AI Surrogate Modeling for Real-Time 3D PPFD Estimation in Greenhouse Digital Twins
MSc Thesis Proposal by: Rustem Izmailov
Date: September 11, 2026
Time: 1:00 pm
Location: Dillon Hall DH253
Abstract:
Greenhouse digital twins require accurate and computationally efficient prediction of three-dimensional photosynthetic photon flux density (PPFD) to support crop modeling, lighting optimization, and future intelligent greenhouse control. While physics-based ray-tracing models can accurately simulate light distribution, their high computational cost limits their use in real-time applications. This research investigates an AI-based surrogate modeling approach for rapid prediction of 3D PPFD distributions in greenhouse crop canopies. The proposed methodology combines a calibrated physics-based digital twin with machine learning, where the digital twin is used to generate synthetic training data representing a wide range of lighting and environmental conditions. To improve applicability to real greenhouse environments, the AI model is further adapted and validated using measurements collected in a research greenhouse at the Agriculture and Agri-Food Canada Harrow Research and Development Centre. The research evaluates the proposed approach in terms of prediction accuracy, computational efficiency, and generalization capability, and compares its performance with conventional physics-based and analytical light modeling approaches. The expected outcome is a physically consistent and computationally efficient surrogate model that enables real-time prediction of greenhouse light distribution and supports the development of next-generation greenhouse digital twins and intelligent decision-support systems.
Keywords: AI Surrogate Modeling, Greenhouse Digital Twin, PPFD Prediction, hybrid learning, simulation-to-reality adaptation
Thesis Committee:
Internal Reader: Dr. Ziad Kobti
External Reader: Dr. Shahpour Alirezaee
Advisor: Dr. Xiaobu Yuan, Dr. Olena Syrotkina
