VIBR: Learning View-Invariant Value Functions for Robust Visual Control - CEA - Commissariat à l’énergie atomique et aux énergies alternatives
Conference Papers Year : 2023

VIBR: Learning View-Invariant Value Functions for Robust Visual Control

Abstract

End-to-end reinforcement learning on images showed significant progress in the recent years. Data-based approach leverage data augmentation and domain randomization while representation learning methods use auxiliary losses to learn task-relevant features. Yet, reinforcement still struggles in visually diverse environments full of distractions and spurious noise. In this work, we tackle the problem of robust visual control at its core and present VIBR (View-Invariant Bellman Residuals), a method that combines multi-view training and invariant prediction to reduce out-of-distribution (OOD) generalization gap for RL based visuomotor control. Our model-free approach improve baselines performances without the need of additional representation learning objectives and with limited additional computational cost. We show that VIBR outperforms existing methods on complex visuo-motor control environment with high visual perturbation. Our approach achieves state-of the-art results on the Distracting Control Suite benchmark, a challenging benchmark still not solved by current methods, where we evaluate the robustness to a number of visual perturbators, as well as OOD generalization and extrapolation capabilities.
Fichier principal
Vignette du fichier
dupuis23a.pdf (4.06 Mo) Télécharger le fichier
Origin Files produced by the author(s)

Dates and versions

cea-04488071 , version 1 (04-03-2024)

Identifiers

Cite

Tom Dupuis, Jaonary Rabarisoa, Quoc-Cuong Pham, David Filliat. VIBR: Learning View-Invariant Value Functions for Robust Visual Control. CoLLAs 2023 - 2nd Conference on Lifelong Learning Agents, Aug 2023, Montréal, Canada. pp.658--682, ⟨10.48550/arXiv.2306.08537⟩. ⟨cea-04488071⟩
78 View
51 Download

Altmetric

Share

More