| makobot collecting data on Moku o Loʻe (Coconut Island), Kāneʻohe Bay, Oʻahu. |

I will never cease to be amazed at the ability of a system that has been working without bugs for six months, tested extensively both in the lab and locally in the field, to suddenly acquire a kernel bug sometime during a six-hour flight to the middle of the Pacific which causes the onboard Ethernet to stop working and the vehicle housings to have to be opened so that the Ethernet ports can be physically accessed to figure out what the hell went wrong. Please, someone explain this to me. I’m baffled.

I blame the ODROIDs. Everything is that stupid computer’s fault.

| cable management is 80% of what I do on any given day. photo by Drew Fulton. |

Anyway. After a summer spent field testing my modified Blue Robotics BlueROV2—affectionately named Makobot after a mako shark—around San Diego, I was at the Hawaiʻi Institute of Marine Biology (HIMB) to collect some datasets on the coral reefs around the island. HIMB is located in Kaneohe Bay on the northern side of Oʻahu, on an island called Moku o Loʻe which is also known as Coconut Island. It’s a beautiful little island that is accessible by a quick boat ride from an unassuming little dock on the main island, and is mostly covered by native jungle interspersed with research facilities and dormitories. There were larged fenced-off lagoons with baby scalloped hammerhead sharks swimming around, researchers feeding microplastics to corals to see how it affects their health, aquariums full of the weirdest sea cucumbers you could possibly imagine. It’s an awesome place.

| happy hammerhead. |

My PhD work is on improving Simultaneous Localization and Mapping (SLAM) in underwater environments, specifically for mapping coral reefs to better understand and protect reef ecosystems. To do this the robot has a pair of downward-facing stereo cameras, which it uses to track features in the environment to figure out its motion and position relative to the environment (localization), while also building up a map of where those features are in the environment (mapping).

| video showing orbslam2 running on the makobot. |

The short video above shows an example of how this works, using the coral reef data collected at HIMB. The video shows the robot using the ORBSLAM2 software to perform localization and mapping, which is a SLAM software that’s widely-used across the robotics community. However, it has many performance deficiencies in the underwater environment, so my task is to build a more robust system that works better with some of the challenges working underwater adds, like less robust features to track, moving scene elements, poor or variable lighting, and low-texture areas.

| flythroughs of 3D models created of coral reefs at HIMB. many thanks to Eric Lo for processing the models in Agisoft and creating the flythroughs of corals in space. |

This video shows some high-resolution 3D models of a few of the coral reefs we mapped at HIMB. These models were created using an off-the-shelf software called Agisoft Photoscan using images from a GoPro camera mounted on the robot. While not really the main focus of the research, they are fun to look at.

| makobot swimming around collecting data. good robot. |

Despite a day of frustration caused by the aforementioned kernel bug and some poor weather and water visibility caused by a hurricane that was passing south of the islands, we had a very successful eight days on the island collecting several datasets and enjoying the island ambiance. Many pineapples were consumed. I will be replacing the ODROIDs immediately upon my return to San Diego.

| honu, mākua beach. |

After this short field trip, I put Makobot on a ship back to the mainland, then headed off to spend a few days camping and snorkeling around Oʻahu as I waited to embark upon an even bigger adventure…