Data-Intensive Autonomous Intelligent Urban Farming for Whole-Plant Yield Optimization




To develop a cost-efficient, replicable, data-centered, autonomous, intelligent urban farming system that can be installed in any specified urban structure.


Partners: SHARE and NTU School of Materials Science and Engineering.

The project is conducted in collaboration with Netatech Engineering Pte Ltd.

Lead PIs:

Prof. Matan Gavish - School of Computer Science and Engineering, the Hebrew University, Jerusalem, Israel

Prof. Ng Kee Woei - School of Materials Science and Engineering, NTU, Singapore


Agricultural food production in urban settings must contend with limited resources: space, radiation, nutrients and water. This is a crucial step in achieving Singapore’s food production independence (e.g., for leafy greens) involves intelligent urban farming systems that are able to maximize crop yield in confined space and constrained resources.


The long-term vision of the research is to develop a cost-efficient, replicable, data-centered, autonomous, intelligent urban farming system that can be installed in any specified urban structure. The system will monitor and moderate culture parameters as external conditions change, and achieve optimal crop yield. The platform will be upscaled in a follow-up research project to full-size vertical or multi-layer horizontal farm with dynamic sunlight mirrors and LED lighting, toward deployment at industrial scale. The overarching vision initiated will increase Singapore’s urban food production capacity and the system itself can be commercialized as a unique “Made-in-Singapore” solution.


Optimizing crop yield will be achieved by minimizing the duration each individual plant spends away from maximum production rate, while providing optimal levels of water and nutrient supply, based on the plant's actual demands. This approach was designed to over come a crucial gap in the prevailing research paradigm: specifically, current research does not emphasize the whole-plant momentary physiological-status as the reference point, and focuses instead on maintaining predetermined levels of ambient conditions (e.g. light and temperature), or on tissue-specific sensors that do not necessarily reflect the whole plant physiological status. This approach represents a partial solution to the dynamic, multi-level challenge of optimized urban farming in general and of yield optimization in particular. Furthermore, existing efforts do not capitalize on innovative technological advances including: advanced data collection; Artificial Intelligence (AI), data science and optimal control; plant remote sensing; advanced organic soil-replacement substrates; precision water/nutrient delivery to the individual plant; and efficient illumination. These recent technological advances are inexpensive to deploy at scale and have an obvious potential to transform urban farming; crucially, advances in information technology now enable us to track and minimize stress of individual plants in real-time-using AI based on statistical models at a resolution previously considered impossible.