Researchers create “Virtual Fish” to revolutionize aquaculture feeding

NORWAY – A Norwegian research team has unveiled an early-stage digital twin model that could transform feeding practices in aquaculture by reducing feed waste and improving fish growth efficiency. 

The initiative, known as FishMet, is being developed at the University of Bergen by Professor Ivar Rønnestad and researcher Sergei Budaev in collaboration with Vestlandets Innovasjonsselskap (VIS).

FishMet functions as a “virtual fish,” integrating biological and environmental data to simulate appetite, digestion, metabolism, and growth. 

Unlike many black-box artificial intelligence systems, the model is built on decades of biological research and neurophysiological feedback loops that regulate appetite and growth.

The system can process inputs such as fish size, feed type, feeding frequency, water temperature, oxygen levels, and behaviour. 

Its outputs include predictions of feed intake, gut transit times, growth efficiency, and even stress or motivation levels.

It can be applied to individual fish or entire populations, making it adaptable for both research and farm-level decision-making.

“We aim to create a transparent digital salmon that combines AI with decades of biological knowledge, serving both as a research tool and a practical aquaculture predictor, especially in situations lacking data,” said researcher Sergei Budaev.

From lab research to proof of concept

The FishMet project originated from years of experimental work by Rønnestad and Budaev’s research groups, who studied the physiological mechanisms of appetite regulation. 

Their work examined how gut-brain signalling, digestion rates and neurohormones shape feeding behaviour in fish. These insights now underpin FishMet’s algorithms.

Early pilot studies have shown promising results. In trials with rainbow trout and Atlantic salmon, the model demonstrated predictive accuracy in estimating gut transit times and growth performance. 

The research team believes that once validated further, the system could help fish farmers cut feed costs, improve production efficiency and lessen environmental impacts linked to overfeeding.

VIS has made the proof-of-concept available for exploratory licensing, creating opportunities for aquaculture companies to test its potential applications in commercial settings. 

Although the project remains at a low Technology Readiness Level, its transparent, biology-driven approach could provide a valuable alternative to conventional machine learning models.

Future potential and sustainability goals

Looking ahead, the researchers hope to expand FishMet to include other species and critical life stages such as smoltification and maturation. 

They see the digital twin as part of a broader move toward precision aquaculture, where advanced models and data-driven tools optimize resources while promoting animal welfare and environmental stewardship.

While still in its infancy, FishMet underscores how fundamental university research can be translated into real-world applications with the help of innovation structures like VIS. 

If further validated and adopted, the “virtual fish” could become a vital tool for the salmon and trout industry, aligning profitability with sustainability in one of the world’s fastest-growing food sectors.

Sign up HERE to receive our email newsletters with the latest news updates and insights from Africa and the World, and follow us on our WhatsApp channel for updates.