ALLO: A Photorealistic Dataset and Data Generation Pipeline for Anomaly Detection During Proximity Operations in Lunar Orbit

Selina Leveugle, Chang Won Lee, Svetlana Stolpner, Chris Langley, Paul Grouchy, Steven L. Waslander, Jonathan Kelly

Keywords: Space robotics, anomaly detection, dataset

TL;DR: A rendered dataset and generation pipeline for anomaly detection in lunar orbit.

Abstract:

NASA’s Artemis missions are paving the way for the return of sustained human operations on and around the Moon. A key challenge for extended space missions is the unprecedented level of autonomy required for system control. This need arises from communication delays and limited human oversight. A critical component of autonomous operation is the ability of external robotic systems—such as future versions of the Canadarm2 currently operating on the International Space Station—to detect hazards from onboard imagery under
extreme and variable lighting conditions. Traditional anomaly detection methods, which rely heavily on large labelled datasets, are not well suited for this setting due to the scarcity of representative imagery and the difficulty of obtaining ground-truth annotations. To address this gap, we introduce ALLO (Anomaly Detection in Lunar Orbit), a novel photorealistic synthetic dataset generated using Blender. The dataset includes 36,385 anomaly-free images and 15,024 anomalous images representing robotic-arm operations in lunar orbit, with realistic variations in illumination, pose, and background. We contribute (i) a photorealistic dataset for anomaly detection, (ii) an open-source Blender-based data generation pipeline for scalable dataset creation, and (iii) a qualitative validation of the rendered imagery against real photographs from the International Space Station (ISS). By releasing both ALLO and its data generation pipeline, we provide a tool for creating realistic space datasets and for developing and evaluating anomaly detection and localization methods for space applications. The dataset and code are available at: https://github.com/utiasSTARS/ALLO.git.

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.

Paper Acceptance: Yes

Submission Number: 11