Research paper
MAGICS: Adversarial RL with Minimax Actors Guided by Implicit Critic Stackelberg for Convergent Neural Synthesis of Robot Safety
MAGICS uses a structured minimax learning game for neural safety-policy synthesis. The study presents local convergence results and evaluates simulated tasks and a quadruped. Local convergence under stated assumptions differs from a global safety guarantee for a deployed robot.
Reviewed 2026-11-05 · Source and bibliography checked
- Original title
- MAGICS: Adversarial RL with Minimax Actors Guided by Implicit Critic Stackelberg for Convergent Neural Synthesis of Robot Safety
- Publisher
- arXiv
- Authors
- Wang, Justin, Hu, Haimin, Nguyen, Duy Phuong, Fisac, Jaime Fernández
- DOI
- 10.48550/arxiv.2409.13867
- First archive posting
- 2024-09-20
- Source language
- en
- Source country
- US
- Edition
- v2
- Access
- Open archive metadata and manuscript
- Publication status
- Preprint
Related resources and engineering notes
Source overview
MAGICS uses a structured minimax learning game for neural safety-policy synthesis. The study presents local convergence results and evaluates simulated tasks and a quadruped. Local convergence under stated assumptions differs from a global safety guarantee for a deployed robot.
Use and review
Original summary of the primary archive abstract and bibliographic record. Specialist technical and native-language review remain pending. The reported method does not establish a deployed safety function or achieved PL.
Sources & editions
MAGICS: Adversarial RL with Minimax Actors Guided by Implicit Critic Stackelberg for Convergent Neural Synthesis of Robot Safety ↗
arXiv · v2
arXiv abstract, authors, submission history and DOI links; first posting 2024-09-20, abstract reviewed in v2. Publisher dates are separate.