Reinforcement Learning (RL) applications in real-world scenarios must prioritize safety and reliability, which impose strict constraints on agent behavior. Model-based RL leverages predictive world models for action planning and policy optimization, but inherent model inaccuracies can lead to catastrophic failures in safety-critical settings. We propose a novel model-based RL framework that jointly optimizes task performance and safety. To address world model errors, our method incorporates an adaptive mechanism that dynamically switches between model-based planning and direct policy execution. We resolve the objective mismatch problem of traditional model-based approaches using an implicit world model. Furthermore, our framework employs dynamic safety thresholds that adapt to the agent’s evolving capabilities, consistently selecting actions that surpass safe policy suggestions in both performance and safety. Experiments demonstrate significant improvements over non-adaptive methods, showing that our approach optimizes safety and performance simultaneously rather than merely meeting minimum safety requirements. The proposed framework achieves robust performance on diverse safety-critical continuous control tasks, outperforming existing methods.
DOI: 10.3233/FAIA251029
Скачать статью (PDF) на сайте конференции (англ.): https://ebooks.iospress.nl/doi/10.3233/FAIA251029
Скачать сборник целиком (PDF) на сайте конференции (англ.): https://ebooks.iospress.nl/volume/ecai-2025-28th-european-conference-on-artificial-intelligence-bologna-including-14th-conference-on-prestigious-applications-of-intelligent-systems-pais-2025
Скачать статью (PDF) или читать онлайн на arXiv.org (англ.): https://arxiv.org/html/2506.04828
ResearchGate: https://www.researchgate.net/publication/396778419_Safe_Planning_and_Policy_Optimization_via_World_Model_Learning
Latyshev A., Gorbov G., Panov A. I. Safe Planning and Policy Optimization via World Model Learning // Frontiers in Artificial Intelligence and Applications: Proceedings of the 28th European Conference on Artificial Intelligence (ECAI 2025). — IOS Press, 2025. — Vol. 413. — P. 1945–1952.