A Sequential Group VAE for Robot Learning of Haptic Representations
Haptic representation learning is a difficult task in robotics because information can be gathered only by actively exploring the environment over time, and because different actions elicit different object properties. We propose a Sequential Group VAE that leverages object persistence to learn and update latent general representations of multimodal haptic data. As a robot performs sequences of exploratory procedures on an object, the model accumulates data and learns to distinguish between general object properties, such as size and mass, and trial-to-trial variations, such as initial object position. We demonstrate that after very few observations, the general latent representations are sufficiently refined to accurately encode many haptic object properties.
| Author(s): | Benjamin A. Richardson and Katherine J. Kuchenbecker and Georg Martius |
| Pages: | 1--11 |
| Year: | 2022 |
| Month: | December |
| BibTeX Type: | Miscellaneous (misc) |
| Address: | Auckland, New Zealand |
| Electronic Archiving: | grant_archive |
| How Published: | Workshop paper (8 pages) presented at the CoRL Workshop on Aligning Robot Representations with Humans |
| State: | Published |
| URL: | https://aligning-robot-human-representations.github.io/docs/camready_11.pdf |
BibTeX
@misc{Richardson22-CORLWS-Sequential,
title = {A Sequential Group {VAE} for Robot Learning of Haptic Representations},
abstract = {Haptic representation learning is a difficult task in robotics because information can be gathered only by actively exploring the environment over time, and because different actions elicit different object properties. We propose a Sequential Group VAE that leverages object persistence to learn and update latent general representations of multimodal haptic data. As a robot performs sequences of exploratory procedures on an object, the model accumulates data and learns to distinguish between general object properties, such as size and mass, and trial-to-trial variations, such as initial object position. We demonstrate that after very few observations, the general latent representations are sufficiently refined to accurately encode many haptic object properties.},
pages = {1--11},
howpublished = {Workshop paper (8 pages) presented at the CoRL Workshop on Aligning Robot Representations with Humans},
address = {Auckland, New Zealand},
month = dec,
year = {2022},
author = {Richardson, Benjamin A. and Kuchenbecker, Katherine J. and Martius, Georg},
url = {https://aligning-robot-human-representations.github.io/docs/camready_11.pdf},
month_numeric = {12}
}