Fish farming in the U.S. is a labor-intensive practice that's falling behind more technologically advanced operations in other countries. But a group of engineers from Florida has invented a new robotics system based on machine-learning algorithms that can modernize the industry and automate water-quality monitoring in aquaculture.

In the last decade, global fish production has shifted so that more than half of all seafood consumed now comes from aquaculture, referring to the farming of fish, shellfish, aquatic plants and other animals, including those for human consumption, as opposed to being wild-caught. The U.S. is considered a minor aquaculture producer compared to top-ranking countries such as China, Indonesia and India. 

There are several different methods of fish farming in practice today, but all need "drastic improvements" in order to be sustainable and economically viable, Bing Ouyang, a professor at Florida Atlantic University and co-inventor of the technology, the patent application for which was published by the U.S. Patent and Trademark Office on March 4, told The Academic Times. 

"The U.S. has such rich water resources. We have lakes and rivers all over the country and coastlines on both sides of the country. And we are running a trade deficit of $16 billion a year, which is one of the largest [trade deficits] in seafood," Ouyang said. "That's why we're really motivated to develop solutions to help fish farmers in the United States to catch up to the rest of the world. Robotics is one key answer for this kind of operation."

The majority of current aquaculture infrastructure in the U.S. uses pond systems that can be labor-intensive, resource-inefficient and expensive to maintain. Offshore aquaculture uses anchored, submerged cages in bodies of water that catch fish native to the area. Once the fish grow to a certain size, they are harvested and sold. Other farms erect tanks in a factory-like setting where fish can be raised in a very controlled environment. 

A combination of these two approaches has become increasingly popular, Ouyang said, by digging ponds into the ground to raise, monitor and harvest fish. Some factors of these environments can still be controlled, such as feeding and water oxygen levels, which are the two most critical aspects of pond-raised aquaculture, according to Ouyang. 

The robotics system invented by the researchers, known as the Hybrid Aerial/Underwater Robotic System, or HAUCS, is designed for this combination method. It can be integrated into vehicles or drones and uses a sensor to gather water quality data for the pond water. Machine learning tools then analyze the data to diagnose the condition of the water and ensure its suitability for fish growth. 

"As a whole, the HAUCS framework is innovative and streamlines continuous monitoring, maintenance and forecasting of next-generation fish farms. The HAUCS framework offers a highly flexible and scalable solution that can be adapted to a diversity of farms to achieve farm-level and pond-level monitoring," Ouyang and co-inventors Paul Wills, Jason Hallstrom and Tsung-Chow Su wrote in the patent application.  

Water quality is most often monitored manually by workers who drive trucks around to each pond on the fish farm, but Ouyang explained that some aquaculture operations already use automated sensors to monitor water quality. The sensors are typically attached to a flotation device and placed directly into the ponds. This method works well for automatically collecting data without needing human involvement, but it also creates problems.

Fish farms can contain hundreds of ponds, and it is usually too expensive to equip each pond with a floating sensor. Additionally, letting the sensors sit in constant contact with water can degrade their performance over time. 

"Another limitation of such a stationary sensor is that only a single location is monitored unless multiple expensive sensor buoys are deployed in each pond. In addition, sensor buoys are an obstruction in the pond during harvest as they must be removed or lifted over the seine," Ouyang said. 

The team's invented system can be mounted onto trucks, autonomous vehicles and drones, offering a range of solutions to meet the needs of different aquaculture structures. With the sensor attached to a drone, it can be flown to each pond and only the sensor payload will be submerged underwater to measure the water quality, eliminating the need for floating devices. 

The authors of the patent application said that compared to available alternatives, their system is more accurate in reporting pond conditions, can be more easily scaled up and offers novel sensing patterns to cover different areas of a large pond. 

Under all weather conditions, the sensing platform is designed to be power efficient to handle multiple bodies of water on an hourly basis, relatively easy to maintain, relatively inexpensive to manufacture, replace or repair, and able to report sensor data to centralized computing devices in real time, according to the patent application.

"A major reason for the low adoption of robotic technology in aquaculture fish farming is the lack of accessibility to technology for the fish farmers," the inventors said, recommending that demonstration sites for their system be established to help educate farmers. 

Ouyang said the focus of the technology for now is on monitoring ponds for dissolved oxygen level, temperature, pH and other environmental water-quality conditions. But in the future, he hopes to expand to monitoring other conditions, such as feeding, fish density and diseases in the fish. 

The project has been funded by the National Institute of Food and Agriculture at the U.S. Department of Agriculture for the last three years. The research team had plans in 2020 to establish partnerships with fish farms in Missouri in order to test their drone prototype, but these efforts were halted due to the COVID-19 pandemic. Ouyang said the collaboration plans will be revisited this year. 

The application for the patent, "Hybrid Aerial/Underwater Robotics System for Scalable and Adaptable Maintenance of Aquaculture Fish Farms," was filed on March 6, 2020, to the U.S. Patent and Trademark Office. It was published on March 4, 2021 with the application number 16/811669. The earliest priority date was May 6, 2019. The inventors of the pending patent are Bing Ouyang, Paul Wills, Jason Hallstrom, Tsung-Chow Su, Martin Richie, Peter Reiff and Yufei Tang. The assignee is the Florida Atlantic University Board of Trustees.

Parola Analytics provided technical research for this story.

Correction: A previous version of this story misidentified the patent application as a patent. The error has been corrected.