Robotics underneath sub-zero waters and outer space

Ocean Worlds is a program focused on exploring the oceans throughout our solar system, not just those on Earth. To achieve this, we combine robotics with newly developed communication technologies needed to explore oceans both on Earth and in space. Investigating the most inaccessible ocean environments also requires advanced communication through different types of ice, slush, and ice-water interfaces—capabilities that did not exist before the start of the project. Exploring remote regions such as the waters beneath the North Pole or underneath the ice on Enceladus demands sophisticated sensing and navigation capabilities for both mapping and missions like searching for life. We have therefore invested in autonomous underwater vehicles (AUVs) as a common platform for exploration on Earth and beyond. The SubZeroSpace project, funded by the Research Council of Norway, aims to consolidate the challenges encountered in the Ocean Worlds program into a unified understanding and technological platform for use both in space and on Earth. We have utilized cutting-edge artificial intelligence that enables us to both see and communicate through ice. This allows us to detect obstacles or other points of interest through ice and the ice-water interface. Furthermore, by using water as an acoustic medium, we can now communicate both vertically and horizontally through ice, creating entirely new opportunities to transmit data from previously inaccessible areas. The technology developed in the project was initially tested experimentally in the lab and later tried out in field conditions on Svalbard and the Juvfonne glacier. This was made possible by Norway’s strong commitment to ocean research in the Trondheimsfjord via the Ocean Space Center, the lab facilities at Mustad Autoline in Gjøvik, and the laboratory and AUV capabilities of Kongsberg Discovery in Horten. These small-scale experiments have been validated in the field by using Svalbard as a strategic platform for testing robotics in ice-covered environments. Another key application for the acoustic technology developed in the project is biomass estimation, where it is used in echo sounders to identify and estimate both fish species and the biomass of fish schools. At the same time, the same methods provide a foundation for space operations, as acoustics can be used to detect signs of biomass in the form of air bubbles or other biologically relevant markers. This opens up new possibilities for detecting and mapping living organisms under extreme conditions, both in Earth’s oceans and in ice-covered environments elsewhere in the solar system. These applications have been developed using hybrid artificial intelligence methods, including an open-source pipeline on GitHub that allows users to build, train, and test machine learning techniques for acoustics on Kongsberg Discovery’s EK platform. This software makes several of our techniques available and enables researchers and industry professionals to adapt our methods and data for their own purposes. The project has been conducted with high ambitions within a capable Norwegian-American consortium consisting of SINTEF, Kongsberg Discovery, NASA-JPL, the University of Oslo, the Norwegian University of Science and Technology, and Mustad Autoline. We believe that the technology developed in this project will have a significant impact on the exploration of the ocean realm in the near future.

Through SubZeroSpace, we pushed the boundary of the application of adopting acoustic sensors along with AI & Machine Learning for both terrestrial and extraterrestrial missions. Using acoustics, the consortium evaluated various acoustic frequencies and established a proof-of-concept wireless acoustic communication through ice at the distance of 200 meters in a setting with real-world heterogeneous elements. Additionally, in a similar real-world setting, the team experimented with the use of acoustic sensors for solid/obstacle detection and collected a novel dataset with promising features for further analysis and modeling. In combination, these newly discovered utilities have the potential to improve the survivability of exploration vehicles in ice, on earth and beyond. In the ocean, the consortium successfully deployed an embedded Machine Learning model using SOTA AI inference hardware on an autonomous vessel simulating the condition of ocean exploration in Europa. The team further demonstrated the viability of applying Unsupervised Machine Learning techniques along with SOTA vision foundation models on proxy datasets to interpret datasets with sparse features representing unknown biomass categories that may be encountered in real world exploration beyond earth. Finally, the team experimented and established the viability of using RGB and simulation to enhance the modeling and synthetic data production for simulating unforeseen condition of data acquisition. In conclusion, within the territorial context, the outcome included a new direction to push the acoustic-based method to quantify biomass for oceanographic mapping and for fisheries, which are a key component for enabling a new generation of methods to model the food chain and the climate from the vantage point of the arctic sea environment. In the extraterrestrial context, it provided a new unified approach around acoustic sensing for both autonomous robotic navigation and organism quantification in environments with ice and ocean

Ocean worlds are a programme pertaining to exploring the oceans of our solar system, not only earth. To do so we seek synergies between robotics and communication required to explore both the earth and ocean worlds in space. Communication through ice as a medium in a vertical column is impossible with current technologies requiring pushing what is currently state of the art. Furthermore, Autonomous Underwater Vehicles does not have enough resolutions in its forward facing sensors to detect small hazards, nor sensors for penetrating ice. Thus, this project will contribute multifaceted research benefiting the robotics-, artificial intelligence- and autonomy communities. We hope that these systems can also be used to help bathymetry mapping and stock estimation in the future. The research in this project will make use of existing datasets gathered from RCN projects such as SaaS (Sonar as a Service), where we will devise a hybrid analytic system based on existing data so that we can improve upon the signal processing in acoustic sensors regarding biomass estimation and detection of organic life. We will use field testing at Svalbard as a strategic platform for future space missions, but also as an instrumented site for robotics in icy environments. Furthermore, we seek close synergies between the large Norwegian commitment to ocean research in Trondheimsfjorden through Ocean Space Centre and the lab facilities at Mustad Autoline. This will allow us to perform near mission situation in controlled environments, where we contribute novel technologies for exploring and communicating in the different scenarios. The results will be disseminated through PhDs, conferences, scientific papers and as commercial hardware and software