Verification of Visual Controllers via Compositional Geometric Transformations

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Estornell, Alexander, Jung, Leonard, Everett, Michael
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866913929584705536
author Estornell, Alexander
Jung, Leonard
Everett, Michael
author_facet Estornell, Alexander
Jung, Leonard
Everett, Michael
contents Perception-based neural network controllers are increasingly used in autonomous systems that rely on visual inputs to operate in the real world. Ensuring the safety of such systems under uncertainty is challenging. Existing verification techniques typically focus on Lp-bounded perturbations in the pixel space, which fails to capture the low-dimensional structure of many real-world effects. In this work, we introduce a novel verification framework for perception-based controllers that can generate outer-approximations of reachable sets through explicitly modeling uncertain observations with geometric perturbations. Our approach constructs a boundable mapping from states to images, enabling the use of state-based verification tools while accounting for uncertainty in perception. We provide theoretical guarantees on the soundness of our method and demonstrate its effectiveness across benchmark control environments. This work provides a principled framework for certifying the safety of perception-driven control systems under realistic visual perturbations.
format Preprint
id arxiv_https___arxiv_org_abs_2507_04523
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Verification of Visual Controllers via Compositional Geometric Transformations
Estornell, Alexander
Jung, Leonard
Everett, Michael
Robotics
Machine Learning
Perception-based neural network controllers are increasingly used in autonomous systems that rely on visual inputs to operate in the real world. Ensuring the safety of such systems under uncertainty is challenging. Existing verification techniques typically focus on Lp-bounded perturbations in the pixel space, which fails to capture the low-dimensional structure of many real-world effects. In this work, we introduce a novel verification framework for perception-based controllers that can generate outer-approximations of reachable sets through explicitly modeling uncertain observations with geometric perturbations. Our approach constructs a boundable mapping from states to images, enabling the use of state-based verification tools while accounting for uncertainty in perception. We provide theoretical guarantees on the soundness of our method and demonstrate its effectiveness across benchmark control environments. This work provides a principled framework for certifying the safety of perception-driven control systems under realistic visual perturbations.
title Verification of Visual Controllers via Compositional Geometric Transformations
topic Robotics
Machine Learning
url https://arxiv.org/abs/2507.04523