Robotics: Science and Systems XVII 2021
DOI: 10.15607/rss.2021.xvii.082
|View full text |Cite
|
Sign up to set email alerts
|

Learning Riemannian Manifolds for Geodesic Motion Skills

Abstract: For robots to work alongside humans and perform in unstructured environments, they must learn new motion skills and adapt them to unseen situations on the fly. This demands learning models that capture relevant motion patterns, while offering enough flexibility to adapt the encoded skills to new requirements, such as dynamic obstacle avoidance. We introduce a Riemannian manifold perspective on this problem, and propose to learn a Riemannian manifold from human demonstrations on which geodesics are natural moti… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
4
1

Citation Types

0
6
0

Year Published

2022
2022
2023
2023

Publication Types

Select...
4
2
1

Relationship

2
5

Authors

Journals

citations
Cited by 10 publications
(6 citation statements)
references
References 46 publications
0
6
0
Order By: Relevance
“…Despite most LfD approaches have overlooked this problem (Paraschos et al 2018;Seker et al 2019;Calinon 2016;Huang et al 2019), recent works have addressed it using probabilistic models (Rozo and Dave 2021;Zeestraten 2018). In our previous work (Beik-Mohammadi et al 2021), we proposed a VAE architecture capable of encoding full-pose trajectories, which is here exploited for learning a variety of real robotic tasks.…”
Section: Learning From Demonstrationmentioning
confidence: 99%
See 2 more Smart Citations
“…Despite most LfD approaches have overlooked this problem (Paraschos et al 2018;Seker et al 2019;Calinon 2016;Huang et al 2019), recent works have addressed it using probabilistic models (Rozo and Dave 2021;Zeestraten 2018). In our previous work (Beik-Mohammadi et al 2021), we proposed a VAE architecture capable of encoding full-pose trajectories, which is here exploited for learning a variety of real robotic tasks.…”
Section: Learning From Demonstrationmentioning
confidence: 99%
“…In our previous paper (Beik-Mohammadi et al 2021), we provided an LfD approach that addressed several of the foregoing problems by leveraging a Riemannian perspective for learning robot motions using task-space demonstrations. We employed a Riemannian formulation to represent a motion skill, in which human demonstrations were assumed to form a Riemannian manifold (i.e., a smooth surface), which could be learned in task space double-struckR3×scriptS3.…”
Section: Introductionmentioning
confidence: 99%
See 1 more Smart Citation
“…In our previous paper (Beik-Mohammadi et al 2021), we provided an LfD approach that addressed several of foregoing problems by leveraging a Riemannian perspective for learning robot motions using task space demonstrations. We employed a Riemannian formulation to represent a motion skill, in which human demonstrations were assumed to form a Riemannian manifold (i.e.…”
Section: Introductionmentioning
confidence: 99%
“…In this paper, we extend our previous work in two different ways. First, inspired by the need of avoiding obstacles at any location of the robot body, we propose a new reactive motion generation method that also leverages the Riemannian approach proposed in (Beik-Mohammadi et al 2021) for joint space skills. To do so, we develop a new VAE architecture that integrates the robot forward kinematics to access task space information of any point on the robot body.…”
Section: Introductionmentioning
confidence: 99%