338752 |
|---|---|
| author | Curtis, Aidan Silver, Tom Tenenbaum, Joshua B Lozano-Pérez, Tomás Kaelbling, Leslie |
| author2 | Massachusetts Institute of Technology. Department of Brain and Cognitive Sciences |
| author_facet | Massachusetts Institute of Technology. Department of Brain and Cognitive Sciences Curtis, Aidan Silver, Tom Tenenbaum, Joshua B Lozano-Pérez, Tomás Kaelbling, Leslie |
| author_sort | Curtis, Aidan |
| collection | MIT |
| description | <jats:p>Generalized planning accelerates classical planning by finding an algorithm-like policy that solves multiple instances of a task. A generalized plan can be learned from a few training examples and applied to an entire domain of problems. Generalized planning approaches perform well in discrete AI planning problems that involve large numbers of objects and extended action sequences to achieve the goal. In this paper, we propose an algorithm for learning features, abstractions, and generalized plans for continuous robotic task and motion planning (TAMP) and examine the unique difficulties that arise when forced to consider geometric and physical constraints as a part of the generalized plan. Additionally, we show that these simple generalized plans learned from only a handful of examples can be used to improve the search efficiency of TAMP solvers.</jats:p> |
| first_indexed | 2024-09-23T08:51:05Z |
| format | Article |
| id | mit-1721.1/150399 |
| institution | Massachusetts Institute of Technology |
| language | English |
| last_indexed | 2024-09-23T08:51:05Z |
| publishDate | 2023 |
| publisher | Association for the Advancement of Artificial Intelligence (AAAI) |
| record_format | dspace |
| spelling | mit-1721.1/1503992023-04-05T03:50:13Z Discovering State and Action Abstractions for Generalized Task and Motion Planning Curtis, Aidan Silver, Tom Tenenbaum, Joshua B Lozano-Pérez, Tomás Kaelbling, Leslie Massachusetts Institute of Technology. Department of Brain and Cognitive Sciences <jats:p>Generalized planning accelerates classical planning by finding an algorithm-like policy that solves multiple instances of a task. A generalized plan can be learned from a few training examples and applied to an entire domain of problems. Generalized planning approaches perform well in discrete AI planning problems that involve large numbers of objects and extended action sequences to achieve the goal. In this paper, we propose an algorithm for learning features, abstractions, and generalized plans for continuous robotic task and motion planning (TAMP) and examine the unique difficulties that arise when forced to consider geometric and physical constraints as a part of the generalized plan. Additionally, we show that these simple generalized plans learned from only a handful of examples can be used to improve the search efficiency of TAMP solvers.</jats:p> 2023-04-04T16:33:43Z 2023-04-04T16:33:43Z 2022 2023-04-04T16:28:10Z Article http://purl.org/eprint/type/ConferencePaper https://hdl.handle.net/1721.1/150399 Curtis, Aidan, Silver, Tom, Tenenbaum, Joshua B, Lozano-Pérez, Tomás and Kaelbling, Leslie. 2022. "Discovering State and Action Abstractions for Generalized Task and Motion Planning." Proceedings of the AAAI Conference on Artificial Intelligence, 36 (5). en 10.1609/AAAI.V36I5.20475 Proceedings of the AAAI Conference on Artificial Intelligence Creative Commons Attribution-Noncommercial-Share Alike http://creativecommons.org/licenses/by-nc-sa/4.0/ application/pdf Association for the Advancement of Artificial Intelligence (AAAI) arXiv |
| spellingShingle | Curtis, Aidan Silver, Tom Tenenbaum, Joshua B Lozano-Pérez, Tomás Kaelbling, Leslie Discovering State and Action Abstractions for Generalized Task and Motion Planning |
| title | Discovering State and Action Abstractions for Generalized Task and Motion Planning |
| title_full | Discovering State and Action Abstractions for Generalized Task and Motion Planning |
| title_fullStr | Discovering State and Action Abstractions for Generalized Task and Motion Planning |
| title_full_unstemmed | Discovering State and Action Abstractions for Generalized Task and Motion Planning |
| title_short | Discovering State and Action Abstractions for Generalized Task and Motion Planning |
| title_sort | discovering state and action abstractions for generalized task and motion planning |
| url | https://hdl.handle.net/1721.1/150399 |
| work_keys_str_mv | AT curtisaidan discoveringstateandactionabstractionsforgeneralizedtaskandmotionplanning AT silvertom discoveringstateandactionabstractionsforgeneralizedtaskandmotionplanning AT tenenbaumjoshuab discoveringstateandactionabstractionsforgeneralizedtaskandmotionplanning AT lozanopereztomas discoveringstateandactionabstractionsforgeneralizedtaskandmotionplanning AT kaelblingleslie discoveringstateandactionabstractionsforgeneralizedtaskandmotionplanning |