NxtGen creates grower-focused tech for smart investments

Demonstration at GreenTech – data infrastructures and data standardisation is one focus of the NxtGen work for greenhouse horticulture.

Increasing difficulty in sourcing labour and a need to boost economic growth following the setback caused by the worldwide pandemic has seen millions of euros released to accelerate the development of ‘hands-free’ crop production in Dutch horticulture.

The support for the NxtGen Hightech programme, which extends to a range of industries important to the Dutch economy, comes from the Netherlands’ national growth fund, established in 2020. Robotics in agriculture and horticulture is one of the areas the programme has singled out for funding, with the work being led by the Dutch high-tech association FME and Wageningen University & Research (WUR), in cooperation with several companies many growers will be familiar with. These include technology specialists Priva and Hoogendoorn, vegetable breeders Rijk Zwaan, and HortiVation, a foundation of ag-tech companies involved in greenhouse design and construction.

Why do we need common standards?

The programme embraces three sectors – open-field crop production, food processing and greenhouse horticulture. “The goal for all is more autonomy; not just robots doing tasks but autonomy with decision-making, so better sensor networks and measuring systems,” says Erik Pekkeriet, who leads the Agro Food Robotics team at WUR.
Data infrastructures and standardisation of data are one focus of the work for greenhouse horticulture. “A lot of innovation stops because of problems integrating the technology with existing greenhouse systems,” he says. “Anything new, such as robotics or data-gathering systems, should be compatible with a grower’s existing management information system, not standing on its own. That’s why we need common standards so that data can be moved between and used by different applications.”

The goal for NxtGen is more autonomy, not just robots doing tasks but autonomy with decision-making, so better sensor networks and measuring systems are needed.

Digital twins

‘Digital twins’ for greenhouse crops is a second strand of research, with artificial intelligence start-up company Sobolt as the lead partner. This is where a computer model predicts how environmental changes or adjustments to any aspect of an individual cropping system, either alone or together, will affect plant growth and development ¬ and then makes the decision on how best to react to those changes without the grower having to intervene.

WUR has already produced growth models for several crops, says Pekkeriet, but while highly advanced, they are never exact enough. “They depend a lot on data about, for instance, light interception, so in practice have to be validated for specific locations as no two greenhouses are identical,” he says.

For a digital twin to predict outcomes with the precision required, it must be continually recalibrated with data from the grower’s actual crop, such as leaf number size and internode length, collected by cameras in the greenhouse. “We want to work out how best to train models to individual greenhouses – and to take account of differences in climate not just across the Netherlands but other countries too, so the models can made available to growers elsewhere,” he says.

“Then the work will be about how to offer growers a convenient package to use in practice so that the system feedback and continual learning do not depend on them for regular inputs. We aim for automated unsupervised machine learning and to ensure it is reliable.”

Handling of picked greenhouse crops

If drones are the information gatherers, robots are the doers, performing labour-intensive tasks like bell pepper picking.

The third area of work for greenhouse horticulture is to improve the way the picked crop is handled, rather than on robotic harvesting itself.

Those robots that have made it to market, or are still in development, are restricted in the quantity of product they can carry, so they need to be frequently off-loaded, says Mr Pekkeriet.

“A fast-moving handling system that can off-load the picking robot often, so it doesn’t have to stop for long, will have other advantages,” he adds. “For instance, you could think about making the harvest robot itself much smaller and the handling system much faster, by continuous unloading, creating a constant stream of picked fruit or cut flowers to the front of the greenhouse for processing and packaging.”

The choice of what technology to concentrate on within WUR’s NxtGen work has been driven by the ag-tech companies, which are members of the consortium. “They are in touch with their customers, so these are all technologies they think growers will want and will be prepared to invest in,” says Pekkeriet. “This is why so much emphasis in the programme is on validation.”

Demonstration greenhouse

Erik Pekkeriet leads the Agro Food Robotics team at WUR.

Much of the progress will be open to view in a demonstration greenhouse, which will give growers an opportunity to share their feedback. “A lot of technology doesn’t get used because it doesn’t work with existing systems, so this will help growers understand what’s happening and how they can adopt these technologies,” he says.

Pekkeriet says tomatoes are the crop that will feature most in the research, but cut flower crops will be included where possible. Dutch growers will not be alone in benefiting from the work as it will become widely available in time, with Dutch technology being so important to horticulture worldwide.

This article was first published in the February 2024 issue of FloraCulture International.