Calibration-First Reward-Component Auditing for Reinforcement Learning Control in Smart Greenhouses

Researchers propose a calibration-first reward audit framework for reinforcement learning control in smart greenhouses, enabling comparison of reward components across simulator training, facility-adapted rollouts, logged competition records, and actuator-rule distillation. The framework decomposes scalar rewards into conditional terms for temperature, CO2, humidity, vapor-pressure deficit, screen, and actuation-proxy, and adapts the GreenLight simulator to the second Autonomous Greenhouse Challenge data.

arXiv:2607.11959v1 Announce Type: new Abstract: Greenhouse reinforcement learning can test climate-control ideas at a speed and scale that is difficult to achieve with crop experiments alone. For smart-greenhouse control, however, a single simulator return is not enough: a grower or control engineer also needs to know when the policy heats, enriches CO2, vents, manages humidity, deploys screens, or uses lamps.We propose a reproducible calibration-first reward audit framework that keeps named greenhouse-control reward components comparable across simulator training, facility-adapted rollouts, logged Autonomous Greenhouse Challenge records, and actuator-rule distillation. In GreenLight-Gym, the framework decomposes the scalar reward into conditional temperature, CO2, humidity and vapor-pressure-deficit, screen, and actuation-proxy terms; adapts GreenLight to the second Autonomous Greenhouse Challenge logged climate traces; and scores the same components on logged greenhouse data.