Lighting control is a key driver of greenhouse productivity and operating costs, yet achieving high-resolution photosynthetic photon flux density (PPFD) regulation requires dense sensor deployments, which are costly to install, calibrate, and maintain. This work introduces a smart lighting digital twin (SLDT) that reduces sensing requirements through a virtual sensor concept, enabling multizone control with minimal instrumentation. The SLDT uses one-third of quantum sensors with a physics model to generate virtual sensor values. Controllers whose parameters are optimized via reinforcement learning (RL) over time use this feedback to respond to sunlight fluctuations and PPFD thresholds. To identify the most suitable control methodology, the SLDT evaluates 18 strategies. The methodology was validated in a greenhouse comprising nine lamps and monitored subzones with distinct PPFD requirements (200, 300, 400 μ mols−1m−2). Using the full Internet of Things (IoT) sensor network, the SLDT achieves a PPFD error of −0.16 μ mols−1m−2 and maintains PPFD within ±4% of the target for 90.47% of the time, while reducing energy consumption by 24.74% compared to a baseline on/off controller. When operating with only one third of the physical sensors, the virtual sensor-driven SLDT preserved high control accuracy, reporting a mean PPFD error of 0.72 μ mols−1m−2, a mean square error (mse) of 542.42, and stable energy savings of 23.72%. Virtual daily light integral (DLI) estimation errors remained within −3.48% to 1.49% across subzones. These results demonstrate high-accuracy, energy-efficient illumination control with significantly reduced sensing requirements, supporting scalable and cost-effective multicrop greenhouse management.

DOI:10.1109/JIOT.2026.3711085