I earned an M.S. degree in Robotics at Oregon State University under Dr. Geoffrey Hollinger in his Robotic Decision Making Laboratory (RDML). While there, my specialization was in marine robotics, wave estimation, and underwater model predictive control.

Along with the RDML, I also supported the OSU College of Earth, Ocean, and Atmospheric Sciences (CEOAS) Glider Research Group and Ocean Observing Initiative under Dr. Jack Barth and Dr. Kipp Shearman. See below for some of the platforms I worked with and some of the RDML research grants that supported me.

Vehicles

Some of the marine platforms I've worked with:

seabotix

SeaBotix vLBV300 ROV

The workhorse of the RDML, Volternus is the lab vehicle most suited to implementing robotic research in the field. It is a tethered ROV actuated by six 100mm brushless DC thrusters and controlled in five degrees of freedom.

  • Payload: 10kg
  • Depth Rating: 300m
  • Low light color camera with 180° Tilt

Webb Research Slocum Glider

slocum

This fully autonomous electric glider uses only changes in density, and internal battery position to "glide" through the water. The result of such low wattage propulsion is a mission profile of about one month or more depending on sensor payload.

  • Depth Rating: 200-1000m
  • Average Speed: 35cm/s
  • Maximum Range: 1500km
  • Sensors: CTD, dissolved O2, ADCP, Turbidity

openrov

  openROV

openROV is an experimental, low-cost underwater robot. Its software was developed completely open source, and the appeal is implementing research applications despite its lesser functionality. It has three thrusters, a 1080p HD camera, LED lights, and is powered by a BeagleBone Black with linux onboard.

University of Washington Seaglider

seaglider

Another glider used by Oregon State CEOAS to collect oceanographic data. Though not yet as developed as the Slocum, SeaGlider's rudderless design and significantly longer mission profile has secured it as a rising star in the glider industry.

  • Depth Rating: 1000m
  • Average Speed: 25cm/s
  • Maximum Range: 4600km
  • Data transmitted after each dive over iridium
  • Sensors: CTD, dissolved O2, Backscatter

Research Supported

Some of the RDML research I supported:

Department of Energy (DOE) Wave Energy: This project is in coordination with the National Northwest Marine Renewable Energy Center (NNMREC) which aims to deploy several Wave Energy Converters (WEC) off the Oregon Coast over the next five years. Our lab is exploring using autonomy to reduce the high cost of maintaining these Energy Arrays. The goal is to be able to use an AUV to inspect infrastructure and manipulate the environment, perhaps by: turning a valve, plugging a wet connector, or even tying a knot to a mooring on the seafloor. All of these tasks currently require a dive team or ROV team to complete and by introducing autonomy would greatly reduce operational costs.

Wave-Predicting Station-Keeper: One of the early tasks when tackling the robotic inspection problem is proper station-keeping. This involves using the vehicle's sensors in unison for effective localization and then controlling the output so that the vehicle remains stationary. Added to this problem is the Pacific Ocean's unchecked wave action. What I am exploring is how to predict an incoming wave field, so that way a controller could anticipate the external forces acting upon the vehicle. To do this, I'm looking at how to apply a neural network neoro-evolutionary learning method to estimate the hydrodynamic properties throughout the water column. This work is funded by the DOE grant listed above.

Office of Naval Research (ONR): Here, we are exploring the decision making benefits of exploration vs. exploitation. This is an active area of research in the path planning community, be it marine or otherwise. The fundamental problem -- when is the best time to act? In the decision making realm, more information gain adds to a robot's observation of its environment, allowing for a more informed decision. Of course, observations come at a cost. The Partially Observed Markov Decision Process (POMDP) seeks to optimize the robot's next action while minimizing computation time and/or cost.

W. M. Keck Foundation Grant: This is a cross-disciplinary research project which aims to install bioacoustic sensors to the OSU College of Earth, Ocean, and Atmospheric Sciences (CEOAS) Webb Slocum Gliders. These sensors are used to detect macrofauna in the coastal ocean. These can be anything from fish, to seabirds, to whales and other large mammals. The team is composed of an oceanographic group for the glider operations, a marine biology group for the actual sensor data, and an RDML group which will handle optimizing the glider's path planning.

Precision Castparts Corporation (PCC): PCC is a local industrial partner which has contracted the RDML to apply Branch & Bound optimization techniques for their Investment Casting process. To do this, we have built a simulated assembly line similar to that used by PCC which outputs an optimized work schedule. This outlines when to begin assembly on each manual loading station.