Safe Reinforcement Learning via Observation Shielding
Files
Date
2023-01-03
Contributor
Advisor
Department
Instructor
Depositor
Speaker
Researcher
Consultant
Interviewer
Narrator
Transcriber
Annotator
Journal Title
Journal ISSN
Volume Title
Publisher
Volume
Number/Issue
Starting Page
6603
Ending Page
Alternative Title
Abstract
Reinforcement Learning (RL) algorithms have shown success in scaling up to large problems. However, deploying those algorithms in real-world applications remains challenging due to their vulnerability to adversarial perturbations. Existing RL robustness methods against adversarial attacks are weak to large perturbations - a scenario that cannot be ruled out for RL adversarial threats, as is the case for deep neural networks in classification tasks. This paper proposes a method called observation-shielding RL (OSRL) to increase the robustness of RL against large perturbations using predictive models and threat detection. Instead of changing the RL algorithms with robustness regularization or retrain them with adversarial perturbations, we depart considerably from previous approaches and develop an add-on safety feature for existing RL algorithms during runtime. OSRL builds on the idea of model predictive shielding, where an observation predictive model is used to override the perturbed observations as needed to ensure safety. Extensive experiments on various MuJoCo environments (Ant, Hooper) and the classical pendulum environment demonstrate that our proposed OSRL is safer and more efficient than state-of-the-art robustness methods under large perturbations.
Description
Keywords
Cyber Operations, Defense, and Forensics, adversarial examples, reinforcement learning, robustness, safety, shielding
Citation
Extent
10
Format
Geographic Location
Time Period
Related To
Proceedings of the 56th Hawaii International Conference on System Sciences
Related To (URI)
Table of Contents
Rights
Attribution-NonCommercial-NoDerivatives 4.0 International
Rights Holder
Local Contexts
Collections
Email libraryada-l@lists.hawaii.edu if you need this content in ADA-compliant format.