Seamless multi-skill learning: learning and transitioning non-similar skills in quadruped robots with limited data
In multi-skill imitation learning for robots, expert datasets with complete motion features are crucial for enabling robots to learn and transition between different skills. However, such datasets are often difficult to obtain. As an alternative, datasets constructed using only joint positions are m...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Frontiers Media S.A.
2025-04-01
|
| Series: | Frontiers in Robotics and AI |
| Subjects: | |
| Online Access: | https://www.frontiersin.org/articles/10.3389/frobt.2025.1542692/full |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850176354405842944 |
|---|---|
| author | Jiaxin Tu Peng Zhai Yueqi Zhang Xiaoyi Wei Zhiyan Dong Lihua Zhang |
| author_facet | Jiaxin Tu Peng Zhai Yueqi Zhang Xiaoyi Wei Zhiyan Dong Lihua Zhang |
| author_sort | Jiaxin Tu |
| collection | DOAJ |
| description | In multi-skill imitation learning for robots, expert datasets with complete motion features are crucial for enabling robots to learn and transition between different skills. However, such datasets are often difficult to obtain. As an alternative, datasets constructed using only joint positions are more accessible, but they are incomplete and lack details, making it challenging for existing methods to effectively learn and model skill transitions. To address these challenges, this study introduces the Seamless Multi-Skill Learning (SMSL) framework. Integrated within the Adversarial Motion Priors framework and incorporating self-trajectory augmentation techniques, SMSL effectively utilizes high-quality historical experiences to guide agents in learning skills and generating smooth, natural transitions between them, addressing the learning difficulties caused by incomplete expert datasets. Additionally, the research incorporates an adaptive command sampling mechanism to balance the training opportunities for skills of various difficulties and prevent catastrophic forgetting. Our experiments highlight potential issues with baseline methods when imitating incomplete expert datasets and demonstrate the superior performance of the SMSL framework. Sim-to-real experiments on real Solo8 robots further validate the effectiveness of SMSL. Overall, this study confirms the SMSL framework’s capability in real robotic applications and underscores its potential for autonomous skill learning and generation from minimal data. |
| format | Article |
| id | doaj-art-ecd581fa0bf7478fabf5bae37d4dc3b6 |
| institution | OA Journals |
| issn | 2296-9144 |
| language | English |
| publishDate | 2025-04-01 |
| publisher | Frontiers Media S.A. |
| record_format | Article |
| series | Frontiers in Robotics and AI |
| spelling | doaj-art-ecd581fa0bf7478fabf5bae37d4dc3b62025-08-20T02:19:16ZengFrontiers Media S.A.Frontiers in Robotics and AI2296-91442025-04-011210.3389/frobt.2025.15426921542692Seamless multi-skill learning: learning and transitioning non-similar skills in quadruped robots with limited dataJiaxin TuPeng ZhaiYueqi ZhangXiaoyi WeiZhiyan DongLihua ZhangIn multi-skill imitation learning for robots, expert datasets with complete motion features are crucial for enabling robots to learn and transition between different skills. However, such datasets are often difficult to obtain. As an alternative, datasets constructed using only joint positions are more accessible, but they are incomplete and lack details, making it challenging for existing methods to effectively learn and model skill transitions. To address these challenges, this study introduces the Seamless Multi-Skill Learning (SMSL) framework. Integrated within the Adversarial Motion Priors framework and incorporating self-trajectory augmentation techniques, SMSL effectively utilizes high-quality historical experiences to guide agents in learning skills and generating smooth, natural transitions between them, addressing the learning difficulties caused by incomplete expert datasets. Additionally, the research incorporates an adaptive command sampling mechanism to balance the training opportunities for skills of various difficulties and prevent catastrophic forgetting. Our experiments highlight potential issues with baseline methods when imitating incomplete expert datasets and demonstrate the superior performance of the SMSL framework. Sim-to-real experiments on real Solo8 robots further validate the effectiveness of SMSL. Overall, this study confirms the SMSL framework’s capability in real robotic applications and underscores its potential for autonomous skill learning and generation from minimal data.https://www.frontiersin.org/articles/10.3389/frobt.2025.1542692/fullmulti-skill learningimitation learningadaptive command samplingself-trajectory augmentationquadrupedal robots |
| spellingShingle | Jiaxin Tu Peng Zhai Yueqi Zhang Xiaoyi Wei Zhiyan Dong Lihua Zhang Seamless multi-skill learning: learning and transitioning non-similar skills in quadruped robots with limited data Frontiers in Robotics and AI multi-skill learning imitation learning adaptive command sampling self-trajectory augmentation quadrupedal robots |
| title | Seamless multi-skill learning: learning and transitioning non-similar skills in quadruped robots with limited data |
| title_full | Seamless multi-skill learning: learning and transitioning non-similar skills in quadruped robots with limited data |
| title_fullStr | Seamless multi-skill learning: learning and transitioning non-similar skills in quadruped robots with limited data |
| title_full_unstemmed | Seamless multi-skill learning: learning and transitioning non-similar skills in quadruped robots with limited data |
| title_short | Seamless multi-skill learning: learning and transitioning non-similar skills in quadruped robots with limited data |
| title_sort | seamless multi skill learning learning and transitioning non similar skills in quadruped robots with limited data |
| topic | multi-skill learning imitation learning adaptive command sampling self-trajectory augmentation quadrupedal robots |
| url | https://www.frontiersin.org/articles/10.3389/frobt.2025.1542692/full |
| work_keys_str_mv | AT jiaxintu seamlessmultiskilllearninglearningandtransitioningnonsimilarskillsinquadrupedrobotswithlimiteddata AT pengzhai seamlessmultiskilllearninglearningandtransitioningnonsimilarskillsinquadrupedrobotswithlimiteddata AT yueqizhang seamlessmultiskilllearninglearningandtransitioningnonsimilarskillsinquadrupedrobotswithlimiteddata AT xiaoyiwei seamlessmultiskilllearninglearningandtransitioningnonsimilarskillsinquadrupedrobotswithlimiteddata AT zhiyandong seamlessmultiskilllearninglearningandtransitioningnonsimilarskillsinquadrupedrobotswithlimiteddata AT lihuazhang seamlessmultiskilllearninglearningandtransitioningnonsimilarskillsinquadrupedrobotswithlimiteddata |