This project envisions the research and development of SENTIENT: Science of resiliENt auTonomy In pErceptually-degraded eNvironmenTs as the new methods and systems that can enable micro aerial robots to resiliently explore, map and characterize industrial environments such as ship ballast tanks or oil & gas facilities. To achieve this goal, the envisioned research first aims to revolutionize how robots perceive their environment, localize within it and map it in detail especially when the conditions are challenging or the environment itself is moving. Particular attention is paid on how diverse and complementary sensors such as cameras, thermal vision and light detection and ranging can be best combined towards one unified way for robot perception and cognition. Building on top of this essential capability, SENTIENT proceeds to identify new ways to enable small and micro aerial robots to plan their trajectories and actions such that they can autonomously map fully and characterize precisely complex industrial environments. To deliver systems that are robust in all conditions we seek to investigate and develop a new science of resilient autonomy for robots that go out of laboratory environments and are deployed in the wild operating in diverse and challenging environments. This new science of resilience will allow to deliver robots that are survivable and can autonomously execute challenging inspection missions in degraded and high-risk environments with precision and completeness without the need for teleoperation or other skilled training. A university-industry partnership is formed to further enable the transition of the new research contributions to technological solutions enabling reliable inspection of ship ballast water tanks, cargo tanks, as well as oil & gas industry tanks and process vessels. Early progress in the project has allowed the development of specialized micro-sized flying robots for autonomous ballast water tank navigation, alongside equipping a small aerial robot with the ability to reconstruct the 3D map of its environment.
This project envisions the research and development of "SENTIENT: Science of resiliENt auTonomy In pErceptually-degraded eNvironmenTs" as the new theory, methodological tools, and field experiments-verified system realization that give rise to a new generation of cognizant small (3kg) and micro (0.25kg) aerial robots capable of resilient autonomy, versatile exploration and inspection inside challenging, GPS-denied, visually-degraded, geometrically complex and dynamic environments such as vessel ballast tanks, cargo tanks, and oil & gas facilities. Motivated by the core hypothesis that there exists a comprehensive science of resilient autonomy for aerial robots seamlessly operating in challenging, high-risk and degraded environments, the envisioned research will be holistically organized around three cross-cutting objectives. In particular, SENTIENT will research on a) designing resilient collision-tolerant aerial robots, b) redundant, resourceful and robust robotic perception for autonomous localization and scene understanding through the fusion of diverse sensing technologies, and c) cognizant informative path planning and machine learning-based navigation.
Capable of resilient autonomy in challenging settings, the SENTIENT robots are driven by important needs of the maritime and energy industry. The project partnership involves the collaboration of NTNU with Scout Drone Inspection (SDI), DNV GL, ALTERA and Equinor, while extensive real-life evaluations in ship ballast tanks, cargo tanks, and oil & gas installations are planned. Capitalizing on the importance of the scientific and application domain, the team has also established collaboration with NASA JPL, UC Berkeley and ETH Zurich. Overall, SENTIENT pushes the frontier in scientific research in robotics and artificial intelligence, while simultaneously building vital competence in industry-needed domains paving the way to offer new opportunities and societal benefits both at short- and long-term horizons.