NOD-TAMP: Generalizable Long-Horizon Planning with Neural Object Descriptors

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Cheng, Shuo, Garrett, Caelan, Mandlekar, Ajay, Xu, Danfei
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917794609627136
author Cheng, Shuo
Garrett, Caelan
Mandlekar, Ajay
Xu, Danfei
author_facet Cheng, Shuo
Garrett, Caelan
Mandlekar, Ajay
Xu, Danfei
contents Solving complex manipulation tasks in household and factory settings remains challenging due to long-horizon reasoning, fine-grained interactions, and broad object and scene diversity. Learning skills from demonstrations can be an effective strategy, but such methods often have limited generalizability beyond training data and struggle to solve long-horizon tasks. To overcome this, we propose to synergistically combine two paradigms: Neural Object Descriptors (NODs) that produce generalizable object-centric features and Task and Motion Planning (TAMP) frameworks that chain short-horizon skills to solve multi-step tasks. We introduce NOD-TAMP, a TAMP-based framework that extracts short manipulation trajectories from a handful of human demonstrations, adapts these trajectories using NOD features, and composes them to solve broad long-horizon, contact-rich tasks. NOD-TAMP solves existing manipulation benchmarks with a handful of demonstrations and significantly outperforms prior NOD-based approaches on new tabletop manipulation tasks that require diverse generalization. Finally, we deploy NOD-TAMP on a number of real-world tasks, including tool-use and high-precision insertion. For more details, please visit https://nodtamp.github.io/.
format Preprint
id arxiv_https___arxiv_org_abs_2311_01530
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle NOD-TAMP: Generalizable Long-Horizon Planning with Neural Object Descriptors
Cheng, Shuo
Garrett, Caelan
Mandlekar, Ajay
Xu, Danfei
Robotics
Artificial Intelligence
Solving complex manipulation tasks in household and factory settings remains challenging due to long-horizon reasoning, fine-grained interactions, and broad object and scene diversity. Learning skills from demonstrations can be an effective strategy, but such methods often have limited generalizability beyond training data and struggle to solve long-horizon tasks. To overcome this, we propose to synergistically combine two paradigms: Neural Object Descriptors (NODs) that produce generalizable object-centric features and Task and Motion Planning (TAMP) frameworks that chain short-horizon skills to solve multi-step tasks. We introduce NOD-TAMP, a TAMP-based framework that extracts short manipulation trajectories from a handful of human demonstrations, adapts these trajectories using NOD features, and composes them to solve broad long-horizon, contact-rich tasks. NOD-TAMP solves existing manipulation benchmarks with a handful of demonstrations and significantly outperforms prior NOD-based approaches on new tabletop manipulation tasks that require diverse generalization. Finally, we deploy NOD-TAMP on a number of real-world tasks, including tool-use and high-precision insertion. For more details, please visit https://nodtamp.github.io/.
title NOD-TAMP: Generalizable Long-Horizon Planning with Neural Object Descriptors
topic Robotics
Artificial Intelligence
url https://arxiv.org/abs/2311.01530