As climate change destabilises global agriculture and demand for fresh produce rises, autonomous greenhouse control is emerging as a critical solution. In the 4th Autonomous Greenhouse Challenge— the first requiring full autonomy from potting to harvest—Team MuGrow (TU Delft, Gardin, and Rijk Zwaan) demonstrated how real-time plant biofeedback can unlock major productivity gains.
Using Gardin’s chlorophyll-fluorescence technology to measure photosynthesis in real time, MuGrow built a grower-inspired control system combining biofeedback, model predictive control, and reinforcement learning. The result: the highest yield, highest quality, and shortest crop cycle in the competition. MuGrow achieved 340 g/pot, 7.3% dry matter, and a 69-day cycle, equal to 44 kg/m²/year—a 45% increase over industry averages and 30% above the challenge’s reference grower. At typical farm-gate prices, this represents €200,000/ha additional annual value, with profits rising up to 60% with optimised harvest timing.
While MuGrow served as a proof of concept, Gardin’s sensor platform and its four Plant Indicators—HEALTH, BALANCE, EFFICIENCY, and PRODUCTIVITY—are already commercially deployed. These indicators give growers direct, real-time insights into plant physiology, helping optimise lighting, temperature, CO₂, and overall cultivation strategy. Growers using Gardin today report improved resilience, earlier stress detection, and more efficient use of resources.
“Chlorophyll fluorescence is one of the most powerful indicators of real-time plant performance,” said Julian Godding, Head of Science at Gardin. “Gardin brings this capability into the modern greenhouse through simple insights that assist day-to-day cultivation.”
Dr. Robert D. McAllister of TU Delft added, “If you cannot measure it, you cannot control it. Gardin’s sensor provided a unique biofeedback signal that made a new class of intelligent control possible.”
As full autonomy continues to advance, Gardin’s technology is already enabling growers worldwide to improve yields, optimise inputs, and future-proof their operations.
To read more on this and more from Gardin, please use the following link
