The article introduces RadDQN, a deep Q-learning architecture aimed at achieving maximum radiation protection in the nuclear industry. The system operates on a radiation-aware reward function and employs unique exploration strategies to balance exploration and exploitation based on radiation exposure. The proposed model outperforms traditional DQN models in terms of convergence rate and training stability.
Publication date: 1 Feb 2024
Project Page: 123arXiv:2402.00468v1
Paper: https://arxiv.org/pdf/2402.00468
