CLAS: Coordinating Multi-Robot Manipulation with Central Latent Action Spaces

Elie Aljalbout, Maximilian Karl, Patrick van der Smagt
Proceedings of The 5th Annual Learning for Dynamics and Control Conference, PMLR 211:1152-1166, 2023.

Abstract

Multi-robot manipulation tasks involve various control entities that can be separated into dynamically independent parts. A typical example of such real-world tasks is dual-arm manipulation. Learning to naively solve such tasks with reinforcement learning is often unfeasible due to the sample complexity and exploration requirements growing with the dimensionality of the action and state spaces. Instead, we would like to handle such environments as multi-agent systems and have several agents control parts of the whole. However, decentralizing the generation of actions requires coordination across agents through a channel limited to information central to the task. This paper proposes an approach to coordinating multi-robot manipulation through learned latent action spaces that are shared across different agents. We validate our method in simulated multi-robot manipulation tasks and demonstrate improvement over previous baselines in terms of sample efficiency and learning performance.

Cite this Paper

BibTeX
@InProceedings{pmlr-v211-aljalbout23a, title = {CLAS: Coordinating Multi-Robot Manipulation with Central Latent Action Spaces}, author = {Aljalbout, Elie and Karl, Maximilian and Smagt, Patrick van der}, booktitle = {Proceedings of The 5th Annual Learning for Dynamics and Control Conference}, pages = {1152--1166}, year = {2023}, editor = {Matni, Nikolai and Morari, Manfred and Pappas, George J.}, volume = {211}, series = {Proceedings of Machine Learning Research}, month = {15--16 Jun}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v211/aljalbout23a/aljalbout23a.pdf}, url = {https://proceedings.mlr.press/v211/aljalbout23a.html}, abstract = {Multi-robot manipulation tasks involve various control entities that can be separated into dynamically independent parts. A typical example of such real-world tasks is dual-arm manipulation. Learning to naively solve such tasks with reinforcement learning is often unfeasible due to the sample complexity and exploration requirements growing with the dimensionality of the action and state spaces. Instead, we would like to handle such environments as multi-agent systems and have several agents control parts of the whole. However, decentralizing the generation of actions requires coordination across agents through a channel limited to information central to the task. This paper proposes an approach to coordinating multi-robot manipulation through learned latent action spaces that are shared across different agents. We validate our method in simulated multi-robot manipulation tasks and demonstrate improvement over previous baselines in terms of sample efficiency and learning performance.} }
Endnote
%0 Conference Paper %T CLAS: Coordinating Multi-Robot Manipulation with Central Latent Action Spaces %A Elie Aljalbout %A Maximilian Karl %A Patrick van der Smagt %B Proceedings of The 5th Annual Learning for Dynamics and Control Conference %C Proceedings of Machine Learning Research %D 2023 %E Nikolai Matni %E Manfred Morari %E George J. Pappas %F pmlr-v211-aljalbout23a %I PMLR %P 1152--1166 %U https://proceedings.mlr.press/v211/aljalbout23a.html %V 211 %X Multi-robot manipulation tasks involve various control entities that can be separated into dynamically independent parts. A typical example of such real-world tasks is dual-arm manipulation. Learning to naively solve such tasks with reinforcement learning is often unfeasible due to the sample complexity and exploration requirements growing with the dimensionality of the action and state spaces. Instead, we would like to handle such environments as multi-agent systems and have several agents control parts of the whole. However, decentralizing the generation of actions requires coordination across agents through a channel limited to information central to the task. This paper proposes an approach to coordinating multi-robot manipulation through learned latent action spaces that are shared across different agents. We validate our method in simulated multi-robot manipulation tasks and demonstrate improvement over previous baselines in terms of sample efficiency and learning performance.
APA
Aljalbout, E., Karl, M. & Smagt, P.v.d.. (2023). CLAS: Coordinating Multi-Robot Manipulation with Central Latent Action Spaces. Proceedings of The 5th Annual Learning for Dynamics and Control Conference, in Proceedings of Machine Learning Research 211:1152-1166 Available from https://proceedings.mlr.press/v211/aljalbout23a.html.

Related Material