Safe Reinforcement Learning via Recovery-based Shielding with Gaussian Process Dynamics Models Conference: Proc. of the 25th International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2026); May 25 – 29, 2026; Paphos, Cyprus #281

Alexander W. Goodall, Francesco Belardinelli · 2026

Read Safe Reinforcement Learning via Recovery-based Shielding with Gaussian Process Dynamics Models Conference: Proc. of the 25th International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2026); May 25 – 29, 2026; Paphos, Cyprus #281 by Alexander W. Goodall, Francesco Belardinelli (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-12444.txt.

Description

Reinforcement learning can optimise control, but it often lacks provable safety guarantees for safety-critical systems. Alexander W. Goodall led this work with one co-author, Francesco Belardinelli, presenting a recovery-based shielding framework for safe RL in unknown, non-linear continuous systems. A backup policy, or shield, joins the learning agent, using Gaussian process uncertainty estimates to predict safety violations.

When danger looms, the shield recovers the system to safe paths, intervening only when necessary. Experience from the shielded agent builds those models, and policy tuning uses internal model-based sampling, allowing free exploration and sample-efficient learning without compromising safety. Tests on continuous control settings show strong performance with strict safety compliance, giving learning agents a provable safety lower bound.

Catalog record

CREATORAlexander W. Goodall, Francesco Belardinelli
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE97 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602-12444

Browse the full library · About the Tower of Babel