Mitigating the Human-Robot Domain Discrepancy in Visual Pre-training for Robotic Manipulation

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Zhou, Jiaming, Ma, Teli, Lin, Kun-Yu, Wang, Zifan, Qiu, Ronghe, Liang, Junwei
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866910904697749504
author Zhou, Jiaming
Ma, Teli
Lin, Kun-Yu
Wang, Zifan
Qiu, Ronghe
Liang, Junwei
author_facet Zhou, Jiaming
Ma, Teli
Lin, Kun-Yu
Wang, Zifan
Qiu, Ronghe
Liang, Junwei
contents Learning generalizable visual representations across different embodied environments is essential for effective robotic manipulation in real-world scenarios. However, the limited scale and diversity of robot demonstration data pose a significant challenge. Recent research has explored leveraging large-scale human activity data for pre-training, but the substantial morphological differences between humans and robots introduce a significant human-robot domain discrepancy, hindering the generalization of these models to downstream manipulation tasks. To overcome this, we propose a novel adaptation paradigm that leverages readily available paired human-robot video data to bridge the domain gap. Our method employs a human-robot contrastive alignment loss to align the semantics of human and robot videos, adapting pre-trained models to the robot domain in a parameter-efficient manner. Experiments on 20 simulated tasks across two different benchmarks and five real-world tasks demonstrate significant improvements. These results span both single-task and language-conditioned multi-task settings, evaluated using two different pre-trained models. Compared to existing pre-trained models, our adaptation method improves the average success rate by over 7% across multiple tasks on both simulated benchmarks and real-world evaluations.
format Preprint
id arxiv_https___arxiv_org_abs_2406_14235
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Mitigating the Human-Robot Domain Discrepancy in Visual Pre-training for Robotic Manipulation
Zhou, Jiaming
Ma, Teli
Lin, Kun-Yu
Wang, Zifan
Qiu, Ronghe
Liang, Junwei
Computer Vision and Pattern Recognition
Robotics
Learning generalizable visual representations across different embodied environments is essential for effective robotic manipulation in real-world scenarios. However, the limited scale and diversity of robot demonstration data pose a significant challenge. Recent research has explored leveraging large-scale human activity data for pre-training, but the substantial morphological differences between humans and robots introduce a significant human-robot domain discrepancy, hindering the generalization of these models to downstream manipulation tasks. To overcome this, we propose a novel adaptation paradigm that leverages readily available paired human-robot video data to bridge the domain gap. Our method employs a human-robot contrastive alignment loss to align the semantics of human and robot videos, adapting pre-trained models to the robot domain in a parameter-efficient manner. Experiments on 20 simulated tasks across two different benchmarks and five real-world tasks demonstrate significant improvements. These results span both single-task and language-conditioned multi-task settings, evaluated using two different pre-trained models. Compared to existing pre-trained models, our adaptation method improves the average success rate by over 7% across multiple tasks on both simulated benchmarks and real-world evaluations.
title Mitigating the Human-Robot Domain Discrepancy in Visual Pre-training for Robotic Manipulation
topic Computer Vision and Pattern Recognition
Robotics
url https://arxiv.org/abs/2406.14235