International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 07 Issue: 09 | Sep 2020 www.irjet.net p-ISSN: 2395-0072 © 2020, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 3911 A Man-Portable Hybrid Autonomous Underwater Vehicle for Antarctic Exploration A Cadena 1 , S Vera 2 1,2 Faculty of Systems and Telecomunicación, Universidad Estatal Península de Santa Elena (UPSE), La Libertad, Ecuador ----------------------------------------------------------------------***--------------------------------------------------------------------- Abstract - The present work describes the development of a Hybrid Autonomous Underwater Vehicle (HAUV) called “Spondylus” that combines the best characteristics of Autonomous Underwater Vehicles (AUV) and Remotely Operated Vehicle (ROV), good hydrodynamics and stability in the water column to perform some tasks of a ROV like get close-range images of specific objects and collect biological samples from the sea floor. The HAUV design is focused on portability for remote field deployments. The HAUV has got Inertial Navigation System (INS), Computer Vision and Artificial Intelligence algorithms implemented on FPGA and ARM processor development boards. The routines are coded by VHDL modules, assembler and C/C++ scripts running on a Linux embedded System. HAUV Spondylus has been deployed in Antarctica at the Ecuadorian Scientific Station Pedro Vicente Maldonado, austral summer 2018-2019. Results from laboratory and sea trials are shown. Key Words: Autonomous Underwater Vehicle, Inertial Navigation System, Computer Vision, Convolutional Neural Networks, FPGA, System-on-Chip (SoC), Antarctica. 1. INTRODUCTION Underwater exploration in Antarctica is a challenging task due the extreme weather conditions and required logistic support. Autonomous Underwater Vehicle (AUV) and Remotely Operated Vehicle (ROV) have been used for oceanographic, biological and geological applications at polar environments [1], [2], [3], [4], [5]. One of the most important fields of study in Antarctica is benthic ecosystems, including their response of near-shore benthos to increasing water temperatures due climate change. Usually scuba divers are employed to get underwater photography and samples of benthic communities on the sea bed [6]. Typically the divers are deployed from small vessels or zodiac boats (inflatable boat) [7]. Scuba divers have got some operational limitations in Antarctica, diving time less than 30 min and immersion depths up to 40 m in water temperature near to 0 o C. Also there is the risk of suddenly changes of meteorological conditions that could implies dealing with winds of 60 knots at sea on a small vessel, compromising the safety of de crew and the divers [8]. Unmanned Underwater Vehicles (UUV), frequently ROV, are employed to sample benthos without scuba divers operational constrains. The vehicle is tethered, a live signal video is send from the installed cameras in the ROV and an operator drives the vehicle and decides with determined scientific criteria what areas require close field imagery and what specimens must be collected from the sea floor [9]. Explore deep-sea benthos in Antarctica with ROV requires a large logistic support: an oceanographic vessel with a five-ton winch for the umbilical cable whose operational cost could be around USD 50000 per day [10]. At the South Shetlands Island, Antarctica most of the scientific stations have got small vessels, zodiac boats and easy-access to deep-sea waters, below 1000 m, but is not possible to deploy large ROVs that can reach deeps more than 1000 m from a zodiac boat. A more affordable alternative are AUVs. They don’t depend on a physical connection to the surface vessel and can carry out a mission below the sea surface without human intervention. Man- portable AUV has got a size less than 2 m and weighted around 30-50 Kg. This kind of underwater vehicle can easily be supported by zodiac boats [11], [12], [13]. However perform the same mission of a ROV, taking close images of a scientific site of interest and collect samples with a robotic arm fully autonomously is a complex and computationally costly task that rely on artificial intelligence, machine learning and other adaptive techniques [14],[15]. Hybrid Autonomous Underwater Vehicle (HAUV) combines the best characteristic of ROV and AUV, high stability in the water column and good hydrodynamics [16]. HAUV can hover in the water column very precisely to get close images and collect samples from the sea floor [17], [18]. Performing complex task in underwater environment requires precise localization, attitude estimation and path tracking for guidance and control. Global Navigation Satellite System (GNSS) signals don’t propagate trough the sea water. The main navigational aid for AUV is the Inertial Navigation System (INS), commonly strapdown INS where the gyroscope and accelerometers are rigidly mounted on the vehicle frame [19]. The INS is complemented by other sensors like sonar, pressure sensor, cameras and Doppler velocity logs [20], [21]. All sources of information for the INS are fused by Kalman Filter (KF) techniques to estimate the velocity, position and orientation. KF usage implies high computational costs using arithmetic floating point, matrix calculations and real time operation [22]. Navigation very close to the sea floor requires a Computer Vision (CV) system to detect and identify obstacles [23]. The images could come from optical cameras and acoustic devices like side-scan sonar [24]. Usually SLAM algorithms are used for this kind of underwater applications [25]. CV and INS can work together to complement each other, updating navigation state vector and generating 3D images [26], [27]. Real time underwater object identification has a great research value for marine