Guardado en:
Detalles Bibliográficos
Autores principales: Gronauer, Sven, Haider, Tom, da Roza, Felippe Schmoeller, Diepold, Klaus
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:https://arxiv.org/abs/2402.04182
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
Tabla de Contenidos:
  • Reinforcement learning algorithms need exploration to learn. However, unsupervised exploration prevents the deployment of such algorithms on safety-critical tasks and limits real-world deployment. In this paper, we propose a new algorithm called Ensemble Model Predictive Safety Certification that combines model-based deep reinforcement learning with tube-based model predictive control to correct the actions taken by a learning agent, keeping safety constraint violations at a minimum through planning. Our approach aims to reduce the amount of prior knowledge about the actual system by requiring only offline data generated by a safe controller. Our results show that we can achieve significantly fewer constraint violations than comparable reinforcement learning methods.