TY - GEN
T1 - Learning Neuro-Symbolic Relational Transition Models for Bilevel Planning
AU - Chitnis, Rohan
AU - Silver, Tom
AU - Tenenbaum, Joshua B.
AU - Lozano-Perez, Tomas
AU - Kaelbling, Leslie Pack
N1 - Publisher Copyright:
© 2022 IEEE.
PY - 2022
Y1 - 2022
N2 - In robotic domains, learning and planning are complicated by continuous state spaces, continuous action spaces, and long task horizons. In this work, we address these challenges with Neuro-Symbolic Relational Transition Models (NSRTs), a novel class of models that are data-efficient to learn, compatible with powerful robotic planning methods, and generalizable over objects. NSRTs have both symbolic and neural components, enabling a bilevel planning scheme where symbolic AI planning in an outer loop guides continuous planning with neural models in an inner loop. Experiments in four robotic planning domains show that NSRTs can be learned very data-efficiently, and then used for fast planning in new tasks that require up to 60 actions and involve many more objects than were seen during training.
AB - In robotic domains, learning and planning are complicated by continuous state spaces, continuous action spaces, and long task horizons. In this work, we address these challenges with Neuro-Symbolic Relational Transition Models (NSRTs), a novel class of models that are data-efficient to learn, compatible with powerful robotic planning methods, and generalizable over objects. NSRTs have both symbolic and neural components, enabling a bilevel planning scheme where symbolic AI planning in an outer loop guides continuous planning with neural models in an inner loop. Experiments in four robotic planning domains show that NSRTs can be learned very data-efficiently, and then used for fast planning in new tasks that require up to 60 actions and involve many more objects than were seen during training.
UR - https://www.scopus.com/pages/publications/85136108749
UR - https://www.scopus.com/pages/publications/85136108749#tab=citedBy
U2 - 10.1109/IROS47612.2022.9981440
DO - 10.1109/IROS47612.2022.9981440
M3 - Conference contribution
AN - SCOPUS:85136108749
T3 - IEEE International Conference on Intelligent Robots and Systems
SP - 4166
EP - 4173
BT - 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS 2022
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS 2022
Y2 - 23 October 2022 through 27 October 2022
ER -