Game of Trawls: towards fully automated intelligent gears to eliminate bycatch in trawl fisheries

For over two centuries, trawling fisheries have operated blindly; catches could not be known until onboard. These activities, which represents over half of seafood protein supply and of fishermen incomes, worldwide, are facing increasing societal pressure for their impacts on the ecosystem and their greenhouse gas emissions. Yet as of today, no alternative exists to maintain the same levels of production while reducing the impacts on the species welfare and ecosystems. We here demonstrate that previously blind trawling activities can now be turned into informed fishing, by developing a trawl system that can automatically detect and exclude bycatch species in real-time, using artificial intelligence (AI). The Game of Trawls system was achieved by coupling acoustic communication, actuators, embedded computer vision and AI trained to recognize both targeted species and unwanted species to release, in real-time. Full size trials were conducted in lab and are now being tested at sea to demonstrate the proof of concept that trawls catch can be monitored and exclusion devices triggered autonomously without human interaction. Several limits remain, in particular with the robustness of species detection by the AI, but these applied technological developments illustrate how wisely operated AI applications may help reaching a sustainable exploitation of marine resources and improving welfare of unwanted catches, offering a new step towards precision fishing.