The automation of optical transient discovery and classification in Rubin-era time-domain astronomy

TL;DR

This paper reviews the current and future automation of transient discovery and classification workflows in optical astronomy, emphasizing the role of machine learning in handling the Rubin Observatory's large data volumes.

Contribution

It provides a comprehensive overview of the current state and future evolution of automated transient workflows in the era of Rubin Observatory data.

Findings

Automation has advanced to fully automated discovery and classification.

Machine learning enables rapid response follow-up.

Rubin Observatory will significantly increase transient discovery rates.

Abstract

Robotic wide-field time-domain surveys, such as the Zwicky Transient Facility and the Asteroid Terrestrial-impact Last Alert System, capture dozens of transients each night. The workflows for discovering and classifying transients in survey data streams have become increasingly automated over decades of development. The recent integration of machine learning and artificial intelligence tools has produced major milestones, including the fully automated end-to-end discovery and classification of an optical transient, and has enabled automated rapid-response space-based follow-up. The now-operational Vera C. Rubin Observatory and its Legacy Survey of Space and Time are accelerating the rate of transient discovery and producing large volumes of data at incredible rates. Given the expected order-of-magnitude increase in transient discoveries, one promising path forwards for optical time-domain astronomy is heavily investing in accelerating the automation of our workflows. Here we review the current paradigm of real-time transient workflows, project their evolution during the Rubin era and present recommendations for accelerating transient astronomy with automation.