Equivariant Reinforcement Learning under Partial Observability

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Nguyen, Hai, Baisero, Andrea, Klee, David, Wang, Dian, Platt, Robert, Amato, Christopher
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909295946235904
author Nguyen, Hai
Baisero, Andrea
Klee, David
Wang, Dian
Platt, Robert
Amato, Christopher
author_facet Nguyen, Hai
Baisero, Andrea
Klee, David
Wang, Dian
Platt, Robert
Amato, Christopher
contents Incorporating inductive biases is a promising approach for tackling challenging robot learning domains with sample-efficient solutions. This paper identifies partially observable domains where symmetries can be a useful inductive bias for efficient learning. Specifically, by encoding the equivariance regarding specific group symmetries into the neural networks, our actor-critic reinforcement learning agents can reuse solutions in the past for related scenarios. Consequently, our equivariant agents outperform non-equivariant approaches significantly in terms of sample efficiency and final performance, demonstrated through experiments on a range of robotic tasks in simulation and real hardware.
format Preprint
id arxiv_https___arxiv_org_abs_2408_14336
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Equivariant Reinforcement Learning under Partial Observability
Nguyen, Hai
Baisero, Andrea
Klee, David
Wang, Dian
Platt, Robert
Amato, Christopher
Robotics
Artificial Intelligence
Computer Vision and Pattern Recognition
Incorporating inductive biases is a promising approach for tackling challenging robot learning domains with sample-efficient solutions. This paper identifies partially observable domains where symmetries can be a useful inductive bias for efficient learning. Specifically, by encoding the equivariance regarding specific group symmetries into the neural networks, our actor-critic reinforcement learning agents can reuse solutions in the past for related scenarios. Consequently, our equivariant agents outperform non-equivariant approaches significantly in terms of sample efficiency and final performance, demonstrated through experiments on a range of robotic tasks in simulation and real hardware.
title Equivariant Reinforcement Learning under Partial Observability
topic Robotics
Artificial Intelligence
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2408.14336