UNH Ocean Seminar

2021 Student Research Update

Vivek Bheeroo, Coral Moreno, Tamer Nada
UNH Ph.D. Students
Friday, Sep. 10, 2021, 3:10pm
Chase 105

Connecting Subsurface Acoustic Fields to Free-Surface Observations in a Multiphase Environment

The aim of this research is to develop a connection between free-surface perturbations and a predictive capability of the acoustic field below. In particular, we are interested in the coherent flows generated within seagrass beds and similar benthic environments. I will present preliminary work on characterizing the free-surface flows. The fundamental findings from this research will improve our ability to monitor dynamic nearshore environments safely, efficiently, and cost-effectively through remote measurements.

Towards Vision-Based Navigation of an Unmanned Surface Vehicle Using Deep Reinforcement Learning

With the increasing interest in using unmanned surface vehicles (USVs) for coastal mapping, there is an increased need to improve USV autonomy and safety of operation due to the challenges imposed by common presence of obstacles, and the complexity of coastal environments. USVs may use various sensors for situational awareness: AIS, radar, LiDAR, and cameras. The sensors are complementary in nature; no single sensor provides a complete solution. Camera systems are the "eyes" of the USV and provide overlapping functionality between long and close detection range regimes.

We tackle the problem of vision-based navigation using deep reinforcement learning (DRL). DRL has proven to be a promising framework for learning-based control of complex behaviors in challenging scenarios from high dimensional raw observations (such as images) as an end-to-end solution. However, due to the challenging nature of DRL algorithms, their applications have been mostly limited to video games and simulated control tasks. The goal is to train a DRL agent, such as a deep Q-network (DQN), to react rapidly to avoid an obstacle in the USV's path based on imagery input from the USV's camera. A Gazebo-ROS simulation was wrapped with OpenAI Gym to make a setup that is suitable for both simulations and real robots, and a DQN agent that was implemented in Tensorflow for training and testing. This talk shows the steps towards a proof of concept of vision-based autonomous navigation of USV using a DQN in simulated maritime environments, as well as a review of the efforts towards implementation with a real USV.

Towards Automated Compilation of Electronic Navigational Charts

Current methods of Electronic Navigational charts production are still strongly human interactive. Significant efforts are needed for distributing, updating, maintaining and consistency checking those ENCs, which heavily impact the efficient generation of the product. The ideal situation is to automate the process and to have charting products that can be generated at the right scale for the user's current situation, and at the point of use. Automated methods for chart compilation are therefore expected to significantly improve the process in terms of speed, scalability, consistency, and quality. However, a number of constraints (e.g., safety of navigation) limits the idea of automating the process. Therefore, the research effort aims to understand, define, document, parametrize, and simulate the compilation process as a prelude to more automated solutions.


Vivek Bheeroo is a 2nd year PhD student in Ocean Engineering, working alongside Prof. Tracy Mandel. His research focuses on submerged vegetative flow and the resulting surface expressions. He earned his M.S. in Coastal & Ocean Engineering from Oregon State University, where he studied long wave activity in a coral-reef lagoon. Prior to that, he earned his B.S. in Civil Engineering at Purdue University. Vivek is from the Island of Mauritius.

Coral Moreno is a doctoral student in Ocean Engineering at the Center for Coastal and Ocean Mapping (CCOM) at the University of New Hampshire. Her research explores perception and navigation solutions for unmanned surface vehicles (USVs) to improve the safety of autonomous seafloor mapping operations. Currently, her research incorporates deep reinforcement learning to develop end-to-end control algorithms that learn directly from sensory data, such as a camera, to navigate a USV through marine environments. Coral earned a B.Sc. in Physics with a specialization in Optics in 2009, and an M.Sc. in Autonomous Systems and Robotics Engineering in 2014 from the Technion - Israel Institute of Technology. During 2015, she divided her time between working as a surveyor and a remotely operated vehicle (ROV) operator with EDT Marine Construction/Marteam LTD in Israel, and being a guest investigator at the Optical Communication Laboratory at Woods Hole Oceanographic Institute (WHOI), until the beginning of her doctorate in 2016. Coral completed the Category A hydrographer training at CCOM. Since 2015 she has participated in various scientific expeditions gaining experience in operating marine robotic systems (ROV and USV) and in seafloor mapping from manned vessels or using a USV. The sea has been an inseparable part of Coral's life, and naturally her love of science, technology, and the sea has led her to be interested in developing marine systems in support of various ocean exploration and marine conservation.

Tamer Nada earned his B.Sc. in Naval Science from the Egyptian Naval College, served as a chief officer in the Coast Guard brigade, then as a Captain of a fast patrol gun boat in the Egyptian Fleet. He started his Hydrographic career by earning his (Cat-B) IHO hydrographic course with honors from the Naval Oceanographic Office in the UK, then participated in numerous survey operations as a Hydrographic surveyor in the Egyptian Hydrographic Department. Following the promotion to a Commander in 2009, he completed his first master's degree in hydrographic surveying from the Egyptian Academy for Science & Technology with the trophy for best thesis in Seismic surveying. Thereafter, in 2010 he earned his second Master of Science from the University of Southern Mississippi while completing his long hydrographic course (Cat-A) with the US Navy. In the Egyptian Hydrographic Department, his last post was the chief of the Egyptian Hydrographic division and was responsible for the production of the Egyptian Electronic Navigation Charts (ENCs). In 2017, he has retired as a Captain from the Egyptian Navy, then worked as a freelancer Geophysical surveyor for FUGRO SAE. In 2018 he started his PhD in the University of New Hampshire and CCOM. His point of research is for a fully automated nautical cartographic solution.