Multi-Level Structure vs. End-to-End-Learning in High-Performance Tactile Robotic Manipulation

Florian Voigt, Lars Johannsmeier, Sami Haddadin
Proceedings of the 2020 Conference on Robot Learning, PMLR 155:2306-2316, 2021.

Abstract

In this paper we apply a multi-level structure to robotic manipulation learning. It consists of a hybrid dynamical system we denote skill and a parameter learning layer that leverages the underlying structure to simplify the problem at hand. For the learning layer we introduce a novel algorithm based on the idea of learning to partition the parameter solution space to quickly and efficiently find good and robust solutions to complex manipulation problems. In a benchmark comparison we show a significant performance increase compared with other black-box optimization algorithms such as HiREPS and particle swarm optimization. Furthermore, we validate and compare our approach on a very hard real-world manipulation problem, namely inserting a key into a lock, with state-of-the-art deep reinforcement learning.

Cite this Paper

BibTeX
@InProceedings{pmlr-v155-voigt21a, title = {Multi-Level Structure vs. End-to-End-Learning in High-Performance Tactile Robotic Manipulation}, author = {Voigt, Florian and Johannsmeier, Lars and Haddadin, Sami}, booktitle = {Proceedings of the 2020 Conference on Robot Learning}, pages = {2306--2316}, 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/voigt21a/voigt21a.pdf}, url = {https://proceedings.mlr.press/v155/voigt21a.html}, abstract = {In this paper we apply a multi-level structure to robotic manipulation learning. It consists of a hybrid dynamical system we denote skill and a parameter learning layer that leverages the underlying structure to simplify the problem at hand. For the learning layer we introduce a novel algorithm based on the idea of learning to partition the parameter solution space to quickly and efficiently find good and robust solutions to complex manipulation problems. In a benchmark comparison we show a significant performance increase compared with other black-box optimization algorithms such as HiREPS and particle swarm optimization. Furthermore, we validate and compare our approach on a very hard real-world manipulation problem, namely inserting a key into a lock, with state-of-the-art deep reinforcement learning.} }
Endnote
%0 Conference Paper %T Multi-Level Structure vs. End-to-End-Learning in High-Performance Tactile Robotic Manipulation %A Florian Voigt %A Lars Johannsmeier %A Sami Haddadin %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-voigt21a %I PMLR %P 2306--2316 %U https://proceedings.mlr.press/v155/voigt21a.html %V 155 %X In this paper we apply a multi-level structure to robotic manipulation learning. It consists of a hybrid dynamical system we denote skill and a parameter learning layer that leverages the underlying structure to simplify the problem at hand. For the learning layer we introduce a novel algorithm based on the idea of learning to partition the parameter solution space to quickly and efficiently find good and robust solutions to complex manipulation problems. In a benchmark comparison we show a significant performance increase compared with other black-box optimization algorithms such as HiREPS and particle swarm optimization. Furthermore, we validate and compare our approach on a very hard real-world manipulation problem, namely inserting a key into a lock, with state-of-the-art deep reinforcement learning.
APA
Voigt, F., Johannsmeier, L. & Haddadin, S.. (2021). Multi-Level Structure vs. End-to-End-Learning in High-Performance Tactile Robotic Manipulation. Proceedings of the 2020 Conference on Robot Learning, in Proceedings of Machine Learning Research 155:2306-2316 Available from https://proceedings.mlr.press/v155/voigt21a.html.

Related Material