Learning Multi-Stage Tasks with One Demonstration via Self-Replay

Norman Di Palo, Edward Johns
Proceedings of the 5th Conference on Robot Learning, PMLR 164:1180-1189, 2022.

Abstract

In this work, we introduce a novel method to learn everyday-like multi-stage tasks from a single human demonstration, without requiring any prior object knowledge. Inspired by the recent Coarse-to-Fine Imitation Learning, we model imitation learning as a learned object reaching phase followed by an open-loop replay of the operator’s actions. We build upon this for multi-stage tasks where, following the human demonstration, the robot can autonomously collect image data for the entire multi-stage task, by reaching the next object in the sequence and then replaying the demonstration, repeating in a loop for all stages of the task. We evaluate with real-world experiments on a set of everyday multi-stage tasks, which we show that our method can solve from a single demonstration. Videos and supplementary material can be found at this webpage.

Cite this Paper

BibTeX
@InProceedings{pmlr-v164-palo22a, title = {Learning Multi-Stage Tasks with One Demonstration via Self-Replay}, author = {Palo, Norman Di and Johns, Edward}, booktitle = {Proceedings of the 5th Conference on Robot Learning}, pages = {1180--1189}, year = {2022}, editor = {Faust, Aleksandra and Hsu, David and Neumann, Gerhard}, volume = {164}, series = {Proceedings of Machine Learning Research}, month = {08--11 Nov}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v164/palo22a/palo22a.pdf}, url = {https://proceedings.mlr.press/v164/palo22a.html}, abstract = {In this work, we introduce a novel method to learn everyday-like multi-stage tasks from a single human demonstration, without requiring any prior object knowledge. Inspired by the recent Coarse-to-Fine Imitation Learning, we model imitation learning as a learned object reaching phase followed by an open-loop replay of the operator’s actions. We build upon this for multi-stage tasks where, following the human demonstration, the robot can autonomously collect image data for the entire multi-stage task, by reaching the next object in the sequence and then replaying the demonstration, repeating in a loop for all stages of the task. We evaluate with real-world experiments on a set of everyday multi-stage tasks, which we show that our method can solve from a single demonstration. Videos and supplementary material can be found at this webpage.} }
Endnote
%0 Conference Paper %T Learning Multi-Stage Tasks with One Demonstration via Self-Replay %A Norman Di Palo %A Edward Johns %B Proceedings of the 5th Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2022 %E Aleksandra Faust %E David Hsu %E Gerhard Neumann %F pmlr-v164-palo22a %I PMLR %P 1180--1189 %U https://proceedings.mlr.press/v164/palo22a.html %V 164 %X In this work, we introduce a novel method to learn everyday-like multi-stage tasks from a single human demonstration, without requiring any prior object knowledge. Inspired by the recent Coarse-to-Fine Imitation Learning, we model imitation learning as a learned object reaching phase followed by an open-loop replay of the operator’s actions. We build upon this for multi-stage tasks where, following the human demonstration, the robot can autonomously collect image data for the entire multi-stage task, by reaching the next object in the sequence and then replaying the demonstration, repeating in a loop for all stages of the task. We evaluate with real-world experiments on a set of everyday multi-stage tasks, which we show that our method can solve from a single demonstration. Videos and supplementary material can be found at this webpage.
APA
Palo, N.D. & Johns, E.. (2022). Learning Multi-Stage Tasks with One Demonstration via Self-Replay. Proceedings of the 5th Conference on Robot Learning, in Proceedings of Machine Learning Research 164:1180-1189 Available from https://proceedings.mlr.press/v164/palo22a.html.

Related Material