Resilient Robotic Autonomy for Underwater Operations in Fish Farms

This project envisions reshaping the underwater operations in dynamic, complex and perceptually degraded environments by developing new knowledge and novel technology to enable resilient autonomy for Unmanned Underwater Vehicles (UUVs). ResiFarm challenges the resilience in one of the most demanding industrial environments such as fish farms. Motivated by the core hypothesis that there exists a broadly defined and unified science for resilient autonomy of UUVs operating in complex, high risk, perceptually-degraded and dynamic environments, the envisioned research will be holistically organized around three cross-cutting objectives that are addressing the current knowledge gaps on: a) resilient multi-modal perception for UUVs operating in dynamic perceptually-degraded environments, b) methods for novel motion planning concepts that enable safe, fish- and structure-aware operations and c) validation of the fully integrated system in field studies. The majority of the operations in fish farms involve the use of novel technologies to address the current and future challenges the industry is facing. To increase the level of automation and contribute to more optimal and efficient operations, it is crucial to develop methods for resilient perception and robust motion planning of UUVs operating in complex underwater environments. Therefore, from 2021 to 2024, several motion planning methods (ResiPlan, RUMP, ResiVis) have been developed and validated in industrial-scale fish farms on SINTEF ACE. The results showcased that robust motion planning methods can be crucial to reaching collision-free navigation in a dynamic fish farming environment. When it comes to resilient perception, the last two years the ResiFarm project targeted the development of new robotic systems with integrated sensors to be utilized to study multi-modal perception methods. Both acoustic and vision sensors have been integrated and achieved SW and HW synch data to investigate robust localization and mapping methods. A novel online self-calibration refractive camera model with application to the underwater domain has been developed. Several publications and datasets have been realized in this project. In the future, the project will focus on extending and validating the investigated methods for resilient perception and the integration of perception and motion planning modules. Final experiments are planned in industrial-scale fish farms in 2025 to demonstrate fully autonomous inspection and intervention operations. By utilising the competence of the interdisciplinary team (SINTEF, NTNU), industry partners' (Eelume, Skarv Technologies) expertise in the underwater robotic and automation domain, and involving international experts in topics relevant to the project (MIT, LSTS, TUM, ETH), our team aims to provide the foundation for the new generation of permanent resident UUVs that co-exist with fish without causing negative impact and autonomously navigate and interact with flexible structures. The result of this project will be the new science and systems to facilitate long-term autonomous underwater operation of UUVs and promote sustainable expansion in fish farms and other maritime industries such as fisheries, subsea oil and gas and offshore wind farms. Overall, ResiFarm will impact research communities, businesses, the public sector and society at large through collaboration between these and with the outmost dedication to developments and demonstrations of novel methods within artificial intelligence, automation and robotics.

This project envisions to reshape the underwater operations in dynamic, complex and perceptually-degraded environments by developing new knowledge and novel technology to enable resilient autonomy for Unmanned Underwater Vehicles (UUVs). ResiFarm challenges the resilience in one of the most demanding industrial environments such as fish farms. Motivated by the core hypothesis that there exists a broadly defined and unified science for resilient autonomy of UUVs operating in complex, high risk, perceptually-degraded and dynamic environments, the envisioned research will be holistically organized around three cross-cutting objectives that are addressing the current knowledge gaps on: a) resilient multi-modal perception for UUVs operating in dynamic perceptually-degraded environments, b) methods for novel motion planning concepts that enable safe, fish- and structure-aware operations and c) validation of the fully integrated system in field studies. By utilising the competence of the interdisciplinary team (SINTEF, NTNU), industry partners' (Eelume, Skrav Technologies) expertise in the underwater robotic and automation domain, and involving international experts in topics relevant to the project (MIT, LSTS, TUM, ETH), our team aims to provide the foundation for the new generation of permanent resident UUVs that co-exist with fish without causing negative impact and autonomously navigate and interact with flexible structures. The result of this project will be the new science and systems to facilitate long-term autonomous underwater operation of UUVs and promote sustainable expansion in fish farms and other maritime industries such as fisheries, subsea oil and gas and offshore wind farms. Overall, ResiFarm will impact research communities, businesses, the public sector and society at large through collaboration between these and with the outmost dedication to developments and demonstrations of novel methods within artificial intelligence, automation and robotics.