Generalizing Movement Primitives to New Situations

Loading...
Thumbnail Image

Access rights

openAccess
acceptedVersion

URL

Journal Title

Journal ISSN

Volume Title

A4 Artikkeli konferenssijulkaisussa

Date

2017

Major/Subject

Mcode

Degree programme

Language

en

Pages

16

Series

Towards Autonomous Robotic Systems - 18th Annual Conference, TAROS 2017, Proceedings, Volume 10454 LNAI, pp. 16-31, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) ; Volume 10454 LNAI

Abstract

Although motor primitives (MPs) have been studied extensively, much less attention has been devoted to studying their generalization to new situations. To cope with varying conditions, a MP’s policy encoding must support generalization over task parameters to avoid learning separate primitives for each condition. Local and linear parameterized models have been proposed to interpolate over task parameters to provide limited generalization. In this paper, we present a global parametric motion primitive (GPDMP) which allows generalization beyond local or linear models. Primitives are modelled using a linear basis function model with global non-linear basis functions. The model is constructed from initial non-parametric primitives found using a single human demonstration and subsequent episodes of reinforcement learning to adapt the demonstrated skill to other task parameters. The initial models are then used to optimize the parameters of the global parametric model. Experiments with a ball-in-a-cup task with varying string lengths show that GPDMP allows greatly improved extrapolation compared to earlier local or linear models.

Description

Keywords

learning from demonstration, generalization, global parametric model, ball-in-a-cup

Other note

Citation

Lundell, J, Hazara, M & Kyrki, V 2017, Generalizing Movement Primitives to New Situations . in Towards Autonomous Robotic Systems - 18th Annual Conference, TAROS 2017, Proceedings . vol. 10454 LNAI, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 10454 LNAI, Springer, pp. 16-31, Towards Autonomous Robotic Systems Conference, Guildford, United Kingdom, 19/07/2017 . https://doi.org/10.1007/978-3-319-64107-2_2