Autonomous aquatic plant management: from sensing to a remote raking boat`

Unwanted, rooting aquatic plants grow in many Dutch waters. They impede flow, disrupt ecosystems and make water management complex. At the same time, that management must be done carefully, with an eye for ecology and biodiversity. Within the cluster Autonomous Green Management, various parties are working on new ways to organize that management more intelligently and efficiently.

"It is a challenge that is much bigger than people often think," says Leon Sterk, quartermaster at KOWW, a partner in the cluster. "You have to remove plants where necessary, but at the same time avoid damaging the ecosystem."

But anyone who wants to remove aquatic plants must first know where they are growing. KOWW is committed to implementing a methodology that was laid down in a NEN standard in 2021 and CEN standard a year later, among water managers. This will create a clear standard for inventorying and removing unwanted aquatic plants.

Through the Platform for Data-Driven Aquatic Plant Management (PDW), drones photograph waterways and send the images to the cloud. An AI model analyzes the photos based on color indexes and growth forms, recognizes different plant species and translates them into polygons in a GIS environment.

"The drone images are run through an AI model," Sterk explains. "That model detects the plants and projects polygons over the spots where they grow. That way, managers know exactly where which plants are growing, and soon the operator will see that in real time on the water." The system is now being used operationally for inventories and as input for implementers.

The next step is an autonomous raking boat that uses this data independently. However, fully autonomous boating in public spaces takes time, both technically and legally. "From inventory to a fully autonomous boat, we are probably five to seven years away," says Sterk. "That's why we are developing the monitoring and the boat in parallel, so we don't lose any time."

The intermediate step is a remote control boat that was recently completed. With it, tests are being conducted to see how the system deals with 4G and 5G connections, sensor integration and communication delay. "These are precisely the learning moments that will soon form the basis for fully autonomous systems," Sterk looks ahead.

The scale of the project also presents another challenge: the collection and processing of large amounts of data. "When we started with the platform, we had about twenty thousand photos," Sterk says. "But when we really got going, that quickly became hundreds of thousands."

Through the EDIH network, KOWW came into contact with the University of Groningen, where the datasets are now being processed on university computing infrastructure. This allows AI models to be trained faster and extended to include new plant species. "Through the cluster, you come into contact with parties you wouldn't easily find otherwise," says Sterk. "That helps enormously in taking these kinds of innovations further."

The inventory is operational, the remote control boat is working and the computing power has been arranged. Step by step, KOWW is building a system in which aquatic plants are automatically recognized and targeted for removal. This is badly needed, because the pressure on water managers is increasing while staffing levels are shrinking. "Autonomous systems are playing an increasing role in this field," says Sterk. "But it starts with knowing well what's growing underwater. If you have that insight, you can also take much more targeted action."

Wondering what else is happening in the Autonomous Green Management cluster?