DIRL: Domain-Invariant Representation Learning for Sim-to-Real Transfer

Ajay Tanwani
Proceedings of the 2020 Conference on Robot Learning, PMLR 155:1558-1571, 2021.

Abstract

Generating large-scale synthetic data in simulation is a feasible alternative to collecting/labelling real data for training vision-based deep learning models, albeit the modelling inaccuracies do not generalize to the physical world. In this paper, we present a domain-invariant representation learning (DIRL) algorithm to adapt deep models to the physical environment with a small amount of real data. Existing approaches that only mitigate the covariate shift by aligning the marginal distributions across the domains and assume the conditional distributions to be domain-invariant can lead to ambiguous transfer in real scenarios. We propose to jointly align the marginal (input domains) and the conditional (output labels) distributions to mitigate the covariate and the conditional shift across the domains with adversarial learning, and combine it with a triplet distribution loss to make the conditional distributions disjoint in the shared feature space. Experiments on digit domains yield state-of-the-art performance on challenging benchmarks, while sim-to-real transfer of object recognition for vision-based decluttering with a mobile robot improves from $26.8 %$ to $91.0 %$, resulting in $86.5 %$ grasping accuracy of a wide variety of objects. Code and supplementary details are available at: https://sites.google.com/view/dirl

Cite this Paper

BibTeX
@InProceedings{pmlr-v155-tanwani21a, title = {DIRL: Domain-Invariant Representation Learning for Sim-to-Real Transfer}, author = {Tanwani, Ajay}, booktitle = {Proceedings of the 2020 Conference on Robot Learning}, pages = {1558--1571}, year = {2021}, editor = {Kober, Jens and Ramos, Fabio and Tomlin, Claire}, volume = {155}, series = {Proceedings of Machine Learning Research}, month = {16--18 Nov}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v155/tanwani21a/tanwani21a.pdf}, url = {https://proceedings.mlr.press/v155/tanwani21a.html}, abstract = {Generating large-scale synthetic data in simulation is a feasible alternative to collecting/labelling real data for training vision-based deep learning models, albeit the modelling inaccuracies do not generalize to the physical world. In this paper, we present a domain-invariant representation learning (DIRL) algorithm to adapt deep models to the physical environment with a small amount of real data. Existing approaches that only mitigate the covariate shift by aligning the marginal distributions across the domains and assume the conditional distributions to be domain-invariant can lead to ambiguous transfer in real scenarios. We propose to jointly align the marginal (input domains) and the conditional (output labels) distributions to mitigate the covariate and the conditional shift across the domains with adversarial learning, and combine it with a triplet distribution loss to make the conditional distributions disjoint in the shared feature space. Experiments on digit domains yield state-of-the-art performance on challenging benchmarks, while sim-to-real transfer of object recognition for vision-based decluttering with a mobile robot improves from $26.8 %$ to $91.0 %$, resulting in $86.5 %$ grasping accuracy of a wide variety of objects. Code and supplementary details are available at: https://sites.google.com/view/dirl} }
Endnote
%0 Conference Paper %T DIRL: Domain-Invariant Representation Learning for Sim-to-Real Transfer %A Ajay Tanwani %B Proceedings of the 2020 Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2021 %E Jens Kober %E Fabio Ramos %E Claire Tomlin %F pmlr-v155-tanwani21a %I PMLR %P 1558--1571 %U https://proceedings.mlr.press/v155/tanwani21a.html %V 155 %X Generating large-scale synthetic data in simulation is a feasible alternative to collecting/labelling real data for training vision-based deep learning models, albeit the modelling inaccuracies do not generalize to the physical world. In this paper, we present a domain-invariant representation learning (DIRL) algorithm to adapt deep models to the physical environment with a small amount of real data. Existing approaches that only mitigate the covariate shift by aligning the marginal distributions across the domains and assume the conditional distributions to be domain-invariant can lead to ambiguous transfer in real scenarios. We propose to jointly align the marginal (input domains) and the conditional (output labels) distributions to mitigate the covariate and the conditional shift across the domains with adversarial learning, and combine it with a triplet distribution loss to make the conditional distributions disjoint in the shared feature space. Experiments on digit domains yield state-of-the-art performance on challenging benchmarks, while sim-to-real transfer of object recognition for vision-based decluttering with a mobile robot improves from $26.8 %$ to $91.0 %$, resulting in $86.5 %$ grasping accuracy of a wide variety of objects. Code and supplementary details are available at: https://sites.google.com/view/dirl
APA
Tanwani, A.. (2021). DIRL: Domain-Invariant Representation Learning for Sim-to-Real Transfer. Proceedings of the 2020 Conference on Robot Learning, in Proceedings of Machine Learning Research 155:1558-1571 Available from https://proceedings.mlr.press/v155/tanwani21a.html.

Related Material