Tag
Proposes a calibration-first reward-component auditing framework for reinforcement learning control in smart greenhouses, enabling decomposition of scalar rewards into interpretable components and validating them across simulated and real-world data.