The Autonomous Greenhouse Challenge, hosted by Wageningen University & Research, continues to push the boundaries of what's possible in AI-driven horticulture. The competition, now in its fourth iteration, challenges growers to cultivate a pre-selected crop with absolutely no human intervention, from climate control to harvesting decisions.
Among the standouts was Team MuGrow, composed of experts from Gardin, TU Delft, Rijk Zwaan and Wageningen University. This group of innovators secured second place in the cultivation category using an unconventional approach: letting the plants themselves dictate the growing process. 
Julian Godding, Lead Data Scientist at Gardin
A new approach to autonomous growing
Team MuGrow's success wasn't just about machine learning (ML) or automated systems, it was about listening to the plants. Their strategy combined using plant biofeedback via the Gardin platform to make real-time growing decisions with machine learning (ML) to predict environmental responses and model predictive control (MPC) to fine-tune performance.
This resulted in record-breaking results:
- Highest yield: 340g of fresh tomatoes per pot
- Fastest growth cycle: Just 69 days from seedling to harvest
- Superior fruit quality: 7.3% dry matter content in cherry tomatoes
- 30% higher profitability than traditional human-led greenhouse methods
"These results are a powerful demonstration that using plant photosynthesis data as feedback on crop strategies can help growers significantly boost both yield and profitability," says Julian Godding, Lead Data Scientist at Gardin.
"We were the only team making decisions based on direct plant biofeedback. Instead of relying solely on pre-programmed climate models, we let the plants tell us what they needed."
Unlike traditional greenhouse setups, which primarily focus on environmental control, Team MuGrow's system responded directly to plant behavior, adjusting lighting, temperature, and nutrient levels in real time. 
The Gardin Pulse dashboard with four Key Plant Indicators: Health, Efficiency, Balance, Productivity. Team MuGrow's control algorithm aimed to control the greenhouse to achieve optimal set points for each indicator.
The power of AI and biofeedback
Greenhouse cultivation is a balancing act. Temperature, humidity, and light conditions shift constantly, and experienced growers usually make adjustments based on intuition. Team MuGrow's AI-driven approach took this to the next level by refining and improving those human instincts with data-driven precision.
Their ML model didn't just predict how plants would react but continuously learned and adapted based on live feedback. But because no model is perfect, model predictive control (MPC) was used to catch deviations and refine decisions.
"Think of it like an expert grower watching over the crops 24/7," Godding explains. "They observe changes, factor in costs and weather, then make adjustments. We did the same thing, except our model did it faster and with more precision."
Sustainability through smart energy management
One of Team MuGrow's standout innovations was its energy-efficient approach, a crucial factor given rising energy costs and the global shift toward decarbonization.
Instead of relying on gas heating, they maximized heat retention from LED lighting, reducing energy waste. By optimizing ventilation and shading, they significantly cut down gas consumption, making the entire operation more sustainable.
"LEDs don't just help plants grow, they generate heat," says Godding. "By strategically managing airflow and shading, we reduced the need for external heating, making the system more energy-efficient."
This approach aligns with the greenhouse industry's growing push toward electrification, proving that high yields and low energy consumption can go hand in hand. 
Detailed view of Efficiency indicator, equivalent to Phi(PSII) which was used to control the lighting, shading and CO2 strategy in the greenhouse.
Lessons from the competition
Designing a fully autonomous growing system is one challenge. Deploying it in under three months, without the opportunity for a full trial run in this specific environment, is another.
"We built and deployed the entire system in just a few months, without the ability to test it in this greenhouse setting beforehand," says Godding. "However, our sensors and technology had already been extensively validated in other environments prior to the challenge."
Despite these challenges, Team MuGrow delivered a high-performance system, while some competitors had to intervene manually, resulting in penalties.
The first-place team from Zhejiang University, China, took a different approach. Instead of relying on biofeedback, they planted at ultra-high density to keep costs low while maintaining efficiency.
"Their strategy worked well for the challenge," says Godding. "But our system was designed for real-world commercial greenhouses, where long-term viability matters just as much as short-term success." 
Gardin sensors monitoring tomato plants in a greenhouse.
The future of autonomous growing
One of the biggest insights from this challenge? The industry is shifting from climate-driven to plant-driven cultivation. "The future isn't just about managing the greenhouse environment, it's about responding to what the plants actually need," says Godding.
This is where Gardin's technology plays a crucial role. Traditional greenhouses rely on fixed schedules and environmental settings, but true autonomy requires real-time plant feedback.
Gardin started in vertical farms to validate their biofeedback technology but it was built to support all farm environments, with its applications having tremendous benefits for traditional greenhouse operations.
"Tomatoes, for example, have a far higher return than crops like lettuce," explains Godding. "This challenge proved that they can be grown efficiently, even in high-tech, autonomous settings."
What's next?
The Autonomous Greenhouse Challenge proved that AI and biofeedback aren't just concepts, but key considerations in the future of farming.
"One of the biggest constraints in agriculture today isn't just energy costs or sustainability regulations, it's the shortage of expert growers. The industry can't scale at its true potential because there aren't enough master growers to go around," says Godding.
With rapid advancements in machine learning, sensors, and automation, fully autonomous greenhouses are closer than ever, and poised to transform food production on a global scale.
For more information:
Gardin
Damiana Price, Head of Marketing
d.price@gardin.ag
www.gardin.ag
Sign up for our daily Newsletter and stay up to date with all the latest news!
Subscribe I am already a subscriber