Think Small, Act Big: Primitive Prompt Learning for Lifelong Robot Manipulation

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Yao, Yuanqi, Liu, Siao, Song, Haoming, Qu, Delin, Chen, Qizhi, Ding, Yan, Zhao, Bin, Wang, Zhigang, Li, Xuelong, Wang, Dong
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866909631289229312
author Yao, Yuanqi
Liu, Siao
Song, Haoming
Qu, Delin
Chen, Qizhi
Ding, Yan
Zhao, Bin
Wang, Zhigang
Li, Xuelong
Wang, Dong
author_facet Yao, Yuanqi
Liu, Siao
Song, Haoming
Qu, Delin
Chen, Qizhi
Ding, Yan
Zhao, Bin
Wang, Zhigang
Li, Xuelong
Wang, Dong
contents Building a lifelong robot that can effectively leverage prior knowledge for continuous skill acquisition remains significantly challenging. Despite the success of experience replay and parameter-efficient methods in alleviating catastrophic forgetting problem, naively applying these methods causes a failure to leverage the shared primitives between skills. To tackle these issues, we propose Primitive Prompt Learning (PPL), to achieve lifelong robot manipulation via reusable and extensible primitives. Within our two stage learning scheme, we first learn a set of primitive prompts to represent shared primitives through multi-skills pre-training stage, where motion-aware prompts are learned to capture semantic and motion shared primitives across different skills. Secondly, when acquiring new skills in lifelong span, new prompts are appended and optimized with frozen pretrained prompts, boosting the learning via knowledge transfer from old skills to new ones. For evaluation, we construct a large-scale skill dataset and conduct extensive experiments in both simulation and real-world tasks, demonstrating PPL's superior performance over state-of-the-art methods.
format Preprint
id arxiv_https___arxiv_org_abs_2504_00420
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Think Small, Act Big: Primitive Prompt Learning for Lifelong Robot Manipulation
Yao, Yuanqi
Liu, Siao
Song, Haoming
Qu, Delin
Chen, Qizhi
Ding, Yan
Zhao, Bin
Wang, Zhigang
Li, Xuelong
Wang, Dong
Robotics
Computer Vision and Pattern Recognition
Building a lifelong robot that can effectively leverage prior knowledge for continuous skill acquisition remains significantly challenging. Despite the success of experience replay and parameter-efficient methods in alleviating catastrophic forgetting problem, naively applying these methods causes a failure to leverage the shared primitives between skills. To tackle these issues, we propose Primitive Prompt Learning (PPL), to achieve lifelong robot manipulation via reusable and extensible primitives. Within our two stage learning scheme, we first learn a set of primitive prompts to represent shared primitives through multi-skills pre-training stage, where motion-aware prompts are learned to capture semantic and motion shared primitives across different skills. Secondly, when acquiring new skills in lifelong span, new prompts are appended and optimized with frozen pretrained prompts, boosting the learning via knowledge transfer from old skills to new ones. For evaluation, we construct a large-scale skill dataset and conduct extensive experiments in both simulation and real-world tasks, demonstrating PPL's superior performance over state-of-the-art methods.
title Think Small, Act Big: Primitive Prompt Learning for Lifelong Robot Manipulation
topic Robotics
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2504.00420